Quality, Patient Safety & Paying for Performance Call for Papers Leaders & Outliers: How Should We Measure Quality & Safety? Chair: Carol Haraden, Institute for Healthcare Improvement Sunday, June 26 • 8:30 am – 10:00 am ●Development of the Medicare Patient Safety Monitoring System Susan L. Abend, M.D., FACP, David R. Hunt, M.D., FACS, Nancy R. Verzier, MSN, RN, Nancy Safer, RN, Janet P. Tate, MPH, Gaston Mbateng, Ph.D. Presented By: Susan L. Abend, M.D., FACP, Qualidigm, 100 Roscommon Drive, Middletown, CT 06457; Tel: (860)6326395; Fax: (860)632-6326; Email: sabend@pol.net Research Objective: Since the Harvard Medical Practice Study 15 years ago, numerous studies have assessed the magnitude of patient injuries occurring in US hospitals, and have estimated the impact of adverse events on health care costs and patient longevity. Assessing the true magnitude and impact of these unintended events has been complicated by variations in study design, event definitions, and data sources. The larger studies use clinician opinion to confirm the presence of an adverse event, (a method demonstrated to have low inter-rater reliability, with correlation coefficients of 0.24-0.34) or use administrative data to identify diagnoses presumed to be unintended consequences of care. The Medicare Patient Safety Monitoring System (MPSMS), developed by CMS under the leadership of the DHHS Patient Safety Task Force, is a cost-effective, standardized surveillance method for assessing hospital-related adverse event rates over time in the Medicare population. The system will be used to provide baseline information for CMS quality improvement initiatives such as the Surgical Care Improvement Project, and to provide annual data to the National Quality Report. The MPSMS can also be used as a tool for hospitals to measure and test changes that create safer therapeutic environments. Study Design: The MPSMS team, supported by Qualidigm (the Connecticut QIO), develops standard definitions of hospital-related adverse events in close collaboration with a Technical Expert Panel (TEP) comprised of major public and private stakeholders, academic experts, and representatives of AHRQ, CDC, FDA and VA. Data elements present in medical records and/or Medicare administrative data, indicating the presence of the adverse event and a closely-associated, antecedent healthcare exposure, are then identified. Adverse events are chosen from AHRQ- developed patient safety measures, FDA reporting systems, CDC recommendations, and information from peer-reviewed literature. Events must meet the following criteria: 1) having easily-acquired data elements, 2) clear association with specific healthcare exposure, 3) high frequency, 4) high likelihood of morbidity or mortality, 5) and potential for preventability, are chosen for development. Algorithms are then created that indicate the presence or absence of a specific event and antecedent healthcare exposure. Algorithms are rigorously tested in an alpha phase for data element findability and ambiguity, in a beta phase for accuracy and interrater reliability, and are then reviewed for face-validity by clinicians from across the country. Population Studied: A random sample of 25,000- 40,000 Medicare fee-for-service patients discharged from general acute-care hospitals are studied annually. The data sources are charts reviewed for Medicare’s Hospital Payment Monitoring Program, as well as related Medicare Part A and B claims data. Eighty charts per month are reabstracted to check reliability. Principal Findings: The kappa statistic, (a measure of reliability that corrects for chance agreement), from variables collected in production, year 1, was 0.94 (95% CI 0.92-0.96). Accuracy of data abstraction against gold standard records was 99.2 + 0.1%. Mean cost of abstracting 10 production measures was $74/record. Conclusions: The MPSMS, the product of a robust publicprivate collaboration, is a reliable, cost-effective tool for measuring adverse events in hospitalized patients, and is the first system able to provide rates of specific, U.S.-hospitalrelated adverse events over time. Implications for Policy, Delivery, or Practice: Because it is standardized and has excellent inter-rater reliability, the MPSMS can be used as a tool by hospitals to assess performance and test process redesign initiatives. While currently designed to find data from conventional medical charts, its data elements can be easily organized into modules within Electronic Health Records. It can be used to assess changes retrospectively, or data elements can be noted prospectively to trap errors or change processes leading to an adverse event. Primary Funding Source: CMS ●Comparative Rankings of Hospital Quality – Does the Data Source Matter? Anne Elixhauser, Ph.D., Bernard Friedman, Ph.D. Presented By: Anne Elixhauser, Ph.D., Senior Research Scientist, Center for Delivery, Organization, and Markets, Agency for Healthcare Research and Quality, 540 Gaither Road, Rockville, MD 20850; Tel: (301)427-1411; Fax: (301)4271430; Email: aelixhau@ahrq.gov Research Objective: To assess the extent to which information on Medicare discharges can be extrapolated to the general inpatient population when comparing hospital quality. Study Design: Cross-sectional analysis using hospital administrative data from the 2001 Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample (NIS), a sample of 986 hospitals from 33 states. Measures were drawn from the AHRQ Quality Indicators (QIs), which use hospital administrative data: 15 Patient Safety Indicators (PSIs) and 12 inhospital mortality measures. Risk adjustment for PSIs used gender, age, comorbidities, and modified DRGs (collapsed across the comorbidity/complication splits); for mortality we used APR-DRGs. To avoid the influence of small numbers of discharges on rankings, we dropped hospitals in the bottom decile based on the frequency of Medicare cases for each measure. For each measure, using Medicare discharges only, we ranked hospitals, ordered them into deciles, and assessed the percentage of hospitals that changed from the highest or lowest rank by at least two deciles when using all discharges. Population Studied: Hospital inpatients from short-term, non-Federal, acute care hospitals. Principal Findings: PSIs: For hospitals in the first decile (i.e., lowest rate of patient safety events) using Medicare discharges, 40% were in the third decile or lower using all discharges for four conditions: postoperative hip fracture, physiologic/metabolic derangement, respiratory failure, and wound dehiscence. About a third of the first decile hospitals based on Medicare patients were in the 3rd decile or lower based on all discharges for: anesthesia complications, death in low mortality DRGS, and foreign body left after procedure. For five PSIs, about one quarter of the first decile hospitals based on Medicare discharges were in the third decile or lower with all patients. Fewer than 20% of the first decile hospitals fell to the third decile or lower for the remaining three PSIs. Improvements in rank were less dramatic than declines. For all but three indicators, fewer than 10% of lowest decile hospitals moved two deciles or more to better relative performance. Inhospital mortality: For one mortality measure (craniotomy), over 30 percent of first decile hospitals based on Medicare patients were in the third decile or lower for all patients. For three of the inhospital mortality measures, 1219% of the first decile hospitals based on Medicare discharges were in the third decile or lower for all payers. For all other inhospital mortality measures, fewer than 10% of first decile hospitals shifted ranking. Improvement by at least two deciles from the worst inhospital mortality ranking was seen for CABG (21.1% of hospitals), craniotomy (24.0%), PTCA (14.3%), and AAA repair (10.3%). Conclusions: We found loose overlap of the top-ranked and worst-ranked hospitals across a number of measures of quality when comparing Medicare-only and all-payer analyses. Depending on the measures used, comparative rankings of hospitals could vary for different populations of discharges. Implications for Policy, Delivery, or Practice: It is critical to pre-test various ranking approaches before they are used to inform patients and clinicians, or to reward high performing hospitals. Primary Funding Source: No Funding ●The Role of Information in Medical Markets: An Analysis of Publicly Reported Outcomes in Cardiac Surgery Mary Beth Landrum, Ph.D., Robert Huckman, Ph.D., David Cutler, Ph.D. Presented By: Mary Beth Landrum, Ph.D., Assistant Professor of Biostatistics, Health Care Policy, Harvard Medical School, 180 Longwood Avenue, Boston, MA 02115; Tel: (617)432-2460; Email: landrum@hcp.med.harvard.edu Research Objective: Increased provider profiling has been urged by many as a fundamental step in medical care reform. Despite the building momentum for this strategy, the underlying premise of profiling – that disseminating information about provider quality will lead to improved quality of care – has yet to be established. We analyze the effect of public reporting using data from the longest-standing effort to measure and report health care quality the Cardiac Surgery Reporting System (CSRS) in New York State. Study Design: We consider how report cards affect the behavior of individual providers, specifically providers who are publicly identified as being significantly better or worse than their peers. Our analysis addresses three intended goals of provider profiling. First, we examine whether report cards lead poor-performing providers to improve their performance relative to their peers, adjusted for changes in patient severity. Second, we consider whether report cards lead to increased volume for highly rated providers and lower volume for poorly rated providers. Finally, we consider how report cards influence the allocation of patients to particular physicians, in particular whether report cards lead to more severe patients being operated on by higher quality physicians. We use longitudinal regression models to examine patterns in volume, average severity and mortality following surgery. To examine how behavior changes after a hospital is publicly labeled as better or worse than their peers, we incorporate a change-point in the model of provider-level outcomes over time. These models hypothesize that a sharper change in hospital behaviors is observed for hospitals publicly labeled as unusual, compared to the general secular trends occurring in hospitals. We compare fixed and random effect models and also examine the use of propensity score models to match hospitals and examine differential trends in the matched samples. Population Studied: Over 100,000 patients who underwent coronary artery bypass surgery (CABG) in one of 34 hospitals in New York State over a 9-year period (1991-1999). Principal Findings: We find no evidence that patient severity changed following a good or bad rating. We find some evidence that public reporting affected both the volume of cases and future quality at hospitals identified as poor performers. Poor performing hospitals lost relatively healthy patients to competing facilities and experienced subsequent improvements in their performance as measured by riskadjusted mortality. However, these results were sensitive to model specification. Conclusions: We show that public reporting of quality information can change provider behavior. We also conclude that researchers need to be careful about how comparison groups are formed for analyses such as these. Implications for Policy, Delivery, or Practice: Understanding the true effects of report cards on access and quality of care becomes increasingly important as more states and other organizations publicly release quality report cards for an increasing number of conditions and as insurers start to use provider-specific quality in determining appropriate reimbursement. Primary Funding Source: NIA ●The Summary Quality Index (SQuId): A Summary Measure for Multiple Quality Indicators in Primary Care Paul Nietert, Ph.D., Ruth G. Jenkins, MS, Andrea M. Wessell, PharmD, Sarah T. Corley, M.D., Steven M. Ornstein, M.D. Presented By: Paul Nietert, Ph.D., Assistant Professor, Biostatistics, Bioinformatics, and Epidemiology, Medical University of South Carolina, 135 Cannon Street, Suite 403, Charleston, SC 29425; Tel: (843)876-1204; Fax: (843)876-1201; Email: nieterpj@musc.edu Research Objective: Primary care quality indicators are helpful tools for translating research into clinical practice. Providing feedback to clinicians on these indicators has been shown to help them make system changes that improve the quality of the care they provide. Currently our research team is providing a geographically diverse group of clinicians with quarterly reports on 78 different quality indicators. All of the health care providers use a common electronic health record, which facilitates data extraction and feedback. We wanted to develop a single composite summary measure that would help the research team evaluate the overall progress of the practices’ quality improvement efforts. The summary measure needed to be both statistically robust and clinically meaningful. Study Design: A total of 92 primary care practices are being followed prospectively from Oct 2002 through Sept 2006. All practices receive an intervention aimed at improving the quality of care provided in the following domains: prevention and management of hypertension, heart disease, stroke, diabetes, and respiratory/infectious disease; cancer screening; immunizations; substance abuse and mental health; nutrition and obesity; and inappropriate prescribing in elderly patients. A summary quality measure was created and calculated for each patient, and patient summary scores were averaged to obtain an overall summary quality index (SQuId) for each practice. The practice level SQuId was tested for normality using a Shapiro-Wilk test. Population Studied: The population studied included over 350,000 patients in these 92 primary care practices. Principal Findings: On average, the patient-level quality measure rose from 25.0% to 30.3% during the first 2 years of the study. A large majority (84.7%) of the practices demonstrated improvement in their SQuId over time. The SQuId was normally distributed across the practices and would serve as an excellent dependent variable in any regression model. Early feedback from clinicians indicates that the SQuId is meaningful and informative. Conclusions: The SQuId is an effective tool to aid in the evaluation of multiple quality indicators. Implications for Policy, Delivery, or Practice: In future research studies and quality improvement efforts that include multiple quality indicators, the SQuId should be considered as a measure of overall quality. Primary Funding Source: AHRQ ●Validating AHRQ PSIs with NSQIP Postoperative Adverse Events Dennis Tsilimingras, M.D., MPH, Patrick S Romano, M.D., MPH, Shukri F Khuri, M.D., William G Henderson, Ph.D., Anne Elixhauser, Ph.D., Amy K Rosen, Ph.D. Presented By: Dennis Tsilimingras, M.D., MPH, Research Health Scientist, Health Services, Bedford VA Hospital, 200 Springs Road, Bedford, MA 01730; Tel: (781) 687-2938; Fax: (781) 687-3106; Email: dtsilimi@bu.edu Research Objective: The Agency for Healthcare Research and Quality (AHRQ) Patient Safety Indicators (PSIs) were developed to identify potential inpatient adverse events using administrative (billing) data. We sought to (1) validate the AHRQ PSIs with postoperative adverse events in the National Surgical Quality Improvement Program (NSQIP), which employs trained nurse data collectors to prospectively collect preoperative patient characteristics and 30-day postoperative outcomes on most major operations in Department of Veterans Affairs (VA) hospitals; and (2) test alternative PSI definitions based on clinical information from NSQIP for the purpose of future refinement of the PSIs. The overall study examines five AHRQ PSIs. As an example, we present results regarding the validation of AHRQ’s “Postoperative Respiratory Failure” definition with NSQIP’s measures of “Unplanned Intubation for Respiratory Failure or On Ventilator >48 hours Postoperatively”. Study Design: VA administrative records of inpatient hospitalizations (from the Patient Treatment File) with valid operating room procedures, excluding cases admitted with respiratory/pulmonary disease, were matched with NSQIP chart-abstracted records based on patient identifiers, facility codes, and dates of service. A hospitalization had a PSI if it met AHRQ’s definition that includes ICD-9-CM codes 518.81 (acute respiratory failure) or 518.84 (acute and chronic respiratory failure). It had a NSQIP event if it met NSQIP’s clinical definition (unplanned intubation or on ventilator >48 hours postoperatively). Using NSQIP as the “gold standard”, we measured the sensitivity and positive predictive value of the PSI using both the current AHRQ PSI definition and alternative ICD-9-CM based definitions. Population Studied: Inpatient discharges from 130 VA hospitals from 10/01/00 through 9/30/01. Principal Findings: From 561,229 inpatient hospitalizations in the Patient Treatment Files, 56,419 were matched with NSQIP data to identify 24,284 elective surgical discharges with valid operating room procedures. From the 24,284 hospitalizations used for the analysis: 1) the sensitivity and positive predictive value for “unplanned intubation or on ventilator >48 hours postoperatively” using AHRQ’s definition were 17% and 74%, respectively; and 2) the sensitivity and positive predictive value for “unplanned intubation or on ventilator >48 hours postoperatively” using NSQIP’s clinical definition based on a combination of ICD-9-CM procedure (codes for intubation/on ventilator) and diagnostic codes were 63% and 66%, respectively. Conclusions: We found a higher sensitivity and a similar positive predictive value when using a modified version of AHRQ’s “Postoperative Respiratory Failure” PSI that includes both diagnostic and procedures codes, rather than only diagnostic codes. This suggests that combining procedure with diagnostic codes may enhance the validity of administrative-data-based indicators, such as the PSIs, as identifiers of potential surgical patient safety events. Implications for Policy, Delivery, or Practice: An efficient screening tool for identifying inpatient safety events (that warrant further investigation) can be constructed when using clinical definitions that incorporate procedure and diagnostic codes. The next release of the AHRQ PSI software will incorporate the findings of this study. Primary Funding Source: VA Call for Papers Crowded House: Crowding, Staffing & Patient Safety Chair: Marie Bismark, Harvard School of Public Health Sunday, June 26 • 10:30 am – 12:00 pm ●A Cluster Randomised Controlled Trial of the Medical Emergency Team System Ken Hillman, Jack Chen, Michelle Cretikos, Rinaldo Bellomo, Daniel Brown, Gordon Doig, Simon Finfer, Arthas Flabouris Presented By: Jack Chen, MBBS, Ph.D., Senior Research Fellow, The University of New South Wales, Simpson Centre for Health Services Research, Locked Bag 7103, Liverpool BC/NSW 1871/Australia, Sydney, 1871; Tel: 612-98286662; Fax: 612-98286111; Email: jackchen@unsw.edu.au Research Objective: It is well documented that patients suffering cardiac arrests or dying on general wards have often received delayed or suboptimal care. We conducted a cluster randomised controlled trial to test the effect of the Medical Emergency Team (MET) system on the incidence of cardiac arrests, unplanned ICU admissions and unexpected deaths in a group of 23 hospitals in Australia. Study Design: We randomised 23 hospitals to either continue operating as normal (n=11) or to introduce a MET system (n=12). The primary outcome was the composite outcome of cardiac arrest, unexpected death or unplanned intensive care unit admission during the six-month study period following MET activation. Principal Findings: The two groups of hospitals were comparable at baseline. From baseline to study period there was a reduction in the rate of cardiac arrests (p=0.003) and unexpected deaths (p=0.01) across both groups combined over time. Introduction of the MET increased the overall calling rate for an emergency team during the intervention period (3.1 vs. 8.7 per 1000 admissions, p< 0.001). The MET was called to 30.4% of patients who satisfied the calling criteria who were subsequently admitted to the ICU. During the study period the incidence of primary outcome was similar in the two groups (5.86 vs. 5.31 per 1000 admissions, p=0.64). Conclusions: The introduction of a MET system did not significantly affect the primary outcome but the rate of cardiac arrests and unexpected deaths decreased across both groups of hospitals during the study. Many eligible patients were not attended by the MET and further research is needed to determine whether better implementation of the MET system can improve outcomes. ●Variation in Volume-Outcome Relationships for Hospitals and Surgeons Performing CABG Surgery Christopher P. Gorton, M.D., MHSA, Jayne L. Jones, MPH, Christopher S. Hollenbeak, Ph.D., David B. Campbell, M.D., Diane L. Arke, MS Presented By: Christopher P. Gorton, M.D., MHSA, Senior Physician Consultant, Pennsylvania Health Care Cost Containment Council, 225 Market Street, Harrisburg, PA 17101; Tel: (717)232-6787; Fax: (717)232-3821; Email: kgorton@phc4.org Research Objective: To describe the variation in risk-adjusted outcomes for individual surgeons and hospitals with similar case volumes; To examine the association between case volume and two outcomes, 30-day mortality rates and 7-day readmission rates, for surgeons and hospitals; To describe trends in utilization of coronary vascular interventions as they relate to volume and outcome. Study Design: Retrospective analysis of administrative and medical record abstracted data reflecting statewide experience from Pennsylvania. Outcomes were modeled using a generalized estimating equations (GEE) framework to account for the correlations between patients treated by the same surgeons and at the same hospitals. Our main outcome measures were: Risk-adjusted thirty-day mortality rates for hospitals and for surgeons; Risk-adjusted seven-day readmission rates for hospitals and surgeons. Population Studied: Data was obtained from the 62 nonfederal Pennasylvania hospitals in which adult isolated CABG surgeries were performed during 2002. CABG surgery was performed by 190 surgeons. Outcomes from 16,435 patients were modeled. Principal Findings: Wide variation was seen between and within outcome measures at all volume levels. In looking at individual hospitals and surgeons with outcomes statistically higher or lower than expected, outliers were observed at both extremes of the volume spectrum. Hospital volume was not independently related to likelihood of patient death after accounting for surgeon volume. Likelihood of patient death within thirty days decreased as surgeon case volume increased (adjusted risk ratio [RR] for each 25 case increment, 0.93; 95% confidence interval [CI], 0.87-0.99; P=0.031). Neither hospital nor surgeon case volumes were associated with seven-day readmission rates. Conclusions: Substantial outcome variations at all provider volume levels render isolated case volume inadequate as a basis for selective referral. The identification of a significant association between case volume and only one of four studied outcome measures shows the importance of multidimensional performance measurement and the need to understand how processes of care affect outcomes. Implications for Policy, Delivery, or Practice: We are convinced that there is no simple proxy measure that will permit appropriate steering of consumers towards certain providers and away from others. We believe the correct way to assess the performance of individual surgeons and hospitals is to measure it, hospital by hospital and surgeon by surgeon. Stakeholders should focus their energies on improving measurement methodologies and standardizing measurement systems. These developments will permit sustained and systematic improvements in the quality of care that cannot be achieved with simple proxies such as surgical case volume. Primary Funding Source: PA Health Care Cost Containment Council ●Hospital Overcrowding is Associated with Increased Seven Day Emergency Admission Mortality: A New Imperative for Patient Safety Peter Sprivulis, MBBS, Ph.D., Julie-Ann Da Silva, B Sc, Ian Jacobs, RN, Ph.D., Amanda Frazer, MBBS, LLB, George Jelinek, MBBS, DM Presented By: Peter Sprivulis, MBBS, Ph.D., Harkness Fellow in Healthcare Policy, General Medicine, Brigham and Women's Hospital, Brigham Circle, Boston, MA 02215; Tel: (617)283 0076; Fax: (617)301 4848; Email: psprivulis@ihi.org Research Objective: Hospital overcrowding causes emergency department dysfunction due to impaired access to beds for emergency hospital admissions. This study examines the relationship between hospital occupancy, emergency department occupancy and emergency admission mortality. Study Design: Analysis of hospital and death records for 62,495 emergency admissions at three tertiary metropolitan hospitals between July-2000 and June-2003. Two, seven and 30 day mortality were evaluated against an Overcrowding Hazard Scale (OZS) based upon hospital and emergency department admitted patient occupancy, after adjustment for age, diagnosis, referral source, urgency and mode of transport to hospital. Population Studied: 68437 emergency hospital admissions at three tertiary metropolitan hospitals in Perth, Western Australia. Principal Findings: A linear relationship between the OZS and 7-day mortality was demonstrated (r = 0.98, 95%CI 0.791.00). An OZS>2 was associated with an increased Day-2, Day-7 and Day-30 death hazard of 1.3 (95%CI 1.1-1.6), 1.3 (95%CI 1.2-1.5) and 1.2 (95%CI 1.1-1.3) respectively. Thirty day deaths associated with an OZS>2 compared to OZS<3 were undifferentiated with respect to age, diagnosis, urgency, transport mode, referral source or hospital length of stay but had longer emergency department durations of stay (RR per hour of ED stay, 1.1, 95%CI 1.1-1.1, P < 0.001) and longer physician waiting times (RR per hour of ED wait, 1.2, 95%CI 1.1-1.3, P = 0.01). Conclusions: Hospital and emergency department overcrowding is associated with increased mortality and may represent a remediable cause of fatal iatrogenic injury. The Overcrowding Hazard Scale may be used to assess the risk to patients posed by hospital and emergency department overcrowding. Reducing overcrowding may improve outcomes for patients requiring emergency hospital admission. Implications for Policy, Delivery, or Practice: Hospitals should examine strategies to improve acute patient flow as well strategies to improve demand and capacity matching for emergency hospital admissions. Primary Funding Source: Australian Health Ministers Advisory Council Grant ●The Relation of Crowded Working Conditions to Patient Safety in Hospitals Joel Weissman, Ph.D., Jeff Rothschild, M.D., Fran Cook, Sc.D., Eran Bendavid, M.D., Melissa Bender, M.D., Peter Sprivulis, M.D., Joann David-Kasdan, RN, Jenya Kaganova, Ph.D, Les Selbovitz, M.D., Scott Evans, Ph.D., Peter Haug, Ph.D., Jim Lloyd, David Bates, M.D. Presented By: Joel Weissman, Ph.D., Associate Proffessor, Institute of Healh Policy, MGH/Harvard, 50 Staniford Street, Boston, MA 02114; Tel: (617)724-4731; Fax: (617)724-4738; Email: jweissman@partners.org Research Objective: There has been a trend towards increasing patient workload for available staff time, skill-mix and physical resources (e.g. accommodation, equipment) in U.S. hospitals. Potential impacts include hospital overcrowding, emergency department inpatient boarding, ambulance diversion, and industrial action by nursing staff to fix patient nurse work load ratios. This study examines whether workload is associated with an increased risk of adverse events (AEs) for patients by evaluating whether the daily rate of AEs vary by workload, controlling for patient complexity and nurse staffing. Study Design: Administrative records were screened for indicators of potential adverse events using previously validated criteria. Trained nurse abstracters reviewed 3446 medical charts (including readmissions) for 2504 screenpositive cases, and prepared case summaries. The summaries were assessed by dual independent physician reviewers in order to determine the presence or absence of AEs, the date on which the event occurred, severity, preventability and the factors contributing to AEs. In a parallel data collection effort, we obtained workload and information on patient complexity for each day of the 365-day study period, including midnight census, throughput (including number of admissions and discharges) RN staff, non-RN staff, patient DRG, nurse assigned acuity, days of diversion, and sociodemographic characteristics. After preliminary analysis, the data for weekends and weekdays were analyzed separately. The day was the unit of analysis. The relationships between the workload variables and the daily rate of adverse events (AEs per patient per day) were assessed using standard correlation techniques. Population Studied: All adults aged 18 to 64 admitted to a major Boston teaching hospital between October 2000 and September 2001, of which n = 8499 cases were selected for study. Principal Findings: The rate of adverse events for adult medical-surgical patients was positively correlated with census (p=.06), number of new admissions to the unit (p=.06), patients per RN staff (p=.04), and patients per non-RN staff (p=.02). Rates of AEs were 12-40% higher in the highest quartile compared with the lowest quartile of days, depending on the workload measure. Similar trends occurred when examining patients in ICUs and non-ICU settings, but fewer differences were significant. Conclusions: High hospital occupancy, high patient throughput, as evidenced by high patient admission rates per day, and high patient:nurse ratios may be associated with higher risks to patients, possibly due to an increased tension between available and required staff time, skill-mix and physical resources. Implications for Policy, Delivery, or Practice: Hospital overcrowding may be a significant contributor to patient harm. Hospitals may need to develop strategies to better match the supply of staff, skill and physical resources to demand in order to optimize patient outcomes and avoid adverse events associated with hospital overcrowding. In addition, as researchers investigate new methods to reduce adverse events, it may be important to consider effects during crowded conditions. Primary Funding Source: AHRQ ●Hospital Financial Performance, Nurse Staffing, and the Outcomes of Patient Care Mei Zhao, Ph.D., Gloria J. Bazzoli, Ph.D., Richard C. Lindrooth, Ph.D., Askar Chukmaitov, MPH Presented By: Mei Zhao, Ph.D., Assistant Professor, Public Health, University of North Florida, 4567 Street Johns Bluff Road, South, Jacksonville, FL 32224-2673; Tel: (904)620-1444; Fax: (904)620-2848; Email: mzhao@unf.edu Research Objective: Hospitals during the late 1990s experienced many financial pressures, which led to declining financial performance and efforts to shore up declining margins by reducing costs. The objective of this paper is to evaluate the relationship between changes in hospital finance performance, nurse staffing, and patient outcomes. Study Design: A panel study design is applied to data from 1995-2000 to examine the effect of hospital financial and staffing incremental changes on patients’ outcome incremental changes. Four risk-adjusted mortality rates and five complication rates generated from AHRQ Inpatient Quality Indicators (IQIs) and Patient Safety Indicators (PSIs) were chosen as the patient outcome. Fixed-effects models and first difference models were used to examine these effects. Population Studied: All nonfederal short term general medical-surgical hospitals in operation between 1995 and 2000 from 11 states that participated in the AHRQ’s HCUP State Inpatient Databases (SID) program. Principal Findings: We found that there is a weak association between nurse staffing and patient outcomes and an insignificant association between hospital financial condition and patient outcomes. This weak correlation or insignificance implies that incremental changes in RN and LPN staffing (or in hospital financial condition) are not large enough to create substantial change in inpatient quality of care or patient safety. We did find significant negative relationships, however, between hospital volume within a major diagnostic category and associated mortality rates after accounting for patient case mix and other factors. Additionally, hospitals treating higher levels of patient severity had lower mortality rates. Conclusions: The findings of this panel study are consistent with recent literature that suggests hospital quality of patient care is only weakly related to changes in hospital nurse staffing, even though the prevailing belief is that increased nurse staffing leads to improved quality of care. In addition, the insignificant effect of hospital financial performance on patient outcomes suggests that hospitals experiencing a decline in financial performance may not be compromising the care they provide patients. Implications for Policy, Delivery, or Practice: In recent years, hospitals have implemented many changes in hospital staffing in an effort to reduce costs and improve financial performance. Our results indicate that these efforts did not result in a deterioration of patient quality of care or safety. At most, our analysis indicates that the impact of hospital staffing changes and financial changes on patient outcomes is limited. Our results indicate that patient outcomes are influenced more by the volume of care a hospital provides and its experience in treating severely ill patients. As such, health policy efforts focused on improving hospital quality should focus more on regionalizing care of certain patients than on implementing minimum staffing levels or enhancing hospital reimbursement. However, even if most of the patient outcome indicators were not significantly affected by staffing levels, it is important to investigate how staffing reductions may affect employee morale and productivity because these may have long term effects on the health system. Primary Funding Source: AHRQ Call for Papers Financing & Quality of Care Chair: Paul Shekell, RAND Sunday, June 26 • 5:30 pm – 7:00 pm ●Hospital Financial Performance and Patient Safety Laurence Baker, Ph.D., David Classen, M.D., Kelly Dunham, MPP, Jeffrey Geppert, JD, Shoutzu Lin, MS, Sara Singer, MBA Presented By: Laurence Baker, Ph.D., Associate Professor, Health Research and Policy, Stanford University School of Medicine, HRP Redwood Building, Room 110, Stanford, CA 94305-5405; Tel: (650)723-4098; Fax: (650)725-6951; Email: laurence.baker@stanford.edu Research Objective: Hospitals in the United States face financial pressures from declining reimbursements and rising costs. One concern that arises is that financial pressures may negatively influence quality of care. For example, financially strained institutions may limit the resources committed to adopting approaches that could prevent adverse events from occurring such as improved staffing ratios, information systems, technology, training, and creating a culture that promotes patient safety. In addition, hospitals attempting to increase revenue or decrease costs might introduce production pressures on staff that adversely impact safety culture and make errors more likely. The objective of this project is to measure the relationship between hospital financial performance and adverse events for patients. Study Design: Longitudinal analysis of the relationship between changes in financial performance and changes in rates of patient safety events. We measure patient safety outcomes at hospitals using the AHRQ Patient Safety Indicators (PSIs), which define a range of adverse patient events identifiable in discharge data, for 1997 and 2000. We use Medicare cost reports and Dun and Bradstreet financial measures to create measures of hospital financial performance and the amount of financial pressure facing hospitals in these years. Measures of hospital financial performance include operating margin (net income/total revenue), capital investment per discharge (total assets/total discharges), profitability (net income/total assets), and Dun and Bradstreet indices for risk of severe financial stress over the next 12 months (financial stress class) and risk of severe payment delinquency over the next 12 months (credit score class). We estimate hospital-level regressions in which the dependent variables are rates of occurrence of adverse events defined by the AHRQ PSIs, and key independent variables are measures of financial performance. Models include hospital fixed effects to control for characteristics of hospitals, year fixed effects to control for time trends, and other controls for hospital characteristics that may change over time like beds and technology. Effects are thus identified by changes in financial performance over time within hospital. Population Studied: Hospitals in 18 states covered by the AHRQ HCUP data, including approximately 1,100 hospitals and 15 million discharges. Principal Findings: On average, hospitals experience a dramatic decline in financial performance between 1997 and 2000. Hospitals that experience a decline in financial performance by our measures also experience statistically significant declines in performance on several Patient Safety Indicators. There are a range of different events in the Patient Safety Indicators. We find that the indicators that are considered the most sensitive to the quality of nursing care experience the greatest decline when financial performance declines. Conclusions: Deteriorating hospital financial performance appears to have a negative impact on patient safety and quality of care. The results also suggest that in recent years hospitals have implemented few efficiency gains that have not had an impact on quality. Implications for Policy, Delivery, or Practice: Policies that tend to increase the financial pressures on hospitals may harm patients, and should be carefully evaluated. Efforts specifically aimed at helping hospitals facing financial pressure to monitor and maintain patient safety may be valuable. Primary Funding Source: AHRQ ●What’s the Return? Assessing the Effect of “Pay-forPerformance” Initiatives on the Quality of Care Delivery Stephen Grossbart, Ph.D. Presented By: Stephen Grossbart, Ph.D., Vice President, Clinical Performance Measurement & Improvement, Quality Management, Catholic Healthcare Partners, 615 Elsinore Place, Cincinnati, OH 45202; Tel: (513) 639-2784; Fax: (513) 639-2762; Email: srgrossbart@health-partners.org Research Objective: Identify the impact of “pay-forperformance” initiatives on the effectiveness of care delivery in a multi-hospital system. This study tests the question: Did hospitals participating in the Center for Medicare and Medicaid Services’ (CMS) “Premier Hospital Quality Incentive Demonstration Project” have significantly different rates of improvement in the quality of care delivered compared to similar hospitals within the same system that did not participate in the voluntary demonstration project? Study Design: The study uses a quasi-experimental design to assess the impact of the Premier Demonstration project. A test group of four acute care hospitals within a single healthcare system that are participating in this demonstration project was compared to a control group of seven hospitals that were not participating in the project. The study limited analysis to three clinical areas that are included in the Premier Demonstration: Acute Myocardial Infarction (AMI), Heart Failure, and Pneumonia. Performance for the first year of the demonstration (federal fiscal year 2004) was compared to a baseline year measured in federal fiscal year 2003 for both participating and non-participating hospitals to determine if the rate of improvement from the baseline year was different between the two cohorts of hospitals. The study used composite score methodology developed by Hospital Core Performance Measurement Project for the Rhode Island Public Reporting Program for Health Care Services and modified by Premier, Inc. for use in their demonstration project. Population Studied: Patients discharged between October 2002 and September 2004 with a diagnosis of Acute Myocardial Infarction, Heart Failure, or Pneumonia from 11 hospitals within Catholic Healthcare Partners. Catholic Healthcare Partners is organized into 10 regional service areas in five states and includes 30 hospitals, several long-term care facilities, housing sites for the elderly, home health agencies, hospice programs, outreach services, medical groups, wellness centers and other organizations that operate diversified healthcare activities. The system is the largest healthcare provider in Ohio and also has hospitals located in Kentucky, Pennsylvania, and Tennessee. The study was limited to 11 hospitals with the highest levels of service within the system. Principal Findings: Hospitals participating in the pay-forperformance demonstration project had significantly higher composite quality scores in each of the three clinical areas studied. For AMI, the composite score for the cohort of participating hospitals was 94.2% vs. 90.7% for nonparticipants (p-value <=.01). For heart failure, the composite score for the cohort of participating hospitals was 86.9% vs. 84.9% for non-participants (p-value <=.01). For pneumonia, the composite score for the cohort of participating hospitals was 87.3% vs. 80.3% for non-participants (p <=.01). More striking was the pace of improvement among the two cohorts. In two of the clinical areas, participating hospitals saw significantly greater improvement in the first year of the demonstration project relative to their baseline performance than the cohort of non-participating hospitals. In the area of AMI, the improvement from baseline to the first year for participating hospitals was 4.3% vs. 3.2% for non-participating hospitals (p-value <= .01). The differential rate of improvement in the area of heart failure was dramatic and significant. The improvement from baseline to the first year for participating hospitals was 19.1% vs. 10.7% for nonparticipating hospitals (p-value <= .01). Only among pneumonia patients did non-participating hospitals slightly outpace the pay-for-performance cohort. In the area of pneumonia, the improvement from baseline to the first year for participating hospitals was 7.2% vs. 8.2% for nonparticipating hospitals (p-value <= .01). Conclusions: Outside of participation in the pay-forperformance demonstration, both cohorts of hospitals were exposed to similar incentives to improve the quality of care for their respective populations. The system shares a common mission and set of values. Public reporting pressures are identical. All hospitals in Catholic Healthcare Partners joined the National Hospital Quality Initiative in October 2003 and therefore each has a common interest to excel in the ten publicly reported measures that are currently published on the CMS web site. Furthermore, in 2004, Catholic Healthcare Partners adopted a corporate strategic goal to perform in the top quartile for all ten publicly reported measures in the areas of AMI, heart failure, and pneumonia. Increased consumer focus, public reporting initiatives, the activities of regulatory and accreditation agencies, payer interest, as well as the commitment of providers to their patients and communities have all contributed to improve the quality of care delivered. Despite facing similar conditions, however, those hospitals in the pay-for-performance project experienced significantly higher rates of overall quality and significantly higher rates of improvement in two of three areas relative to the control group of non-participating hospitals. The alignment of financial incentives with performance has increased the rate of quality improvement resulting in an overall higher quality of care. Implications for Policy, Delivery, or Practice: This analysis provides strong evidence that pay-for-performance initiatives significantly improved performance among a cohort of participating hospitals. The study provides evidence to support the conclusions regarding payment incentive found in 1) Institute of Medicine. _Crossing the Quality Chasm: A New Health System for the 21st Century_. Washington, DC, National Academy Press, 2001 and 2) MedPAC. _Report to the Congress: Variation and innovation in Medicare_. June 2003. The study helps to demonstrate support for policies that call for the alignment of incentives with performance. Primary Funding Source: No Funding ●Achieving High Quality-Low Cost Hospital Performance: The Effects of Market and Organizational Characteristics H. Joanna Jiang, Ph.D., Bernard Friedman, Ph.D., James W. Begun, Ph.D. Presented By: H. Joanna Jiang, Ph.D., Social Scientist, Center for Delivery, Organization and Markets, Agency for Healthcare Research and Quality, 540 Gaither Road, Rockville, MD 20850; Tel: (301)427-1436; Fax: (301)427-1430; Email: jjiang@ahrq.gov Research Objective: Hospital performance historically has been measured one-dimensionally. Recent efforts at payingfor-performance have yet to embrace the goal of achieving both higher quality and lower cost. The purpose of this study is twofold: 1) to develop methods to identify hospitals based on performance in both quality and cost of care, and 2) to examine market and organizational characteristics that affect quality and cost performance. Study Design: Hospitals were classified into performance quartiles for both quality and cost. Quality was measured by a weighted composite of 10 risk-adjusted morality indicators covering common medical conditions and surgical procedures. Cost per discharge was computed by applying cost-to-charge ratios to hospital discharge abstracts and adjusting for principal diagnosis, severity of illness, and the area wage index. The likelihood of achieving high quality-low cost performance in relation to market and organizational characteristics was examined through logistic regression for the year 1997 and the year 2001, respectively. Population Studied: Over 1350 nonfederal, general acute hospitals in 10 states (AZ, CA, CO, FL, GA, IA, IL, NY, TN, WI) that have organizational data available from the American Hospital Association and discharge data available from the Healthcare Cost and Utilization Project. Data on market characteristics were extracted from the Area Resources File and InterStudy HMO County Surveyor. Principal Findings: Only a small proportion of hospitals obtained high quality-low cost performance – being in the lowest quartiles for both mortality and cost (8.04% in 1997 and 8.14% in 2001). Hospital competition and number of HMOs in the market were positively associated with the likelihood of achieving high quality-low cost performance (p<.01 for both years). HMO penetration also had a positive effect but only was significant for 1997 (p<.01). Among hospital characteristics, for-profit status and system membership showed positive relationships with the likelihood of having high quality-low cost performance (p<.05 for both years). No effects were found for service specialization and nurse staffing. (Higher RN proportion was significantly associated with lower mortality rate but higher cost, resulting in minimal impact on achieving high quality-low cost performance.) Significant interactions (p<.05) were found between market and organizational characteristics. The positive effect of HMO penetration on having high quality-low cost performance was stronger for large, non-teaching hospitals; likewise, the positive effects of hospital competition and HMO penetration were greater for public hospitals; but for system hospitals, these effects were negative. Conclusions: Both market and organizational characteristics have significant impact on a hospital’s likelihood of obtaining high quality-low cost performance. Strong market forces appear to have positive effects on hospital performance, and hospitals with certain characteristics may have distinct advantages to achieve high quality-low cost performance. Implications for Policy, Delivery, or Practice: Payment reforms are usually designed to encourage improvements in either quality or cost separately. If the aim were to encourage hospitals to attain high performance in both quality and cost, realistic expectations should recognize the market forces and organizational differences that are conducive or counteractive to the goal. Primary Funding Source: AHRQ ●The Effect of Palliative Care on Hospital Costs of Terminal Hospitalization Joan Penrod, Ph.D., Partha Deb, Ph.D., Carol Luhrs, M.D., Cornelia Dellenbugh, MPH, Tsivia Hochman, MS, Sean Morrison, M.D. Presented By: Joan Penrod, Ph.D., Research Health Scientist, TREP/GRECC, Bronx VA Medical Center, 130 Kingsbridge Road, Bronx, NY 10468; Tel: (718)584-9000 x3814; Fax: (718)741-4211; Email: joan.penrod@mssm.edu Research Objective: Hospital-based interdisciplinary palliative care (PC) consultation is increasingly being used to meet the growing demand for high quality care for patients with chronic and life-limiting illnesses. PC focuses on management of pain and other symptoms, identification of goals of care, psychological support and care coordination. Despite the prevalence of hospital-based PC teams, only a few studies have examined their effect on hospital costs. We examine total direct, ancillary, pharmacy and ICU costs associated with receipt of PC compared to usual care (UC) during terminal hospitalization. Study Design: This is a retrospective, observational study of hospital costs of 314 decedents at two urban VA hospitals. We use data from VHA’s cost system and from review of patients’ medical records. A generalized linear model (GLM) with the gamma distribution and identity link was used in the analyses. Age, gender, co-morbidity, complications, principal diagnosis, and whether the patient was medical or surgical were included as covariates in all the models. To control for selection bias, we also used propensity scores to match patients for whom the probability of receiving PC was similar. We used a Probit to generate a propensity score for receiving PC as a function of covariates above and nearest neighbor and stratification matching to compare average treatment costs. We also used a two-part model (Probit followed by GLM) to examine ICU costs. Population Studied: Veterans who died from non-trauma causes at two urban VA hospitals in 2003. We identified 82 decedents who received PC consultation and 232 who did not. Principal Findings: PC and UC patients did not differ by age, gender, hospital length of stay, or number of hospitalizations in the prior year. PC patients had fewer complications, slightly fewer comorbidites and were more likely to have cancer compared to UC patients. After controlling for covariates, PC decedents’ average direct total and ancillary costs were $192 (95% CI -$300 to - $84, p < .0001) and $97 (95% CI -$142 to $51, p < .0001) less per day, respectively, compared to UC patients. Pharmacy costs were not significantly different between the groups ($1.41, 95% CI -$20 to $22, p = .90). PC decreased the likelihood of ICU use by .23 percentage points (95% CI -36 to - .09). ICU costs for users were not significantly different for PC compared to UC patients ($138, CI -$166 to $442, p = .37). The results were similar using a variety of propensity score matching methods. Conclusions: PC consultation was associated with lower hospital costs during terminal hospitalization. Costs may be lower because PC teams help patients and families establish clear and realistic goals of care and assist them to select treatments that meet those goals, including less use of ICU. Implications for Policy, Delivery, or Practice: A recent metaanalysis found PC care is associated with better symptom control and higher family satisfaction. Our results suggest that PC is also associated with lower costs. Taken together, the findings suggest positive consequences with expansion of PC for hospitalized patients. Primary Funding Source: VA ●Delegation of Care to Support Personnel in Outpatient Physical Therapy: Implications for Quality and Efficiency Linda Resnik, Ph.D., PT, OCS, Zhanlian Feng, Ph.D., Dennis Hart, Ph.D., PT Presented By: Linda Resnik, Ph.D., PT, OCS, Assistant Professor, Community Health, Brown University, 2 Stimson Avenue, Providence, RI 02912; Tel: (401)863-9214; Fax: (401)863-3489; Email: Linda_Resnik@Brown.edu Research Objective: There is a growing trend in healthcare to supplement or substitute cheaper health care personnel for more highly skilled and costly health care providers As an example, there is a trend in nursing to use licensed practical nurses (LPNs) and on-the-job-trained support personnel such as nurses aides in place of more highly trained registered nurses (RNs). Similar staffing trends have occurred in physical therapy where physical therapist assistants (PTAs) and therapy aides are used in place of licensed physical therapists (PTs). To our knowledge, no prior research has examined the impact of utilization of support personnel on outcomes or efficiency of physical therapy. The purpose of this study was to examine the relationship between care delegation of care to support personnel, utilization of physical therapy services, and patient outcomes. Study Design: Patients completed Health Related Quality of Life (HRQL) surveys at intake and discharge. Clinicians recorded number of visits and reported percentage of time spent with each provider, ie, aide, PTA and PT. Three independent variables representing delegation of care were tested: More than 50 percent time with PTA (yes/no), percent time, and percent time with an aide. Dependent variables were: efficiency of care, measured by the number of visits per treatment episode, and patient outcome, measured by discharge HRQL. Two multilevel linear regression models were developed. The first model predicted number of visits in the treatment episode. The second model predicted the HRQL score. In the multilevel models patients were nested within therapists, therapists nested within facility, and facilities nested within states. Models controlled for confounders at the patient level (gender, age, impairment type, OHS intake score, acuity, surgical history, payer type, employment), at the facility level (ratio of full-time equivalent PTAs and PTs on the staff of the clinic), and at the state level (4 types of regulation governing PT). Population Studied: The sample was 63,900 patients treated by 2466 therapists, from 39 states in 2000 and 2001 drawn from 351 clinics who participated with the Focus On Therapeutic Outcomes, Inc. (Knoxville, TN). Principal Findings: After controlling for patient, therapist, clinic and state factors, lower percent time seen by a PT and greater percent time seen by a PTA or aide was associated with more visits per treatment episode. Patients seen less than 75% time by a PT, more than 50% time with a PTA and 125% time with an aide had 4.5 more visits as compared to patients seen primarily by a PT. Patients seen by the PT less than 25% of the time had OHS score almost 2 points lower than those seen by PTs 76-100% of the time. Conclusions: Patterns of care delegation are associated with efficiency of services. More time with the PT and less time with PTA and aides is more efficient. Patterns of care delegation impact patient outcomes. Better patient outcomes occurred when the PT spent more time with the patient. Implications for Policy, Delivery, or Practice: Although delegation of care to physical therapy support personnel may extend the productivity of PTs, it appears to result in less efficient and effective services. More research is needed to confirm these findings, and to explore the impact of care delegation patterns in other settings. Primary Funding Source: AHRQ Call for Papers Culture Change, System Change & Quality Improvement Chair: Susan Edgman-Levitan, Massachusetts General Hospital Tuesday, June 28 • 8:30 am – 10:00 am ●Adverse Event Reporting Laws and Medical Errors Laurence Baker, Ph.D., Kelly Dunham, MPP, Jeffrey Geppert, JD, Michelle Michalek, BA Presented By: Laurence Baker, Ph.D., Associate Professor, Health Research and Policy, Stanford University School of Medicine, HRP Redwood Building, Room 110, Stanford, CA 94305-5405; Tel: (650)723-4098; Fax: (650)725-6951; Email: laurence.baker@stanford.edu Research Objective: Improving patient safety is an important goal for the U.S. healthcare delivery system. One specific issue is the rate of preventable errors, and resulting harm to patients. In an effort to reduce the incidence of preventable errors, twenty-two states passed legislation requiring that hospitals report the occurrence of preventable errors. The objective of this project is to study the relationship between state legislation and medical errors to determine if these laws influence the incidence of preventable medical errors that result in patient death across the United States. Study Design: Longitudinal analysis of the relationship between the adoption of adverse event reporting laws and changes in rates of adverse events. We use multiple cause-ofdeath data from 1984 to 1998 to create four adverse event indicators measuring death from a preventable error, based on the methodology used to create the AHRQ Patient Safety Indicators (PSIs). We use data from the Institute of Medicine and state legislative records to identify states that passed reporting laws, the timing of passage and/or implementation, and other law characteristics. We estimate state-level regressions in which the dependent variables are population rates of mortality from the four preventable errors and the key independent variable is the existence of adverse event reporting legislation. We examine the overall effects of laws as well as short term (1-3 years after passage) and long term (4+ years after passage) effects. Work in progress will model the effects of laws with different degrees of specificity and different penalty structures, and possibly other characteristics. Population Studied: Since we use population rates for persons aged 1-85 in all 50 states and the District of Columbia, the population studied is implicitly the entire population of the United States in this age range. Our data contain 26,747 deaths where the proximate or immediate cause was one of the AHRQ preventable errors. Principal Findings: Summing deaths from all four causes, regression point estimates suggest that passage of adverse event reporting laws is associated with a decline in mortality rates from the studied events. Taking individual measures separately, 2 of the 4 measures show declines in mortality rates with passage of reporting laws and 2 show increases. Statistical significance is often limited in these preliminary analyses, so conclusions need to be drawn with caution. Further work to improve model precision is ongoing. Some results also suggest that the effect was largest in the years immediately following adoption of the reporting law, and remained but declined somewhat in subsequent years. Conclusions: Legislation requiring the reporting of adverse events may influence mortality rates from some preventable errors. Drops in mortality rates may be due to actual reduction in the occurrence of medical error, or to the coding of medical error, as the proximate or contributing cause of death. The increases in mortality rates with law passage that we observe are harder to understand and will require further analysis. Implications for Policy, Delivery, or Practice: Continued attention to adverse event reporting laws is warranted. Efforts to raise awareness and promote reporting may be beneficial. Primary Funding Source: AHRQ ●Improving Complex Systems: Top Performing Hospitals in Door-to-Balloon Times for Patients with Acute Myocardial Infarction Elizabeth Bradley, Ph.D., Leslie A. Curry, Ph.D., David Berg, Ph.D., Robert L. McNamara, M.D., MHS, Martha J. Radford, M.D., Harlan M. Krumholz, M.D., HS Presented By: Elizabeth Bradley, Ph.D., Associate Professor, Epidemiology and Public Health, Yale School of Medicine, 60 College Street, New Haven, CT 06520; Tel: (203)785-2937; Fax: (203)785-6287; Email: elizabeth.bradley@yale.edu Research Objective: Hospitals are under increasing pressure to meet clinical guidelines with public reporting of performance indicators by the Centers for Medicare & Medicaid (CMS) and the Joint Commission for Accreditation of Healthcare Organizations (JCAHO). Recent research has demonstrated substantial hospital-level variation in performance on quality indicators; however, little is known about how top performing hospitals are achieving their success. We sought to describe common strategies in hospitals with top performance in a key quality indicator for acute myocardial infarction: door-to-balloon time. This quality indicator reflects a complex clinical process that requires coordination among several departments and disciplines, is time-sensitive, and is critical to patient survival. Study Design: We conducted a qualitative study using indepth interviews of clinical and administrative staff at hospitals with exceptional performance in door-to-balloon time for patients with ST-segment elevation myocardial infarction (STEMI) treated with PCI. This method is well suited for investigating complex interactions among people and processes that are common in performance improvement and organizational change. We used the constant comparative method of qualitative data analysis facilitated by NUD-IST software. Population Studied: We included hospitals participating in the National Registry of Myocardial Infarction, a large registry of hospitalizations for myocardial infarction in the U.S., that provided at least 50 patients with STEMI with PCI between January 1, 2001 and December 31, 2002 (n = 151). Among hospitals with median door-to-balloon times within the 90 minute clinical guidelines (n=35), we ranked hospitals by their improvement during the previous 4 years and visited those with the greatest improvement. We performed site visits and in-depth interviews until we reached theoretical saturation, which occurred after 11 hospitals and 122 interviews. Principal Findings: Several key themes emerged to characterize the top performing hospitals. First, top performing hospitals had implemented substantial process redesign efforts to eliminate unneeded steps and delays in the clinical process. Second, top performing hospitals developed rapid data feedback mechanisms both to monitor performance and to identify areas for further improvement. Third, top performing hospitals demonstrated effective management of several key paradoxes: 1) standardized protocols yet flexibility in adapting these protocols to fit individual patient and changing organizational circumstances; 2) strong clinical champions along with extensive interdisciplinary teamwork; 3) broad organizational goals yet narrowly-focused performance targets; and 4) formal, explicit channels of communication as well as informal, implicit methods of communication. Fourth, even these top performing hospitals struggled with interdepartmental and interdisciplinary conflicts, which challenged improvement efforts; participants worked through these “speed bumps” by regularly and explicitly reminding groups and individuals of their common goal of improving patient care. Conclusions: We find a set of organizational features and strategies that are common to hospitals with top performance in a complex clinical process. Systems improvement, with rapid data feedback, within an organizational culture that promoted common goals and fostered multiple strategies to achieve those goals were dominant themes among these institutions. Implications for Policy, Delivery, or Practice: Policy and management efforts to improve hospital quality must involve not only innovative system design and data feedback but also sophisticated understanding of organizational culture. Achieving top performance may require effective management of paradoxical processes inherent in transforming organizations. Primary Funding Source: National Heart, Lung, Blood Institute; Patrick and Catherine Weldon Donaghue Medical Research Foundation ●Are Difference in Quality Between Physician Groups Explained by Organizational Characteristics or Use of Quality Improvement strategies? Ateev Mehrotra, M.D., MPH, Meredith Rosenthal, Ph.D., Arnold M. Epstein, M.D., MA Presented By: Ateev Mehrotra, M.D., MPH, Research Fellow, Department of Health Care Policy, Harvard Medical School, 180 Longwood Avenue, Boston, MA 02115; Tel: (617)548-7517; Email: amehrotra@partners.org Research Objective: Despite continuing evidence demonstrating large variations in health care quality, little is known about what explains these variations. In this study we looked at variations in the quality of primary care across physician groups. We examined the association of organizational characteristics (e.g., size, use of an electronic medical record (EMR)) and the use of quality improvement strategies (e.g., disease registries, reminder systems) with the quality of primary care delivered to beneficiaries of PacifiCare, a large California health maintenance organization. Study Design: We obtained information on organizational characteristics and quality improvement strategies through structured interviews with PacifiCare-contracted physician groups. Quality data were obtained from PacifiCare´s audited quality performance reports (Quality Index and Provider Profile), which are based on administrative data. Quality was measured by determining what percentage of eligible patients received the following: mammography, Pap smear, Chlamydia screening, childhood immunizations, diabetic eye exam, inhaled steroid for asthmatics, beta-blocker after myocardial infarction, and ACE inhibitor for CHF. An overall quality score for each physician group was created by adding standardized individual scores. We tested whether 16 organizational characteristics and 12 quality improvement strategies were associated with quality scores. Those significant in univariate analyses were included in a multivariate linear model. A model was created for each of the 8 individual quality measures as well as the overall quality score. Population Studied: 119 physician groups in California who care for over 1.5 million PacifiCare enrollees. Principal Findings: In univariate analyses, medical groups had statistically significant (p < 0.05) higher quality scores on 4 of the 8 individual quality measures as well as in overall quality as compared with IPAs. In multivariate regression, medical groups had significantly higher scores in overall quality and in 3 of the 8 individual quality measures as compared with IPAs (19.5% higher mammogram rate, 20.4% higher Pap smear rate, and 11.4% higher diabetic eye screening rate). The other covariates in the models were not consistently associated with quality scores. Increasing fraction of physician salary based on productivity was predictive of higher overall quality. Higher volume of patients was predictive of higher mammogram and childhood immunization rates and older age of group was predictive of higher Chlamydia screening rates. The use of an EMR, quality bonuses for individual doctors, reminder systems for patients, or feedback of quality performance to individual physicians were not statistically significant predictors of quality in the multivariate model. Medical groups were more likely to use an EMR (36.8% vs. 1.6%) and have quality improvement programs (e.g., contact patients to remind them of mammograms, 73.7% vs. 27.8%). Yet less than 25% of the quality differences between medical groups and IPAs were explained by organizational characteristics and use of quality improvement strategies. Conclusions: As compared to IPAs, medical groups have significantly higher scores on many quality measures. Although medical groups are more likely to use electronic medical records and have quality improvement strategies, these characteristics and strategies explain only a fraction of the differences in quality. Implications for Policy, Delivery, or Practice: If these findings are confirmed in other studies, policymakers should implement laws and regulations that encourage the formation of medical groups. Primary Funding Source: CWF ●Do Quality Improvement Organizations Improve the Quality of Hospital Care for Medicare Beneficiaries? Claire Snyder, MHS, Gerard Anderson, Ph.D. Presented By: Claire Snyder, MHS, Doctoral Candidate, Health Policy & Management, Johns Hopkins Bloomberg School of Public Health, 624 North Broadway, 6th Floor, Baltimore, MD 21205; Tel: (410)732-7113; Fax: (410)955-0470; Email: csnyder@jhsph.edu Research Objective: This study explored whether Quality Improvement Organizations (QIOs) improve the quality of hospital care for Medicare beneficiaries. Hospitals participate with the QIOs on a voluntary basis, and this study investigated (1) the characteristics of hospitals that participate with the QIOs and (2) whether hospitals that participate with the QIOs improve the quality of care for Medicare beneficiaries more than hospitals that do not. Study Design: This was a retrospective comparative study that assessed the performance of hospitals on 15 dichotomous quality of care indicators associated with improved outcomes in the prevention or treatment of 5 clinical areas (acute myocardial infarction, atrial fibrillation, heart failure, pneumonia, stroke). The quality of care was evaluated using data from a cross sectional sample of Medicare beneficiary medical records abstracted in 1998 (baseline) and a separate sample of medical records abstracted in 2000-2001 (follow-up). First, the characteristics (bedsize, profit status) and baseline performance of participating hospitals were compared to non-participating hospitals. Then, the improvement in performance on the 15 quality indicators by hospitals that actively participated with the QIOs was compared to the improvement by hospitals that did not participate with the QIOs while controlling for hospital bedsize and profit status and patient age, sex, and race. Population Studied: Four QIOs with responsibility for five states (Maryland, New York, Nevada, Utah, Washington) and the District of Columbia (DC) provided their data for this study. Approximately 700-800 medical records per state in each of the 5 clinical areas were abstracted at baseline (1998) and at follow-up (2000-2001). Principal Findings: Hospitals that voluntarily participated with the QIOs were more likely to be not-for-profit and larger than non-participating hospitals (p<0.05). At baseline, in analyses controlling for hospital and patient characteristics, there were statistically significant (p<0.05) differences between participating and non-participating hospitals on 5/15 quality indicators, with participating hospitals performing better on 3/5. There were no statistically significant differences in change from baseline to follow-up between participating and non-participating hospitals on 14/15 quality indicators controlling for hospital and patient characteristics. Participating hospitals improved more on the pneumonia immunization indicator than non-participating hospitals (p=0.005). For the 14 non-statistically significant indicators, participating hospitals improved more than non-participating hospitals on 8, and non-participating hospitals improved more than participating hospitals on 6. Conclusions: There are some important differences in hospital characteristics and baseline performance of hospitals that participated with the QIOs compared to those that did not. Participating hospitals improved statistically significantly more than non-participating hospitals on only 1/15 quality indicators. Thus, hospital participation with the QIOs was not associated with greater improvement in the quality of care for Medicare beneficiaries. Implications for Policy, Delivery, or Practice: The Medicare program invests $200 million annually in the quality improvement activities of the QIOs. This study does not support the effectiveness of the QIOs in improving the quality of care in the inpatient setting. Further research to assess the QIOs’ effectiveness is needed, and the QIOs’ quality improvement efforts may require modification. Primary Funding Source: AHRQ ●Organizing for Quality: Journeys of Improvement at Leading Healthcare Organizations in the US & UK James L. Zazzali, Ph.D., MPH, Peter Mendel, Ph.D., Paul Bate, Ph.D., Glenn Robert, Ph.D. Presented By: James L. Zazzali, Ph.D., MPH, Assosiate Policy Researcher, RAND, 1776 Main Street, Santa Monica, CA 90407; Tel: (310)393-0411; Email: zazzali@rand.org Research Objective: To present cross-site and cross-national findings of a recently completed 18-month international research project to understand the ability of healthcare organizations to sustain quality improvement (QI), and to identify best practices in change management related to the introduction and implementation of QI programs and processes in large healthcare systems in the US and UK. Study Design: Mixed methods with a multilevel approach. Within each organization we conducted interviews with senior leaders (i.e., a “macro-system” perspective) and used this data to construct a survey that was administered to the clinical and administrative staff in one high performing department (i.e., a “micro-system” perspective). The survey measured perceptions of the importance of key factors related to sustaining QI efforts, the degree to which these factors were met in the organization, the organizational culture of the micro-system, and the respondents’ level of QI training and QI team experience. Population Studied: Ten leading health care organizations in the US & UK (6 US sites and 4 UK sites). Survey of microsystem staff: 477 respondents in the US & UK (48% response rate). Principal Findings: Respondents from the two countries were in agreement as to the two top most important factors related to sustaining QI efforts: “training and time so we can improve the quality of patient care and service,” and the need for “senior management to make improving the quality of patient care a priority.” In general, the staff in the US micro-systems generally rated the various factors related to sustaining QI efforts as significantly more important than the UK respondents. Further, the staff in the US sites reported significantly higher levels of QI training (51.7% for the US and 15.3% for the UK) and higher levels of QI team experience (36.1% served on a QI team in the US, whereas 11.3% reported this in the UK). The cultures of the micro-systems were somewhat dissimilar across the two countries, with the US sites exhibiting a more “task-oriented” culture than the UK sites. Conclusions: Respondents from the two countries were in agreement at to some of the top factors related to sustaining QI efforts, but there exist cross-national differences in how these two countries organize for quality. Implications for Policy, Delivery, or Practice: There are significant opportunities for intra-national as well as crossnational learning for health care organization in the US and the UK with regard to how they organize and sustain their QI efforts. Primary Funding Source: The Nuffield Trust, RAND Corporation Call for Papers Quality & Safety for All: Caring for Vulnerable Populations Chair: Paul Shekelle, RAND Tuesday, June 28 • 10:30 am – 12:00 pm ●Racial Differences in Impact of HMO Coverage of Diabetes Blood Glucose Monitors on Initiation of SelfMonitoring Connie Mah, MS, Dennis Ross-Degnan, Sc.D., Alyce S. Adams, Ph.D., Stephen B. Soumerai, Sc.D. Presented By: Connie Mah, MS, Pre-Doctoral Research Fellow, Department of Ambulatory Care and Prevention, Harvard Medical School and Harvard Pilgrim Health Care, 133 Brookline Avenue, 6th Floor, Boston, MA 02215; Tel: (617) 509-9955; Fax: (617) 859-8112; Email: cmah@fas.harvard.edu Research Objective: Despite being at higher risk for poor health outcomes, black diabetic patients receive lower quality of care and face greater barriers to self-management compared to white patients. These differences exist even in managed care settings, where system-related differences in access to care are diminished. Insurance coverage of patient self-management technologies like blood glucose selfmonitoring equipment may help to reduce race-related barriers to effective care. We examine whether providing free home glucose monitors in a health maintenance organization (HMO) had greater impacts among black compared to white diabetic patients. Study Design: Using five years (two years pre-policy and three years post-policy) of computerized medical record and claims data, we used segmented longitudinal survival analysis to examine racial differences in rates of initiation of selfmonitoring blood glucose (SMBG) following free insurance coverage of self-monitoring equipment in 1994. To control for fixed (age, gender, SES, BMI, pre-policy HbA1c test) and timedependent covariates (monthly diabetes medication use, monthly glycemic levels), a piecewise extended Cox model produced estimated hazard ratios comparing rates of SMBG initiation between blacks and whites, before and after the change in coverage. We modeled time in months until first SMBG (1+ test strip dispensed), censoring at the end of the study period or at time of first insulin use. Population Studied: 2,275 continuously enrolled adult (18+ years) Boston-area diabetic patients with black or white racial identification receiving care within 14 multi-specialty health centers of Harvard Vanguard Medical Associates and insured by Harvard Pilgrim Health Care. Principal Findings: There were racial differences in rates of initiating SMBG among patients using oral medications at baseline, but not among insulin users. Among oral medication users, unadjusted Cox models showed that SMBG initiation rates prior to policy among black diabetic patients were not statistically different from white patients (Hazard Ratio=0.864, p=0.271). However, blacks had nearly 30% higher post-policy initiation rates compared to whites (HR: 1.296, p=0.023). Insulin use and prior month HbA1c level were strongly associated with rate of SMBG initiation (HR: 1.436, p<0.0001; HR: 1.457, p<0.0001, respectively). Controlling for these time-dependent covariates and fixed effects, blacks were as likely to initiate SMBG as whites prior to the coverage policy (HR: 0.980, p=0.931) but more likely after the policy (HR: 1.546, p=0.075). However, the pre-post change in these hazard rates was not significant (p=0.152). Conclusions: Our findings suggest that black diabetes patients may have increased rates of initiating glucose selfmonitoring more rapidly than white patients following insurance coverage of home monitors, but that this change is explained in part by higher average HbA1c values and greater insulin use. Although race may not be an independent predictor of the effects of covering glucose self-monitors in this setting, the policy appeared to be effective in triggering patients to self-manage, particularly among blacks who may potentially face larger barriers to quality care. Implications for Policy, Delivery, or Practice: Coverage of diabetes self-monitoring devices is currently mandated in over 38 state Medicaid programs. These results show the potential of these policies narrowing racial gaps in self-management and quality of care among racial minority groups. Primary Funding Source: AHRQ, Harvard Pilgrim Health Care Foundation ●Is JCAHO Accreditation Associated with Better Patient Outcomes in Rural Hospitals? Laura Morlock, Ph.D., Lilly Engineer, MBBS, MHA, Cyrus Engineer, MHA, MHS, Maureen Fahey, MLA, Rebecca Clark, BA, Andrew Shore, Ph.D. Presented By: Laura Morlock, Ph.D., Professor, Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, 624 North Broadway, Baltimore, MD 21205; Tel: (410) 955-5316; Fax: (410) 955-6959; Email: lmorlock@jhsph.edu Research Objective: Fewer than 60% of rural hospitals seek accreditation from the Joint Commission on Accreditation of Healthcare Organizations (JCAHO) in contrast to 95% of urban hospitals. In a recent survey of rural hospitals, 79% of respondents indicated that the cost of accreditation was a major deterrent. It is not known whether JCAHO accreditation is associated with quality indicators based on patient outcomes in rural hospitals. The objective of this analysis was to examine relationships among JCAHO accreditation, financial status and selected quality indicators based on inhospital mortality. Study Design: This analysis utilizes data for a three year time period for a random sample of approximately half of the rural hospitals in five states. Information on the most recent accreditation status was obtained from the JCAHO. Indicators of financial performance were derived from the Medicare Cost Report Files; total margins averaged over fiscal years 19992001 were used to measure financial status. The AHRQsponsored Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases for five states were used to construct Inpatient Quality Indicators. We used three of these risk-adjusted indicators to assess clinical performance, including In-hospital Mortality for Acute Myocardial Infarction (AMI), In-hospital Mortality for AMI patients (excluding transfer cases), and In-hospital Mortality for Congestive Heart Failure (CHF). We selected these indicators because the majority of both JCAHO accredited and non-accredited rural hospitals met a volume threshold of at least 30 patients with these conditions over the three year time period. Our analyses were based on multivariate regressions using Generalized Estimating Equations to adjust for the clustering of hospitals within states. Population Studied: Patients discharged from 106 rural hospitals in Arizona, Colorado, Florida, Iowa and Wisconsin during 1999-2001. Principal Findings: Among our sample of 106 rural hospitals, 54 (51 percent) were accredited by the JCAHO. Rural hospital size as measured by number of beds (p<.0001), and better financial status as measured by financial margin (p<.01), were both associated with the likelihood of having sought and obtained JCAHO accreditation. After controlling for financial status, JCAHO accreditation was significantly associated with lower in-hospital mortality rates for AMI (p<.01), for AMI excluding patient transfers (p<.05), and for CHF (p<.0001). Because accreditation was associated with hospital size, these relationships were examined separately for hospitals larger than 50 beds. Among these rural hospitals, accreditation was significantly associated with in-hospital mortality for AMI patients excluding transfers (p<.01) and for CHF patients (p<.01). Conclusions: JCAHO accreditation was associated with better outcomes for patients with AMI and CHF treated in rural hospitals after taking into consideration hospital size and financial status. Implications for Policy, Delivery, or Practice: These findings, if corroborated by other studies, suggest that accreditation status may be a useful indicator of the quality of hospital clinical performance for rural health care consumers, payers and regulators. These findings may also be encouraging to the many rural hospitals with resource constraints who are considering the costs and benefits of embarking on the process of JCAHO accreditation. Primary Funding Source: AHRQ ●The Usual Source of Care for Medicare Beneficiaries and Whether they Receive Preventitive Services Hoangmai Pham, M.D., MPH, Debra Schrag, M.D., MPH, Marie Reed, MSH, J. Lee Hargraves, Ph.D., Peter B. Bach, M.D., MAPP Presented By: Hoangmai Pham, M.D., MPH, Senior Health Researcher, Center for Studying Health System Change, 600 Maryland Avenue SW, Suite 550, Washington, DC 20002; Tel: (202) 554-7571; Fax: (202) 484-9258; Email: mpham@hschange.org Research Objective: Rates of preventive services remain below national goals. Quantifying the association between characteristics of patients´ usual source of care, and variation in receipt of these important quality indicators can help focus quality improvement efforts. We examine the association between characteristics of a nationally representative sample of physicians and their practices, and the likelihood of patients they treated receiving six necessary preventive services. Study Design: Data on 8,517 physicians from the 2000-01 Community Tracking Study (CTS) Physician Survey were linked to Medicare ambulatory care claims data on 110,201 beneficiaries they treated in 2001. Beneficiaries 65 years and older were eligible if the physician with whom they had the plurality of visits for evaluation and management services was a CTS primary care physician. Outcome variables reflected whether eligible beneficiaries received six recommended preventive services. Independent variables reflected physician sex; training and experience; practice type and revenue sources; and reported access to information technology for generating preventive care reminders and accessing guidelines. Multivariable logistic regressions adjusted for beneficiary race, sex, zip code income, local level of education, urban/rural location, percentage of visits that were with their usual source of care (USOC) physician, and Klabunde comorbidity score, taking into account beneficiary clustering. We used the attributable risks calculated from these multivariable models to estimate how changes in physician and practice characteristics would affect the probability that beneficiaries receive these six necessary preventive services. Population Studied: Nationally representative sample of physicians and a representative sample of Medicare beneficiaries they treated. Principal Findings: 3,360 CTS primary care physicians served as the USOC for 24,581 beneficiaries. Eligible beneficiaries received the necessary services at rates of 47.9% for diabetic eye exams, 55.9% for HbA1c monitoring, 46.7% for mammography, 5.8% for colon cancer screening, 46.5% for influenza vaccination, and 8.0% for pneumococcal vaccination. Receipt of services was suboptimal for beneficiaries at all income levels. Beneficiaries with board certified physicians as their USOC were more likely than those treated by non-certified physicians to receive HbA1c monitoring (57.1% vs. 48.8%, p<0.05), mammograms (48.5% vs. 36.5%, p<0.01), colon cancer screening (6.1% vs. 4.0%, p<0.001), and influenza (47.4% vs. 41.7%, p<0.05) and pneumococcal (8.3% vs. 6.5%, p<0.01) vaccinations. Similarly, beneficiaries cared for by physicians who graduated from U.S./Puerto Rican/Canadian medical schools had higher rates of receipt of five services, than those treated by graduates of other international medical schools (p<0.05 for all comparisons). The advantage for beneficiaries with general internist vs. family/general practitioner USOCs was seen in diabetic eye exams (50% vs. 44.9%, p<0.05), mammograms (50.5% vs. 42.4%, p<0.001), colon cancer screening (6.6% vs. 4.9%, p<0.01), and pneumococcal vaccination (8.5% vs. 7.4%, p<0.05). Beneficiaries with USOCs in group practices were more likely to receive all six necessary services than those in other types of practices, as were beneficiaries cared for in practices with lower relative Medicaid revenues (p<0.05 for all comparisons). Physician sex, years in practice, and access to information technology had limited associations with receipt of services. Results were robust in multivariable analyses, and under different definitions of usual source of care. Conclusions: Although delivery of necessary preventive care to Medicare beneficiaries does not meet national targets, linkage of physician survey and patient data reveal provider characteristics consistently associated with superior care. Such information may help tailor quality improvement efforts to different physician populations and care settings. Implications for Policy, Delivery, or Practice: Such information may help tailor quality improvement efforts to different physician populations and care settings, and inform quality assurance efforts and choice of physician. Primary Funding Source: RWJF ●Perceptions of Specific Clinician Behaviors Linked to Health Care Quality Karen Shore, Ph.D., Roger Levine, Ph.D., Judy Mitchell, MS, Margarita Hurtado, Ph.D., Steve Garfinkel, Ph.D., Kristin Carman, Ph.D. Presented By: Karen Shore, Ph.D., Sr. Research Scientist, Health Program, American Institutes for Research, 1791 Arastradero Road, Palo Alto, CA 94304; Tel: (650)843-8121; Fax: (650)858-0958; Email: kshore@air.org Research Objective: To develop a comprehensive taxonomy of the components of quality ambulatory health care, based on both patient and provider perspectives. Study Design: We conducted in-depth telephone interviews with clinicians (physicians, nurse practitioners, and physician assistants) and patients to elicit examples of specific behaviors that represent good or poor quality care based on actual office visits. We reviewed and analyzed clinician and office staff actions reported to affect the quality of care in these encounters, and then we produced a taxonomy of behaviors that constitute quality care. This work is part of a larger study to develop measures of health care quality in ambulatory settings. Population Studied: We collected over 3,000 “critical incidents” (specific behaviors) from a total of 37 clinicians and 170 patients in two different geographic areas of the US. Patient respondents at each of the two sites were divided approximately equally among four racial/ethnic groups. Principal Findings: We identified nine major categories of incidents/behaviors, including: clinical skills (representing 34% of all incidents); rapport (28%); health-related communication (21%); office practices, office and ancillary staff (8%); ensuring that patients get needed care (5%); accessibility/availability (3%); information seeking (2%); “above and beyond” (behaviors that represented more effort than expected) (1%); and ethical behavior (<1%). Preliminary results show that Asian respondents and those less than 50 years old were more likely to report incidents related to treating patients with courtesy and respect (e.g., seeing patients on time, giving patient undivided attention). Hispanic respondents were more likely to describe incidents related to receiving care from ancillary staff (nurses, therapists, technicians, etc.) and incidents related to accessibility/availability (e.g., having weekend hours, returning phone calls or emails, encouraging patients to come in or call for follow-up care). Females were more likely to report incidents related to ensuring that patients get needed care (e.g., contacting other providers on behalf of patients) and information seeking (e.g., obtaining relevant medical information about patient, obtaining information needed to treat patient). Further analyses are being conducted on these data, and related analyses will be conducted comparing responses of patients to those of providers. Conclusions: Individuals with different demographic characteristics appear to value different aspects of health care quality. To ensure patient satisfaction with the quality of care they receive, providers may need to modify their behaviors or emphasize different aspects of the encounter based on the demographic characteristics of the patient. Implications for Policy, Delivery, or Practice: This study presents critical information regarding what patients and providers value in a health care encounter. It also contributes to health care providers’ knowledge and understanding of the extent to which the concept of health care quality differs according to patients’ demographic characteristics. This information can be used in quality improvement efforts, and in developing policies and practices that effectively address the areas of greatest concern to patients. Primary Funding Source: AHRQ ●Prescribing Rates of Drugs to be Avoided in the Elderly in Managed Care Lok Wong, M, Russell Mardon, Ph.D., Arlene Bierman, M.D., Philip Renner, MBA Presented By: Lok Wong, M, Senior Health Care Analyst, Quality Measurement, National Committee for Quality Assurance, 2000 L Street NW, Suite 500, Washington, DC 20036; Tel: (202)955-1784; Fax: (202)955-3599; Email: wong@ncqa.org Research Objective: To evaluate the extent to which Medicare enrollees in managed care received drugs to be avoided in the elderly. Drugs to be avoided in the elderly are defined in the study as drugs classified as never or rarely appropriate by Zahn (2001). Study Design: Retrospective pharmacy claims data analysis in 9 health plans in the U.S. Percentages of elderly enrolled throughout the year who received at least one drug or at least two different drugs considered high-risk and harmful to the elderly, reported by plan, age, gender and across the study population were calculated. Population Studied: Over 433,000 Medicare enrollees ages 65 and older in 9 health plans across the United States in 2002 and 2003. The number of enrollees enrolled throughout a year in each plan ranged from 7,500 to 187,000. Principal Findings: Nearly half a million elderly enrollees received more than 500,000 prescriptions for drugs considered never or rarely appropriate in the elderly. About one third of Medicare enrollees (35.2%) received at least 1 prescription for a drug to be avoided in the elderly, classified as never and rarely appropriate by Zahn (plan range 13.2 % to 53.2%). Less than a third (27.5%) of enrollees got at least one rarely appropriate drugs (plan range 10.8% to 44.1%). About 7.7% of elderly enrollees got at least one never appropriate drug (plan range 2.4% to 9.1%). Overall women were more likely than men to receive at least one never appropriate drug (8.3% vs. 6%) or rarely appropriate (drugs 28.6% vs. 25.7%). A subset of Medicare enrollees (average 6%, plan range 1.1% to 9.3%) received at least 2 different drugs to be avoided in the elderly. Overall women were more likely than men to get two or more different drugs to be avoided (6.7% vs. 4.3%). Elderly enrollees ages 85 and older were less likely than enrollees ages 65-74 years (4.2% vs 6.1%) to receive two or more prescriptions. Differences are statistically significant. Conclusions: Previous literature had suggested lower (about 1-20%) prescribing rates for drugs to be avoided in the elderly in managed care. This study shows variable and high rates of potentially harmful prescribing in the elderly managed care population which warrants quality improvement and monitoring. A HEDIS performance measure can highlight this patient safety issue and encourage health plans to develop interventions to prevent the elderly from receiving these drugs. The extent of the quality problem is further highlighted by patients receiving multiple types of drugs to be avoided. Additional attention on women and younger enrollees aged 65-74 is warranted by higher prescribing rates in these populations. Implications for Policy, Delivery, or Practice: This study highlights the need to monitor and reduce rates of harmful drug prescribing in the elderly enrolled in managed care in order to ensure patient safety and avoid adverse drug events. Primary Funding Source: CMS Related Posters Poster Session B Monday, June 27 • 6:15 pm – 7:30 pm ●Comparison of Weighted versus Dichotomous Thresholds for Diabetes Performance Measurement in Veterans Health Administration (VHA) Facilities David Aron, M.D., MS, Leonard M. Pogach, M.D., MBA, Mangala Rajan, MBA Presented By: David Aron, M.D., MS, Director, Cleveland VA Medical Center 14(W), VA HSR&D Center for Quality Improvement Researach, 10701 East Boulevard, Cleveland, OH 44106; Tel: (216)421-3098; Fax: (216)231-3427; Email: david.aron@med.va.gov Research Objective: To compare health care system performance using weighted versus dichotomous measures. Study Design: We obtained de-identified VHA chart abstraction data for all veterans identified as having diabetes between October 1, 2000 and September 30, 2001 with one or more visits in the previous 12 months. We constructed 3 dichotomous performance measures based upon 2001 American Diabetes Association guidelines: <8% for A1c, <140/80 mm/Hg for blood pressure (BP), and <130mg/dl for Low density lipoprotein cholesterol (LDL-C). We utilized published estimates from the Centers for Disease Control and Prevention to calculate Quality Adjusted Life Years (QALYs) gained from reduction in A1c from 8 to 7%; BP from 154/86 to 144/82 mm/Hg; and LDL-C from 192 to 142 mg/dl using the following age strata: 25-34; 35-44; 45-54; 55-64; 65-74; 75-84; and 85 and above. Similar to dichotomous performance measurement, values above the upper limit for any parameter received no QALYs gained; values below the lower limit received maximal QALYs gained. QALYs gained for each risk factor were calculated for each patient and expressed as a percent of maximal QALYs possible. These results were aggregated by facility for each risk factor. In addition, a summary measure combining all three risk factors was calculated. Similarly, a summary measure combining the three dichotomous measures was calculated as the average of the three measures. If a test was not performed it was counted as above the upper limit for both the dichotomous and weighted measure. Population Studied: all veterans identified as having diabetes between October 1, 2000 and September 30, 2001 with one or more visits in the previous 12 months.The population was largely male (97.5%) and older (53.7% > 65). Principal Findings: There was substantial variation among the 141 facilities for both weighted and dichotomous measures. Mean performance as assessed by dichotomous measures was 62% {range 48-75%} for A1c, 56% {37-70%} for BP, 82% {66-91} for LDL-C, and 67% {55-76%} for the summary measure. By comparison mean performance as assessed by weighted measures (% maximal QALY gained) was 45 {31-60%} for A1c, 73 {55-84%} for BP, 86 {68-97%} for LDL-C, and 75 {59-84%} for the summary measure. Correlation (Spearman rank) between the two types of measures for summary score was very good (R= 0.75). Of the 14 poorest performers (bottom 10%) assessed by dichotomous measures, when assessed with weighted measures 12 remained in the bottom 10% and 2 moved to the 3rd quartile (below median). Of the 14 best performers (top 10%) by dichotomous measures, when assessed with weighted measures, 6 remained in the top 10%, 2 moved to the 2nd quartile, 5 moved to the 3rd and 1 moved to the 4th. Conclusions: Mean A1c adherence was overestimated by 17% and BP underestimated by 17% with the dichotomous measure relative to the weighted measure. Both methods similarly identified the worst performing facilities. However, there was marked discordance in the rankings of best performers. Implications for Policy, Delivery, or Practice: Since the use of QALYs may more accurately reflect population health, we propose that performances measures be based upon QALYs rather than dichotomous thresholds. As a continuous measure, QALYs may be perceived as fairer because “partial credit” can be given. Such measures may be more effective in organizational improvement efforts. Primary Funding Source: VA ●The Quality Olympics: A System-Wide Approach to Improving Diabetes Care David Aron, M.D., MS, Scott Ober, M.D., MBA, Aleece Caron, Ph.D., Michelle Davidson, MEd, David C. Aron, M.D., MS Presented By: David Aron, M.D., MS, Director, Cleveland VA Medical Center 14(W), VA HSR&D Center for Quality Improvement Researach, 10701 East Boulevard, Cleveland, OH 44106; Tel: (216)791-3800x4947; Fax: (216)231-3427; Email: david.aron@med.va.gov Research Objective: Unacceptable levels of adherence to recommended preventative care have prompted physician managers to search for interventions to improve outcomes, eg; clinical reminders, audit and feedback. Key components necessary for successful continuous quality improvement (CQI) include a strong and innovative quality leadership together with top hospital management and a current reporting system that can be accessed by local managers. The competition among Community-based Outpatient Clinics (CBOCs) termed the “CBOC Olympics,” was designed to provide local managers with such a tool. This competition provided broad dissemination of real-time data and was tied directly to network performance measures, as well as providing positive incentives to improve adherence and enhancing multidisciplinary teamwork. The goal of this project was to develop and evaluate a CQI intervention to improve adherence among CBOC primary care providers to clinical reminders for their patients with diabetes mellitus(DM). Study Design: We created a inter-clinic competition to provide continuous feedback to clinicians and staff about every CBOCs level of adherence to clinical reminders. Specifically, we tracked 6 DM measures: eye exam, foot exam, lipid screening, hypertension screening, hemoglobin A1c monitoring, and hemoglobin A1c control. VHA network leadership determined target measures for satisfactory and exceptional performance. Population Studied: Data from primary care physicians from 13 VA CBOCs in Northeast Ohio were reviewed monthly from June 2002 through November 2004 to determine guideline adherence to diabetes clinical reminders. CBOCs were divided into 3 tiers in order to adjust for patient volume.Data was analyzed using Statistical Process Control (SPC). Site visits were performed monthly to all clinics by an interdisciplinary senior management team. Monthly Olympic winners from each tier were disseminated to all employees and were based solely on adherence to clinical reminders. Characteristics of successful CBOC Olympic winners are included. Principal Findings: Baseline results were high; every CBOC achieved satisfactory adherence in 2002. There was an overall trend of improvement; the average overall initial adherence to clinical reminders was 81% and increased to 83%. This increase has been maintained. There was also a decrease in overall variation to adherence rates. Between June 2002 and May 2003, the Upper and Lower Control Limits (UCL and LCL) were 0.955 and 0.693 respectively. This variation was reduced between July 2003 and October 2004: the UCL was 0.95 and the LCL was 0.71. However, there was considerable variation among the CBOCs. Of the 13 CBOCs, 7 showed significant, maintained improvement. The average adherence rate for the first and second years ranged from 76.6% to 85.0% and from 80.9% to 89.7% respectively. Examination of tier results indicate that the smallest and largest tiers made improvements, but the middle tiers adherence declined. Characteristics of successful CBOCs include enthusiastic physician managers, multiple opinion leaders, good interdisciplinary collaboration, and ownership of the data. Conclusions: Successful implementation the CQI initiative depended on continuous reinforcement, clarity of objectives, and effective opinion leaders. The system as a whole reduced variation in adherence to clinical reminders, but the level of adherence desired was not achieved. Implications for Policy, Delivery, or Practice: System redesign may be necessary to achieve optimal adherence rates to DM clinical reminders. Primary Funding Source: VA ●Leaping the Chasm: Reducing Inpatient Mortality at Academic Medical Centers Raj Behal, M.D., MPH Presented By: Raj Behal, M.D., MPH, Medical Director, Clinical Effectiveness & System Redesign, University HealthSystem Consortium, 2001 Spring Road Suite 700, Oak Brook, IL 60523; Tel: (630)954-4892; Fax: (630)954-5879; Email: rbehal@uhc.edu Research Objective: To reduce risk adjusted inpatient mortality at academic medical centers and to understand organizational factors associated with higher than expected mortality rates. Study Design: Sequential quality improvement projects were conducted at 3 US academic medical centers. Key steps included: 1) analysis of risk adjusted comparative data to identify populations with higher mortality; 2) evaluation of coding and documentation practices and their impact in risk adjustment; 3) interviews with physicians, nurses, pharmacists, house officers, quality improvement staff, hospital and clinical leadership; 4) assessment of standardization of care using evidence-based practices; 5) study of systems factors and clinical processes; 6) identification of best practices at other centers; 7) presentation of findings and recommendations to senior leadership. Population Studied: Inpatient adult and pediatric inpatients at 3 academic medical centers in the United States. Principal Findings: The following contributory factors were identified: hand-offs with inadequate information exchange; incidence of preventable complications; failure to recognize and act upon early signs of patient deterioration; impeded patient flow into and out of intensive care units; suboptimal selection and optimization of operative cases; and, variations in care and use of evidence-based therapies. Follow up review of outcomes showed that the first hospital reduced the standardized mortality rate from 167 to 110; the second hospital reduced the SMR from 108 to 92; the third hospital reduced mortality for sepsis and cardiac bypass surgery by over 50%. Conclusions: Systematic identification of factors contributing to inpatient mortality and subsequent interventions targeting these factors can effectively reduce excess mortality. Leadership focus on quality of care, effective use of data, physician leaders' participation in improvement, and effective structures for quality improvement were critical for success. Implications for Policy, Delivery, or Practice: Redesign of care processes and systems is important for radical improvement in mortality in academic medical centers. Assigning accountability based on organizational goals is critical for sustaining results. Interventions targeting mortality should be encouraged by payers and policy makers. Primary Funding Source: No Funding Source ●Use of Incentives to Stimulate Physician Behavior Change to Improve Quality: Physician Reactions and Potential Barriers Sandra Berry, MA, Stephanie Teleki, Ph.D., Cheryl L. Damberg, Ph.D., Chau Pham, MA, Sandra Berry, MA Presented By: Sandra Berry, MA, Senior Behavioral Scientist, Health, RAND, 1776 Main Street, Santa Monica, CA 90407; Tel: (310)393-0411 x7779; Fax: (310)393-4818; Email: sandra_berry@rand.org Research Objective: To understand physicians’ opinions of and experiences with financial incentives for quality, and identify barriers to using incentives to drive quality improvements. Study Design: Structured telephone interviews and focus groups with Blue of California (BCC) Preferred Provider Organization (PPO) physicians who are part of BCC’s Physician Quality and Incentive Program (PQIP). PQIP is a RWJ Rewarding Results demonstration project. All BCC PPO physicians who met measurement criteria (n=11,000) were scored on a set of clinical, generic prescribing, and administrative measures and provided a comparative performance report card. A subset (n=1,200) in Northern California was also eligible for a financial bonus. One-hour telephone interviews were conducted with 25 physicians (15 men, 10 women; 16 pilot, 9 non-pilot; majority aged 50-59). Additionally, two focus groups (n = 24 doctors) were held in the pilot area (18 men, 6 women; all pilot; majority aged 4059). A wide range of specialties was represented. Participants were asked about their opinions of health care quality and financial incentives, the quality of care they provide and help needed to improve, and experiences with financial incentives for quality, especially PQIP. Population Studied: A convenience sample of BCC PPO physicians. Principal Findings: Physicians reported that their primary objective in practicing medicine is “doing what is right” for the patient. However, many obstacles were cited: limited time, coding challenges, reimbursement delays/reductions, noncompliant patients, lack of systems to track care, unrealistic expectations, and personal sacrifice. They agreed selfmonitoring is important and stated they did this by talking/comparing themselves to colleagues, attending conferences, reading journals, and following guidelines. Overall, physicians felt financial incentives are a good idea because “physicians are underpaid;” however, they did not believe it possible to accurately measure quality and tie money to quality performance, feel the bonus is money already owed, and sponsors’ motives are suspect. Most feel that opinions of and comparisons to colleagues are more important than money and that they would respond to comparative information; they noted that money will only change behavior if measures are reasonable and easy to implement, and the amount offered represents minimally 5-10% of their annual salary. Many said they are overwhelmed with instructions from the many plans with which they contract; their reaction is to “tune out” the information and “just practice the best I can.” Nearly all physicians were against public reporting, citing inaccurate measurement and misinterpretation by patients, which could damage physician careers. Physicians felt that EMRs and reminder systems are key to improving quality, and that measurement by medical specialty societies would be more accepted. Awareness of any quality incentive program was minimal, with some physicians accidentally discarding bonus checks. Conclusions: While many physicians indicated that they would readily accept financial incentive payments, they remain skeptical about measurement accuracy, the intentions of program sponsors, and the likelihood of these programs to modify their behavior. Implications for Policy, Delivery, or Practice: Pay-forperformance programs must take physicians’ perspectives into account, given that their buy-in is a crucial component of success. Primary Funding Source: RWJF, California HealthCare Foundation ●Are Hospitals that Refuse to Participate in Adverse Events Studies Different from Those that Accept? Regis Blais, Ph.D., Marie-Laure Boursiquot, M.D. Presented By: Regis Blais, Ph.D., Professor, Department of Health Administration, University of Montreal, PO Box 6128, Station Centre-ville, Montreal, H3C3J7; Tel: (514) 343-5907; Fax: (514) 343-2448; Email: regis.blais@umontreal.ca Research Objective: Patient safety has become a major preoccupation in health care systems. One common indicator of patient safety is the rate of adverse events (AE) among hospital patients. Researchers who want to conduct AE studies have to get permission from hospital authorities to access patient charts. Some hospitals may refuse to give that permission. If those that refuse provide poorer quality care than those that accept, study results may not be valid. The objective of this study was to compare hospitals that refused to participate in one AE study to those that accepted, in terms of hospital structure and services, patients' characteristics and some patient outcomes. Study Design: A sample of hospitals from Quebec (Canada) were invited to participate in an AE study: 18 accepted and 5 refused. The two groups of hospitals were compared on their structure and services (number of beds, services and admissions per year, average length of stay, proportion of day surgery), patients' charateristics (age, sex, case severity) and outcomes (in-hospital death, 30-day and 90-day readmission). Data mainly came from discharge abstract databases. Population Studied: Population studied included 14 large community hospitals and 9 teaching hospitals located within 260km of Montreal, and the 293000 patients who were hospitalized in these hospitals over a year. Principal Findings: The two groups of hospitals did not differ in terms of structure or services. The patients of hospitals that refused to participate in the AE study were nearly 5 years older on average, but the difference did not reach statistical significance. The proportion of females and the average patient case severity did not differ. In-hospital mortality was higher (significant) in non participating hospitals, as was readmission (non significant). Multivariate regression showed that after adjusting for patients' characteristics, refusing to participate in the AE study was not associated with in-hospital mortality or readmission. Conclusions: Hospitals that refuse to participate in AE studies may have different patients' case mix and worse outcomes than those that accept. These could contribute to their refusal. AE studies with high refusal rates could be biased toward better outcomes than the reality. Implications for Policy, Delivery, or Practice: Health care settings sampled in AE studies should be representative of all settings. To gain collaboration from potentially low health care performers, the purpose of AE studies should be quality improvement and not punishment. Primary Funding Source: Quebec ministry of Health ●Methods to measure the Impact of Nurse-to-patient staffing ratios at the micro level Terry Capuano, RN, MSN, MBA, Joanna Bokovoy, RN, DrPH, Nancy Davies-Hathen, RN, MSN, CCRN, Elizabeth Karoly, RRT, MBA, Martin Everhart, BA, MPA, Paulette Kennedy, RN, MSN, Carol Cyriax, RN, BSN, Tina Dalessandro, RN, Stephen Bogar, BS, Lisa Romano, RN, MSN, Stephanie Pacelli, BS, Patricia Matula, RN, MSN Presented By: Joanna Bokovoy, DrPH, Director, Healthcare Research, Clinical Services, Lehigh Valley Hospital, 1245 South Cedar Crest Boulevard, Allentown, PA 18049; Tel: (610)402.1813; Fax: 610.402.1408; Email: joanna.bokovoy@lvh.com Research Objective: To Develop methods that will allow researchers to effectively study the impact of nurse staffing on nurse and patient satisfaction and patient outcomes at a micro level in a Magnet Hospital setting. Study Design: Prospective, longitudinal study of four similar medical-surgical units – one with a capped nurse-to-patient staffing ratio and the other three with usual staffing ratios. Methods and processes were developed to effectively evaluate the unit level the impact of a capped nurse-to-patient staffing ratio on nurse and patient satisfaction and nurse-sensitive patient outcomes. Population Studied: Staff and patients on four similar medical-surgical units. Principal Findings: Since many factors besides a capped nurse patient staffing ratio can impact quality patient and staff outcomes, those factors had to be identified and addressed. Choosing a Magnet Hospital environment was one method, and a second method was to build on the improvements defined from our use of a validated model to evaluate the impact of our nursing delivery a year before our study began. An effective process was developed to maintain nurse to patient ratios on the intervention medical-surgical unit at 1:4-5, not to exceed 6 patients on days/eve shifts, and 1:7, not to exceed 10 on nights. A variety of measures were developed or refined to determine the effects of this staffing model on nurse-sensitive patient outcomes, patient/family and staff satisfaction, LOS, as well as other quality of care measures. Included in these were relevant contextual information, detailed staffing profiles (includes numbers, Benner stage of clinical competence, qualifications, etc.), available workload data (includes dependency/complexity), reliable and valid quality assurance/outcome measures and comprehensive costing data. Our overall approach appears methodologically robust, reliable, and replicable. Conclusions: Nurse staffing ratio issues can be effectively measured on the micro level. Implications for Policy, Delivery, or Practice: Defining the role that nurse-patient ratios do or do not play in quality of care is important because of current policy to mandate nursepatient ratios. Primary Funding Source: No Funding Source ●Impact of Personal Digital Assistant Devices on Medication Safety in Primary Care James D. Bramble, Ph.D., Kimberly Galt, Pharm.D., Ann Rule, Pharm.D., Wendy Taylor, BBL, Mark Siracuse, Ph.D., Bartholomew Clark, Ph.D., Wayne Young, Pharm.D. Presented By: James D. Bramble, Ph.D., Associate Professor, Creighton Health Services Research Program, Creighton University Medical Center, 2500 California Plaza, Omaha, NE 68178; Tel: (402)280-4129; Fax: (402)280-4809; Email: jbramble@creighton.edu Research Objective: This study examines the impact of personal digital assistants (PDAs) use on potential prescribing errors in primary care office-based practices. PDAs offer great potential for improving quality at the point-of-care. The PDA offers the opportunity to access drug information, enter data, and print information pertaining to the patient including the prescription. PDA use allows for easy access to electronic information at the point-of-care that has the potential to reduce prescribing errors. Additionally, computer generated prescriptions reduce legibility problems. Specifically, this study will measure the occurrence of potential preventable prescribing related errors. Study Design: Using a randomized controlled trial we examined the impact of PDAs on potential prescribing errors. The intervention group received training to use the PDA for prescription generation and drug information retrieval. After successful completion of initial training of the intervention group, they were encouraged to use the PDA in their prescribing process. However, use was voluntary, resulting in prescriptions being generated by PDA and traditional methods based upon providers’ daily behaviors. Control group physicians maintained their traditional method of prescription generation. We assessed the face of the prescription and compared baseline and post-intervention data (500 prescriptions each) on errors of legibility, commission, omission, and interpretation. Population Studied: We recruited 78 primary care physicians (85% family practitioners, 15% internal medicine) from 30 primary care offices in the Omaha metropolitan statistical area and one rural clinic. The average age was 42 with 33% female and the majority family practitioners (85%). Principal Findings: The intervention group generated 43% of their post intervention prescriptions using the PDA. Potential errors associated with legibility decreased from 9.1% to 2.7%. The remaining legibility error we attribute to the less than 100% use of the PDA in the prescribing process. We also observed significant reduction in errors of all categories of omission such as failing to include the patient’s age (including the patient’s age increased from 4.5 to 40.8%). Use of abbreviations on the face of the prescription, often not interpretable by the patient, also decreased. Use of abbreviations for the route of administration decreased from 63.2% to 36.7% while use of abbreviations for frequency of administration decreased from 85.8% to 50.7%. Two types of errors increased including the vagueness of instructions and identifying the wrong technique for drug administration. Conclusions: This project introduced the use of PDAs and corresponding applications that have the potential to enhance patient safety by reducing prescription errors (e.g., legibility errors). Additionally, the study showed that these improvements are achievable within the typical small to medium primary care office based practice. This work also reveals the existence of barriers to achieving 100% success (i.e., not all prescriptions were written on the PDA by the intervention group). Despite, not having full adoption of using the PDA, there was an overall improvement in patient safety. Implications for Policy, Delivery, or Practice: This research generated empirical evidence demonstrating the contribution of PDAs to potentially reduce medical errors and improve patient safety in the primary care physician office-based practice environment. These findings inform policy makers and administrators of the benefit of adopting hand held technology. Primary Funding Source: AHRQ ●Beneficial Effect of Physical Therapy in Lower Extremity Trauma Renan Castillo, MS, Ellen J. MacKenzie, Ph.D., Michael J. Bosse, M.D., LEAP Study Group Presented By: Renan Castillo, MS, Doctoral Student, Health Policy and Management, Johns Hopkins University Bloomberg School of Public Health, 624 North Broadway, Room 544, Baltimore, MD 21205; Tel: (410) 614-4024; Fax: (410) 6142797; Email: rcastill@jhsph.edu Research Objective: Despite the fact that physical therapy services are routinely prescribed following lower extremity trauma, there is little evidence that these services are beneficial to the patient. The current study examines the effect of physical therapy use on a range of measures of physical impairment in a cohort of lower extremity trauma patients. Study Design: Patients (N=382) from a larger study of severe leg trauma were interviewed by a research coordinator and examined by an orthopedic surgeon and a physical therapist during the initial hospitalization, and at 3, 6, 12 and 24 months post discharge. Two measures of unmet need for physical therapy were defined – based alternatively on need as assessed by the orthopedic surgeon and the physical therapist. Specifically, the orthopedic surgeon and a physical therapist independently evaluated each patient during follow up and were asked whether they felt the patient would benefit from physical therapy between that time point and the next follow up visit. Patients who were classified as needing physical therapy services over the given time period and who reported receiving no physical therapy at the end of that time period (next follow up time point) were classified as having unmet need as for that follow up segment (from the independent perspectives of both the orthopedic surgeon and the physical therapist). Multivariate regression techniques were used to compare improvement in measures of physical impairment between the groups with met and unmet need over the 3-6, 6-12 and 12-24 month periods. Population Studied: Patients in the current analysis constitute a subgroup of a larger study conducted to assess the outcomes of amputation or reconstruction following limbthreatening lower extremity trauma below the distal femur. Only patients treated by reconstruction were selected for this analysis. Principal Findings: A range of measures of physical impairment were used, including: percent impairment in knee and ankle range of motion, reciprocal stair climbing pattern, gait deviations when walking, self selected walking speed greater than 4 feet per second, and the Functional Independence Measure (FIM)... Patients with unmet need for physical therapy as assessed by the physical therapist were statistically significantly less likely to improve in all five of the selected domains of physical impairment than patients whose physical therapy need was met. These results held true even after adjustment for patient socio-demographic, personality, and social resources, as well as injury and treatment characteristics, reported pain intensity, and impairment level at the beginning of the study time period. Patients with unmet need for physical therapy as evaluated by an orthopedic surgeon were generally worse off than patients with met need, but in only one of the five selected impairment measures (range of motion)was the difference statistically significant. Conclusions: The results are consistent with a beneficial effect of physical therapy following lower extremity trauma. Furthermore, the results suggest that physical therapists may be more likely to identify patients who could benefit from such services than orthopedic surgeons. Implications for Policy, Delivery, or Practice: The results point to a need for improved standards for the prescription of physical therapy services, and highlight the importance of involving a physical therapy professional in the prescribing process. Primary Funding Source: National Institutes of Health ●Using Patient Safety Indicators to Identify Outlier Hospitals: A Picture is Worth 1000 Statistics Cindy Christiansen, Ph.D., Peter Rivard, MHSA, Shibei Zhao, MPH, Susan Loveland, MAT, Dennis Tsilimingras, M.D., MPH, Amy K. Rosen, Ph.D. Presented By: Cindy Christiansen, Ph.D., Associate Professor, Health Services, Boston University, 200 Springs Road, Building 70, Bedford, MA 01730; Tel: (781)687 2915; Fax: (781) 87 3106; Email: cindylc@bu.edu Research Objective: Patient Safety Indicators (PSIs), developed by the Agency for Healthcare Research and Quality (AHRQ), are useful for identifying potential in-hospital patient safety events. However, characteristics inherent to indicators make it difficult to interpret PSI results when trying to understand rates from multiple indicators. Our objectives were to 1) compare Bayesian and average-ranking methodologies for selecting hospitals that have extremely high or low rates and 2) improve presentation of hospital-level information from PSI analyses. Study Design: We used a retrospective one-year cohort to calculate hospital-level PSI counts (numerators), acute-care hospitalizations (denominators), observed, expected, and AHRQ-smoothed rates (estimates of rates that are both riskadjusted and smoothed across multiple years of data) for 16 PSIs using Department of Veterans Affairs (VA) administrative data from fiscal year 2001 (FY’01) and AHRQ PSI software (version 2.0). From observed (O) to expected (E) ratios and Bayesian models, distributions representing “true” O/E ratios for each indicator at 118 hospitals were determined. For the 6 most frequent PSIs, we used simulation methods to obtain hospital-level posterior densities for the six-indicator combination of O/E ratios. We compared rankings of median ratios from the posterior densities to average rankings of smoothed rates for hospitals that ranked in the top 10 (or bottom 10) by either method. Graphical methods were developed to facilitate communication of results. Population Studied: Patients discharged from Veterans Affairs Medical Centers (VAMCs) between October 1, 2000 and September 30, 2001. The sample consists of 430,552 hospitalizations incurred by 281,423 patients at 118 VAMCs. Principal Findings: Both methods selected 7 of the same hospitals in the top 10 and 7 in the bottom 10. Six hospitals chosen by the Bayesian but not by the average-ranking method had rankings that varied widely across indicators. By averaging ranks, the strength of evidence for having extreme O/E ratios is lost; hospitals with ranks that vary widely, i.e., those having some very high and some very low ranks for the 6 indicators, are not identified as extreme. The Bayesian method retains the evidence by accounting for the certainty in the estimates of the individual indicators (using the posterior densities) and incorporates known correlation across PSIs. Using the Bayesian method we can conclude, with more certainty than not, that the top 10 hospitals had 23% to 46% fewer PSIs than expected, and the bottom 10 had 39% to 92% more than expected. Posterior density graphs demonstrate how the certainty of the estimates affects the combination ratio and add another dimension when translating the results to policy-makers. Conclusions: Bayesian analyses provided more information, appropriately recognized the uncertainty in estimates of performance, and, with the help of graphics, were as easy to understand as results from selection based on average rankings. Implications for Policy, Delivery, or Practice: Translating patient safety analyses into concise, useable information is both science and art. Bayesian models and graphics are tools that all health care systems should consider when conducting patient safety research and when translating results to consumers and policy-makers. Primary Funding Source: VA ●Consumer Satrisfaction with Primary Care Provider Choice and Associated Trust Ming Ying Lisa Chu-Weininger, Ph.D., Rajesh Balkrishnan, Ph.D., Lu Ann Aday, Ph.D., Mark A. Hall, JD Presented By: Ming Ying Lisa Chu-Weininger, Ph.D., Doctoral Student, Management, Policy, and Community Health, University of Texas School of Public Health, 2120 El Paseo, Unit 408, Houston, TX 77054; Tel: (713) 797-1795; Email: Ming.Ying.L.Chu-Weininger@uth.tmc.edu Research Objective: Trust is important in medical relationships and for the achievement of better health outcomes. Developments in managed care in the recent years are believed to affect the quality of healthcare services delivery and to undermine trust in the healthcare provider. Physician choice has been identified as a strong predictor of provider trust but has not been studied in detail. Consumer satisfaction with primary care provider (PCP) choice includes having or not having physician choice. The study specific aims were: (1) to determine variables related to the factors: consumer characteristics and health status, information and consumer decision-making, consumer trust in providers in general and trust in the insurer, health plan financing and plan characteristics, and provider characteristics that may relate to PCP choice satisfaction; (2) to determine if the factors in aim one are related to PCP choice satisfaction; and (3) to analyze the association between PCP choice satisfaction and provider trust, controlling for potential confounders. Study Design: Cross-sectional survey Population Studied: Analyses were based on secondary data from a random national telephone survey in 1999, of residential households in the United States which included respondents aged over 20 and who had at least two visits with a health professional in the past two years. Among 1,117 eligible households interviewed (response rate 51.4%), 564 randomly selected to respond to insurer related questions made up the study sample. Principal Findings: Analyses using descriptive statistics, and linear and logistic regressions found continual effective care and interaction with the PCP beyond the medical setting most predictive of PCP choice satisfaction. Consumer characteristics, information and decision-making, provider characteristics and plan characteristics were factors found to predict consumer’s report of enough PCP choice. The information and decision making factor and provider characteristics factor were also found to predict provider trust and renamed ‘continuity of care related’ and ‘provider related’ in the revised framework. All in all, PCP competencies were significant in predicting PCP choice satisfaction and provider trust. Conclusions: It appeared that PCP choice satisfaction as a predictor of provider trust overlapped with other predictors of provider trust, such as communication between the provider and consumers. Although it may sound an aggressive task, to examine all the known predictors of provider trust in one study in the future. It may provide a holistic view of what’s relatively important in predicting provider trust and in sustaining trust in the long term. Findings alluded to the importance of PCP competency training to include interpersonal (which also entails cultural) competency. Implications for Policy, Delivery, or Practice: This study has implications for the development of health services information. In the mist of growing consumerism, relevant, quality, cognitive, and accessible provider specific information would be helpful to the consumer. The goal of information is for consumers to make a fair choice, without too much complication to the decision making process. That is why it makes sense to study ‘trust in provider specific information.’ Provision of useful information may be a better business strategy compared to hard core media marketing. Primary Funding Source: RWJF ●Variations in Quality Outcomes Among Hospitals in Different Types of Health Systems, A Panel Study Askar Chukmaitov, M.D., MPA, Gloria J. Bazzoli, Ph.D., Mei Zhao, Ph.D. Presented By: Askar Chukmaitov, M.D., MPA, Research Associate/Ph.D. Candidate, Health Administration, Virginia Commonwealth University, 1008 East Clay Street, P.O. Box 980203, Richmond, VA 23298; Tel: (804) 827-1811; Fax: (804) 828-1894; Email: achukmaitov@vcu.edu Research Objective: Although prior research has found differences in costs and financial performance across different types of hospital systems, there has been no systematic study of variations in patient quality of care or safety indicators across different systems. Our study examines whether five main types of health systems - centralized, centralized physician/insurance, moderately centralized, decentralized, and independent - as well as other hospital characteristics are associated with differences in quality of patient care. Study Design: Data were assembled for 6 years (1995 – 2000) from multiple sources. We used 4 AHRQ risk adjusted inpatient quality indicators (IQIs) and 5 risk-adjusted patient safety indicators (PSIs) as dependent variables. In addition to examining hospital system type, we also examined hospital characteristics including compliance with selected JCAHO standards. The type of system to which a hospital belongs may influence its ability to meet these standards as well as influence its quality of care. Random effects models were used in the analysis. Population Studied: All short term general medical-surgical hospitals from 11 states (AZ, CA, CO, FL, IA, MD, MA, NJ, NY, WA, and WI) in operation between 1995 and 2000 that participated in the AHRQ’s HCUP State Inpatient Databases (SID) program. Principal Findings: Overall, our findings have not identified a clear pattern (neither worsening nor improvement) in quality performance across different types of health systems. However, selected quality and patient safety indicators show significant variation for selected system types. Hospitals in moderately centralized health systems may provide better care for AMI patients. Hospital membership with centralized physician/insurance system type may be associated with worsening outcomes for post-operative pulmonary embolism or deep vein thrombosis. Independent hospital systems are worse performers in terms of decubitus ulcer in comparison with all other types of health systems. In addition, hospital compliance with the JCAHO standard for availability of patient specific information is associated with better outcomes for pneumonia patients. Conclusions: Structural differences in health systems have not affected quality performance in a definite way during the 1995 – 2000 period. Organizational characteristics of moderately centralized health systems, such as moderate centralization of authority at the system level and high differentiation of services at the hospital level, may be contributing to better inpatient quality performance. Low-tomoderate centralization combined with little differentiation of services, as in independent hospital systems, may be resulting in worsening of patient safety performance. Hospital compliance with selected JCAHO standards may lead to improvements in the process of care delivery and better clinical outcomes. Implications for Policy, Delivery, or Practice: Hospital systems hold much potential for hospitals in improving patient quality of care and safety because they provide a laboratory for studying the health care process and sharing lessons across multiple institutions. Our findings suggest that hospital and health system managers may have not systematically reaped the benefits of this potential in improving quality performance of their organizations. Based on our findings, we recommend that future studies use a combination of IQIs and PSIs when examining institutional quality of care because both provide different and complementary information. Primary Funding Source: AHRQ ●Racial and Ethnic Disparities in Patient Safety: Findings from the 2004 National Healthcare Disparities Report Elizabeth Dayton, MA, Ernest Moy, M.D., MPH, Roxanne Andrews, Ph.D., Anna Poker, MS, RN Presented By: Elizabeth Dayton, MA, Junior Service Fellow, Department of Health and Human Services, Agency for Healthcare Research and Quality, 540 Gaither Road, Rockville, MD 20850; Tel: (301)427-1320; Fax: (301)427-1341; Email: edayton@ahrq.gov Research Objective: Our objective is to describe racial/ethnic disparities in patient safety in American healthcare, drawing on research from the 2004 National Healthcare Disparities Report, a congressionally mandated annual report analyzing differences in healthcare by race/ethnicity and several other factors. Study Design: Data come from two nationally representative sources: the Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases (SID) and the Medical Expenditure Panel Survey (MEPS). A sample of 7 million all-payer hospital discharge abstract data from 22 SID (states with good quality race data) applied to the AHRQ Patient Safety Indicator software was used to develop national estimates for hospital patient safety events. The MEPS collects annual household data via computer-assisted surveys. Each MEPS respondent undergoes five rounds of interviews over a two and a half year period, and a sample of respondents’ data is linked to information from their employers and medical and insurance providers. The MEPS sample size is about 10 thousand households. Twenty patient safety measures are analyzed. Data for non-Hispanic blacks, non-Hispanic Asians and Pacific Islanders, and Hispanics are compared with data for nonHispanic whites revealing components of patient safety where minorities receive better, about the same, or worse care than whites. Population Studied: The national U.S. population (MEPS) and discharges from US community hospitals (HCUP). Principal Findings: For a quarter of measures minorities have poorer patient safety than non-Hispanic whites. For over a quarter, minorities have better patient safety than whites. However, for nearly half of measures minorities and whites have fairly equal patient safety. More specifically, rates of iatrogenic pneumothorax (puncture of the lung) are lower among Hispanics than whites, deaths from complications potentially resulting from care are higher among Asians and Pacific Islanders (APIs) than whites, rates of deep vein thrombosis (blood clots forming in the legs following surgery) and postoperative pulmonary embolus (blood clots traveling from the legs to the lungs following surgery) are higher among blacks and lower among APIs than whites, and rates of postoperative septicemia (life-threatening invasion of the bloodstream by microorganisms) are higher among blacks and Hispanics than whites. Conclusions: There are significant disparities in patient safety between minorities and whites in the United States. For some measures minorities have worse patient safety, for others they have better patient safety, and for many measures no racial/ethnic difference is observed. Analyzing and understanding differences in patient safety may prove a valuable tool for improving healthcare. Implications for Policy, Delivery, or Practice: Understanding patient safety differences may guide improvement efforts and inform policy: differences emphasize the best and poorest performing components of healthcare, suggesting best practices and components of care that would most benefit from intervention. Promoting awareness of disparities is an important first step towards their elimination. Primary Funding Source: AHRQ ●Effect of Short Stay Patients on the Measurement of Nursing Sensitive Quality Indicators Nancy Dunton, Ph.D., Byron Gajewski, Ph.D. Presented By: Nancy Dunton, Ph.D., Research Associate Professor, School of Nursing, University of Kansas Medical Center, 3901 Rainbow Boulevard Mail Stop 4043, Kansas City, KS 66160; Tel: (913) 588-1456; Fax: (913) 588-4531; Email: ndunton@kumc.edu Research Objective: The valid measurement of nursing sensitive quality indicators is fundamental to quality improvement activities in acute care hospitals. Short stay patients (e.g., same day surgery patients and observation patients) add to nurses’ workloads, but are not captured in midnight censuses. Further, the growth of managed care and improvements in medical procedures have increased the prevalence of short stay patients. This presentation establishes the effect of including short stay patient hours in two quality performance indicators endorsed by the National Quality Forum: nursing hours per patient day (NHPPD) and the patient fall rate. Study Design: The National Database of Nursing Quality Indicators (NDNQI), a project of the American Nurses Association, has collected quarterly data on nursing sensitive quality indicators since 1998. Until the fourth quarter of 2003, the number of patient days was measured by the midnight census. Since that time, hospitals have had the opportunity to report on care hours for short stay patients. This analysis was based on a correlational design comparing five specifications of patient census, used in the denominators of NHPPD and the fall rate: midnight census, midnight census plus actual short stay hours, midnight census plus an estimate of short stay hours, actual hours for all patients, and the average of multiple censuses throughout the day. The various indicator specifications were examined for five unit types: critical care, step down, medical, surgical, and combined medical-surgical. Population Studied: The analysis was based on NDNQI data for the third quarter of 2004 from nearly 500 hospitals across all 50 states. The patient care unit was the level of analysis and over 4,700 units were included in the analysis. Principal Findings: Nearly half (48%) of the units used midnight census, 14% used midnight census plus actual hours, 5% used midnight census plus an estimate of short stay hours, 6% used actual hours for all patients, and 9% used an average of multiple censuses throughout the day. The interaction between unit type and patient day method was significant for NHPPD (F=4.198). The simple main effects for patient day method were significant for surgical and combined med-surg units. The interaction between unit type and patient day method for fall rates was not significant. The main effect for patient day method on fall rates was significant (F=14.55). Conclusions: Incorporating the presence of short stay patients significantly altered the NHPPD and fall rate indicators, reflecting an improved measurement of nursing workload. The four data capture methods for incorporating short stay patient hours produced different results. In particular, the midnight census plus estimate of short stay hours method and the average of multiple daily census measures may not be as valid as methods that incorporate actual hours. Implications for Policy, Delivery, or Practice: Standardizing the measurement of actual patient load in acute care hospitals will improve the validity of national nursing quality indicators, providing hospitals with improved tools for targeting quality improvement activities. Moreover, improving the validity of patient day measurement will enable improved specification of multivariate models used to examine the influence of nurse staffing on patient outcomes. Primary Funding Source: American Nurses Association ●Instruments for Measuring Outcomes in End-of-Life Care: An Evidence-Based Review Sydney Dy, M.D., MSc Presented By: Sydney Dy, M.D., MSc, Assistant Professor, Health Policy & Management, Johns Hopkins Bloomberg School of Public Health, Room 609, 624 North Broadway, Baltimore, MD 21205; Tel: (410)614-4047; Fax: (410)955-0470; Email: sdy@jhsph.edu Research Objective: (for the Southern California EvidenceBased Practice Center) Assessing end-of life care requires accurate, reliable, and comparable measures of patient and caregiver experience. Our aims were to explore psychometric properties of available instruments for evaluating end-of-life care and to describe the use of these measures in the end of life literature. Study Design: As part of an evidence-based review of end-oflife care, we identified measurement instruments used to define outcomes in interventions studies. We also identified interventions related to advance care planning, continuity, symptoms, caregiver burden, and satisfaction, and described the use of measures in these studies. Population Studied: Sources for our systematic review included Medline, the Cochrane Database of Reviews of Abstracts of Effects (DARE), the National Consensus Project for palliative care, and several recent systematic reviews including the Toolkit of Instruments to Measure End-of-Life Care (TIME) and reviews from Health Canada and the National Institute for Clinical Excellence (NICE), United Kingdom. Searches were limited to published articles in the English language involving adult human subjects. Principal Findings: From 24,423 total citations, we identified one high-quality recent systematic review that recommended 35 measures across 11 domains (TIME), and we identified an additional 48 measures with published psychometric data. Few of the instruments had been developed or evaluated in palliative care populations; most were developed specifically for cancer. Few were validated across different settings important for end-of-life care (e.g. hospitals, nursing homes, and hospices). Almost none have been compared across different populations. Eighty-one unique intervention studies met the inclusion criteria and underwent a detailed review. Twenty-one were related to advance care planning, 20 to continuity, 18 to symptoms, 13 to caregiver concerns, and 12 to satisfaction. These studies used 97 separate instruments to measure outcomes. There was little overlap in the instruments used across studies: 80/97 were used in only one study, and only 5 were used in 4 or more studies. Conclusions: Instruments used in end-of-life intervention studies lack sufficient evaluation across diseases, stages of illness, settings, and populations applicable to end-of-life care. Many intervention studies in end-of-life care did not use highquality instruments to measure outcomes. We identified little overlap in measurement instruments across studies. Implications for Policy, Delivery, or Practice: The lack of appropriate and uniform measures for evaluating interventions in end-of-life care limits the ability to synthesize this literature and its implications for im-proving health care. Further research should emphasize improving measures and standards for evaluating outcomes in end-of-life intervention studies, as well as further psychometric evaluation of existing measures. Primary Funding Source: AHRQ, National Institute for Nursing Research ●Risky Concomitant Medication Dispensing in Ambulatory Care Jennifer Elston Lafata, Ph.D., Lonni Schultz, MS, Ph.D., K. Arnold Chan, M.D., Sc.D., David H. Smith, MHA, Ph.D., Marsha A. Raebel, Pharm.D., Marianne Ulcickas Yood, MPH Presented By: Jennifer Elston Lafata, Ph.D., Director, Center for Health Services Research, Henry Ford Health System, 1 Ford Place, Suite 3A, Detroit, MI 48202; Tel: (313)874-5454; Fax: (313)874-7137; Email: jlafata1@hfhs.org Research Objective: To estimate rates of risky concomitant medication dispensing among ambulatory care patients. Study Design: We used automated membership, prescription drug and medical claims data to compile information on patient demographics, comorbidities and dispensings of medications of interest during baseline (07/01/1999 – 12/31/1999) and follow-up (01/01/2000 – 06/30/2001). Identification of risky medication pairs was based on FDA warning labeling, Hansten and Horn Ratings and input from practitioners practicing within the 10 participating health plans. Two definitions of concomitant use were used: same day dispensings and dispensings with overlap according to the dispensed “days supply.” Among enrollees with a dispensing during follow up for warfarin, digoxin, cyclosporine, lovastatin/simvastatin, other statins, and nitrates, we report the proportion and number with a concomitant dispensing of a risky medication pair. Population Studied: National cohort of 953,889 health plan members aged 18 years or older who were continuously enrolled with prescription drug coverage during baseline. Principal Findings: Among these health plan members, 110,802 were dispensed at least one of the 6 object drugs/drug classes of interest. The mean age of those dispensed a medication of interest ranged from 49 years (cyclosporine users) to 72 years (digoxin users), and males represented between 51% (digoxin users) and 100% (nitrate users) of the cohort. The proportion (number) of object drug/drug class users who received a risky concomitant dispensing was 1.3% (N=257) nitrate users, 8.9% (N=2,553) other statin users, 19.0% (N=10,321) lovastatin/simvastatin users, 24.3% (4,660) warfarin users, 26.4% (N=3,958) digoxin users, and 29.4% (N=210) cyclosporine users when concomitant use was defined using days supply overlap. When only same day dispensings were considered the proportions (numbers) were 0.3% (N=53), 2.7% (N=784), 9.2% (N=4,988), 8.7% (N=1,675), 15.8% (N=2,373) and 18.5% (N=132), respectively. Thus, although the rate of risky dispensing is greatest among patients receiving cyclosporine, the greatest absolute number of risky dispensings occurred among those dispensed lovastatin or simvastatin. For warfarin users, the most frequent risky concomitant dispensing was with NSAIDs (days supply/same day: 55.1%/46.1%). Among those dispensed digoxin, 68.4%/76.5% were also dispensed verapamil/ditiazem/bepridil. Verapamil/diltiazem was also the most frequent risky concomitant dispensing among those receiving cyclosporine (64.3%/69.7%) and lovastatin or simvastatin (52.8%/69.6%). Conclusions: Most research in the area of risky concomitant medication use has focused on co-prescribing in the hospital setting. We found evidence of risky concomitant medication dispensing among ambulatory care patients. In some cases, this occurs despite the availability of low risk alternatives (e.g., Tylenol as opposed to NSAID use among warfarin users, and other blood pressure medication as opposed to verapamil or diltiazem use among digoxin, cyclosporine and lovastatin or simvastatin users). Implications for Policy, Delivery, or Practice: Because using prescription claims data to monitor concomitant prescribing in the outpatient setting is relatively simple and inexpensive, an opportunity exists to augment current benchmarking and other quality improvement efforts to better understand how risky concomitant medication dispensing translates into adverse events and ultimately to improve medication safety in the outpatient setting. Primary Funding Source: AHRQ ●Inappropriate Prescribing for Medicare Managed Care Beneficiaries in the Last Year of Life Cheryl Fahlman, BSP, MBA, Ph.D., Danielle Doberman, MPH, M.D., Joanne Lynn, M.D. Presented By: Cheryl Fahlman, BSP, MBA, Ph.D., Senior Research Associate, Health Research and Educational Trust, 325 7th Street NW, Washington, DC 20004; Tel: (202)6262365; Fax: (202)626-2689; Email: cfahlman@aha.org Research Objective: The primary aim of this study is to assess violations of the Beers’ criteria for inappropriate prescribing in the elderly for a group of Medicare beneficiaries in their last year of life (LYOL) in a large national managed care organization (MCO) using a pharmacy benefits manager. This research provides the first claims-based data on Beers’ violations by disease and demographics at the end of life. Study Design: This research used drug claims and enrollment data collected between January 1998 and December 2000, supplemented by the Medicare denominator file. We based the analysis on the 2002 list of Beers’ criteria. We analyzed the relationship between socio-demographic descriptors, insurance characteristics, cause of death, and violations of the Beers’ criteria. Logistic regression techniques were used to estimate the determinants of Beers’ criteria violations. Population Studied: Four thousand six hundred and two beneficiaries qualified for the study, based on dying between 1999-2000, being continuously enrolled for two years before death and having at least one prescription filled in the LYOL. Principal Findings: Two thousand three hundred and forty five beneficiaries (52%) had at least one claim in the LYOL that violated a Beers’ criterion. Fully 13% experienced more than one unique Beers’ violation, and 5% have three or more violations during the study period. The most common violations were the use of propoxyphene (13.5%), followed by doxazosin (2.8%), and amitriptyline (2.1%). Based on total claims, cancer patients were most likely to receive propoxyphene (35.3%) followed by patients with a heart condition (29.6%). For amitriptyline and doxazosin, patients with a heart condition receive the greatest number of prescriptions (35.7% and 32.3% respectively) and cancer patients received the second highest amount (25.2% and 23.8% respectively). A large proportion of the potentially inappropriate prescribing involves psychoactive drugs (ten out of the most frequent seventeen). Sometimes, what meets criteria for a violation may in fact be appropriate treatment, such as promethazine for cancer patients. The logistic model showed fewer Beers’ criteria violations associated with one or fewer comorbidities (31.9%), being male (35.1%), being nonwhite (51.7%), and having died in 2000 (18.1%). Beers’ violations were more common if the beneficiary has five or more comorbidities (36.7%) or has employer sponsored retiree benefits (27.7%). Conclusions: This study showed that many beneficiaries have prescriptions that violate a Beers’ criterion, though most received only one inappropriate agent. Most violations involved psychoactive medications. With more prescriptions comes a higher risk of receiving inappropriate medications. Implications for Policy, Delivery, or Practice: The overall rate, 52%, is higher than most other studies of community dwelling elderly and inappropriate prescribing, probably reflecting this particularly high-utilizing portion of the population, those sick enough to die. Given their frail health status, the level of prescription use, and the presence of a pharmacy benefits manager, the level of inappropriate prescribing may be surprising. Assessing whether prescribing of these drugs reflects errors or is actually appropriate is worth investigating. Primary Funding Source: RWJF ●Assessing Outpatient Quality of Care Brian Gallagher, MS, Edward Hannan, Ph.D., Liyi Cen, MS Presented By: Brian Gallagher, MS, Senior Healthcare Analyst, Health Policy, Management and Behavior, School of Public Health, University at Albany, One University Place, Rensselaer, NY 12144; Tel: (518)402-4091; Fax: (518)402-0414; Email: bxg04@health.state.ny.us Research Objective: Examine the quality of care in the outpatient setting by examining complication rates of selected procedures. Study Design: Patients who had an outpatient procedure in New York State during 2002 were matched to New York State inpatient records. Patients who were admitted to the inpatient setting within 30 days with a complication relating to the outpatient procedure were identified. Rates of complications by selected procedure were calculated. Population Studied: Patients who had an outpatient procedure in New York State during 2002 and who were reported to the New York State Department of Health in the outpatient hospitalization administrative data system, the Statewide Planning and Research Cooperative System (SPARCS). Principal Findings: Of the 1,138,913 outpatient discharges in New York State during 2002, 28,784 (.025%) were admitted to inpatient care within 30 days of the outpatient discharge. Of these readmitted patients, a subset can be identified as suffering complications. The rate of complications varies by procedure performed in the outpatient setting, but were all relatively low. Outpatient procedures which had the highest complication rates included tonsillectomy (ICD-9-CM procedure code 28.2) with a 2.3% complication rate, arteriovenosomy (ICD-9-CM procedure code 39.27), with a 1.9% complication rate, and insertion of devices (ICD-9-CM procedure code 86.06, 86.07), which had a 1.3% complication rate. High volume procedures including cataract (ICD-9-CM procedure codes 13.1, 13.2, 13.3, 13.4, 13.5, 13.6), with 95,991 procedures, had a complication rate of .11%, knee (ICD-9-CM procedure code 80.6), with 36,676 procedures, .36% complication rate, and non-diagnostic intestine (ICD-9-CM procedure codes 45.0, 45.3, 45.4, 45.5, 45.6, 45.7, 45.8, 45.9), with 62,606 procedures, .5% complication rate, had low rates of complications. Conclusions: The rate of complications from outpatient procedures as measured by inpatient treatment is low. It does not appears that there is a major problem with the quality of surgical care delivered in the outpatient setting. Implications for Policy, Delivery, or Practice: The complication rate for selected outpatient procedures in New York State as measured by readmission to the inpatient setting is low, but lessons can be learned from examining the types of complications which do occur and the type of patients who are most at risk for post operative complications. Primary Funding Source: AHRQ ●Measuring the Use of Spirometry in Affirming a Diagnosis of Chronic Obstructive Pulmonary Disease (COPD) Min Gayles Kim, MPH, Fernando Martinez, M.D., Russ Mardon, Ph.D., Phil Renner, MBA Presented By: Min Gayles Kim, MPH, Senior Health Care Analyst, Quality Measurement, National Committee for Quality Assurance, 2000 L Street NW Suite 500, Washington, DC 20036; Tel: (202)955-1731; Fax: (202)955-3599; Email: gayles@ncqa.org Research Objective: To explore the potential for a performance measure to assess the appropriate use of diagnostic instrument (spirometry) to confirm airway obstruction and a diagnosis of COPD. Despite numerous clinical guideline recommendations to incorporate spirometry in the assessment and diagnosis of COPD and staging of disease severity, health plan data shows that most members with COPD do not undergo spirometry. Study Design: Observational study conducted in five health plans. Using administrative data, an incident case of COPD was specified using a negative diagnosis history period of two years. Spirometry use rates assessed use within the two year negative diagnosis period and the six months following the first incidence of COPD diagnosis in administrative claims as an indicator for confirmation of airway obstruction and presence of disease. Population Studied: Five health plans participated in the study by providing patient-level administrative and medical record data. The enrollments of these plans included commercial, Medicare, and Medicaid product lines across several geographical regions of the U.S., and ranged in size from 52,000 to over 820,000 members. Principal Findings: Using administrative records, the planspecific frequency of new cases of COPD ranged from 0.56 per 1000 to 3.52 per 1000 commercial plan members, and averaged 28.41 per 1000 Medicare members. Of members with a new COPD diagnosis, the average plan rate for spirometry use was 32% (range 26% to 37%). Spirometry use did not vary across product lines but showed lower rates in men versus women and in the older age groups. A higher percentage of spirometry tests (60%) occurred in the physician office compared to a pulmonary function lab (39%). The denominator validation in medical record averaged 65% with a range of 30% to 100%. Validation was higher in the Medicare population (73%). Contrary to perceived notion that many spirometry tests happen in the physician office without a claim generated, no spirometry tests were found in the MR only without a corresponding administrative claim in two plans and relatively few in the remaining two plans indicating administrative data are reliable for capturing spirometry tests. Conclusions: This study showed low rate of overall spirometry use with greater use in younger population. Administrative data provides reliable information on spirometry use in COPD diagnosis. The low rate of spirometry use may reflect ongoing controversy over the usefulness of this diagnostic tool despite consensus in guidelines for its use. Implications for Policy, Delivery, or Practice: Measurement of spirometry use is feasible and may encourage greater research into its role in diagnosis and management of COPD. Primary Funding Source: Other, Supported in part by educational grant from Boehringer-Ingelheim & Pfizer ●Do Malpractice Claims Influence Physician Outcomes and Practice Patterns? Gilbert Gimm, Ph.D., MBA Presented By: Gilbert Gimm, Ph.D., MBA, Health Care Systems, University of Pennsylvania, 110 Sandra Road, Wilmington, DE 19803; Tel: (267)252-8797; Email: gilber20@wharton.upenn.edu Research Objective: This paper examines how accurately the tort liability system is able to identify physicians with adverse outcomes, and whether malpractice claims have an influence on physician behavior. Specific variables of interest are the number of adverse outcomes, volumes, C-section rates, and Medicaid patient mix. Study Design: Using a panel dataset that links inpatient deliveries between 1992-2000 in the state of Florida with malpractice claims, I examine a series of OLS regressions that control for observable patient and physician characteristics, and for fixed effects. Population Studied: Individual physicians in the state of Florida between 1992-2000. Principal Findings: The tort liability system does a relatively poor job of identifying physicians with a higher percentage of adverse outcomes, controlling for volumes. I find strong evidence that physicians experience a 10% lower volume, on average, in response to a malpractice claim. And I find evidence that physician have fewer adverse outcomes in response to injuries, not to the claim itself. No significant effects were found with respect to C-section rates or Medicaid patient mix. Conclusions: The tort liability system is not very effective at improving the quality of care. Also, malpractice claims have a real effect on individual physician behavior in the form of a lower volume of deliveries, which may be due to a supply-side factors (i.e. greater cautiousness of practice, referring more difficult cases to others, or time spent complying with an investigation), or a demand-side factor (i.e. patient demand and reputation effects). Implications for Policy, Delivery, or Practice: The malpractice liability system, which is designed to deter negligent care, may not be very effective as a policy tool for improving the quality of care. Futher research is needed to determine the overall welfare effects associated with the reduction in delivery volume, and whether substitution of cases is occuring between physicians. Primary Funding Source: AHRQ ●Implementation of Infection Control Guidelines in U.S. Hospitals:A Pilot Study Karen Goldsteen, MPH, Ph.D., , , Julie A. Jefferson, RN, MPH, Raymond L. Goldsteen, MS, DrPH, , , , , , Presented By: Karen Goldsteen, MPH, Ph.D., Clinical Associate Professor, Graduate Program in Public Health, State University of New York, Stony Brook, Health Sciences Center, Level 3, Room 071, Stony Brook, NY 11794; Tel: (631) 4446658; Fax: (631)444-3480; Email: karen.goldsteen@stonybrook.edu Research Objective: To indicate the extent to which evidencebased infection control guidelines are implemented in U.S. hospitals. Study Design: We interviewed Infection Control Practitioners (ICPs) in 20 acute care hospitals in the U.S. We selected five hospitals each from the Northeast, Southeast, Midwest, and West. One of the five hospitals in each region was chosen for its excellent reputation in infection control. Of the remaining four hospitals, two were selected for their very high score and two for their very low score on a measure of overall hospital performance developed by HealthGrades. A telephone interview, using a structured questionnaire developed by the authors, was conducted in February-March 2004. The main outcome measure was the Infection Control Performance Indicator developed by the authors and based on reported implementation of Category I guidelines in thirteen areas of infection control: surgical site infections, intravascular catheter-related infections, construction and environmental risks, isolation precautions, disinfection and sterilization, health care-associated pneumonia, nosocomial transmission of methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant enterococcus (VRE), emerging antimicrobial resistance, hand hygiene, Legionnaire’s Disease and dialysis unit, bloodborne pathogens, surveillance, and employee health policies. Population Studied: U.S. teaching and community hospitals with more than 200 beds. Principal Findings: No hospital was 100% compliant with the infection control guidelines measured by our Infection Control Performance Indicator. The highest rate of compliance was 88%, and the lowest compliance rate was 53%. Just onequarter of the hospitals we interviewed were 80-88% compliant. About half were 70-79% compliant. Conclusions: The findings suggest that the evidence-based guidelines for infection control are not fully implemented in most hospitals, and many hospitals fall very far short of achieving full implementation. Implications for Policy, Delivery, or Practice: The U.S. might significantly reduce the hospital-acquired infection rate by ensuring that current, evidence-based infection control guidelines are fully implemented. Further work is needed to determe how to: (1) reduce internal barriers to full implementation of guidelines; and (2) develop policies in the external environment that facilitate infection control within hospitals. Primary Funding Source: Long Island Community Foundation – Lilo and Gerard Leeds Fund ●Patient Characteristics and the Impact of Mortality Indicator Definition on Provider Performance Measurement Christopher Gorton, M.D., MHSA, Jayne L. Jones, MPH, Diane L. Arke, MS Presented By: Christopher Gorton, M.D., MHSA, Senior Physician Consultant, 225 Market Street, Harrisburg, PA 17101; Tel: (717)232-6787; Fax: (717)-232-3821; Email: kgorton@phc4.org Research Objective: To determine differences in patient characteristics (including cause of death) between subpopulations included in two commonly used mortality measures, in-hospital mortality and 30-day mortality. Study Design: Retrospective analysis of a Pennsylvania statewide inpatient dataset in which thirty-day mortality and cause of death were identified by linking Pennsylvania Health Care Cost Containment Council inpatient data with Pennsylvania Department of Health death certificate data. Risk factor selection and expected rates were calculated using multivariate logistic regression models to determine hospital risk-adjusted mortality rates. We divided Pennsylvania residents who died after CABG surgery into three distinct subpopulations (Group 1: death during index hospitalization <=30 post-operative days; Group 2: death during index hospitalization >30 post-operative days; Group 3: death after discharge from index hospital <= 30 post-operative days) and studied patient characteristics including cause of death. Population Studied: In 2002, 14,830 Pennsylvania residents had isolated CABG in 62 non-federal Pennsylvania hospitals. Principal Findings: 289 individuals died during the index admission (in-hospital mortality rate 1.95%). 347 individuals died within thirty days of their procedure (30-day mortality rate 2.34%). This difference is statistically significant (p=0.01). Inhospital mortality aggregates deaths in the index hospitalization that occur both <= 30 post-operative days (Group 1) and >30 post-operative days (Group 2), while 30day mortality aggregates deaths <=30 post-operative days that occur both during (Group 1) and after (Group 3) discharge from the index hospitalization. Group 1, the largest subpopulation, contains 257 Pennsylvania residents. These individuals make up the majority (74.1%) of the deaths captured in both measures. Differences in observed rates for in-hospital and 30-day mortality arise from differing patient characteristics of individuals in Groups 2 and 3. We observed highly significant differences in patient characteristics when comparing Groups 2 and 3: Atlas™ Probability of Death on admission (7.50% vs. 3.73%; p=0.003); the incidence of heart failure diagnosis before/during the index admission (59.38% vs. 3.33%; p<0.0001); the rate of cardiac causes of death (37.50% vs. 65.55%; p=0.003); and the rate of renal causes of death (9.38% vs. 0.0%; p=0.002). We also observed highly significant differences when Group 3 was compared to Group 1: Atlas™ Probability of Death on admission (3.73% vs. 5.72%; p<0.0001); the incidence of heart failure diagnosis before/during the index admission (3.33% vs. 41.25%; p<0.0001); and the rate of pulmonary embolism as cause of death (5.56% vs. 0.78%; p=0.003). Finally, we observed a number of moderately significant differences (p<0.05) between Group 2 and Group 1. Conclusions: Definition of outcome indicators is critical in the measurement and reporting of provider performance. Seemingly minor changes in definition produce significant changes in apparent provider performance. These changes can arise from significant differences in the patient characteristics of the sub-populations that are aggregated to produce the different mortality measures. Outcomes in distinct mortality sub-populations may or may not be related to the research objectives of particular investigations or initiatives. Implications for Policy, Delivery, or Practice: Researchers, policy makers, clinicians and public health practitioners must carefully choose outcome measure definitions that relate directly to their public health policy objectives and quality improvement goals. Primary Funding Source: PA Health Care Cost Containment Council ●Geographic Variation and Chronic Illness in Pediatric Pneumonia Admissions Christopher Gorton, M.D., MHSA, Jayne L. Jones, MPH Presented By: Christopher Gorton, M.D., MHSA, Senior Physician Consultant, Pennsylvania Health Care Cost Containment Council, 225 Market Street, Harrisburg, PA 17101; Tel: (717)232-6787;Email: kgorton@phc4.org Research Objective: To compare county rates of hospital admissions for pediatric pneumonia; To assess the contribution of chronic conditions to county and statewide pediatric pneumonia admission rates. Study Design: Retrospective analysis of a Pennsylvania statewide inpatient dataset containing 5,429 pediatric pneumonia admissions during the last three quarters of 2003 and the first quarter of 2004,of which 4,948 (91.1%) were included in the study. We divided the admissions into two universes (all pneumonia and pneumonia excluding coded comorbid chronic conditions) and calculated pediatric pneumonia admission rates for each Pennsylvania county. We excluded transfers from other hospitals. We obtained county child population demographics from US Bureau of Census data and county Medicaid enrollment from PA Department of Health data. We adjusted raw county pneumonia admission rates using indirect standardization to account for differences in mean age in months; percent male; percent black race; and percent Medicaid enrollment. Population Studied: All PA-resident children aged two months through 17 years who were admitted to acute care hospitals for a primary diagnosis of pneumonia (DRG 081 or 091) during the 12-month study period. Principal Findings: The average total charges per admission were $10,171. The Pennsylvania statewide admission rate for all pneumonia was 156.3 admissions per 100,000 children (adm/100kkids). There was substantial county-to-county variation: County admission rates for all pneumonia ranged from 77.0 adm/100kkids to 457.6 adm/100kkids. 5 of 6 regional health districts had one or more counties in the highest admission rate quintile; 4 of 6 had one or more counties in the lowest quintile. Six of 13 highest quintile counties were geographically contiguous with one or more lowest quintile counties. Pennsylvania’s two largest urban counties fell in the middle quintile, as did several sparsely populated rural counties. Similar geographic patterns were seen among the 2,851 admissions that remained in the second universe after the exclusion of 2097 records (42.4%) coded for co-morbid chronic conditions. The Pennsylvania statewide admission rate for pneumonia without chronic conditions was 90.0 adm/100kkids. County admission rates for pneumonia without chronic conditions ranged from 18.3 adm/100kkids to 350.3 adm/100kkids. 62 of 67 counties (93%) remained in the same or an adjacent admission rate quintile after children with chronic conditions were excluded. On average, the county admission rates for pneumonia without chronic conditions were 58.1% of their admission rates for all pneumonia (95% CI: 31.5 – 84.6%). Conclusions: County pediatric pneumonia admission rates vary widely, even among geographically contiguous and demographically similar counties. Excluding children with comorbid chronic conditions to control for varying community disease burden does not substantially alter county rank order or the pattern or degree of variation in admission rates in our study. Implications for Policy, Delivery, or Practice: Hospitalization is an intrusive intervention that has substantial physical and emotional impacts on children and their families and significant costs to public and private payors. Researchers, policy makers and decision makers should explore the sources of unnecessary variation in pediatric pneumonia admission rates and seek mechanisms to eliminate it. Primary Funding Source: PA Health Care Cost Containment Council ●Associations Between Public Reporting of Health Outcomes and In-Hospital Mortality Christopher Gorton, M.D., MHSA, Ying P. Tabak, Ph.D., Christopher S. Hollenbeak, Ph.D., Jayne L. Jones, MPH, Karen G. Derby, BA, Richard S. Johannes, M.D., MS Presented By: Christopher Gorton, M.D., MHSA, Senior Physician Consultant, Pennsylvania Health Care Cost Containment Council, 225 Market Street, Harrisburg, PA 17101; Tel: (717)232-6787; Email: kgorton@phc4.org Research Objective: There are currently few studies that have attempted to determine whether outcomes for patients in public reporting (PR) of health outcome states differ from those in non-public reporting (non-PR) states. We report here on the first phase of a multi-part study to address this question. The objective of this study was to determine whether in-hospital mortality rates for six important principal diagnoses differed between patients from hospitals in Pennsylvania (PA), a PR state, and those from hospitals in non-PR states. Study Design: We identified six high-frequency, high-mortality principal diagnoses for which risk-adjustment models have been previously reported. We used a large clinical database to analyze discharges from 200 hospitals from PA and 42 hospitals from other non-PR states (non-PA). Using clinical data on admission (history, physical findings, vital signs, laboratory results) imported electronically or abstracted from chart review, we used logistic regression models fit to 20002001 data and validated with 2002-2003 data to identify 17 to 32 variables for each condition predictive of mortality at the p<0.05 level. In order to minimize selection bias and to balance distribution of covariates, we used these models to calculate in-hospital mortality propensity scores for each discharge and then created propensity-score matched PA and non-PA cohorts for each condition. We then computed the odds ratio for in-hospital mortality between the matched cohorts. Population Studied: The data set included a total of 757,947 adult admissions in 2000-2003 for acute myocardial infarction (AMI; n=107,280), congestive heart failure (CHF; n=261,349), hemorrhagic stroke (n=17,414), ischemic stroke (n=98,361), pneumonia (n=208,237), and sepsis (n=65,306) from 242 acute care hospitals. Propensity score matching yielded 62,947 matched pairs for analysis. Principal Findings: Overall, the median age for the six conditions ranged from 75 to 77. The crude mortality was 3.9% (CHF), 4.5% (pneumonia), 6.3% (stroke), 10.6% (AMI), 17.9% (sepsis), and 31.2% (hemorrhagic stroke). After propensity score matching, PA patients showed significantly lower mortality for AMI (OR: 0.74, CI: 0.66-0.84), CHF (OR: 0.73, CI: 0.65-0.81), hemorrhagic stroke (OR: 0.73, CI: 0.55-0.98), ischemic stroke (OR: 0.78, CI: 0.67-0.90), and pneumonia (OR: 0.75, CI: 0.67-0.83). For sepsis, mortality among PA patients was not significantly different from non-PA patients (OR: 0.98, CI: 0.84-1.2). The predictive power for all models was excellent with C-statistics ranging from 0.78-0.91 for the propensity-score-matched cohorts. Conclusions: These preliminary findings suggest that public reporting of health outcomes is associated with improvements in in-hospital mortality in at least one state that provides public reports of health outcomes. Implications for Policy, Delivery, or Practice: Policy makers should consider enhancing public reporting within current PR states and funding research to evaluate its effectiveness. Future research should further examine the impact of public reporting by assessing additional conditions; outcomes such as 30-day mortality, readmissions and length of stay; and outcome trends over time. Primary Funding Source: PA Health Care Cost Containment Council and Cardinal Health ●Statewide Experience in In-Hospital Complications of Bariatric Surgery Christopher P. Gorton, M.D., MHSA, Jayne L. Jones, MPH Presented By: Christopher P. Gorton, M.D., MHSA, Senior Physician Consultant, Pennsylvania Health Care Cost Containment Council, 225 Market Street, Harrisburg, PA 17101; Tel: (717)232-6787; Email: kgorton@phc4.org Research Objective: To determine rates of in-hospital complications occurring during admissions for bariatric surgery. Study Design: Retrospective analysis of a Pennsylvania statewide inpatient dataset containing administrative data on all individuals who underwent bariatric surgery (principal procedure 43.7, 43.89, 44.31, 44.39, 44.69, or 45.91) for a diagnosis of obesity (principal diagnosis 278.00 or 278.01 with DRG of 288, 292 or 293) during 2003. We developed a list of 324 candidate complication ICD-9 diagnosis and procedure codes and grouped them into organ system based complication groups. We performed descriptive statistical analysis of the demographics of the population undergoing bariatric surgery and calculated rates of individuals experiencing specific complication codes and falling into various complication code groups. Population Studied: In 2003, 7,231 patients underwent bariatric surgery in Pennsylvania general and specialty acute care facilities. The mean age of the patients was 43.0 years (median 43 years; range 15 to 78 years; inter-quartile range 35 to 51 years). The patients were 83% female. Principal Findings: The majority of patients underwent Rouxen-Y Stomach Bypass (N=6,795; 94.0%), while the remainder underwent either Gastric Banding/Vertical Banded Gastroplasty (N=401; 5.5%) or Biliopancreatic Diversion with/without Duodenal Switch (N=35; 0.5%). Fourteen individuals died during the index hospitalization (0.2%). Over twenty percent of discharges (N=1491; 20.6%) were coded with one or more of 107 ICD-9 diagnosis or procedure codes indicative of a complication. Patients experienced the following rates of complication groups during or after surgery: Procedure/Medical Care-related, N=483, 6.68%; Respiratory, N=342, 4.73%; Cardiovascular, N=330, 4.56%; Metabolic Disturbances, N=262, 3.62%; Digestive System, N=230, 3.18%; Hernias, N=184, 2.54%; Infection, N=124, 1.71%; and Neurological, N=14, 0.19%. 331 individuals (4.58%) had complications from 2 or more complication groups, including 10 of the 14 individuals who died. Patients 65 years and older had significantly more complication groups coded than those under 65 (35.0% vs. 20.6%; p=0.0008). Post-operative length of stay for patients without coded complication groups averaged 3.0 days, compared to 3.9 days for individuals with complications from 1 complication group and 8.7 days for individuals with complications from 2 or more complication groups. Charges for patients without coded complication groups averaged $31,744, compared to $39,971 for individuals with complications from 1 complication group and $76,107 for individuals with complications from 2 or more complication groups. All these differences were significant at the level of p< 0.0001. Neither hospital nor surgeon bariatric case volume was significantly correlated with numbers of patients having complications from 1 or more complication groups. Conclusions: Bariatric surgery was widely performed in Pennsylvania in 2003. While the in-hospital mortality rate was quite low, significant numbers of individuals experienced complications. Individuals who experienced complications had significantly longer post-operative lengths of stay and significantly higher charges. Each individual who died had complications from at least 1 complication group and the majority of decedents had complications from 2 or more groups. Implications for Policy, Delivery, or Practice: Patients considering bariatric surgery need to be aware of the risks of complications as well as the potential benefits. Primary Funding Source: PA Health Care Cost Containment Council ●Medication Errors in Nursing Homes: A State's Experience Implementing a Reporting System Sandra B. Greene, DrPH, Charlotte E. Williams, MPH, Richard Hansen, Ph.D., Kathleen D. Crook, MPA, Roger Akers, MSIS, Timothy S. Carey, M.D., MPH Presented By: Sandra B. Greene, DrPH, Research Associate Professor, Health Policy and Administration, Cecil G. Sheps Center for Health Services Research, University of North Carolina at Chapel Hill, 725 Airport Road, CB # 7590, Chapel Hill, NC 27599; Tel: (919) 966-0993; Fax: (919) 966-1634; Email: SandraB_Greene@unc.edu Research Objective: Medication errors are prevalent in nursing home settings due in part to the large number of medications prescribed and the at-risk nature of the population. In an effort to improve patient safety North Carolina passed a Medication Error Quality Initiative requiring nursing homes to annually report errors occurring in the facility, including information on causes, the type of staff involved, types of drugs and impact on patients. This paper describes the implementation process for the first annual report and summary findings. Study Design: We developed a medication error reporting form that nursing homes were encouraged to use throughout the reporting period to capture individual error events. At the end of the reporting period the nursing homes were directed to a secure web site where they entered aggregate information concerning their facility’s errors. A medication error was defined as any preventable medication-related event that adversely affects a patient in a nursing home and is related to professional practice, or healthcare products, procedures, and systems, including prescribing, prescription order communications, product labeling, packaging and nomenclature, compounding, dispensing, distribution, administration, education, monitoring and use. Population Studied: All licensed nursing homes in North Carolina, excluding nursing home beds within hospitals, are subject to the legislation. Principal Findings: All 384 facilities successfully reported, with over 96% using the web-based tool. The remainder faxed their reports. A total of 10,920 medication errors were reported for this first report, which included 9 months of data. Three facilities reported no errors, while one reported 1,648 errors. The median number of errors was 15. The most frequent type of error reported was a dose omission (55.6%). Twenty-five percent involved other dosage related errors: wrong dose (8%), extra dose (7.5%) under dose (5.4%) and overdose (4.5%). The vast majority of errors (79%) occurred among patients who were cognitively impaired and not able to direct their own care. In 5.9% of the errors, the patient was affected in some way. The most predominant effect of the medication error was “inadequate effect of the product” which is consistent with the large number of dose omissions. With respect to patient outcomes, 91.1% of the errors did not reach the patient or reached the patient and caused no harm; 8.1% required further monitoring; 0.5% of errors caused temporary harm. Twenty-three patients (0.2%) required emergency treatment because of a medication error, and for one patient the medication error resulted in permanent harm. Conclusions: Self reported medication errors vary widely among nursing homes. Facility variations may be due to differences in a facility’s comfort level in reporting errors, a facility’s interpretation of what is an error as well as actual variations in errors incurred in the medication use process. Implications for Policy, Delivery, or Practice: Mandatory reporting of medication errors should be viewed as a way to encourage education and process changes at the facility level, to reduce subsequent errors and improve patient safety. We will work with facilities over the next years to improve reporting. Primary Funding Source: North Carolina Department of Health and Human Services, Division of Facilities Services ●Hospital Factors Associated with Uterine Rupture and Neonatal Birth Trauma in Women with History of Prior Cesarean Delivery Kimberly D. Gregory, M.D., MPH, Lisa M. Korst, M.D., Ph.D., Ida R. Shihady, MPH, Moshe Fridman, Ph.D., Arlene Fink, Ph.D., Linda Burnes Bolton, Dr.Ph., RN, FAAN Presented By: Kimberly D. Gregory, M.D., MPH, Director, Maternal Fetal Medicine, Obstetrics & Gynecology, CedarsSinai Medical Center, 8700 Beverly Boulevard, Suite 160W, Los Angeles, CA 90048; Tel: (310) 423-5420; Fax: (310) 4230140; Email: gregory@cshs.org Research Objective: Nationally, the risk of uterine rupture is 0.7% for women attempting vaginal birth after cesarean (VBAC), and the risk is associated with patient-specific clinical factors. This study attempts to identify hospital level risk factors for uterine rupture and newborn birth trauma and specifically evaluates clinical policies that might be associated with low risk of adverse maternal and neonatal outcomes. Study Design: We conducted structured interviews of nurse managers at hospitals in California reporting greater than 50 deliveries in 2002 regarding their hospital resources and clinical policies. We developed a “willingness to intervene” construct to define hospitals with practices that supported nurses calling physicians for specific clinical scenarios. This construct was based on responses to 3 questions using a 4point Likert scale (1=Very Likely to Call Physician, 4=Very Unlikely to Call Physician). We obtained maternal and newborn linked hospital delivery discharge and birth certificate data for 2002 from the Office of Statewide Health Planning & Development. We report the odds ratio for uterine rupture and newborn birth trauma based on hospital structural (hospital ownership, teaching status, delivery volume, 24 hour anesthesia) and clinical variables (allow VBAC Y/N, “willingness to intervene”), and case mix adjustments (age group, medical complications, route of delivery) for singleton liveborn deliveries using hierarchical logistic regression techniques. This study was IRB approved. Population Studied: Hospitals in California providing care to women with a history of prior cesarean (and their newborns) during 2002. Principal Findings: The response rate was 84% (225/268 hospitals completed the interview), and 74% of hospitals allow VBAC. There were 400,702 singleton linked records, of which 14% (55,154) were to women with a history of prior cesarean delivery. The VBAC attempt rate was 26%, the VBAC success rate was 17%. The uterine rupture rate was 0.3%. Among hospitals that allow VBAC, the following variables were associated with uterine rupture on univariate analyses: hospital ownership, teaching status, VBAC allowed, and “willingness to intervene.” After controlling for patient clinical factors, hospital type, VBAC allowed, and “willingness to intervene” were still significantly associated with uterine rupture. State hospitals had increased risk and churchaffiliated hospitals had decreased risk relative to for-profit hospitals. The odds of uterine rupture in hospitals with clinical practices that support “willingness to intervene” (median=2; range 1-4) were 1.46 times lower (95% confidence region, 1.11-2.12) per unit increase in the “willingness to intervene” factor. After excluding births with neonatal anomalies/malignancies, 51,638 deliveries were used to model neonatal trauma. The odds of neonatal trauma in hospitals with clinical practices that support “willingness to intervene” were 1.13 times lower (95% confidence region, 1.01-1.26) per unit increase in this factor. Conclusions: Previous studies have focused on patientspecific risks for uterine rupture including obstetric management. Current recommendations have emphasized availability of hospital resources to maximize patient safety for women attempting VBAC. Implications for Policy, Delivery, or Practice: Our study suggests that there may be specific nursing practices that contribute to VBAC safety, risk of uterine rupture, and neonatal birth trauma. Primary Funding Source: AHRQ, RO1HS11334 ●Factors Affecting Patient Outcomes in Ambulatory Surgery Brian Harahan, BA, Maureen Smith, M.D., MPH, Ph.D., Ann Hundt, MS, Ph.D., Scott Springman, M.D., Pascale Carayon, Ph.D. Presented By: Brian Harahan, BA, Trainee, Population Health Sciences, University of Wisconsin-Madison Medical School, 610 Walnut Street, 707 WARF Building, Madison, WI 53726; Tel: (608)263-2880; Fax: (608)263-2820; Email: bjharahan@wisc.edu Research Objective: Over 60% of surgical procedures in the U.S. are performed in ambulatory settings. This growing trend places increasing amounts of responsibility on patients to self-manage their follow-up care as more complex procedures are performed and more complicated patients get these procedures. Increased age and comorbidity may predispose a patient to adverse outcomes. Our objective is to explore the effect that age and comorbidity have on patient response to post-surgical symptoms. Study Design: Data were obtained from patients, N=342, undergoing ophthalmic, open joint, otolaryngological, or intraextra-abdominal ambulatory surgery from four outpatient care centers in a mid-sized midwestern community. Medical record abstraction was used to verify eligibility and collect data on patient comorbidities. Nurses queried patients by telephone at least seven days after discharge on the incidence of twenty-one possible symptoms and whether the patient contacted a provider for each symptom (Hundt et al., 2004). Symptoms were categorized prior to analysis into high risk (e.g., shortness of breath, chest pain), moderate risk (e.g., eating problems), and low risk (e.g., coughing) based on expert opinion. Data on comorbidities were summarized using the Charlson Comorbidity Index. The dependent variable was, “Did you attempt to contact your regular doctor, surgeon's office or the outpatient surgery center, for each symptom.” Analysis was conducted at the symptom level accounting for clustering within patients. Logistic regression was used to examine the relationship of the dependent variable to symptom risk, age and the Charlson index. The final model controlled for gender, education, type of procedure, and presence of a permanent caretaker. Variables examined but not included in the final model (P-value > 0.20) were self-reported health status, ASA status, patient information and preparedness for the surgery, and type of anesthesia. Population Studied: 342 patients aged 18-95, 39% greater than 65, responded to the survey. 150 patients, 30% of whom were greater than 65 years old, were selected for analysis because they contacted a health provider for a new or worsening symptom. 284 new or worsened symptoms were reported at the surgery center or after discharge. Principal Findings: Twenty-four percent of symptoms were reported by respondents older than 65 years. The probability of provider contact increased significantly for high risk compared to low risk symptoms (odds ratio=3.76, 95% confidence interval=1.53-9.28). The data are suggestive that patients over the age of 65 years are less likely to contact their provider for high risk symptoms, but this study lacked the statistical power to permit a firm conclusion. There was no evidence that patients with higher comorbidity levels were more likely to contact a provider for high risk symptoms. Conclusions: High-risk symptoms were more likely to lead to contact with a provider when compared to low risk symptoms, with suggestive evidence that older patients were less likely to respond appropriately to high risk symptoms. Implications for Policy, Delivery, or Practice: Older patients are likely at higher risk for adverse events related to postsurgical complications and further research should evaluate whether older patients are less likely to respond appropriately to high risk symptoms. Primary Funding Source: AHRQ ●Mitigating Factors in Near Miss Events Reported to a Patient Safety Reporting System: A Qualitative Study From the ASIPS Collaborative Daniel Harris, Ph.D., Wilson D. Pace, M.D., Douglas H. Fernald, MA Presented By: Daniel Harris, Ph.D., Senior Project Director, Center for Healthcare Research, The CNA Corporation, 4825 Mark Center Drive, Alexandria, VA 22311; Tel: (703) 824-2283; Fax: (703) 824-2511; Email: harrisd@cna.org Research Objective: Medical error reporting systems are advocated as strategies for improving healthcare quality and safety; however, a growing concern exists regarding how analysis of reports collected by these systems can be translated into improved healthcare practices. We analyzed mitigating factors described in near miss events reported to one such system to see if focusing on “what went right” in these events can identify potential intervention strategies to break the chain of events that lead to patient harm. Study Design: The three-year Applied Strategies for Improving Patient Safety (ASIPS) demonstration project developed a patient safety reporting system (PSRS) to collect anonymous and confidential narrative medical error reports voluntarily submitted by clinicians and staff in two PracticeBased Research Networks (PBRNs) in Colorado. We qualitatively coded factors contributing to and mitigating the cascade of events in the 608 incident reports submitted during the first two years of the project, and how these factors affected whether the incident did or did not result in patient harm. Using Atlas.ti software, we analyzed mitigators by locus, agent, and agency in different types of reported events, and compared harmful events with no-harm near miss events to identify how mitigators intervened in the event process to protect patient safety. Population Studied: Over 600 error reports submitted by primary care providers, other clinical staff, and non-clinical staff from 34 primary care practices affiliated with two PBRNs serving a largely Medicaid patient base in urban and rural communities in Colorado. Principal Findings: We identified three major categories of mitigating factors: actions deliberately undertaken by specific actors with the intent of mitigating an incident, systems factors whereby the operation of existing procedures or technological systems mitigates the incident, and serendipity that leads to mitigation by chance without deliberate action or system performance. These factors are not mutually exclusive and can build on each other. We also identified primary subcategories of each mitigating factor, in which types of incidents they typically occur, and where in the chain of events they typically occur. We found that communication chains and transition points in long event chains are particularly rich opportunities for error and mitigation of error. Finally, we found that the existence of various mitigating and contributing factors differentiated between no-harm and harmful incidents, respectively, that were otherwise similar and shared attributes in common. Comparing mitigators and contributors revealed system vulnerabilities and suggested interventions. Conclusions: A wide range of factors acted to mitigate the cascade of events surrounding medical errors and to prevent errors from reaching or harming patients. While some mitigators appear across event types, others tend to be more associated with specific types. Implications for Policy, Delivery, or Practice: Medical error reporting systems provide opportunities for taking advantage of “the gift of failure.” Focusing on the mitigating factors of what went right in near miss events rather than exclusively on what went wrong expands the opportunity to identify interventions that can address system vulnerabilities that lead to patient harm. Primary Funding Source: AHRQ ●Reorganizing Community Leg Ulcer Care: A Pre-Post Implementation Study Margaret B. Harrison, RN, Ph.D., Ian D. Graham, Ph.D., Karen Lorimer, RN, MscN, Elaine Friedberg, RN, MHA, Tadeusz Pierscianowski, MBBS, FRCPC, FAAD, Tim Brandys, MD, FRCSC Presented By: Margaret B. Harrison, RN, Ph.D., Associate Professor, Faculty of Health Science, Schoolf of Nursing & Community Health and Epidemiology, Queen's University, 78 Barrie Street, Kingston, Ontario, Tel: (613) 533-6000 ext 74760; Fax: (613) 533-6331; Email: harrisnm@post.queensu.ca Research Objective: The objective was to determine the health outcomes and efficiencies of the impact of implementating a nurse-led service supporting guidelinedriven practice for the popoulation with leg ulcers. Study Design: A prospective, one year pre-post design was used to evaluate the new service model which involved both organizational and clinical changes. Outcomes included healing rates at 3 months, and resource use (nursing visits, supplies). Population Studied: Individuals newly referred to homecare for management of their leg ulcer in one large homecare authority (mixed urban-rural setting of approximately one million people) were eligible to participate in the study. Study inclusion criteria included: English or French speaking, having an ulcer (open wound) below the knee of venous or mixed etiology (venous-arterial). Community leg ulcer care was delivered in either the individual’s home or in one of two nurse clinics. Principal Findings: There was a statistically significant improvement in 3-month healing rates increasing overall from 23.1% in the pre-period (18/78) to 55.6% (100/180) in the postperiod. The median number of nursing visits declined from 37 to 25 over the period(z -2.047, p = 0.041) while the median supply cost per case went from $1923 to $406 per case (z 2.828, p = 0.005). Conclusions: Reorganization of care for the population with leg ulcers was associated with improved healing and efficient use of nursing visits. More people could be cared for with the same allocation of home care dollars. Implications for Policy, Delivery, or Practice: The significant impact that the intervention had was due to two major elements; using a quality guideline for clinical care and importantly re-organization with a service model to support evidence-based practice with community leg ulcer care. As an exemplar of a population requiring complex community care, the approach we describe using population-based planning and evidence-based practice, may have broader application. Primary Funding Source: Ontario Ministry of Health and Long Term Care ●Improving Patient Safety in Hospitals: Contributions of High Reliability Organizations and Normal Accident Theory Michael Harrison, Ph.D., Michal Tamuz, Ph.D. Presented By: Michael Harrison, Ph.D., Senior Research Scientist, Center for Delivery, Organization, and Markets, Agency for Healthcare Research & Quality, 540 Gaither Road, Rockville, MD 20850; Tel: (301)427-1434; Fax: (301)427-1430; Email: mharriso@ahrq.gov Research Objective: To identify the distinctive contributions of two leading organizational approaches to safety in highhazard industries. To specify conditions under which these theories apply to hospitals. Study Design: A qualitative, interview-based study of learning from medication errors in hospitals provides examples for our review of empirical and theoretical studies of high reliability organizations (HRO) and normal accident theory (NAT) in hospitals and high-hazard industries. Population Studied: Three tertiary care teaching hospitals (n=342 healthcare professionals and administrators) Principal Findings: (1)Developing Culture: HRO emphasizes that organizational culture influences safety. Researchers advocate developing a “culture of mindfulness” through noticing hazards. Unfortunately, planned cultural change is usually slow, hard to manage, and risky. Interviews with clinicians illustrate how turnover, employment arrangements, and professional education undermined development of mindful cultures. (2)Designing Organizational Structures and Technologies: NAT highlights how organizational structure and technology contribute to errors by intensifying interactive complexity and coupling. Hospitals are characterized by interactive complexity, but traditionally their clinical processes have been loosely coupled. Loose coupling leads to inefficiency, but provides opportunities to catch errors before they cause harm. Tradeoffs in loose coupling were evident in interviews with providers who order, dispense, and administer medications. When adverse events do occur to individual patients in loosely-coupled health systems, they do not usually spread. Recent technological developments, such as computerized order entry, tighten links among units and functions (e.g., ordering and dispensing medication). Although these information technologies improve efficiency and routine safety, they may also increase the chances of errors spreading across patients and units. (3)Specifying Conditions for HRO Solutions: HRO advocates continuous training, but training cannot solve problems from poorly designed systems. NAT suggests that other HRO solutions might actually reduce hospital safety. For instance, adding redundancies by increasing human or technological checks can make processes harder to understand and reduce clinicians’ vigilance. In interviews, pharmacy managers observed that redundancy could lead to complacency and medication errors. (4)Implementing Changes: NAT and HRO agree on some safety solutions, but differ on how to implement them. Both perspectives recognize the importance of rewarding safety-promoting activities. HRO assumes that safety cultures readily create appropriate reward systems. NAT sees developing reward systems as problematic. In practice, employees are subject to conflicting expectations and rewards. As NAT anticipates, in interviews nursing managers reported being subject to conflicting incentives for reducing errors and encouraging error reporting. Conclusions: NAT provides a more realistic image of hospital operations than HRO and reveals barriers to implementing HRO solutions. The cultural changes envisioned by HRO may slowly yield safety benefits; NAT highlights the immediate political impact of competing goals and conflicting reward systems. Implications for Policy, Delivery, or Practice: HRO advocates the long-term advantages of promoting a culture of mindfulness and immediate benefits of conducting safety audits. NAT underscores the potential contribution of safety monitoring and promotion by external stakeholders (e.g., JCAHO) to tip the internal balance towards patient safety. NAT cautions against adding redundancies and warns of possible risks of tightly coupling operations through information technology or efficiency measures. Primary Funding Source: AHRQ, Aetna Foundation's Quality Care Research Fund ●Developing and Validating an Algorithm to Identify Pregnancy Episodes in an Integrated Health Care Delivery System Mark C. Hornbrook, Ph.D., Evelyn Whitlock, M.D., MPH, Cynthia Berg, M.D., MPH, William Callaghan, M.D., MPH, Carol Bruce, MPH, Rachel Gold, Ph.D., MPH Presented By: Mark C. Hornbrook, Ph.D., Cnief Scientist, Center for Health Research, Kaiser Permanente Northwest, 3800 North Interstate Avenue, Portland, OR 97227-1110; Tel: (503) 335-6746; Fax: (503) 335-2428; Email: mark.c.hornbrook@kpchr.org Research Objective: To validate a computerized algorithm for grouping appropriate ambulatory and inpatient care encounters into pregnancy episodes. The algorithm was compared with the gold standard of mothers medical records. Study Design: We developed a computerized algorithm to identify key parameters of all types of pregnancy episodes, including live births, stillbirths, therapeutic abortions, spontaneous abortions, ectopic pregnancies, and trophoblastic disease. To assess the algorithm’s accuracy, we applied it to a sample of automated medical records of women of childbearing age who were members in 1998-2001 of an integrated health care delivery system. The algorithm searches for diagnosis, procedure, and Diagnosis Related Groups as well as test results associated with pregnancy. It uses a logical hierarchy to define the episode parameters: start and stop dates, pregnancy outcomes, outcome procedure, gestational age, fetal count, and phases of pregnancy. We validated the algorithm’s assessment of the pregnancy histories of a sample of subjects with assessments obtained by blinded medical records abstractors using actual medical records. Population Studied: Members of Kaiser Permanente Northwest, a prepaid group practice HMO with about 455,000 members, constituted our study population. Within this population, the algorithm searched for evidence of pregnancy episodes among all women of childbearing age, 12 to 55 years old, who were KPNW members for at least 42 continuous days at any time during the study period, January 1, 1998, through December 31, 2001. The sample size was 511 cases. We included only those episodes fully contained within the study period and the mothers health plan eligibility period. Principal Findings: The tested version of the algorithm correctly identified 97 percent of pregnancy episode end dates and achieved exact or close agreement for 99 percent of pregnancy outcome types. It achieved 96 percent agreement on gestational age among episodes agreeing on both end date and outcome type. Using the results obtained from the validation study, we modified the algorithm and improved the classification results. Conclusions: Using comprehensive medical record and administrative data to retrospectively identify the beginning and end of pregnancy regardless of pregnancy outcome allows for defining a window to explore maternal pregnancy complications. Understanding these complications may facilitate design and improvement of obstetrical care in health care systems. Our pregnancy episode grouper makes use of more information than various commercial episode grouper software. Implications for Policy, Delivery, or Practice: While more than six million women in the US become pregnant each year, efforts to assess the extent and nature of morbidities experienced during pregnancy have been limited to analyzing hospitalization data. Hospitalization data likely capture only severe and intrapartum morbidities and do not identify whether morbidities appeared gradually over pregnancies. These shortcomings are salient because recent changes in medical practice have resulted in increased treatment and better outcomes for maternal complications in outpatient settings. A pregnancy episode framework is needed to measure morbidity profiles in the prepartum, intrapartum, and postpartum phases. Primary Funding Source: CDC ●Identification of Actionable Items for Systems Improvement through Executive WalkroundsÓ in a Pediatric Hospital Lisa Horowitz, Ph.D., MPH, Melissa McClay, MPH, Allan Frankel, M.D., James Mandell, M.D., Eileen Sporing, MSN, RN, Don Goldmann, M.D. Presented By: Lisa Horowitz, Ph.D., MPH, Quality Improvement Consultant, Department of Patient Safety and Quality, Children's Hospital Boston, 300 Longwood AvenueWolbach 001, Boston, MA 02115; Tel: (617)355-7447; Fax: (617) 730-0632; Email: lisa.horowitz@childrens.harvard.edu Research Objective: To use comments elicited during Patient Safety Executive Walkrounds to identify themes and actionable items in order to resolve potential safety concerns. Study Design: As part of a multi-center study, the Children’s Hospital Executive Walkrounds Team (EWT) visited a different unit in a large pediatric teaching hospital for one hour three times per month. Three rotating executives conducted semistructured interviews with groups of staff including staff physicians, physician trainees, nurses and pharmacists, using standardized questions and explicitly encouraging staff to share concerns free of retribution. To identify key themes and actionable items, all staff comments were recorded verbatim by a scribe during the rounds and entered as text into a database. Comments were then numbered and coded by two observers into thematically distinct categories, and summarized in an Action Plan – a key tool for ensuring follow up of actionable items. One week later a Debrief Session including the EWT and clinical leaders from the specific unit was convened to discuss each item in the Action Plan, determine actionability, and assign responsibility for resolving the item. 12 weeks later, staff who participated in Executive Walkrounds received feedback and progress-to-date in the form of the updated Action Plan. Each comment was then labeled “completed,” “in progress,” or “tracked in database,” depending on the action taken. Population Studied: Three rotating senior hospital executives and 550 multidisciplinary staff members, including staff physicians, physician trainees, nurses, and pharmacists, of a large urban teaching hospital participated from February through December 2004. Principal Findings: Twenty-seven Executive Walkrounds were conducted in unique clinical areas. Attendance at Executive Walkrounds ranged from 8-47 (mean 12) staff per round. Of 384 items, 14 thematically distinct categories were identified by 2 observers, with an agreement rate of 90%. The top five themes were: Workload/Resources (89% of Executive Walkrounds), Communication (78%), Facilities (78%), Information Systems (63%), and Documentation (52%). Pediatric-specific concerns, such as child safety in medical exam rooms and sterilizing toys in the waiting area, were present in 15% of the comments. In a sub-sample of the first 12 Executive Walkrounds, 83% of items incorporated into the database were considered “actionable” and assigned to clinical leaders for resolution. At the end of 12 weeks, 32% of the actionable items were deemed “completed.” Efforts to resolve the remainder of the items are ongoing. Conclusions: Concerns regarding workload, resources and communication were identified frequently by staff as impacting patient safety. Actionable items were identified and addressed through the Action Plan. One third of the items were resolved comprehensively. Implications for Policy, Delivery, or Practice: Conducting Patient Safety Executive Walkrounds is feasible and effective in a large tertiary care teaching hospital, and allowed this institution to recognize and address systems issues with the potential to compromise high quality healthcare delivery. The Action Plan is an important tool that helps in assigning accountability for resolving staff concerns, allows for ease in feeding back information to staff, and enables the Executive Walkrounds Team to track what would have been otherwise unmanageable data. Primary Funding Source: No Funding Source ●First Look at a Composite Safety Measure Robert Houchens, Ph.D., Chinliu Zhan, M.D., Ph.D., Rosanna Coffey, Ph.D., Edward Kelley, Ph.D. Presented By: Robert Houchens, Ph.D., Director, Statistical Informatics, Medstat, 5425 Hollister Avenue, Suite 140, Santa Barbara, CA 93111-2348; Tel: (805) 681-5867; Fax: (805) 6815899; Email: bob.houchens@thomson.com Research Objective: To develop a composite (aggregate) safety index using as components the Agency for Healthcare Research and Quality’s (AHRQ) Patient Safety Indicators (PSIs). This composite is designed as a test for developing composite measures for the National Healthcare Quality Report and the National Healthcare Disparities Report. Study Design: We applied AHRQ’s PSI software to five years of the Health Care Utilization Project (HCUP) Nationwide Inpatient Sample (NIS), 1998 – 2002. We used estimates of excess mortality and excess length of stay (LOS) attributable to patient safety events, similar to those developed by Zhan and Miller (JAMA, 2003), as relative weights for different types of safety events to produce a composite measure of inpatient safety. We also developed standard error estimates. We used these estimates to assess trends in inpatient safety nationwide, and to statistically compare patient safety performances across various demographic groups. Population Studied: Inpatients discharged from short-term community hospitals in the United States between 1998 and 2002. Principal Findings: The patient safety composites have sound statistical properties and they are easy to understand by broad audiences. The composite indices showed substantial variation in inpatient safety over time and across patient groups, in terms of both excess mortality and excess LOS. “Drill-down” analyses further revealed many important component PSIs driving the overall safety differences. Conclusions: Composite indices are feasible, meaningful, and useful in inpatient safety assessment, especially for trend analyses and for comparing the level of inpatient safety among demographic groups. Implications for Policy, Delivery, or Practice: It is difficult to use the numerous existing patient safety measures to describe inpatient safety in the United States. Composite measures represent important communication tools. Such indices could help to assess overall progress in safety improvement, to identify vulnerable subgroups, and to evaluate safety improvement strategies. Primary Funding Source: AHRQ ●Advance Care Planning: Does it Make a Difference? Ronda Hughes, Ph.D., MHS, RN, Anne Wilkinson, Ph.D., Karl Lorenz, M.D., Joann Lynn, M.D. Presented By: Ronda Hughes, Ph.D., MHS, RN, Senior Health Scientist Administrator, Center for Primary Care, Prevention, & Clinical Partnerships, AHRQ, 540 Gaither Road, Rockville, MD 20850; Tel: (301) 427-1578; Fax: (301) 427-1595; Email: rhughes@ahrq.gov Research Objective: To examine the research literature on end-of-life care to determine the impact of advance directives (including living wills and do not resuscitate orders) on those at the end-of-life, using studies involving advance-care planning. Study Design: As part of a systematic review of end-of-life care, we reviewed articles that evaluated advance directives and other aspects of advance care planning at the end of life. Population Studied: Sources for our review included Medline, the Cochrane Database of Reviews of Abstracts of Effects (DARE), the National Consensus Project for palliative care, and several recent systematic reviews from both Health Canada and National Institute for Clinical Excellence (NICE), United Kingdom. The searches were limited to published articles in the English language (1990 – 2004), involving adult human subjects. Research was included if it targeted patients or families, not only reporting about clinicians. Principal Findings: From 24,423 total citations, 1,274 articles were reviewed, including 95 systematic reviews, 134 intervention, and 682 observational studies. We identified 10 systematic reviews, 12 intervention studies, and 17 observational studies on patient or caregiver satisfaction. The effectiveness of advance directives and advance care planning is supported by little reliable scientific evidence of improving outcomes. Targeted communication and planning can improve patient and family satisfaction. However, high quality research designs have not been used consistently and, when applied, have shown quite modest effects, even upon increasing the rate of making decisions in advance. Whether improved advance care planning actually improves the experience for patients and their families has only thin and equivocal evidence. The evidence provides several key insights involving advance care planning. First, advance care planning should reflect changing preferences and circumstances because patient’s preferences can change over the course of their illness. Second, when clinicians and families understand and agree with a patient’s preferences and prognosis, patients are more likely to experience preferred outcomes. Third, physical and psychosocial support for patients and their families is needed and can improve communication and decision-making among clinicians, patients, and families. Fourth, interventions limited to one type of strategy and one site of care as well as having few study subjects are not likely to change care patterns or have long term impact; which describes about half of the studies reviewed. Conclusions: This systematic review identified a very large, diverse literature reflecting the tremendous growth and importance of the field of end of life care over the last decade. This review illuminates strengths of the field as well as opportunities for research. We identified equivocal evidence supporting the use of advance directives. Advance care planning requires a well-informed patient or their surrogate to make decisions about future care so that treatments undertaken during a future period of decisional incapacity will still be in accord with the patient’s preferences. Implications for Policy, Delivery, or Practice: Studies examining the effect of advance care planning on subsequent health care utilization. Inconsistent evidence supports advance directives, but evidence does support advance care planning to support patient’s preferences and as a means to improve patient-clinicians communication. Primary Funding Source: No Funding Source ●Valued Aspects of Ambulatory Health Care: Comparing Patients and Health Care Providers Based on Office Visit Accounts Margarita Hurtado, Ph.D., MHS, Karen Frazier, BA, Anna Levin, BA, Erica Eisenhart, BA, Karen S. Shore, Ph.D. Presented By: Margarita Hurtado, Ph.D., MHS, Principal Research Scientist, American Institutes for Research, 10720 Columbia Pike, Suite 500, Silver Spring, MD 20901; Tel: (301) 592-2215; Fax: (301) 593-9433; Email: mhurtado@air.org Research Objective: To compare patient and provider perceptions regarding high quality ambulatory care based on office visit accounts, and examine if the quality aspects identified differ between good and poor visits. This is part of a larger study to identify quality improvement levers based on specific provider behaviors and CAHPS survey data. Study Design: We conducted semi-structured telephone interviews with patients and health care providers to obtain detailed personal accounts of specific office visits they considered of good or poor quality. After transcription, we used Atlas.ti to code the text regarding aspects of quality care, stage of encounter, provider type and nature of the experience (positive or negative). We defined the initial coding scheme based on processes of care in ambulatory settings and previous research, and refined it based on the narratives. Intercoder reliability was 89%. Using content and frequency analysis, we examined the relative saliency of quality-related behaviors at three levels: office visit, individual behaviors and respondent. Population Studied: 82 adult patients and 16 health care providers (physicians and nurse practitioners) described 536 office visits (417 and 91, respectively, mean = 5.5). Office visit accounts included 2505 specific behaviors. Principal Findings: Both patients and health care providers described office visits where behaviors related to clinical skills (e.g., problem resolution, skill in performing procedures, history-taking, appropriate prescribing) were the most frequently mentioned aspect of quality (73% of visits). Providing health-related information was also mentioned by patients and providers in equal proportions, but less often (43% of visits). For patients, visit accounts included behaviors related to interpersonal and communication skills (e.g., careful listening, clear explanations, courtesy, respect) almost as frequently (64% of visits) as clinical skills, but for providers, they were among those less frequently mentioned (16% of visits). In contrast, aspects related to coordination of care were mentioned more frequently by providers (41% of visits) than by patients (29% of visits). Access-related aspects (e.g., waiting time, visit duration, financial access) were mentioned in about half of the visit accounts by patients and one-third of those by providers. Aspects related to shared decision-making were seldomly mentioned by either patients (15% of visits) or providers (10% of visits). Rarely mentioned were administrative and office environment aspects. Most behaviors mentioned occurred during the consultation phase of the visit. We did not find any major differences in the aspects of care mentioned in accounts of good quality visits as compared to poor ones. Conclusions: Both patients and providers place a high and equal emphasis on clinical skills when describing good and poor quality office visits. To a lesser degree, both also value the provision of health-related information. The largest gap between the two is with respect to the provider’s interpersonal and communication skills, which were mentioned 4 times more frequently in patient accounts than in provider accounts. Implications for Policy, Delivery, or Practice: This study provides further evidence that quality improvement of ambulatory care should emphasize provider interpersonal and communication skills. Increased awareness of the importance of shared decision making among providers and patients is also needed. Primary Funding Source: AHRQ ●Patient Safety in Primary Care: the Relationship Between Human Error Consequence and Physician Discretion Sylvia Hysong, Ph.D, Richard Best, Ph.D., Frank Moore, Ph.D. Presented By: Sylvia Hysong, Ph.D, Research Health Scientist, VERDICT, Audie L. Murphy Veterans Memorial Hospital, 7400 Merton Minter Boulevard -11C6, San Antonio, TX 78229; Tel: (210)617-5300 x5229; Email: shysong@verdict.uthscsa.edu Research Objective: This research is part of a larger project, the prime objective of which is to systematically analyze the work of primary care health delivery in the VA. Key among the objectives is to create optimization models to demonstrate potential staffing mix redesigns for increased efficiency, employee satisfaction, and patient satisfaction. The present study specifically examines the relationship between the consequence of error in task performance and the degree of discretion in performing the task. Study Design: We used Functional Job Analysis (FJA Fine & Cronshaw, 1999), to empirically examine task statements as the smallest identifiable units of work analysis. Task statements collectively describe the work of primary care, and individually can be analyzed for levels of complexity across multiple dimensions (e.g., data, reasoning, autonomy, human consequence for error (HEC), etc). Primary care personnel from five VA medical centers participated in two-day focus groups to generate task statements descriptive of primary care in the VA. After editing and rating the statements per FJA protocol, statements were distributed to all primary care personnel in 9 sites (including the original five) for verification; participants indicated the frequency and duration of each task that they personally performed. Population Studied: We analyzed the jobs of physicians, advanced practitioners, RNs, LVNs, health technicians, clerks, and pharmacists. Each focus group contained 6-8 subject matter experts from each job title across 5 sites. Surveys were then distributed to the entire complement of primary care employees across the 9 sites. Principal Findings: Approximately 65% of the tasks performed by physicians have potentially serious consequences if errors are made, while only 1 – 4% of the tasks performed by RNs, LVNs, Health Techs, and clerks carry the same severity of error consequences. These data suggest the most consequential tasks also emphasize the importance of decision-making processes: HEC, reasoning, and data requirements of physician tasks were significantly correlated (r’s = .65, .51, and.74 respectively). On average, tasks with very high HEC ratings were also below the midpoint on discretion, thus suggesting that the most consequential physician tasks also leave little room for judgment. This relationship pattern also generally held true for RNs, LVNs, and Health Techs. Conclusions: Physicians have a high amount of work with potential for harm compared to other primary care personnel, thus potentially increasing error probability. The VA has addressed this concern via strategies that minimize subjective decision-making. For example, practice guidelines are promoted as prescriptions for ensuring high quality of care with maximum patient safety. Our research supports this observation, though results are preliminary at present. Implications for Policy, Delivery, or Practice: The inverse relationship between worker discretion and human error consequence is reflective of a trend towards regulating clinical practice through guidelines and pathways as decision-making tools. However, health care settings are characteristically complex, adaptive systems that co-evolve with their environment. Thus, algorithms or practice guidelines may be helpful but insufficient strategies for healthcare quality and patient safety. Initiatives that help cultivate good clinical decision-making may be a complementary strategy for improving patient safety. Task analyses can help clinics make better, evidence-based decisions about specific strategies and initiatives. Primary Funding Source: VA ●Resident Physicians Report on Adverse Events and Their Causes Reshma Jagsi, M.D., DPhil, Barrett T. Kitch, M.D., MPH, Debra Weinstein, M.D., Eric Campbell, Ph.D., Matthew Hutter, M.D., Joel Weissman, Ph.D. Presented By: Reshma Jagsi, M.D., DPhil, Resident Physician, Radiation Oncology, Massachusetts General Hospital, Cox 3, 100 Blossom Street, Boston, MA 02114; Tel: (617)470 2091; Fax: (617)726 3603; Email: rjagsi@partners.org Research Objective: Resident physicians are front-line providers of medical care in inpatient and ambulatory settings, and thus have a unique vantage point from which to comment upon patient safety. This study explored residents’ experiences regarding patient safety and perceived causes of adverse events. Study Design: A questionnaire assessed residents’ experiences with adverse events (AEs), mistakes, and nearmisses, as well as opinions regarding the causation of these events. Data were gathered about most recent event, and the number of events in the last week. Data also were gathered on post-graduate year, specialty, chronic fatigue, and characteristics of the last week of practice including setting of rotation, patient load, and work hours. Population Studied: Responses were obtained from 821 residents and fellows in the 76 accredited programs at the Massachusetts General Hospital and Brigham and Women’s Hospital in 2003 (response rate = 57%). Analyses were limited to the 689 responses from trainees in a primarily clinical year of training. Principal Findings: Over half (55%) of the trainees in this study reported ever having a patient under their care who suffered an AE. The most common types of AE were procedural and medication-related. Over two-thirds were considered serious. Respondents in surgical specialties who reported events described them as significant or worse 83% of the time, compared with 66% for medical and 67% for hospital-based specialties (p=.002). Of the most recent AEs, 24% were attributed to mistakes, and trainees felt at least partially responsible for 77% of these mistakes. The most common reasons for the mistakes as perceived by trainees were working too many hours (19% of mistakes), inadequate supervision (20%), and problems with hand-offs (15%) In the last week, 114 respondents (18%) reported having a patient under their care suffer an AE, 42 (6%) reported AEs involving a mistake for which they felt responsible, and 141 (23%) reported near-miss incidents for which they felt responsible. In multivariate analyses, significant predictors of AEs in the past week were inpatient rotation, procedural specialty, and duty hours > 80 in the past week (all p<.05). Predictors of near-miss errors were inpatient rotation, PGY-1 status, and days of fatigue in the past month (all p<.05). Conclusions: Information obtained from front-line trainees suggest that adverse events are commonly encountered and often associated with errors. By showing a direct correlation between fatigue and errors in a broad spectrum of specialties other than medicine, and in settings outside of the ICU, these findings add to evidence from studies by Landrigan, et al, and Lockley, et al, that suggested that long work hours may lead to increased attentional failures and errors. Implications for Policy, Delivery, or Practice: Eliciting residents’ perspectives is important because residents may perceive events, actions, and causal relationships that chart reviewers or observers cannot. This study provides essential data for designing patient safety interventions and provides evidence supporting recent policies aimed at reducing resident work hours. Primary Funding Source: Anonymous Donor ●U.S. Providers Reported Use of the Human Papillomavirus (HPV) Test for Recommended and NonRecommended Indications: Results from a Nationally Representative survey, 2004 Nidhi Jain, M.D., MPH, Kathleen Irwin, M.D., NPH, Rheta Barnes, MSN, MPH, Daniel Montano, Ph.D, Danuta Kasprzyk, Ph.D, Crystal Freeman, Ph.D., MPH Presented By: Nidhi Jain, M.D., MPH, Medical Epidemiologist, Division of STD Prevention, Centers for Disease Control and Prevention, 1600 Clifton Road, Mailstop E-80, Atlanta, GA 30333; Tel: (404)639-1823; Fax: (404)6398607; Email: ncj0@cdc.gov Research Objective: Human papillomavirus (HPV) infection is associated with genital warts, cervical intraepithelial neoplasia (CIN) and anogenital cancers. In the U.S., 20 million people are infected with HPV and 5.5 million new infections are diagnosed annually making it the most common sexually transmitted infection at an economic burden of 4 billion dollars. Recently, CDC and several national clinical organizations recommended a recently licensed test for HPV for providers to test patients for two reasons: to guide the management of women with borderline abnormal (“ASCUS”) Pap test results and to use as an adjunct to Pap testing for women aged =30 years to monitor persistent HPV infection which may lead to cervical cancer. How U.S. providers are currently using the HPV test is not known. Study Design: Providers received self-administered mail surveys May-August, 2004 addressing their HPV testing practices. Percentages were weighted for disproportionate sampling and non-response by specialty. Population Studied: U.S. physician and mid-level provider specialties who provide primary care or are likely to manage patients with genital warts were selected randomly from national registries and master files. Sampled physician specialties included internal and adolescent medicine, family/general practice, obstetrics/gynecology, urology, and dermatology as well as certified nurse-midwives, physician assistants, and nurse-practitioners (n=6906). Principal Findings: The adjusted response rate was 81%. Of all respondents, 17% were unaware of the new HPV test; of those aware of the test, 47% reported ever testing their patients for HPV. Obstetricians/gynecologists (92%) and nurse-midwives (88%) were more likely to report HPV testing compared to primary care providers (31%-56%) and specialists (5-10%). Some clinicians reported HPV testing for uses not recommended by guidelines including: upon request of female (58%) or male (32%) patients; for female (32%) or male (21%) patients with anogenital warts, for female (34%) or male (18%) patients with non-HPV sexually transmitted infections, and to check infection status of female (30%) and male (13%) patients as a primary screening test for HPV infection. Among the subset of providers who performed Pap tests for cervical cancer screening (n=2930), 74% reported HPV testing women with “ASC-US” Pap results for management (recommended use), 48%-67% reported testing patients with higher-grade Pap abnormalities (nonrecommended use), and 21% reported HPV testing women as an adjunct to Pap tests with 12% of providers usually/always testing women >30 years (recommended use) and 15% of providers usually/always testing women <30 years (nonrecommended use). Obstetrician/gynecologists and nursemidwives most commonly reported HPV testing for nonendorsed reasons. Conclusions: In this nationally representative sample of clinicians, most U.S. providers are not currently testing their patients for HPV. Of those who do, many report using the HPV test for reasons consistent with recent guidelines, but some providers are testing for other non-recommended reasons. Testing practices for HPV varied by clinician specialty and non-recommended testing uses were most common in those specialties more likely to report HPV testing. Implications for Policy, Delivery, or Practice: Survey findings should be used to develop/update clinical training materials, decision support tools and patient educational materials to encourage HPV testing that is consistent with national guidelines. Primary Funding Source: CDC ●Failure of ICD-9-CM Coding to Identify Chronic Kidney Disease in Patients with Diabetes Elizabeth Kern, M.D., Donald R. Miller, Sc.D., Chin-Lin Tseng, DrPH, Miriam Maney, MS, David Aron, M.D., MS, Leonard Pogach, M.D., MBA Presented By: Elizabeth Kern, M.D., VA Quality Scholar, Education, Cleveland Department of Veterans Affairs Medical Center, 10701 East Boulevard, Cleveland, OH 44106; Tel: (216) 368-6129; Fax: (216)368-5824; Email: eok@case.edu Research Objective: Diabetes and chronic kidney disease (CKD) are increasingly prevalent co-morbid diseases that are independent risk factors for poor outcomes, especially in patients with cardiovascular disease. There is evidence that CKD is often under-diagnosed. However, methods of casemix adjustment for quality performance measures frequently depend on ICD-9-CM diagnosis codes in administrative records to identify co-morbid disease. The objective of this study was to determine how frequently and accurately ICD-9CM diagnosis codes in administrative records identify CKD in individuals with diabetes. Study Design: We used the 4-variable (age, sex, race, serum creatinine) Modification of Diet in Renal Disease predictive equation to estimate glomerular filtration rate (eGFR) for a national, retrospective cohort of veterans with diabetes. We calculated eGFR from two or more serum creatinine measurements over the span of one year beginning 9/30/1998. Using eGFR, we classified individuals for presence or absence of CKD according to National Kidney Foundation guidelines. Over 70% of veterans were Medicare beneficiaries. Searching linked VA and Medicare inpatient and outpatient administrative records, we determined the extent to which individuals with CKD were recognized by various algorithms of ICD-9-CM diagnosis codes. Population Studied: The population studied was 561,088 veterans with diabetes who were alive as of 9/30/1998, without a diagnosis of end-stage renal disease or dialysis for two years prior. Principal Findings: We were able to determine eGFR for 263,730 individuals (47.0%). The crude prevalence of CKD was 32% (83,338/263,730), declining to 29% when adjusted to the US diabetic population over age 40. CKD rose sharply with age, from a crude rate of 3.8% for those under 40 years old, to 46.1% in those 70 years and older. Using an algorithm of nine ICD-9-CM codes indicating kidney failure, only 20.2% (16,864/83,338) of individuals with CKD were able to be identified by diagnosis codes in both inpatient and outpatient records over the time span of one year (true positives). The false positive rate was 1.2% (1,034/180,392). Expanding the algorithm to include any of 79 renal-related diagnosis codes increased identification of true positives to 42.4% (35,318/83,338), while the false positive rate increased to 6.8% (12,317/180,392). Conclusions: CKD is a common co-morbidity in older veterans with diabetes. Despite having two or more laboratory encounters for serum creatinine tests within a one year time span, individuals with CKD generally failed to be coded for their disease in either VA or Medicare reimbursed encounters. However, if present, ICD-9-CM codes were highly specific for the presence of true CKD. Implications for Policy, Delivery, or Practice: The failure to code for CKD in populations of patients with diabetes and CKD may adversely affect the validity of quality performance measures that depend on ICD-9-CM diagnosis codes in administrative records to identify co-morbid disease. Primary Funding Source: VA ●Quality and Patient Safety in Telephone Medicine Shersten Killip, M.D., MPH, Carol L Ireson, Ph.D., RN, Steven T Fleming, Ph.D., Margaret M Love, PhD, Carol M McLay, BSN MPH, Whitney J Katirai, BA Presented By: Shersten Killip, M.D., MPH, Associate Residency Director FP, Family and COmmunity Medicine, University of Kentucky, K 302 KY Clinic 0284, Lexington, KY 40536; Tel: 859-323-5917; Fax: 859-323-6661; Email: skill2@email.uky.edu Research Objective: The purpose of this project was to analyze the telephone medicine systems at an academic health center from a systems perspective, looking for potential threats to patient safety. The research questions included: (1) Are there currently gaps in our telephone care system which may contribute to patient adverse events? and (2) Can a systems based evaluation method be developed to identify key components in patient safety as it relates to telephone medicine? Study Design: This study used both quantitative and qualitative design elements. Telephone interviews were conducted with a sample of 64 patients over 10 weeks. Analyses of the outcomes from each telephone call, including systems-level outcomes (was the patient called back, did the message get to the primary care doctor, etc) and patient-level outcomes (checking that the patients felt better/followed instructions/did not suffer any adverse events) were done. Interviews were conducted with all residents and attendings at the University of Kentucky Family Practice in person. Process data from these interviews were analyzed for existing and actual weaknesses in the system. Population Studied: The populations studied in this project are UKFP faculty, the UKFP residents, and the English, Spanish, or French-speaking adult patient population of the UKFP Clinic. The University of Kentucky Medical Center is an academic health center that draws on a wide area with the principal catchment area being the eastern half of the state. The Family Practice Residency is staffed with 9 faculty and 18 family practice residents. Principal Findings: Preliminary findings identified significant gaps in our telephone system, as well as patient satisfaction issues. Almost 10% of phone calls went unanswered, and nearly a third of patients rated their experience as poor. The majority of patients understood and followed the physician’s recommendations which were a visit to the emergency room or an over the counter medication. Less than a third of patients sought no further care after their telephone call. An unexpected finding was that the majority of callers were proxies for the actual patients (i.e. mothers calling for children), adding yet another layer of complexity to telephone medicine. Conclusions: The telephone medicine system at the UKFP has system and patient satisfaction problems. Problems with patient safety were implied by failure to answer calls, but few actual adverse events were reported. This is not surprising given the small sample size and the level of intervention usually done over the phone; however, the level of potential harm was quite high. Implications for Policy, Delivery, or Practice: Telephone medicine is a large part of ambulatory patient care, and it has rarely been investigated for patient safety issues as they relate to the telephone systems. This study will allow us to redesign our telephone system to better care for patient safety. A future project will evaluate the results of our intervention, then disseminate it to several related sites. Should this intervention be successful, and reproducible at other sites, the implications for the practice of telephone medicine could be enormous. Primary Funding Source: AHRQ ●Gestational Pyelonephritis as an Indicator of the Quality of Ambulatory Maternal Health Services Lisa Korst, M.D., Ph.D., Carolina Reyes, M.D., Moshe Fridman, Ph.D., Michael C. Lu, M.D., MPH, Calvin J. Hobel, M.D., Kimberly D. Gregory, M.D., MPH Presented By: Lisa Korst, M.D., Ph.D., Associate Professor, Pediatrics, USC Keck School of Medicine, 4650 Sunset Boulevard, MS #30, Los Angeles, CA 90027; Tel: (323) 671.7623; Fax: (323)906.8043; Email: LKORST@chla.usc.edu Research Objective: Inpatient conditions that might be avoided through improved outpatient services are called Ambulatory Care Sensitive Indicators, and they include pyelonephritis in non-pregnant adults. No such indicators have been developed for pregnant women. Because early screening and treatment of asymptomatic bacteriuria should prevent most cases, the purpose of this study is to examine whether hospital-specific rates of gestational pyelonephritis may serve as a measure of the quality of ambulatory maternal care. Study Design: The California Office of Statewide Health Planning and Development provided a data set linking the birth certificate to maternal and newborn delivery records for the 1997 calendar year. Antepartum records of hospital admissions for these women were included. For clarity, a “low-risk” study population was created, and patients with maternal, fetal and placental morbidities were largely excluded, as were women without first trimester prenatal care. We generated hospital-specific infection rates using a Bayesian hierarchical logistic regression model. Population Studied: All low-risk women and their newborns who delivered in California in 1997. Principal Findings: We identified 281,031 low-risk women, of whom 1,853 (0.66%) had at least one inpatient admission for gestational pyelonephritis. The model suggested only two significant risk factors: MediCal as a payer (OR = 1.62 compared with all other payers), and African-American race (OR = 1.24 compared with Caucasian race). Women with pyelonephritis were more than twice as likely to deliver preterm. Adjusted rates of gestational pyelonephritis for the 291 hospitals in the sample ranged from 0.25% to 2.34%. Conclusions: These findings suggest that because of its preventability and its consequent related morbidity, gestational pyelonephritis meets clinical requirements as a quality indicator for ambulatory maternal care. Hospitalspecific pyelonephritis rates also meet technical requirements for quality indicators. Implications for Policy, Delivery, or Practice: The use of such rates would provide an opportunity for hospitals to improve patient outcomes through partnership with admitting obstetricians in the development of policies and procedures for women at risk. Primary Funding Source: Maternal-Child Health Branch of the California Department of Health Services, and the Agency for Healthcare Research and Quality ●The National Database of Nursing Quality Indicators: Developing Measures of Pediatric Nursing Quality Susan R. Lacey, Ph.D., RN, Susan F. Klaus, RN, MSN, Janis B. Smith, RN, MSN, Karen S. Cox, Ph.D., RN Presented By: Susan R. Lacey, Ph.D., RN, Associate Professor, College of Nursing, University of Alabama in Huntsville, Sparkman Drive, Huntsville, AL 35899; Tel: (256) 824-2438; Email: laceys@uah.edu Research Objective: This research sought to identify and pilot test nurse sensitive pediatric indicators for inclusion in the National Database of Nursing Quality Indicators (NDNQI), the American Nurses Association's repository for national unit-based benchmarking that measures nursing's direct impact on quality linked to staffing. Study Design: 1) Exploratory, descriptive approach including literature review, consultation with experts, and comment from clinical and administrative leaders in pediatric care. 2) National multi-site pilot test to examine the reliability of data collection tools. Population Studied: This study included pediatric patients from 26 units across the United States; 13 pediatric, medicalsurgical, 6 pediatric intensive care, and 7 neonatal units. Principal Findings: This study found the indicators developed - 1) pain assessment/intervention/reassessment cycle and 2) peripheral intraveneous infiltration assessment - nurse sensitive and feasible. Full scale nation-wide implementation is now in progress for all participating NDNQI sites. Conclusions: This work, coupled with other quality initiatives, helps establish comparison benchmarks for pediatric patients and supports evidence of nursing’s impact on quality patient outcomes. ANA and NDNQI continue to lead the way in nurse sensitive quality initiatives. Implications for Policy, Delivery, or Practice: These new pediatric indicators will 1) widen the effort to protect our most vulnerable patients from neonates to adolescents and 2) further the scientific study between patient outcomes and staffing. Primary Funding Source: This study was performed under contract to the American Nurses' Association ●Comparing and Assesing Risk Adjustment Methods for Quality Report Yue Li, Ph.D. Candidate, MS, Andrew W. Dick, Ph.D., Dana B. Mukamel, Ph.D. Presented By: Yue Li, Ph.D. Candidate, MS, Community and Preventive Medicine, University of Rochester Medical Center, 601 Elmwood Ave., Box 644, Rochester, NY 14642; Tel: (585)260-8195; Fax: (585)461-4532; Email: yue_li@urmc.rochester.edu Research Objective: Quality report-cards provide information about the relative performance of providers based on riskadjusted health outcomes. The validity of these performance reports depends on the ability of statistical modeling techniques to correctly adjust for differences in patient mix, and on the correct specification of quality measures. Current practice is to model risk assuming an additive provider effect (quality) on patient risks and outcomes, rather than allowing for interaction between them. Quality measures are defined arbitrarily as either additive (O-E difference: difference between Observed and Expected outcomes) or multiplicative (O-E ratio). The difference measure is consistent with the additive statistical risk adjustment model while the ratio specification is not. Furthermore, the exact relationship between quality, patient risks and patient outcomes is usually unknown. This study was designed to examine the consequences of misspecified risk-adjustment model and quality measures. We investigated whether different specifications of risk-adjustment models and quality measures return consistent quality rankings, eg, identification of high or low quality-outliers. Study Design: We developed hospital quality measures based on mortality by estimating logistic and linear probability regression models on the 2002 New York State Cardiac Surgery Reporting System (CSRS) data. This dataset includes detailed clinical information for patients undergoing isolated coronary artery bypass graft (CABG) surgeries. We assumed either additive or multiplicative fixed effects of quality in model estimation, and constructed corresponding additive or multiplicative quality measures based on mortality rate or odds of mortality. Kappa statistics were used to test for agreement between these measures. We also performed computer simulations, based on the empirical data, to explore how the commonly used risk-adjustment method (based on conventional logistic regression) performs for varied patterns of quality effects. Population Studied: All 16120 patients undergoing isolated CABG surgeries at 36 New York State hospitals in 2002. Principal Findings: All risk-adjustment models calibrated approximately equally well to the data (Hosmer-Lemeshow statistics = 8.2-10.0, P = 0.27-0.41), and had excellent discrimination (c-statistics > 0.80). Among the 36 hospitals offering the surgery, four were designated as quality outliers by all quality measures. Three others were flagged as such by some, but not all, measures. Kappa statistics showed fair (kappa = 0.2 – 0.4) to substantial (kappa > 0.8) agreement between measured quality of the hospital. In the simulation we found that when the risk-adjustment was performed on a small sample (eg, number of patients in a hospital <100), only 60-70 percent of the true quality-outliers can be identified for some simulated specifications of quality effects; when the sample size was relatively large (eg, >500 per hospital), over 90% of the true outliers can be identified for all simulated quality effects. Conclusions: Existing risk-adjustment methods potentially oversimplify the patterns of the interaction between quality of care and patient characteristics and thus cannot estimate quality accurately. This problem is mitigated when quality measures are based on a large sample of patients. Implications for Policy, Delivery, or Practice: The effectiveness of health care quality reports depends on the accuracy of the performance information they convey. Appropriate statistical modeling is essential to ensure valid translation of outcomes data, especially of small sample, into meaningful quality measures. Primary Funding Source: No Funding Source ●Nursing Factors and Surgical Outcomes in VHA Elliott Lowy, Ph.D., Sales, Anne E, Ph.D., RN, Nancy Sharp, PhD, Gwendolyn Greiner, MPH, Yu Fang Li, Ph.D., RN Presented By: Elliott Lowy, Ph.D., HSR&D, VA, 1100 Olive Way, Seattle, WA 98101; Tel: (206)277-4789; Fax: (206)7685343; Email: elliott.lowy@med.va.gov Research Objective: We report on the association between nursing factors, including staffing levels and skill mix, and patient outcomes after surgery, including overall in-hospital mortality and in-hospital failure to rescue. Failure to rescue measures mortality of patients who experience serious, possibly nursing-sensitive, complications after surgery. Study Design: We present an observational, cross-sectional study using archival data. We used the DSS department budget and cost report (ALBCC) for number of nursing personnel hours per month by type of nurse, and we used responses to a survey of all VHA nursing personnel in 124 facilities, mailed 2/03-6/03. Patient data, for surgeries in VHA facilities between February and June 2003, come from the National Surgical Quality Improvement Program, and include patient characteristics, pre-existing conditions, complications, and outcomes of surgery. Data on nursing units come from the DSS TRTIPD files, a new DSS extract file that can link inpatients to nursing units. All analyses were conducted at the nursing unit level. Population Studied: All patients who stayed on an acute med-surg unit at any of 124 VHA facilities between February 1, 2003 and June 30 2003. Principal Findings: With 510 inpatient units in the sample, overall mean unit-level mortality rate was 4.0% (SD 12.3%); mean failure to rescue rate was 20.1% (SD 27.5%). RN nursing hours per patient day (RNHPPD) are positively and significantly correlated with unadjusted overall mortality, probability of experiencing one of the serious complications after surgery, and failure to rescue. Non-RN hours per patient day (OTHPPD) are negatively and significantly correlated with overall mortality, probability of experiencing one of the serious complications, and conditional mortality. A likelihood score for complications was created, based on age, gender, race, Charlson index, and ICU stay; after risk adjustment for this likelihood, RNHPPD and OTHPPD are no longer significantly correlated with either mortality or failure to rescue. Conclusions: Risk adjustment is critical. Conducting patientlevel analyses is essential for power, but on unit-level analysis, there may be no significant relationship between nursing hours and patient outcomes. Implications for Policy, Delivery, or Practice: Final results are pending, but this is the first large scale study of nursing factors and patient outcomes with unit-level data. A finding of no significant relationship might call into question results of prior studies suggesting that nursing hours are an important factor in patient mortality outcomes. Primary Funding Source: VA ●Organizational Factors and Human Factors Related to Harmful Medical Event Outcomes in 23 Academic Medical Centers – Evidence using Electronic Medical Error-Event Reporting Systems for Targeting Patient Safety Programs Sandra Magnetti, MS, DrPH, Raj Behal, M.D., MPH Presented By: Sandra Magnetti, MS, DrPH, Post doctoral fellow, Clinical Practice Advancement Center, University HealthSystem Consortium, 2001 Spring Road, Oak Brook, IL 60523; Tel: (630) 954-1719; Fax: (630) 954-5879; Email: magnetti@uhc.edu Research Objective: To identify human and organizational factors, which are significantly related to harmful vs. nonharmful medication events. In response to the IOM ‘s report that at least half a million preventable injuries and 44,000 to over 98,000 preventable deaths occur in hospitals each year, at a cost exceeding $30 billion due to medical errors, we utilized a web-based hospital event reporting system to identify human organizational factors that are related to patient harm. Study Design: We analyzed medication error-event data (representing approximately 25% of total medical events) reported over a three-year period submitted by 25 academic medical centers, which participated in a collaborative, on-line, web based event reporting system. Front-line clinical staff entered information on the events and scored the level of harm using a progressive 10-point scale ranging from near miss to patient death due to the event. Unit managers had the option to review the report and comment on factors that contributed to the event. The input screen format allowed them to select one or more of 36 human and organizational factors, depending on their perception of the event. We used the nurse unit manager’s report to correlate these human and organizational factors with a harm/no harm score. The score was developed by dichotomizing the progressive 10-point scale into those events that caused harm to the patient ranging from temporary harm requiring extra intervention to patient death versus those events that caused no harm to the patient. Chi-Square statistics were used to evaluate these relationships. Population Studied: Medication error-event data from inpatients aged 18 or older were selected from 23 academic medical center hospitals’ web based reporting systems for years 2001-2004. Principal Findings: We found that of the 24,809 reported medication events, 10.49% were related to some degree of harm or discomfort to the patient. The remaining 90.51% of the reported events did not reach the patients or require additional treatment. Twelve of the 36 Human and Organizational factors were significantly related to the harmful events; a p value <= 0.001 was used to determine significance for these factors using Chi Square analysis. These events were significantly related to factors centering around each of six areas: TEAM FACTORS: “communication problem between providers”, WORK ENVIRONMENT: “distractions/interruptions”, “high noise level”; STAFF FACTORS: “staffing insufficient”, “system for covering patient care”; TASK FACTORS: “inexperienced staff”, “inadequate resident supervision”, “order entry system problem”, “emergency situation”, “cardiac/respiratory arrest situation”; ORGANIZATION/MANAGEMENT: “bed availability”; and PATIENT CHARACTERISTICS: “patient understanding”. Conclusions: This study reveals that there are specific human and organizational factors related to medication error-events, which, either necessitated additional treatment measures, were life threatening, or resulted in death. Implications for Policy, Delivery, or Practice: This project provides information as to which events are more associated with harm to the patient and could be used to target errorevent prevention programs. This study also provides empirical evidence as to the use of and importance of the development of standardized web-based reporting systems. Primary Funding Source: No Funding Source ●Functional Performance and the Continuum of Rehabilitation Care Chetan Malik, MBBS, Carl V. Granger, M.D., Carol M. Brownscheidle, Ph.D. Presented By: Chetan Malik, MBBS, Clinical Instructor, Dept. of Rehabilitation Medicine, University at Buffalo, 4351 Chestnut Ridge RD #8, Amherst, NY 14228; Tel: (716)8177867; Email: cmalik@buffalo.edu Research Objective: Analyze the relationships among admission FIM(TM) ratings, etiological diagnosis and utilization of home-based rehabilitation services. Study Design: The FIM(TM) instrument allows a variety of health care professionals to assign a rating to a patient’s level of function in 18 motor and cognitive tasks that represent basic activities of daily life. FIM(TM) ratings range from 0-126 with higher values representing more independence or fewer hours of help in performing daily functions. Based upon the admission FIM(TM) ratings patients were divided into 4 groups: Group 1: <=60 (N=56), Group 2: >60 to <= 90 (N=177), Group 3: >90 to <=100 (N=208) and Group 4: >100 (N=424) respectively representing >=4, <4 to >=1, <1 to >0 and zero hours of help needed. The FIM(TM) database was searched for records of patient receiving home-based rehabilitation. Admission FIM(TM) ratings of patients receiving rehabilitation services at home were examined to determine potential relationships to utilization of physical therapy (PT), occupational therapy (OT), and home health aide (HHA) services. Net rehabilitation charges were also compared among the four groups. Kruskall-Wallis test was used to compare mean PT, OT and HHA visits among the four FIM(TM) groups. Five etiological groups (based on ICD codes) were represented in large enough numbers to permit statistical analysis: musculoskeletal (N=254), circulatory, including cardiac and stroke (N=199), poisoning and injuries, including motor vehicle accidents (N=194), infection (N=44), and nervous system disorders (30). Population Studied: 865 patients receiving home-based rehabilitation were studied. Mean age was 75 years (SD=12; range 13 to 98 years); 66% were females, 33% were males. Data on gender was not available for 1% of patients. Principal Findings: Higher FIM(TM) ratings were associated with fewer PT visits (17, 15, 12 and 11 respectively for Groups 1 to 4 p<0.001), OT visits (12, 7, 3 and 2 respectively for Groups 1-4, p<0.001), HHA visits (21, 10, 6 and 3 for Groups 1-4 respectively, p<0.001) and lower mean net rehabilitation charges ($ 3883, 2581, 1592 and 1324 for Groups 1-4 respectively, p<0.001). Patients with circulatory and neurological disorders required more OT visits (7 and 9 respectively) as compared to patients with infections (4 visits), poisoning and injuries (3 visits), and musculoskeletal diseases (1 visit). Number of HHA visits was 1 for patients with musculoskeletal disorders and 2 in the remaining four impairment groups. Number of PT visits was 13 for patients with circulatory disease and poisoning and injuries and 12 for the other three groups. Conclusions: Higher FIM(TM) ratings are associated with decreased utilization of home-based rehabilitation services, and lower rehabilitation charges. Functional performance at the onset of home-based rehabilitation care appears to be a better determinant of future in-home rehabilitation services than etiological diagnosis. Implications for Policy, Delivery, or Practice: In 2002, the FIM(TM) instrument was adopted by the Centers for Medicare and Medicaid Services (CMS) for inpatient rehabilitation assessment under the prospective payment system. The results of the present study indicate the value of the FIM(TM) instrument as an assessment tool in the home care venue of the continuum of post-acute care settings. Primary Funding Source: No Funding Source ●Towards a Framework for Long-Term Maintenance of Quality Measures: Results of a Survey of Leading Measures Developers Soeren Mattke, M.D., D.Sc., Soeren Mattke, Cheryl Damberg, Ph.D. Presented By: Soeren Mattke, M.D., D.Sc., Scientist, RAND, 1200 South Hayes Street, Arlington, VA 22202; Tel: (703)4131100; Fax: (703)413-8111; Email: mattke@rand.org Research Objective: In recent years, quality measures have gained increasing importance not only for research but also for policy and practice, such as in pay-for-performance or public reporting initiatives. Their new role requires developers and users to ensure that measures continue to meet rigorous standards in the long run. However, little is known about effective and efficient approaches to measures maintenance. Study Design: We conducted a survey of ten nationally recognized measures developing organizations and selected researchers about their current polices and procedures as well as desirable properties for a comprehensive system for measures maintenance. We consolidated their input into a briefing document for panel discussions with all respondents to arrive at consensus recommendations for a measures maintenance framework. Population Studied: Five measures developers, two provider and three purchaser organizations. Six were private sector organizations; two governmental agencies and two selfregulatory institutions. Principal Findings: All organizations had procedures for measures maintenance, but the degree of formalization of the procedures varied. Three key functions for a measures maintenance system emerged: Ad hoc review to deal with unexpected problems with a measure, annual maintenance to incorporate changes in coding conventions and regular reevaluation to thoroughly review measures in pre-defined intervals. Importance, scientific soundness, feasibility and usability were universally used as evaluation criteria. The panel discussions yielded a consensus set of recommendations for relationships between maintenance functions, evaluation criteria and measures disposition. Conclusions: A sufficient degree of implicit consensus was found among leading measures developers to arrive at a consensus framework for policies and procedures for measures maintenance. Implications for Policy, Delivery, or Practice: This framework offers a starting point for the development of measures maintenance systems. While different organization may choose to implement it in a way that is most consistent with their mission and structure, it provides guidance as to which components should be included. Primary Funding Source: CMS ●Challenging Accepted Wisdom on Rules, Blame and Safety Culture: a Hospital Operating Theatres Case Study. Ruth McDonald, Ph.D., Justin Waring, Ph.D., Stephen Harrison, Ph.D. Presented By: Ruth McDonald, Ph.D., Research Fellow, National Primary Care Research and Development Centre, University of Manchester, Williamson Building, Oxford Road, Manchester, M13 9PL; Tel: +44 (0) 161 275 3535; Fax: +44 (0) 161 275 7600; Email: ruth.mcdonald@man.ac.uk Research Objective: The current orthodoxy within patient safety research and policy is characterised by a faith in rulesbased systems, which limit the capacity for individual discretion, (and hence fallibility) and the eradication of ‘blame cultures’, which are assumed to discourage clinicians from reporting errors. This study examines how different groups of hospital staff (doctors and managers) view guidelines and errors, in order to assess both the extent to which this orthodoxy is accepted by hospital staff and the likely barriers to implementing these policy solutions. Study Design: An ethnographic study of a hospital operating theatres department in an acute care hospital. In addition to observation of activities and documentary analysis, 39 semistructured interviews were conducted. Interviews were taperecorded and transcribed verbatim. Population Studied: Clinicians and managers working within the operating department of a large teaching hospital in the North of England. Interview participants were consultantgrade surgeons from a range of specialties (general surgery, urology, ENT, maxillofacial, cardiothoracic, gastrointestinal, ophthalmic, orthopaedic), consultant anaesthetists and departmental managers. Principal Findings: We identified two different and distinct sets of narratives, with the accounts given by medical staff challenging the belief of mangers that surgical work can be governed by formal rules and procedures. Whilst doctors rejected written rules, they adhered to the unwritten rules of what constitutes acceptable behaviour for members of the medical profession. This includes eschewing guidelines and refusing to pass judgement on other medical colleagues. In contrast to ‘cognitive bias’ based explanations of physician behaviour, which depict doctors as being overoptimistic, we found that doctors expected to make mistakes and accepted them as inevitable. The fear of blame was less of a barrier to reporting than the belief that since mistakes were inevitable and all patients and contexts are different, then the ability to learn from mistakes is negligible. The manager narrative views mistakes as preventable, places faith in formal rules and processes and lays blame at the door of unruly medical staff. Reducing this propensity to blame individuals is a remote prospect and is by no means guaranteed to increase error reporting, particularly if doctors see mistakes as inevitable and do not recognise failure to comply with guidelines as an error or violation. Conclusions: Current ‘solutions’ intended to improve patient safety may understate the impact and persistence of different norms, values and attitudes amongst different groups of staff on behaviour in the hospital setting. Rather than a safety culture with shared values and norms, it may make more sense to talk of different identities (managerial and medical) and the different attitudes to risk, blame and error associated with these. Implications for Policy, Delivery, or Practice: Solutions based on standardisation, systematisation and developing a safety culture are likely to prove difficult to implement and may fail to deliver the benefits anticipated. Primary Funding Source: Department of Health (England) ●Physician Attitudes Toward Pay-for-Performance Programs: Development and Validation of a Measurement Instrument Mark Meterko, Ph.D., Gary J. Young, Ph.D., Bert White, MBA, James F. Burgess, Ph.D., Dan Berlowitz, M.D., Matthew R. Guldin, MPH Presented By: Mark Meterko, Ph.D., Manager, Methodology & Survey Unit, Center for Organization, Leadership & Management Research, VHA HSR&D, VA Medical Center (152M), 150 South Huntington Avenue, Boston, MA 02130; Tel: (617)278-4433; Fax: (617)232-6140; Email: mark.meterko@med.va.gov Research Objective: Financial incentive programs for physicians have become increasingly prevalent as a mechanism for reducing care process variation, improving clinical outcomes and reducing costs. Empirical support for the efficacy of such programs is scarce, however, due in part to the lack of a conceptual model of the features of incentive programs that are salient to physicians, and to the lack of a psychometrically sound measure of physician attitudes toward those features. The goal of the present study was to develop, test and evaluate such a measure. Study Design: The study involved three phases: (a) the development of a conceptual model of the factors relevant to physicians’ perceptions of and reactions to financial incentives; (b) the pilot testing of a preliminary questionnaire representing those constructs through focus groups with physicians at three medical practices; and (c) the large-scale mail administration of the revised survey and thorough psychometric assessment of the proposed measures. Population Studied: Focus groups were conducted at three eastern Massachusetts medical practices selected to represent a variety of relationships between physician groups and payers involved in incentive programs. Each group involved 13-15 primary care physicians and lasted 40-60 minutes. Information from this phase was used to modify a draft questionnaire in several crucial ways. The revised instrument was subsequently distributed to 2497 private and group practice physicians in New York State and Massachusetts who were affiliated with two of the seven sites participating in a larger national evaluation study of financial incentives. Principal Findings: A total of 798 physicians (32%) responded. Of these, 632 (79%) indicated that they were aware of a particular quality target and associated incentive and subsequently answered the key core survey questions. The aware respondents were randomly split into a derivation sample (n=316) and a validation sample (n=316). Exploratory factor analysis in the derivation sample suggested that six underlying dimensions could account for 57% of the variance in the 26 core items: awareness, credibility, control, cooperation, lack of unintended consequences, and impact. Multi-item scales representing these six constructs were then computed in the validation sample, and these data were subjected to a multi-trait analysis. Ninety percent of the itemto-hypothesized scale correlations demonstrated convergent validity, and 92% of item-to-other scale correlations demonstrated discriminant validity. Internal consistency reliabilities of the hypothesized scales ranged from .69 to .80. and were substantially greater than any of the correlations among the scales themselves. Only one scale (lack of unintended consequences) showed any substantial ceiling or floor effects (14% of respondents at the theoretical maximum). Conclusions: Six dimensions – awareness, credibility, control, cooperation, lack of unintended consequences, and impact – appear to be important for understanding physician reactions to quality target and financial incentive programs. Physician perceptions of these dimensions can be measured with reliability and validity using a relatively brief self-report questionnaire. Implications for Policy, Delivery, or Practice: There is a clear need for further systematic empirical research regarding the efficacy of linking financial incentives to quality targets as a way to influence physician practice behavior. The existence of a conceptual model of the salient features of targets and incentives, and of a valid and reliable measure of physician attitudes toward those features and of their self-reported impact, should contribute to the advance of that research agenda. Primary Funding Source: AHRQ, RWJF ●Beyond Staffing: Variation in Administratively Mediated Variables in Hospitals Ann F. Minnick, Ph.D., RN, FAAN, Lorraine Mion, Ph.D., RN, FAAN, Mary Johnson, Ph.D., RN, Catharine Catrambone, DNSc, RN Presented By: Ann F. Minnick, Ph.D., RN, FAAN, Independence Professor and Associate Dean-Research, College of Nursing, Rush University, 600 South Paulina Street, 1062A AAC, Chicago, IL 60612; Tel: (312) 942-6990; Fax: (312) 942-3038; Email: Ann_ F_Minnick@rush.edu Research Objective: To describe administratively mediated inputs (capital, employment, organizational, and labor) that comprise US hospital nurse working conditions. Determination of the interactive impact of working conditions on outcomes has been limited because most data sets contain only staffing data or patient variables. This has made it impossible to study the interrelationship of these inputs and to examine their overall impact patient outcomes. The extent inputs other than labor quantity varies by unit and by hospital has not been demonstrated. Such determination is needed to guide safety and quality efforts. Study Design: Descriptive. 130 work place inputs suggested by literature review, a national panel and the Minnick and Roberts clinical behavior production model were studied based on potential(1) explanatory power for specific outcomes and (2) impact on multiple outcomes. Capital inputs included those which may influence (1) the development of unsafe patient behavior (2) ability to observe patients (3) staff knowledge about the patient (4) visual and auditory stimuli. Examples include bed monitors, distances from bedside to supplies, computerization and room configuration. Employment terms included temporal terms (e.g. shift length) and work models (e.g. nursing, physician and support staff deployment). Organizational variables included worker empowerment and work process design. Labor variables included work force characteristics and labor quantity. The investigators conducted all observations, physical measurements, reviews of staffing data and interviews with unit personnel on-site in 2003-5 Population Studied: General acute care hospitals (mean occupancy 100 or more) regardless of ownership, teaching or profit status were included in the randomization procedure that encompassed the New York , Chicago, Houston, DallasFt. Worth, Denver and Phoenix metropolitan areas. Within each of the 40 hospitals, a general ICU, a medical and a surgical floor were studied . Principal Findings: Variation in administratively mediated variables exceeded variation in staffing quantity. ICU and nonICUs varied from each other on almost every variable; within hospital variation on non-ICU units equaled that noted across hospitals. There was little variation in characteristics of registered nurse assignment and use of ancillary nursing personnel on day shift in ICU. Large within unit variation was evident; e.g. distance to clean supplies varied on one unit from 3 feet to more than 200. The relationship of some inputs’ presence or absence suggests a potential unit resource categorization. Conclusions: (1) Most ICUs possess labor saving and safety capital inputs (2)Major improvements on non-ICU units are needed. (3) Assuming that administratively mediated variables within a hospital are similar is unwarranted. Implications for Policy, Delivery, or Practice: (1) Outcome production studies, including those determining the role of nursing labor, must account for these inputs and how they vary by unit. It is possible that these other inputs may mask the true impact of nursing labor and/or that nursing labor is a proxy for some input combination. (2) Determining the role of these inputs in producing outcomes could provide direction for input allocation. (3)As the US hospital system embarks on another cycle of bed replacement, avoiding many of the observed design flaws may be possible. (4) New typologies to categorize the work models are needed. Primary Funding Source: NIA ●Failure to Rescue or Failure to Measure What's Important? No Shortcuts to Assessing Quality James Moriarty, BS, BA, James Naessens, MPH, Dawn Finnie, MPA, Matthew Johnson, BA Presented By: James Moriarty, BS, BA, Data Analyst, Health Care Policy & Research, Mayo Clinic, 200 1st Street SW, Rochester, MN 55905; Tel: (507) 538-3973; Fax: (507)284-1731; Email: moriarty.james@mayo.edu Research Objective: To investigate the usefulness of the Failure to Rescue indicator defined by the Agency for Healthcare Research and Quality (AHRQ) Patient Safety Indicators (PSI’s) for monitoring hospital quality. Failure to Rescue has been defined as the death of patients who experience certain complications during hospitalization. Each complication has specific criteria which exclude a patient from the population at risk. Study Design: A retrospective cohort study was used to assess issues of interest. AHRQ PSI’s were applied to hospital administrative data. Failure to rescue rates were compared between complication categories. Analyses were also performed based on whether the condition putting the patient into the eligible population was present upon admission or was acquired during care. Trends over time of the Failure to Rescue population were assessed. Population Studied: The patients included in this study were all inpatients discharged from January 1, 1996 to December 31, 2003 at St. Marys Hospital (SMH) and Rochester Methodist Hospital (RMH), the two facilities of Mayo Clinic Rochester, both of which provide a full range of hospital care. Patients aged 75 or greater, who are excluded in the AHRQ measure, were kept in this study. The number of inpatients per year ranged from 52,793 in 1996 to 64,705 in 2003. Principal Findings: There were a total of 21,432 patients over the eight year timeframe. The AHRQ Failure to Rescue indicator is composed of six subgroups of complications: acute renal failure, deep vein thrombosis or pulmonary embolism (DVT/PE), pneumonia, sepsis, shock and/or cardiac arrest, and gastrointestinal (GI) hemorrhage and/or acute ulcer. The mortality rates for the six subgroups were: acute renal failure, 23.0%; DVT/PE, 9.3%; pneumonia, 13.5%; sepsis, 20.4%; shock/cardiac arrest, 58.0%; and GI hemorrhage/acute ulcer, 10.4%. The proportion of patients falling into each of the six subgroups at RMH differed from SMH with higher proportions of DVT/PE at RMH (25.1% vs. 18.1%) and higher proportions of pneumonia at SMH (35.1% vs.23.3%). Whether or not the condition is pre-existing or acquired can also have an effect on mortality. The mortality rate for patients diagnosed with acute renal failure classified as having the condition upon admission was 20.8%, but those patients that acquired the condition after admission had a mortality rate of 27.3% (p <.001). Conclusions: The results of this study show that the AHRQ Failure to Rescue measure could be improved to give a more accurate representation of the quality of patient care. These results should also be applied to the Failure to Rescue measure used by the National Quality Forum for their Core Measures of Nursing Care Performance which includes five of the six complication subgroups on major surgical patients with similar exclusion criteria. Implications for Policy, Delivery, or Practice: Hospitals with a greater percentage of patients in these higher mortality groups or with more patients with pre-existing complications may affect overall mortality rate for the Failure to Rescue PSI. This may lead to bias in comparing institutions on the measure. More study needs to be performed to attain reliable and valid quality measures. Primary Funding Source: No Funding Source ●The Scope of the End of Life: a Systematic Review of the Literature Richard Mularski, M.D., Joanne Lynn, M.D., MA, MS, Ronda Hughes, Ph.D., MHS, Karl Lorenz, M.D., MSHS, Sally Morton, Ph.D., Paul Shekelle, M.D., Ph.D. Presented By: Richard Mularski, M.D., Staff Physician, Fellow in Health Services Research, Medicine, VA Greater Los Angeles Healthcare System, 11301 Wilshire Boulevard, mailcode 111G, Los Angeles, CA 90073; Tel: (310) 478-3711; Fax: (310) 268-4933; Email: Richard.Mularski@med.va.gov Research Objective: Improving end-of-life care requires a definition of the population to whom we apply interventions and assess patients’ and caregivers’ experiences. A definition of the end-of-life population across the myriad of chronic progressive diseases has eluded consensus in the field. We aimed to systematically review literature that might inform a conceptual and operational definition of the end-of-life population. Study Design: As part of an evidence-based review on end-oflife care, we systematically reviewed articles that evaluated prognosis or trajectories of illness at the end of life. Inclusion criteria included English language, publication period from 1990 to 2004, and adult human subjects. Articles were selected using pre-determined criteria to inform the definition of end of life or prognosis at the end of life. Population Studied: Sources for our review included Medline, the Cochrane Database of Reviews of Abstracts of Effects (DARE), the National Consensus Project for Palliative Care, several recent systematic reviews from Health Canada and National Institute for Clinical Excellence (NICE - United Kingdom), and input from a technical expert panel. Our exclusions included articles addressing survival estimates for acute hospitalizations or disease processes, articles enrolling patients already in hospice or palliative care programs, and articles evaluating isolated prognostic factors or the natural history of a cohort. Principal Findings: From 24,423 total citations, we identified 348 potentially relevant articles, 63 of which ultimately met inclusion / exclusion criteria. We identified four themes for specification of the end-of-life period: (1) active dying, (2) patient readiness or acceptance of treatment limitations, (3) severity of illness, and (4) poor prognosis. The literature base failed to provide a uniform definition of the scope the end of life. Conclusions: In general, there are multiple and varied definitions of the end-of-life population. None of the constructs identified had substantial empirical validation of their defining characteristics. Prognostication models and clinician estimates lack precision and generalizability for counseling patients and families or for discriminating appropriate health care interventions. The prognostication literature does not support a clear definition for the end-of-life nor does it provide sufficient estimates for a statistical likelihood of dying within a specific time period. Implications for Policy, Delivery, or Practice: Our review identified several constructs and varied definitions of the scope of the end-of-life experience. The absence of a consensus definition or rigorous validation of prognostic methods poses a significant barrier to advancing the field of end-of-life care and evaluating interventions to improve patients’ and families’ experiences at the end of life. Research is needed to characterize the implications of alternative conceptual and operational definitions of the end of life. Novel approaches should consider methods other than attempting to determine a natural divide between everyday existence and the end-of-life period or attempting to determine a statistical divide based on the likelihood of dying within a certain time period. With increased use and testing of theoretical models for the end-of-life experience and consensus about the end-oflife scope, the field can be further advanced and the quality of care for the dying improved. Primary Funding Source: AHRQ, NINR ●Development of a Risk Adjustment Methodology for ICU Length of Stay Using Administrative Data Helen Neikirk, MA, Raj Behal, M.D., MPH, Liping Lu Presented By: Helen Neikirk, MA, Program Director, Analytic Services, Clinical Informatics, University HealthSystem Consortium, 2001 Spring Road, Suite 700, Oak Brook, IL 60563; Tel: (630) 954-2462; Fax: (630) 954-5879; Email: neikirk@uhc.edu Research Objective: To develop a risk adjustment methodology for Length of Stay (LOS)in the intensive care unit (ICU) using administrative data. Study Design: An administrative data base comprised of patient level data for all inpatient admissions from 106 academic medical centers and teaching hospitals was used to select patients who were admitted to an ICU during 2003. Cases with ICU LOS greater than 300 days were excluded from the study resulting in a total sample of 141,301 cases. A log transformation was performed to normalize the distribution of ICU LOS and linear regression analysis was conducted on the log-transformed dependent variable. Independent variables included age, sex, race, admission source transfer from another acute care hospital, admission source transfer from skilled nursing facility (SNF) or long term care facility (LTC), emergency admission status, mechanical ventilator use, tracheostomy, 30 specific comorbid conditions defined by AHRQ, and other diagnoses associated with increased severity of illness. The data set was randomly split into two equal subsets and a model was generated on one half and validated on the other half. Population Studied: In-patients aged 18 and older from 106 academic medical centers and teaching hospitals who were admitted to an intensive care unit in 2003. Principal Findings: R-square value of the final model was 0.33 with a maximum VIF of 1.79. Variables found to be highly significant predictors of increased ICU LOS included tracheostomy, small bowel surgery, sepsis, craniotomy, mechanical ventilation, weight loss/malnutrition, lymphoma, congestive heart failure, chronic anemia, hemorrhagic stroke, acute respiratory failure, and psychosis. Variables found to be highly significant predictors of decreased ICU LOS included diabetes, shock, coma, emergency admission, and transfer from SNF or LTC. The model results were applied to the entire data set to generate an expected ICU LOS for each case and an observed to expected ICU LOS ratio was calculated for each hospital. The hospital-specific bserved to expected ratios ranged from 0.57 to 1.53. Conclusions: This study demonstrates the usefulness of administrative data in risk adjustment for ICU LOS. The regression technique was able to account for a significant amount of variation with clinically meaningful variables derived from ICD-9 diagnosis and procedure codes and other administrative data elements. Application of the model revealed considerable variation in risk adjusted ICU LOS among hospitals. Implications for Policy, Delivery, or Practice: Widespread concern about quality, cost, and efficient utilization of health care resources indicate the need for cost effective methods for outcome evaluation. ICUs account for approximately 7 percent of hospital beds and 20 to 30 percent of hospital costs in the U.S. Clinically meaningful methodologies for risk adjustment that use administrative data will be important tools for evaluation of efficient use of ICUs in hospitals. Primary Funding Source: No Funding Source ●Making it Safe: The Effects of Leadership Inclusiveness and Professional Status on Psychological Safety and Learning in Health Care Teams Ingrid Nembhard, MS, Amy C. Edmondson, AM, Ph.D. Presented By: Ingrid Nembhard, MS, Ph.D. Candidate, School of Business Administration, Harvard University, Soldiers Field Road, Morgan Hall T20, Boston, MA 02163; Tel: (617) 495-1488; Email: inembhard@hbs.edu Research Objective: Most hospitalized patients require care from a range of disciplines, making teamwork in the routine delivery of care essential. Moreover, hospitals attempting to improve the quality of care often rely on interdisciplinary teams to carry out improvement projects. The present study investigates factors that promote team engagement in quality improvement (QI) work, despite the salience of a welldocumented status hierarchy in medicine. We aim to contribute to the understanding of the role of status in shaping perceptions of psychological safety (the belief that it is safe to speak up with questions, concerns or suggestions), and how these effects may be locally modified by leader behavior to facilitate team engagement in QI work. Study Design: We surveyed staff in 23 neonatal intensive care units (NICUs) that were members of a quality improvement collaborative. The NICUs were recruited by phone and email directed to the leader of each improvement team. We then invited all team members to participate in our survey via an invitational letter distributed in accordance with procedures outlined in their hospital IRB approval (i.e. email or staff mailboxes). We offered two versions of the survey (paper and web-based) to allow participants to choose their preferred format. Participants responded to items that measured their impressions of psychological safety, leadership inclusiveness (words and deeds exhibited by leaders that invite and appreciate others’ contributions) and team engagement in QI work, and reported their profession and other demographic information. Population Studied: Multidisciplinary health care teams from NICUs pursuing quality improvement. In all, 1,440 health care professionals (46% percent of team members contacted) from 23 NICUs in the United States and Canada completed the survey, for a NICU response rate of 58%. Of the respondents, 1,229 persons declared their profession: 100 physicians, 998 nurses, and 131 respiratory therapists. Principal Findings: Not only do high status individuals have higher psychological safety than low status individuals, but also cross-disciplinary teams differ in mean safety. Moreover, there is an interaction between status and team, which appears related to the degree of leadership inclusiveness in the team. Greater inclusiveness minimizes the effect of status on psychological safety, and vice versa. With more leadership inclusiveness comes greater psychological safety, which in turn predicts greater team engagement in QI work. Conclusions: Status does predict psychological safety, but it need not be deterministic. Aspects of team membership, especially the degree of leadership inclusiveness influence individuals' perceptions of safety as well. This is noteworthy because we can alter teams to create safety, a precursor of staff engagement in QI work. Implications for Policy, Delivery, or Practice: Our results provide reassurance that traditional status differences in health care may be overcome - to the benefit of quality improvement efforts. We identify a strategy for improving the climate within cross-disciplinary teams in health care (i.e. leadership inclusiveness), and provide an initial specification of the antecedents for QI work in health care. Our findings imply training leaders to be inclusive to foster psychological safety may be a critical antecedent of effective quality improvement. Primary Funding Source: Harvard Business School Division of Research ●Improving Quality of Care in Addiction Treatment Maria Orlando, Ph.D., Suzanne Wenzel, Ph.D., Pat Ebener, MA, James Dahl, Ph.D., Donna Farley, Ph.D., Wallace Mandell, Ph.D. Presented By: Maria Orlando, Ph.D., Behavioral Scientist, Behavioral and Social Sciences Group (BHS), The RAND Corporation, 1776 Main Street, Santa Monica, CA 90401; Tel: (310)393-0411, 6604; Fax: (310)260-8150; Email: orlando@rand.org Research Objective: Measurement of quality of care and targeted Quality Improvement interventions has received little attention in addiction treatment. The RAND Corporation and Phoenix House, the largest provider of therapeutic community treatment in the nation, have collaborated on developing methods to monitor and improve quality of care in therapeutic communities. This collaboration has produced the Dimensions of Change Instrument (DCI) to assess treatment process, and a study of how programs can use DCI information in a Continuous Quality Improvement (CQI) framework to enhance quality of care. This presentation reports principal findings on the measurement of treatment process, and how quantitative data can be used by program staff to plan targeted interventions. Study Design: The DCI was administered to an entry cohort of 993 clients (519 adults in 10 programs; 474 adolescents in 8 programs) who were assessed longitudinally throughout treatment. Retention and post-treatment outcomes data were also collected. Six treatment programs were subsequently selected for participation in a quality improvement demonstration. Staff of the 6 programs were presented with DCI scores and retention rates for clients in their programs compared with norms based on other programs’ data. Staff of 4 selected experimental programs received instruction in CQI and help in preparing and implementing quality improvement action plans based on the longitudinal client process data from their own program. Experimental programs are being monitored via regular telephone calls and site visits. Clients in all 6 programs are completing the process instrument pre and post demonstration period, allowing for observation of the impact of the CQI process. Population Studied: Adult and adolescent clients and programs represent diverse regions of the U.S. and of the population of clients in Phoenix House programs. Study participants are primarily members of ethnic minority groups (African American or Hispanic), and a majority has been referred to treatment from the criminal justice system. Principal Findings: Analyses to-date support the validity of the DCI (e.g., measurement of process is invariant across client subpopulations; the instrument is sensitive to change in residents during treatment; scores are associated with retention). Staff acceptance of the ongoing quality improvement activities demonstrates that DCI data can be understood by program staff with researcher-facilitated discussion groups, the data can be used by staff to develop targeted quality improvement action plans, and staff enthusiastically carry out these plans within their own programs. Site visits now underway are yielding detailed information on the implementation of quality improvement action plans within each of the programs. Conclusions: This study represents an innovative step forward in enhancing quality of care in addiction treatment. It furthermore is demonstrating that health services researchers and addiction treatment providers can collaborate to ensure that complex data can be applied to informed program-level decision-making to improve quality of care. Implications for Policy, Delivery, or Practice: The chief implication for addiction services delivery and practice is that researchers and providers can collaborate successfully to enhance understanding of quality of care and to improve quality of care. Our development of a new process instrument and its successful application to a quality improvement effort should prove valuable to the addiction services field generally. Primary Funding Source: NIDA ●Patient Safety Attitudes of Trainee Physicians: A cross sectional survey Gareth Parry, BSc, MSc, Ph.D., Lisa Horowitz, Ph.D., J Bryan Sexton, Ph.D., Donald Goldmann, M.D. Presented By: Gareth Parry, BSc, MSc, Ph.D., Reader in Health Services Research, Health Services Research, University of Sheffield, Regent Court, 30 Regent Street, Sheffield, S1 4DA; Tel: 44 114 2220798; Fax: 44 114 2724095; Email: g.parry@sheffield.ac.uk Research Objective: To measure the patient safety attitudes of trainee physicians at an academic pediatric center. Study Design: Cross sectional survey of Trainee physicians’ self-reported patient safety attitudes as measured using the 51 item Safety Attitudes Questionnaire (Inpatient Version) and a specific 23 item Trainee Survey. The Safety Attitudes Questionnaire can be summarised into six domains: perceptions of management, safety climate, teamwork climate, working conditions, stress recognition and job satisfaction. The Trainee Survey can be summarised into six domains: clinical supervision and support, communication with senior clinical staff, handoffs and multiple services, orientation of new personnel, role identification during codes and support following an adverse event. Population Studied: 295 trainee physicians based at an academic pediatric health center in January 2004. Principal Findings: From 295 trainee physicians surveyed, responses were obtained from 209 (71%); residents and fellows accounted for 29.2% and 70.8% of responses respectively. From the Safety Attitudes Questionnaire trainee physicians rated most positively job satisfaction (58.8% indicating a very positive response), safety climate (51.3%) and working conditions (51.5%). They rated less positively teamwork climate (only 16.1% indicating a very positive response), and stress recognition (14.1%). From the Trainee Survey trainee physicians rated most positively clinical supervision and support (47.7% indicating a very positive response) and communication with their immediate senior physician (28.9%). They rated less highly support following an adverse event (8.2% indicating a very positive response) handoffs and multiple services (7.7%) and role identification during codes (4.0%). Conclusions: Trainee physicians are relatively comfortable with their supervision and ability to care independently for patients but may be less confident about their communication and teamwork with other health care providers and thus their ability to care interdependently for patients. Implications for Policy, Delivery, or Practice: Traditionally resident and fellow programs aim to train physicians to become independent care providers. To improve patient safety, physicians may also need to work interdependently with fellow health care professionals within their own and other departments. Consideration could be given to enhancing training programs so that they produce physicians who can work not only independently but also interdependently. Senior role models will be important in convincing trainees that quality improvement and patient safety are as high a priority as learning traditional medical skills. Arguably, most junior and senior faculty in academic institutions do not have these skills themselves (i.e., they are not formally trained and evaluated on collaboration), and it is difficult to see how they will serve as role models and teachers without considerable education. Continuing education programs, possibly involving experts from outside the healthcare industry, are needed. Initiatives to improve patient safety attitudes of trainee physicians must be accompanied by real change in the systems in which they work, so that interdependent working is encouraged and rewarded. When trainee physicians make errors, hospitals must provide clear, blame-free reporting mechanisms and transparent systems to support the involved clinicians. Accordingly, clinicians involved in successful interdependent recovery from errors should be celebrated as local heroes. Primary Funding Source: CWF ●Linking Data to Decision-Making: The Use of Outcomes Management Systems in Child/Adolescent Mental Health Treatment Settings Vaishali Patel, MPH, Ph.D. Candidate, Anne W. Riley, Ph.D. Presented By: Vaishali Patel, MPH, Ph.D. Candidate, Ph.D. Candidate, Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, 6 Char Street, Edison, NJ 08820; Tel: (732)261-8323; Email: vapatel@jhsph.edu Research Objective: The IOM’s ‘Crossing the Quality Chasm’ report cites the need to harness information technology to inform care and measure performance as a means of driving quality improvement. Yet, little is known about the “human factors” involved in the implementation of Outcomes Management Systems (OMS) and how outcomes data can inform decision-making at various levels of the organization. We conducted a qualitative study to identify organizational conditions that support and hinder the use of outcomes data, and to describe the experiences of different types of staff using OMS in child and adolescent out-of-home care agencies in Maryland. Study Design: Using a multiple case study design we purposefully selected 2 Residential Treatment Centers (RTC) and 2 Treatment Foster Care (TFC) programs from 10 programs representing 4 different organizations that were using the same Internet-based OMS that collected the following data: demographic, family history, behavioral/social issues, psychiatric diagnoses, treatment history, services, and functioning rating scale. Within each case, clinicians, managers/clinical supervisors, quality improvement personnel and executive directors were interviewed using tailored semistructured field guides. Quality improvement and treatment team meetings were observed to describe decision-making processes. Grounded theory and case study methods guided the analysis of the transcribed interviews and field notes describing the meetings. Population Studied: Thus far, 23 interviews with staff members have been conducted across 3 programs (2 TFC and 1 RTC), representing 10 clinicians/case managers, 4 quality improvement managers, 7 Unit Directors/clinical supervisors, and 2 Executive Directors (one of whom is responsible for 2 of the 3 programs). Interviews are being conducted at a 4th program, an RTC, to confirm initial findings. Six meetings (both quality improvement and treatment team meetings) have been observed. Principal Findings: All agencies struggled with a key step in their efforts to use an OMS to inform decision-making: generating meaning from the data. Organizational efforts emphasized data collection and entry of outcomes data rather than analysis and interpretation of the data, resulting in a repository of data that was not used, and was considered meaningless by staff. Staff from all levels of the organization reported experiencing: difficulties interpreting the data and applying it to their work; frustration regarding the limited amount of feedback provided; and concerns regarding how the data might be misused. Issues related to poor implementation of OMS also impacted staff buy-in and subsequent use of outcomes data for decision-making. Other types of data were more valued and used to inform group decision-making processes. Conclusions: In order for outcomes data to be useful for decision-making, a needed initial step is to work with staff to assess how outcomes information can guide current planning and clinical decision-making. This critical aspect of data utilization is no less challenging than implementing the data entry and collection of outcomes data but involves different skills and resources. Implications for Policy, Delivery, or Practice: As states and accreditation organizations seek to mandate the use of OMS within child mental health treatment settings measures should be taken to ensure that adequate infrastructure and knowledge level are in place within organizations so that the data can be actively used for quality improvement. Primary Funding Source: NIMH ●Using Quality Improvement Techniques to Improve Nursing Engagement and Bedside Care Marjorie Pearson, Ph.D., MSHS, Lynn Soban, Ph.D., RN, Valda V. Upenieks, Ph.D., RN, Patricia H. Parkerton, Ph.D., MPH, Jack Needleman, Ph.D. Presented By: Marjorie Pearson, Ph.D., MSHS, Social Scientist, RAND Corporation, 1776 Main Street, Santa Monica, CA 90401; Tel: (310) 393-0411x7566; Fax: (310) 260-8155; Email: mpearson@rand.org Research Objective: To assess the change implementation activities of hospitals participating in the pilot phase of the Transforming Care at the Bedside (TCAB) initiative, a collaborative-based effort to improve the work environment on medical/surgical units and thereby improve the quality of patient care and nurse engagement. Study Design: An observational study of implementation data reported by the hospitals participating in a pilot quality improvement (QI) collaborative led by the Institute for Healthcare Improvement and funded by The Robert Wood Johnson Foundation. Linking patient and nurse outcome objectives, the TCAB QI framework encourages multiple change strategies to improve four major elements of medical/surgical operations: 1) workplace vitality, 2) reliability and safety, 3) patient-centeredness, and 4) value and efficiency. Participating organizations are encouraged to test new ideas as well as proven practices, and suggested strategies are characterized by the quality of the existing evidence base (as applied to medical/surgical settings). The UCLA/RAND TCAB evaluation team is utilizing this framework to code all change activities reported in the hospitals’ monthly collaborative reports and designing key study variables to measure implementation intensity and framework fidelity. These variables will be used to assess the feasibility of changes at the medical/surgical unit level, variation in such changes, and their relationship to both nursing and patient outcomes. Population Studied: Thirteen hospitals, geographically and organizationally diverse, selected as innovators and early adopters (based on criteria such as Magnet hospital status and Baldridge Awards) to participate in the 2-year pilot phase of the TCAB initiative. Principal Findings: Preliminary findings suggest that most organizations were able to initiate numerous changes early in the collaborative. They undertook a total of 152 different change activities during the first 3 months of the collaborative. Changes per individual hospital ranged from 0 to 26. The organizations emphasized two of the four TCAB framework elements in their initial QI efforts. 42% of the change activities they undertook in the first three months of the initiative were directed toward improving the patientcenteredness of care and 38% towards efficiency and value. Considerably less initial emphasis was placed on changes to increase reliability (12%) and workforce vitality (9%). The vast majority of changes tested were new ideas or strategies not yet fully tested in medical/surgical settings; 7% fit the TCAB framework designation of medical/surgical unit best practices. Conclusions: As participants in this ambitions initiative to change the work environment as well as the quality, safety, and patient-centeredness of care, TCAB pilot hospitals initially have emphasized two of four TCAB priorities (patient centeredness and increased value) and relied primarily on new ideas not yet fully tested. Subsequent research will assess their ability to expand into the areas of reliability and vitality, as well as the relationship of these activities to changes in medical/surgical care processes and patient and nurse outcomes. Implications for Policy, Delivery, or Practice: The TCAB pilot collaborative offers promising potential for contributing to the evidence base for medical/surgical management and increasing our understanding of whether and how QI techniques can be applied to changing the work environment as well as care processes. Primary Funding Source: RWJF ●Can Pharmacogenomics Improve Drug Safety? Kathryn Phillips, Ph.D., David Bates, M.D., Jennifer Haas, M.D., Brian Alldredge, PharmD, Amalia Issa, Ph.D., Haiden Huskamp, Ph.D. Presented By: Kathryn Phillips, Ph.D., Professor, Clin Pharm/Institute for Health Policy Studies, UCSF, 3333 California Box 0613, San Fransisco, CA 94143; Tel: (415)5028271; Fax: (415) 502-0792; Email: kathryn@itsa.ucsf.edu Research Objective: The recent, highly publicized safety issues with antidepressants and Cox-2 NSAIDS have prompted calls for greater use of genetic information to individualize drug therapies (“pharmacogenomics”). Pharmacogenomics has the potential to improve drug safety both by enabling the development of safer drugs and by tailoring prescribing for individuals so that they may receive therapies that optimize the likelihood of benefit while minimizing the risk of adverse effects. Yet, there have been many questions raised about if and when pharmacogenomicbased drugs and diagnostics will be of value and how the FDA can facilitate the use of pharmacogenomics to improve drug safety. We examine how pharmacogenomics can be used to improve drug safety and review current and proposed efforts to achieve this goal. Study Design: We conducted a review of academic, industry, and government literature (N=187); reviewed key FDA guidance documents; conducted interviews with key academic, industry, and government thought leaders through our role as advisors to the FDA (N=15); and reviewed the published literature Population Studied: NA Principal Findings: We analyzed the following issues: (1) What data are (a) available and (b) needed to evaluate the use of pharmacogenomics to improve drug safety? (2) What is the (a) current and (b) proposed role of the FDA in using pharmacogenomics to improve drug safety? We then specifically evaluate these issues as they relate to antidepressants and Cox-2 NSAIDS. We found an alarming lack of existing data about how to evaluate the use of pharmacogenomics to improve drug safety, including a lack of data on the incidence of adverse drug events (ADEs) and how these are associated with genetic variation. These limitations include the lack of a national, systematic, and comprehensive ADE database; the lack of relevant and comparable data in drug package inserts; and the lack of data on ADEs for individual drugs. More fundamentally, few data are available suggesting to what extent ADEs are due to genetic variation. Thus, simply obtaining more data on ADEs will not resolve this problem. The FDA is taking a proactive approach to exploring how to use pharmacogenomics, including the development of several key guidance documents that will be released in the near future. These documents include guidance on pharmacogenomic data submissions; guidance on the co-development of pharmacogenomic based drugs, biological products and diagnostic tests; and an initiative on addressing the drug “pipeline problem”. Conclusions: As illustrated by the recent Cox-2 cases, improving the safety of drugs is a vital priority. As illustrated by the recent drug safety cases, improving the safety of drugs is a vital priority. Pharmacogenomics has the potential to improve drug safety by providing more tailored regimens and thus improving health care outcomes and decreasing costs; however, many challenges remain to achieve these goals. Adequate data are not presently available regarding ADE rates, common pathophysiologies amongst adverse events, and genetic determinants of ADEs to permit an evaluation of the potential for pharmacogenomics to improve drug safety or impact healthcare costs. Implications for Policy, Delivery, or Practice: Regulatory policies implemented by the FDA may directly impact the availability and accessibility of data by creating incentives or disincentives for example to share data or to collect certain types of data. The current FDA initiatives are expected to increase the amount of PGx data that are used and to make such data more widely available; however, more remains to be done. The challenges in using pharmacogenomics to improve drug safety will increasingly require interdisciplinary efforts and require health services researchers to play a more visible role. Primary Funding Source: NCI prevalence of error at specific process-level steps for ED management of allergy data Study Design: A prospective observational study of parent/child dyads presenting to an urban tertiary care ED. Consecutive dyads were identified at triage; parents were subsequently approached to complete a structured gold standard interview (GSI) to judge the allergy history. Parents were asked "Has your child ever had a problem or reaction to a medication that he/she was given?" A yes or not sure answer was further explored to determine if an immune-based reaction was causative. Records inclusive of nursing forms and medication order sheets were abstracted. The main outcome was the rate of false positive (FP) and false negative (FN) data specific to medication allergies documented at nursing triage and on medication order forms. Population Studied: English and Spanish-speaking parentchild dyads presenting for ED care at an urban tertiary care hospital. Principal Findings: We completed 256 observations over 23 days in summer 2004; 211 of 256 (82.4%) parents completed the GSI. 33 of 59 medication problems/reactions were judged immune-based; 20/33 were Type I hypersensitivity reactions. At triage, twenty-seven of 33 instances of allergy were correctly documented (18.2% FN rate). 8 of 186 patients without allergies to medications were incorrectly noted to have an allergy (4.3% FP rate.) 111 subjects had at least one medication ordered. 3/111 patients with no true allergy had incorrect data recorded on an order form (2.7% FP rate) and 4/111 patients with a true allergy had no allergy recorded (3.6% FN rate). No medication contraindicated by allergy history was ordered. Parents reported that 23/201 (11%) physician interviewers did not review allergies. 43 patients received a medication prior to GSI; parents reported 6/43 (14%) treating nurses did not review the allergy history prior to giving medication. Conclusions: Current ED practice produces both false negative and false positive data for the topic of allergies to medications. These errors may perpetuate based on limited subsequent interviews with parents and incorrect actions during the process of ordering and delivering medications. Implications for Policy, Delivery, or Practice: A "front-end" health information technology solution that captures allergy data directly from parents could improve the valid capture and dissemination of critical data that influences safe prescribing. Primary Funding Source: AHRQ, Local hospital award ●Getting the Data Right: Evaluating Information Accuracy in Emergency Medicine Stephen Porter, M.D., MPH, MSc, Shannon Manzi, PharmD, Diana Volpe, RN, Anne Stack, M.D. ●A Comparison of Formal Consensus Development Methods Used for Developing Clinical Guidelines Rosalind Raine, BSc, MBBS, MSc Ph.D. FFPH, Andrew Hutchings, BSc, MSc, CPFA, Colin Sanderson, MA, MSc, PhD, Nick Black, M.D., FFPH, FRCS, FRSA, DRCOG, DCH Presented By: Stephen Porter, M.D., MPH, MSc, Assistant Professor, Division of Emergency Medicine, Children's Hospital Boston, 300 Longwood Avenue, Boston, MA 02115; Tel: (617)355-6624; Fax: (617)730-0335; Email: stephen.porter@childrens.harvard.edu Research Objective: Information management in the emergency department (ED) setting is prone to risk. Allergies to medication represent a critical element of a patient's medical history that must be accurately captured and communicated to ensure safety. Our objectives were: 1) to identify gaps in ED system performance for capture and integration of patients allergy histories, and, 2) to identify the Presented By: Rosalind Raine, BSc, MBBS, MSc Ph.D. FFPH, MRC Clinician Scientist/senior lecturer in health services research, Public Health and Policy, London School of Hygiene & Tropical Medicine, Keppel Street, London, WC1E 7HT; Tel: (0) 207 927 2038; Fax: (0) 207 580 8183; Email: rosalind.raine@lshtm.ac.uk Research Objective: To compare two consensus development methods commonly used for developing clinical guidelines (Delphi and nominal group technique) in terms of the judgments produced, closeness of consensus, concordance with research evidence and reliability. Study Design: Groups rated the appropriateness of four treatments (cognitive behavioural therapy, behavioural therapy, brief psychodynamic interpersonal therapy and antidepressants) for three conditions (irritable bowel syndrome, chronic fatigue syndrome and chronic back pain), in the context of various clinical and social cues. Main outcome measures were the mean differences in group ratings and extent of agreement about those ratings, odds ratios for concordance of group ratings with research evidence, and weighted kappa statistics for reliability of ratings between groups. A group's rating for a scenario was the median of individual ratings and extent of agreement was the mean absolute deviation from the median. Population Studied: 213 general practitioners and mental health professionals in four Delphi and four nominal groups. Participants were randomly sampled from UK national clinical databases. Principal Findings: The effect of consensus method on final ratings varied with therapeutic intervention. Within-group agreement (consensus) was closer in the nominal than in the Delphi groups in both rounds. Except for brief psychodynamic interpersonal therapy (p<0.001) there was no difference in the improvement in agreement. There was no difference between groups in their concordance with research evidence (odds ratio 1.13, 95% confidence interval 0.79-1.61). The Delphi method was more reliable (Kappas 0.88 and 0.89 compared with 0.41 and 0.65 for nominal groups), largely due to the difference in group size. Conclusions: The advantages of nominal groups (more consensus; greater understanding of reasons for disagreement) could be combined with the greater reliability of the Delphi approach by developing a hybrid method. Implications for Policy, Delivery, or Practice: Clinical practice guidelines are extensively used to improve the quality of health care. The methods used to develop guidelines vary widely but there is a shift towards basing them on a combination of the best available scientific evidence and consensus judgements obtained by formal, explicit methods. Little is known about the relative advantages of the Delphi method, which is conducted by mail, compared with the Nominal Group Technique which includes a facilitated meeting. Our study is the first to compare directly the reliability of Nominal and Delphi groups for developing clinical guidelines. Our results are being used to inform the future development of guidelines by the UK National Institute of Clinical Excellence, whose aim is to develop guidelines to reduce disparities by improving quality of care for all. Primary Funding Source: Medical Research Council UK ●Switching Health Plans and the Role of Perceived Quality of Care Julie Rainwater, Ph.D., Patrick S. Romano, M.D., MPH, Jorge S. Garcia, M.D., MPH Presented By: Julie Rainwater, Ph.D., Research Scientist, Center for Health Services Research in Primary Care, University of California Davis, 2103 Stockton Boulevard, Sacramento, CA 95817; Tel: (916)734-5265; Fax: (916)734-2732; Email: jarainwater@ucdavis.edu Research Objective: Pay-for-performance initiatives should foster value-based health plan choices as plans compete on both cost and quality of care. While several studies find that increases in costs drive consumers to switch plans, the role of other factors, including beliefs about quality of care are less clear. This study prospectively evaluates factors related to health plan switching. Study Design: An observational study linked to the 2002 Open Enrollment (OE) period of California’s largest purchasing organization was conducted. Members were surveyed before Open Enrollment. After OE, health plan selection data was supplied by the employer and respondents were resurveyed to find out why they switched. Population Studied: 2,500 members of the California Public Employees Retirement System (CalPERS). Pre-OE questionnaires were completed by 69% of the total sample; 81.3% of the pre-OE respondents completed the post-survey. Principal Findings: Overall, 6% of members surveyed voluntarily switched health plans during OE. The most frequently cited reasons for switching were to lower costs, copayments, or out-of-pocket expenses (40%), or wanting a better choice of doctors (34%). Switch rates varied from 11.1% among members who were facing monthly premium increases of $25-$49 to 2.8% among those facing smaller increases. Switch rates varied from 9.5% among persons aged 40 years or younger to 1.1% among those over 70, and from 8.7% among persons with annual household income between $50,000 and $75,000 to 1.2% among those whose household income was less than $30,000. Respondents with a history of switching plans (12.1%) or medical groups (9.5%) in the past three years, or with an obstetric-related hospitalization in the past year (16.7%), were especially likely to switch. Kaiser members were significantly less likely to switch plans than members of other HMOs (1.0% versus 11.1%). Controlling for expected premium increase, pre-OE plan satisfaction was significantly associated with switching. Switchers were more likely to be dissatisfied with their care or plan, to have had problems getting needed care, and to have had treatment delays while waiting for approvals from their current plan. Switchers were more likely to report feeling susceptible to poor care if they stayed with their current plan, indicating that, if they stayed with their current plan, they anticipated problems with having to find a new doctor or with obtaining referrals or tests. Respondents who reported hearing negative stories about their plan in the media over the past year were significantly more likely to switch plans. Conclusions: Costs are the main driver of health plan switching, but perceived susceptibility to continued poor care, negative media reports, and dissatisfaction with providers and plans also play a role. Implications for Policy, Delivery, or Practice: Mediareported negative information about health plan quality did stimulate increased switching, suggesting a role for public reporting in driving the health care system toward higher quality. However, respondents' own experiences were the dominant quality-related predictors of switching. Health plans and providers that operate in the pay-for-performance environment should continue to be responsive to consumer experiences and their beliefs about what constitutes quality of care. Primary Funding Source: AHRQ ●The Relationship of Hospital Organizational Culture to Patient Safety Indicators Peter Rivard, MHSA, Cindy L. Christiansen, Ph.D., Shibei Zhao, MPH, Susan A. Loveland, MAT, Amy K. Rosen, Ph.D. Presented By: Peter Rivard, MHSA, Research Associate, Center for Healthcare Quality, Outcomes, and Economics Research, Veterans Administration, 200 Springs Road (152), Bedford, MA 01730; Tel: (781)687-3573; Fax: (781)687-2227; Email: rivardp@bu.edu Research Objective: There are increasing efforts to understand, measure, and improve the organizational culture of patient safety. While numerous studies have examined the relationship between organizational culture and quality improvement, little attention has been focused on the relationship between culture and hospital patient safety. This paper examines the association between one aspect of hospitals’ organizational culture, as indicated by shared values, and rates of potential patient safety events. There are three competing hypotheses from organization theory: a culture where hierarchy and bureaucracy are important is (1) positively associated with safety because it supports the necessary integrating role of formal structures; (2) negatively associated with safety because it inhibits timely response to new threats to patient safety and other aspects of organizational learning for safety; or (3) not directly associated with safety. Study Design: Potential patient safety events were measured using the Agency for Healthcare Research and Quality (AHRQ) Patient Safety Indicators (PSIs). Bayesian models were used to estimate observed-to-expected ratios for the six most frequent indicators; ratios were then combined, using simulation methods and posterior densities, to create a single combined indicator for each hospital. We used the Zammuto & Krakower (1991) Competing Values instrument, which indicates shared values in terms of the relative dominance of each of four archetypes: hierarchical/bureaucratic, group, entrepreneurial, and rational cultures. Data were obtained from a previously administered employee survey. To examine the relationship between the shared-values measure and combined PSI rates, we ran bivariate correlations and linear regressions; control variables included hospital bed size and teaching status. Population Studied: We studied 115 Veterans Administration inpatient facilities. PSI rates were calculated using administrative data from 2001. Culture surveys were administered to hospital employees in 2000. Principal Findings: Results support hypothesis 3: we found no relationship between culture and patient safety measures. Bivariate correlation between the hierarchical/bureaucratic shared values measure and the combined patient safety indicators was r = -0.05 (p = 0.58); multivariate linear regression of the combined patient safety indicators on the hierarchical/bureaucratic shared values measure showed a small, non-significant relationship (Adjusted R-square = 0.016; p = 0.76). Conclusions: There appears to be no direct association between the importance of hierarchy and bureaucracy in a hospital’s culture and that hospital’s performance on patient safety indicators. Based on organization theory, we speculate that this relationship may be mediated by the extent to which a given hospital’s structures of hierarchy and bureaucracy can be characterized as enabling, as opposed to coercive. Implications for Policy, Delivery, or Practice: While a great deal has been done to apply knowledge about safety culture from other industries to healthcare settings, this study suggests that we have more to learn about the role of hierarchical and bureaucratic values in a hospital organizational culture of patient safety. It may be more important for patient safety leaders to focus on the particular functions of hierarchy and bureaucracy rather than on their extent or overall importance in the hospital setting. Primary Funding Source: VA ●Practice Pattern Analysis – Quality Improvement Though Intervention Deborah Rogers, BS, Soyal Momin, MS, MBA, Alan Denham, Raymond Phillippi, Ph.D., Judy Slagle, RN, MPA, Ken Patric, M.D. Presented By: Deborah Rogers, BS, Senior Research Analyst, Health Services Research, BlueCross BlueShield of Tennessee, 801 Pine Street, Chattanooga, TN 37402; Tel: (423)755-5889; Fax: (423)755-5100; Email: Deborah_Rogers@bcbst.com Research Objective: The primary objective of this study was to develop a Practice Pattern Analysis addressing the quality of care delivered to the Medicaid population. The secondary objective was to improve the practice patterns of the providers through intervention. Study Design: The Practice Pattern Analysis (PPA) was designed to address three issues: access to care, utilization, and quality of care. Access to care measures consisted of PCP Visits and Specialist Visits per member per year. Utilization measures included were Admissions/1,000, Emergency Room Visits/1,000, Outpatient Surgeries/1,000, Presciptions/1,000, Percentage of Generic Drugs Dispensed, and Out-Of-Network Referrals. Quality of care measures incorporated percentages of Mammography Screenings, Cholesterol Screenings, and Well Child Visits where applicable. Primary Care Practitioners (PCPs) were provided a comparison of their panel access, utilization, and quality of care to that of their peer group for the Medicaid population, with the peer group being the physician specialty. The PPA also provided the Regional Medical Directors and Regional Clinical Network Analysts with a snapshot of practitioner performance within his/her region. Data was examined on a quarterly basis with a three-month run out to complete the data. Outliers, (i.e. those with Out-OfNetwork Referrals, Specialist Visits, or Emergency Room Visits/1,000 above the 90th percentile,) were then provided with a detailed listing of members for that measure. The Regional Clinical Network Analyst or Regional Medical Director then contacted the practitioner to discuss the practice pattern. Additionally, the Practice Pattern Analyses were subdivided between Adult and Pediatric populations to provide an accurate median for each specialty. Specialty Medians for each measure, presented in graphical format, allowed the provider to quickly gage his/her status, as well as make a comparison to his/her peers within his/her specialty. A pre-post study design was utilized to evaluate the impact of the PPA. Population Studied: To be included in the study, a physician must have had a panel size of 50 or more members for the quarter. The population was approximately 850 practitioners overseeing the care of approximately 465,000 Medicaid members. Principal Findings: A comparison of second quarter 2003 and second quarter 2004 PPAs was conducted. Favorable declines were noted in the Outpatient Sugeries/1,000, Admissions/1,000 and Out-Of-Network Referrals. Advantageous increases were noted in the PCP Visits Per Member Per Year, as well as the percentage of Well Child Visits and Mammography Screenings. Adverse findings were noted as increases in the Emergency Room Visits/1,000 and the Specialist Visits Per Member Per Year, and the decline in Cholesterol Screenings. Conclusions: The PPA continues to be an effective tool for monitoring provider performance and improving quality of care for the Medicaid population. Negative behaviors and patterns can be positively identified and addressed by the appropriate staff in an ongoing improvement process Implications for Policy, Delivery, or Practice: 1. PPAs can be used to decrease the variability in provider behavior. 2. PPAs can be used to support the Network Development Process. 3. PPAs should be implemented using a multi-disciplinary approach. Primary Funding Source: BlueCross BlueShield of Tennessee ●Violation of Patients' Trust Expectations: What Determines Disrupted Trust in Physicians and Plans Marsha Rosenthal, MPA, Ph.D. Presented By: Marsha Rosenthal, MPA, Ph.D., Center for Gerontology and Health Care Research, Brown University, 2 Stimson, Providence, RI 02912; Tel: (401)863-3401; Email: marsha_rosenthal@brown.edu Research Objective: This study examines how older patients lose trust following a problem with health services. Although researchers have examined patients’ trust in doctors and health plans, they have not studied what causes patients to lose trust, nor what restores trust if it has been disrupted. The subject takes on added significance with managed care, as patients navigate a system that relies on their informed decisions about quality of care. This paper examines patients’ expectations of physicians and plans, and how a violation of those expectations is evaluated by patients to decide if a doctor or plan is trustworthy. The research further aims to identify how patients’ loss of trust is influenced by 1) enrollment in a managed care plan; 2) plan ownership (nonprofit vs. for-profit status); 3) problems with care, coverage or access to health services. Study Design: A national telephone survey of consumer trust in health care was conducted in 2002. Multivariate analyses (OLS, logistical regression, ordered logit) were used to analyze responses about patients’ expectations and experiences with physicians and plans. In 2003, semi-structured interviews were conducted with volunteers, expanding on the questions in the survey. These responses were examined for repeated categories and themes, and for insights into patients’ perspectives. Population Studied: Three hundred eighty-one Medicare beneficiaries were studied from a national survey of 5000 adults. In-depth interviews were conducted with 107 Medicare beneficiaries at six sites in New Mexico and New Jersey. Principal Findings: Survey and interview responses provided new insights into how patients make decisions about whether physicians or plans are trustworthy. Qualitative data revealed that patients have “hidden expectations” beyond those typically stated on surveys. If these are violated, there is loss of trust in doctors and plans. However, patients feel restored trust if doctors or plans communicate effectively and work to resolve the issue. Survey data showed that patients in managed Medicare and traditional Medicare lose trust for similar reasons. Patients who have problems with care do not lose trust in their doctors, but patients who have problems with coverage lose trust in their plans (p<.01). However, patients who believe they are in a nonprofit plan have greater expectations that they can trust their doctors (p<.05) and plans (p<.01); patients are more likely to maintain trust when they have an adequate choice of plans (p<.01). Conclusions: Managed care per se does not influence patients’ decisions about trusting doctors or plans. However, plans’ nonprofit status connotes more trustworthy behaviors to patients. Among such behaviors, communication matters most in maintaining patient trust, along with having a choice of plans. Implications for Policy, Delivery, or Practice: Patients’ loss of trust can affect their evaluation of quality of health services, plan selection, health care utilization and compliance. 1) As the Medicare Advantage program moves forward, federal policy should take account of the importance that adequate choice of plans plays maintaining patient trust. 2) Physicians and plans should take account of the value of communication and problem resolution in maintaining patient trust. Primary Funding Source: AHRQ, Aspen Institute ●The Severity of Medication Errors Identified by a Bar Code Point of Care Administration System Julie Sakowski, Ph.D., Jeff Newman, M.D., MPH, Tim Schiro, PharmD, Beverly Michaelsen, MBA, Patricia Zrelak, Ph.D., RN CNRN, Krystin Dozier, RN Presented By: Julie Sakowski, Ph.D., Senior Health Services Researcher, Institute for Research and Education, Sutter Health, 345 California Street Suite 2000, San Francisco, CA 94104; Tel: (415)296-1808; Fax: (415)296-1844; Email: sakowsj@sutterhealth.org Research Objective: Bar code point of care medication administration (BPOC) systems are effective at mitigating medication administration errors. Previously, we found that the BPOC implemented in one hospital system prevented errors in 1.1% of administrations, but an additional 2.8% of administrations deviated from the order written. However, the clinical impact these errors could have on patients is unclear, making it difficult to draw conclusions about the impact of BPOC systems on patient safety. The purpose of this study is to assess the potential clinical significance of the prevented and observed errors identified by a BPOC system. Study Design: This study utilized a previously validated technique to estimate the potential severity of medication errors without knowing patient outcomes. A panel of pharmacists, nurses, and physicians reviewed medication administration error scenarios and rated the potential severity of the errors between 0-10, where 0 signified no effects and 10 indicated probable death. A severity index was calculated for each scenario using the average rating from all reviewers. The medication error scenarios were compiled from BPOC generated error reports. Errors included early and wrong doses and administration of medications not ordered or discontinued (N=127,117, 134, 46). Both observed and prevented errors were included in the sample, and reviewers were blinded to prevent potential bias. Descriptive statistics including the number of error events receiving each rating score and the mean severity were calculated. Reduced form linear regressions were performed to estimate the associations among prevented and observed errors, type, and severity. Population Studied: Medication errors occurring in adult inpatients at 6 community hospitals affiliated with a large nonprofit health system in California. Principal Findings: Among our 424 error scenarios (N=304 observed, 120 prevented) we found that only 10% (95% CI = 7% - 12%) might potentially result in any adverse effects (N=32 observed, 10 prevented errors). The most severe errors involved attempted administration of the wrong drug (N=2) and the wrong dose (N=1). Regression analyses indicate that ‘dose early’ errors are less severe than no order in system, order discontinued, and wrong dose errors. We were not able to detect any difference in severity or type classification breakdown between prevented and observed errors. Conclusions: While 90% of the medication errors detected by our BPOC system were minor, the frequency of more serious errors was similar among those prevented by and observed despite BPOC. Applying our findings to the 3,000,000 medication administrations delivered via BPOC in our network suggests 33,000 errors were prevented by BPOC; 3,300 of which would have resulted in moderate to severe adverse events. Implications for Policy, Delivery, or Practice: Our findings support the potential benefits of BPOC. However, the number and severity of errors occurring despite its implementation suggests that there are opportunities to improve the use of these systems. We recommend users concentrate on developing processes to prevent occurrences of no order in system, wrong dose, or order discontinued errors, especially for the high-alert drugs most likely to cause harm including anticoagulants, insulin, opiates and narcotics, injectable potassium chloride, and sodium chloride (solutions greater than 0.9%) to maximize the impact on patient safety. Primary Funding Source: No Funding Source ●Nursing Factors and Patient Outcomes in VHA Anne Sales, MSN, Ph.D., Nancy Sharp, Ph.D., Gwendolyn Greiner, MSW, MPH, Yu-Fang Li, Ph.D., Elliott Lowy, Ph.D. Presented By: Anne Sales, MSN, Ph.D., Research Scientist, Health Services Research and Development, VA Puget Sound Health Care System, 1660 South Columbian Way, Seattle, WA 98108; Tel: (206)764-2068; Fax: (206)764-2935; Email: ann.sales@med.va.gov Research Objective: We report on the association between nursing factors, including staffing levels and skill mix, and patient outcomes, including overall inpatient mortality and conditional mortality, similar to failure to rescue for a wider group of patients among inpatients in Veterans Health Administration (VHA) hospitals. Conditional mortality measures mortality of patients who experience serious complications that may be sensitive to inpatient nursing care. Study Design: Retrospective observational study. We used VA cost accounting data for number of nursing personnel hours per month by type of nurse and skill mix, at the nursing unit level, in 125 VA medical centers. Data on nursing process are from a survey of all nursing personnel in these 125 VA medical centers. Data on patient characteristics, pre-existing conditions, complications, and outcomes of admission for all patients admitted to VHA inpatient care from 2/03 to 6/03 come from the VA inpatient Patient Treatment Files. All analyses were conducted at the patient level, including nursing unit and facility level variables. We stratified by ICU and nonICU admitting unit and estimate the model in two steps, first predicting patient probability of experiencing a major complication, then probability of death conditional on experiencing a complication, controlling for probability of complication. Results are preliminary with analysis continuing, including correction for the complex multi-level error pattern. Population Studied: We included 14,719 VHA patients first admitted to an ICU and 78,193 patients first admitted to an acute medical/surgical unit. Principal Findings: Overall mortality for ICU patients was 6% with conditional mortality 16%. Overall mortality for acute med/surg patients was 3% with conditional mortality 7%. In the multivariate regression predicting probability of death after experiencing a complication, several nursing process and facility variables were significant with predicted probability of complication increasing the risk of death very significantly in both ICU and acute patients. Among ICU patients, increased total nursing hours per patient day were related to increased risk of death (OR 1.31, p=0.03) while increased RN hours per patient day were related to decreased risk of death (OR 0.71, p=0.01). Among acute patients, the inverse was found: total nursing hours were related to decreased risk of death (OR 0.79, p=0.01) while RN hours were related to increased risk of death (OR 1.30, p=0.01). Skill mix was significantly associated with decreased risk for acute patients but not for ICU patients. Conclusions: Preliminary results appear to indicate different patterns of association of nurse staffing with patient outcomes for ICU compared to non-ICU acute patients. Prior large-scale studies have not distinguished between patients initially admitted to ICU vs. non-ICU care. These results may indicate different approaches to nurse staffing patterns in intensive vs. acute care. Implications for Policy, Delivery, or Practice: This is the first large-scale study of nursing factors and patient outcomes using unit level rather than facility level data and is the first opportunity to explore differences among these two settings. Nurse staffing continues to be a critical factor in providing optimal patient care. Understanding the effects of different settings is essential. Primary Funding Source: VA ●Coverage of Asthma Care in Employer-Sponsored Health Insurance Plans Marissa Scalia, MPH, Seymour Williams, M.D. Presented By: Marissa Scalia, MPH, Fellow, Air Pollution and Respiratory Health Branch, Centers for Disease Control and Prevention, CDC/NCEH/APRHB, 1600 Clifton Road MS E17, Atlanta, GA 30333; Tel: (404)498-1879; Fax: (404)498-1088; Email: mscalia@cdc.gov Research Objective: Providing quality asthma care may reduce excess medical expenditures and increase productivity for plan enrollees with asthma. Our study evaluated the coverage of quality asthma care in a convenience sample of employer-sponsored health insurance plans. Study Design: We identified the standards of asthma care as defined by the National Asthma Education and Prevention Program (NAEPP) and assessed the coverage of the associated activities found in health insurance benefits summaries received from three employers. Health benefits managers clarified information missing from or ambiguous in the summaries. We also compared the smoking cessation benefits to the benefit design recommendations made by the Centers for Disease Control and Prevention’s (CDC) Office on Smoking and Health. Population Studied: Summaries of 22 plans were provided by two Fortune 500 companies and one government agency. Our review identified eight preferred provider organizations (PPOs), seven fee-for-service with PPO option plans, six health maintenance organizations, and one consumer-driven plan. Principal Findings: All of the plans cover most of the recommended interventions. In two of the plans, neither spacers nor chambers are covered by either the pharmacy benefit or as medical supplies. Disease management (DM) programs for asthma are available in fourteen of the plans. Education as a stand-alone service, including asthma education, is excluded by fourteen of the plans. Fifteen of the 22 plans cover nicotine replacement therapy and Bupropion for smoking cessation as part of the pharmacy benefit or smoking cessation benefit. Each of six plans has a smoking cessation program. One of these six offers counseling alone; four offer counseling combined with nicotine replacement therapy and Bupropion; and one offers Bupropion and nicotine replacement but no counseling. Conclusions: All of the 22 plans reviewed cover most of the recommended equipment and activities of quality asthma care; however, some activities of care are inconsistently covered by these health insurance plans. Asthma education is not covered in more than half of the plans because education is excluded, although some education may be delivered while providing another service (e.g., during an appointment). Although over half of the plans include Bupropion and nicotine replacement therapy, less than a quarter of the plans offer a smoking cessation program that includes counseling, and fewer offer a multifaceted smoking cessation program. Implications for Policy, Delivery, or Practice: Our review of these plans reveals at least three areas where improvement in coverage would satisfy NAEPP recommendations. First, NAEPP recommends providing spacers and chambers, which if covered will improve the delivery of medications, especially for certain high-risk groups (i.e., children). Second, since the guidelines emphasize a need for self-management, covering asthma education as a stand-alone service should be considered. Third, smoking cessation should be the foremost priority for coverage in these health insurance plans since the health improvements and savings for both enrollees and employers would be substantial. Although not addressed in the NAEPP guidelines, improved asthma care might also be achieved by increasing the number of plans with asthma DM programs if these programs meet all the NAEPP guidelines. Primary Funding Source: CDC ●Regulation and Quality Improvement in Invasive Cardiology Facilities Charles Schade, M.D., MPH, John G. Brehm, M.D., FACP, David Lomely, BS Presented By: Charles Schade, M.D., MPH, Medical Epidemiologist, Scientific Support, West Virginia Medical Institute, 3001 Chesterfield Place, Charleston, WV 25304; Tel: (304) 346-9864; Fax: (304) 346-9863; Email: cschade@wvmi.org Research Objective: (1) To determine whether invasive cardiology procedures can be safely performed in facilities without on-site cardiac surgery serving patients of substantial cardiac risk in a rural state; and (2) to validate the reliability of routine clinical data reporting in support of safety monitoring. Study Design: Time series Population Studied: Three newly-licensed invasive cardiology acilities in West Virginia performing angioplasty without onsite cardiac surgery, each performing procedures (85% angiography, 15% percutaneous transluminal angioplasty) on about 1,000 patients per annum. Under contract with the West Virginia Health Care Authority, we monitored data these facilities submitted to the American College of Cardiology National Cardiovascular Data Registry (NCDR) in 2003 and 2004. We audited 30 records per facility submitted following the third quarter of 2003, using a trained nurse abstractor to collect NCDR data elements from the original patient charts. Each facility received a detailed report of the results of its first year of operation and audit findings, with recommendations for improvement. Principal Findings: Results are now available for 6 consecutive quarters (all prior to facility feedback), and results will be available for the first post-feedback quarter before the meeting. Patients seen in the facilities were generally of high cardiac risk (21.5% obese [BMI>35]; 28.9% with history of diabetes; 63.1%, hypertension; 64.2%, hypercholesterolemia; 22.9%, current smoking). Complications occurred sporadically, with total significant adverse events remaining less than the relevant ACC benchmark (4.5%) statewide in every quarter, and in each facility for all quarters but one. Complication rates have not varied significantly among facilities, though there are insufficient data to assess long term trends. Audits confirmed unexplained variations in certain demographic items (e.g., smoking prevalence) among hospitals, and demonstrated generally reliable submission of complication data (most agreed 100% of the time). Conclusions: ?Experience to date with invasive cardiology procedures in facilities without onsite cardiac surgery in a rural state with a high prevalence of cardiac risk factors has demonstrated complication rates not higher than expected based on comparable NCDR data. Facility reports reliably reflect events actually documented in medical records. Implications for Policy, Delivery, or Practice: Feedback to facilities following audits may present an opportunity to improve documentation and care quality. Primary Funding Source: West Virginia Health Care Authority ●Incorporating Health Care Quality within Health Antitrust Law Helen Schneider, Ph.D. Presented By: Helen Schneider, Ph.D., Economist, Petris Center, UC-Berkeley and Los Alamos National Lab, D-3, Mail Stop F607, Los Alamos, NM 87545; Email: hc72@cornell.edu Research Objective: To examine the impact of hospital mergers and competition on the post-merger quality of care. Study Design: Regression analysis is used to examine independent associations between hospital mergers and health care outcomes. We use Coronary Artery Bypass Graft (CABG) mortality to measure health care quality. In our further analysis, broader quality indicators of patient experience and of hospital quality rankings will be considered. Population Studied: California general acute care hospitals that report to the Office of Statewide Health Planning and Development (OSHPD). The study separately examines the effect of mergers on nonprofit hospitals. Principal Findings: Preliminary findings show no post-merger quality improvements. This result does not vary by hospital ownership. Conclusions: There is no significant association between hospital mergers and competition and health care quality as measured by the CABG mortality rates. Implications for Policy, Delivery, or Practice: Antitrust policy usually treats price as proxy for consumer welfare although this claim is not supported by empirical evidence. Therefore, quality should be incorporated into the antitrust analysis. If mergers lead to higher prices and lower quality, thus lower social welfare, the case for the antitrust challenge of hospital mergers is strengthened. Primary Funding Source: No Funding Source ●Patient and Caregiver Satisfaction with Care at the End of Life: A Review of the Evidence Lisa R. Shugarman, Ph.D., Karl Lorenz, M.D., Joanne Lynn, M.D., M.S., Richard A. Mularski, M.D. Presented By: Lisa R. Shugarman, Ph.D., Associate Health Policy Researcher, RAND Corporation, 1776 Main Street, PO Box 2138, Santa Monica, CA 90407-2138; Tel: (310) 393-0411 x.7701; Fax: (310) 393-4818; Email: Lisa_Shugarman@rand.org Research Objective: Both patient and caregiver satisfaction are key metrics of healthcare performance at the end of life. In order to evaluate progress in the field of end-of-life care and clarify research priorities we examined the research literature on end-of-life care to determine the association of other processes and outcomes of care with patient and caregiver satisfaction. Study Design: As part of a systematic review of end-of-life care, we reviewed articles that evaluated patient and caregiver satisfaction with care provided at the end of life. Population Studied: Sources for our review included Medline, the Cochrane Database of Reviews of Abstracts of Effects (DARE), the National Consensus Project for palliative care, and several recent systematic reviews from both Health Canada and National Institute for Clinical Excellence (NICE), United Kingdom. The searches were limited to published articles in the English language (1990 – 2004), involving adult human subjects. Principal Findings: From 24,423 total citations, 1,274 articles were reviewed, including 95 systematic reviews, 134 intervention, and 682 observational studies. We identified 10 systematic reviews, 12 intervention studies, and 17 observational studies on patient or caregiver satisfaction. The preponderance of the literature supports the effectiveness of palliative care for improving patient and caregiver satisfaction. Subjective measures of the end-of-life care experience include satisfaction and quality-of-care measures that overlap significantly. Compared to global or nonspecific measures, satisfaction instruments that assess focused aspects of care or that were specifically developed for palliative care settings have been most useful in demonstrating intervention effects. Possibly for that reason, effects of interventions on satisfaction have been somewhat inconsistent. Measures of satisfaction that are more specific and strongly related to explicit intervention aims or processes (e.g., communication, pain control, practical support and enhanced caregiving) have demonstrated greater sensitivity to change and support a process-outcome relationship among these variables. The relationship of other processes or attributes of care (e.g., treatment of symptoms other than pain, spiritual support, continuity and coordination of care) to satisfaction is less evident in the literature, although such relationships are supported qualitatively. The ability to demonstrate relationships between these aspects of care and satisfaction may be partially related to challenges in defining spiritual support as an intervention and measuring spiritual support and continuity of care. Conclusions: Our systematic review identified a very large, diverse literature reflecting the tremendous growth and importance of the field of end of life care over the last decade. This review illuminates strengths of the field as well as opportunities for research. We identified evidence supporting the association of satisfaction with pain management, communication, practical support, and enhanced caregiving; however, little is known about the association of satisfaction with non-pain symptoms, spiritual support and continuity. Implications for Policy, Delivery, or Practice: Studies evaluating satisfaction should use specific measures that reflect processes of care, and studies are needed to examine the relationship of satisfaction to less studied processes such as non-pain symptoms, spiritual support, and continuity. Primary Funding Source: AHRQ, National Institute of Nursing Research ●Creating a Culture of Safety in Hospitals Sara Singer, MBA, Anita Tucker, DBA Presented By: Sara Singer, MBA, Doctoral Student, Health Policy, Harvard University, 303-1 Sherman Hall, Boston, MA 02163; Tel: (617)495-5047; Fax: (617)496-4397; Email: ssinger@hbs.edu Research Objective: We aimed to provide a framework of actions through which healthcare leaders can instill strong safety cultures in hospitals. We also offer case studies of two hospitals to illustrate the framework and to enable leaders to evaluate their own efforts to create a culture of safety. Study Design: Drawing on relevant findings from organizational, leadership, and high reliability organization literature, we developed a managerial framework for creating a culture of safety. To illustrate our framework, we present qualitative and quantitative data from two hospitals. We gathered qualitative data through semi-structured interviews with senior leadership and front-line staff as well as on-site observation of safety-related meetings. In addition, to provide a quantitative measure of our framework, we used survey data from 6538 respondents from 44 hospitals participating in a nationwide study of patient safety climate. Exploratory factor analysis of survey questions related to senior leaders’ role in patient safety resulted in a 7-item measure of patient safety leadership (Cronbach’s alpha = 0.89). We analyzed the average levels of problematic responses, i.e. respondents who reported strongly disagreeing or disagreeing with items that asked about positive safety practices, to compare the two hospitals’ patient safety leadership. Population Studied: We studied senior leaders and other personnel from two hospitals that differed in size, location, organization, and safety leadership. Survey participants at each hospital included 100% of senior managers and physicians and a 10% random sample of all other personnel. Principal Findings: Strong safety culture leadership required five actions: 1. setting and communicating a clear, compelling safety vision, 2. focusing on system issues, 3. leading by example, 4. valuing and empowering personnel, and 5. being dissatisfied with the status quo. In addition, leaders needed to apply these rules clearly and consistently. The case studies demonstrated that safety vision, a systems approach, and a willingness to act in support of safety were necessary but insufficient for achieving a strong safety culture. In addition to those three actions, strong safety leadership required consistently empowering others to act on behalf of safety and a constant quest for process improvement. The cases highlighted the benefits of storytelling as a means of engaging personnel and relationship training for encouraging individuals to speak up. Survey results supported qualitative findings. Conclusions: Creating a compelling vision of safety and being charismatic are important characteristics of safety leadership. However, creating a strong safety culture also requires leaders to empower others to act on behalf of safety and to overtly value front line employees’ expertise and contributions in this area. In addition, safety leadership demands a constant quest for process improvement, even when admitting the need for change poses a legal risk. Implications for Policy, Delivery, or Practice: Developing leadership skills to achieve strong safety culture is particularly challenging in an industry that enforces hierarchy, individual skill, and individual accountability. As financial pressures on hospitals create starker tradeoffs between safety and productivity, the need for the leadership skills explored in this paper may become even more acute. Primary Funding Source: AHRQ, Fishman Davidson Center ●Emergency Room – Next Frontier for Performance Measures Kurt Stevenson, M.D., MPH, Karen Benson-Huck, Ph.D., ARNP Presented By: Kurt Stevenson, M.D., MPH, Lead Medical Director, Healthcare Improvement and Research, Qualis Health, 720 Park Boulevard, Suite 120, Boise, ID 83712; Tel: (208) 389-5021; Fax: (208) 343-4705; Email: kurts@qualishealth.org Research Objective: To assess the feasibility of measuring and reporting emergency room (ER) care for selected clinical conditions by utilizing retrospective medical record abstraction. Study Design: Data were abstracted from ER medical records for patients seen for 1) Acute myocardial infarction (AMI) 2) Asthma 3) Closed head injury with subdural hematoma 4) Traumatic pain 5) Visceral pain 6) Penetrating extremity trauma and 7) Pulmonary embolus (P.E.). The conditions were prioritized for study by experts using a modified Delphi consensus process from a list of 56 potential areas for quality improvement in ER care. Demographic data/information about general care (such as documenting allergies) as well as data for elements of care for a specific clinical condition were collected and analyzed, with reports provided to the Centers for Medicare & Medicaid Services (CMS) and to participating ER’s for their use in quality improvement. Population Studied: 978 patients age 18 and over from 15 states (Arizona, California, Florida, Georgia, Illinois, Indiana, Iowa, Maine, Michigan, North Carolina, New Hampshire, Pennsylvania, Texas, Washington, and Wisconsin) with ER visits in critical access, rural, and urban hospitals from January 1, 2001 through July 30, 2004. Principal Findings: Opportunities for improving care were identified for each area: *AMI - Average time from ER arrival to an EKG was 84 minutes (median time 19.5 minutes), while the recommendation is 10 minutes. Only 38% of AMI patients diagnosed in the ER eligible for Beta blockers received the drug in the ER. *Asthma - Beta2 agonists were used only 76% of the time; yet this is a drug that can best provide symptomatic relief for most patients. *Closed Head Injury/Anticoagulant status - Elderly patients on anticoagulant drugs because of conditions such as atrial fibrillation can have unrecognized intracranial bleeding from closed head injuries. Yet, an assessment of anticoagulant status was documented in only 75% of patients who were diagnosed with a subdural hematoma. *P.E. – This condition may be missed in the ER. Even so, tools that assess for P.E. risk factors were used in only 7 individuals out of 119 P.E. patients. *Pain - For patients with hip fractures, the mean time from ER arrival to pain medication was 80 minutes; the time to medication for urinary tract calculi patients was 57 minutes. Conclusions: It is possible to assess the quality of care in the ER using medical records data to identify gaps in care, opportunities for improving care, and to make changes in ER processes and treatment that can improve patient outcomes. Additional efforts to standardize ER terminology, documentation, and reporting are needed in order to be able to compare care across settings. Implications for Policy, Delivery, or Practice: CMS is implementing performance measures for ERs in critical access hospitals starting in August, 2005. It is expected that performance measures will eventually include all ERs in the U.S. that treat Medicare patients, which will impact care delivered in the ER and reimbursement policy across the country. Primary Funding Source: CMS ●Quality Improvement in Rural Hospitals: Results from the Idaho Critical Access Hospital Collaborative Kurt Stevenson, M.D., MPH, Sharon Wilson, RN, BS, Jane Burgman, RN, James W. Moore, MS, Janel Galbraith, RN, Mary Sheridan, RN Presented By: Kurt Stevenson, M.D., MPH, Lead Medical Director, Healthcare Improvement and Research, Qualis Health, 720 Park Boulevard, Suite 120, Boise, ID 83712; Tel: (208)389-5021; Email: kurts@qualishealth.org Research Objective: Determine the impact of group learning, as outlined by the Institute for Healthcare Improvement (IHI) Breakthrough Series Collaborative model, on healthcare performance improvement in a cohort of rural critical access hospitals (CAH) in Idaho. Study Design: Following the IHI model, CAH in Idaho were recruited to participate in three learning sessions followed by action periods and a final outcomes conference. Hospitals were trained on evidence-based measures for acute myocardial infarction (AMI) and community-acquired pneumonia (CAP) and were provided tools for measurement in the first learning session. They developed plans for testing specific interventions and shared successes and failures at each subsequent learning session. All data were self-reported and assessed in a pre-post intervention study design. Population Studied: Fourteen CAHs participated in this demonstration project that was conducted from January 2004 to October 2004. All patients with either AMI or CAP were included. Additionally, all eligible patients >65 years were assessed for administration of pneumococcal vaccination. Principal Findings: There was 100% participation by CAHs in all aspects of the project. All teams shared their work, ideas, intervention tools, lessons learned, and barriers to improvement. Among required measures, the proportion of AMI patients who received timely administration of aspirin increased from 78.1% to 92.4%; the proportion of CAP patients who received antibiotics within 4 hours improved from 77.9% to 90.7%; and the proportion of CAP patients with blood cultures within 24 hours improved from 15.1% to 40.0%. Among optional measures, the proportion of CAP patients who received initial antibiotic therapy consistent with current guidelines improved from 89.7% to 100%; the proportion of CAP patients with arterial oxygenation assessment during the first 24 hours improved from 98% to 100%; and. the proportion of smoking patients who were counseled about smoking cessation improved from 0% to 53.7% for AMI patients and from 8.9% to 100% for CAP patients. The total proportion of eligible patients >65 years who received the pneumococcal immunization improved from 29.1% to 53.8%. Conclusions: There were significant quality gaps in the baseline performance for many of the quality indicators examined in this cohort of rural hospitals. The hospitals, however, were motivated to participate in a shared learning collaborative model and demonstrated significant improvements in care for almost indicators tested. Implications for Policy, Delivery, or Practice: Rural hospitals, although representing a significant portion of healthcare delivery in the US, are often excluded from research studies due to low patient census and small numbers of practitioners. Based on the positive outcomes in this study, rural hospitals should be considered in future research studies and performance improvement projects. Assessments of ideal performance improvement methodology may be needed to address the unique characteristics of these hospitals. Primary Funding Source: CMS ●Improving Patient Safety and Quality of Care Through Bar Coding Solutions Patsy Sublett, MSN, RN,-BC Presented By: Patsy Sublett, MSN, RN-BC, Clinical Systems Manager, Danville Regional Health System, 142 South Main Street, Danville, VA 24541; Tel: (434)799-4517; Email: sublettp@drhsi.org Research Objective: Alarmed by the reports regarding the increased number of patients that die every year from medical errors, the staff at Danville Regional Health System (DRHS) organized a multidisciplinary team to address the specific issue of medication errors and analyze the organization’s medication administration process enterprise-wide. The team identified three objectives: 1) Streamline workflow; 2.) Reduce medication errors; and 3.) Improve patient safety. Study Design: Using Siemens Medical Solutions medication management solution (Med Administration Check (MAC)), DRHS set out to achieve its goals. MAC is an application that helps nurses verify and document medication administration at the point of care. Using wireless computers containing electronic medical records and wireless bar code scanners, nurses scan the bar code on a patient’s ID badge, which displays an online medication administration work list on the computer. The bar code on the medication is scanned, and the system automatically checks the scanned medication vs. the medication ordered. The computer then verifies that the five “rights” of administration are accurate – right medication, right dose, right time, right route, and right patient. If there is a discrepancy—for example, the medication that was scanned is the incorrect medication for the patient about to receive it— an alert appears on the computer screen, warning the nurse that one of the five rights is being breached, preventing a potentially fatal error. Implementation began in October 2002. Population Studied: DRHS implemented the MAC solutions from Siemens throughout its 350-bed facility. Principal Findings: Results were immediate. DRHS documented an average of 116 potential medication errors avoided each month, and an average annual savings of $800,000 due to increased efficiency. In addition the facility has eliminated all hard copy paper charting because all medication distribution is charted online. Conclusions: DRHS achieved the goals outlined in its original program. Implementing healthcare information technology (IT) solutions within the hospital arena to help manage medication distribution can be successful. There has also been improvement in workflow and the working relationships between nurses and pharmacy staff has improved. Overall, DRHS demonstrated that patient controversy (death, injury) caused by medication errors can be controlled and potentially eliminated. Implications for Policy, Delivery, or Practice: The future of the healthcare industry is indeed based on the ability to implement healthcare IT solutions such as bar coding for more efficiency, better patient care and cost effectiveness. Bar coding and medication management are key for reaching these goals. The implementation and success of the project at DRHS can be used as an example for the industry. Primary Funding Source: No Funding Source ●Toward a Clinically Plausible and Cost-Effective Risk Adjustment System for Pay-for-Performance: Development and Validation of Mortality Predictive Models Using Automated Clinical Data Ying Tabak, Ph.D., Karen Derby, BA, Yadong Yang, DS Presented By: Ying Tabak, Ph.D., Senior Biostatistician, Research, CardinalHealth (MediQual), 500 Nickerson Road, Marlborough, MA 01752; Email: ying.tabak@cardinal.com Research Objective: Pay-for-performance has been promoted by many stakeholders. One obstacle to adopt this approach is the lack of a clinically plausible risk adjustment system that can be implemented on a large scale at low cost. Methods using billing data lack precision and are poorly accepted by physicians. Methods using clinical data from chart abstraction are too costly to be practical. The objective of this study was to develop and validate admission-based severity adjustment models for heart failure, pneumonia, ischemic stroke, acute myocardial infarction, and septicemia mainly using automated laboratory data (LAB) for clinical severity measures supplemented with electronically collected UB92 data for demographics, discharge status, and comorbidities. The values of vital signs and altered mental status, that were not currently automated, were also analyzed. Study Design: In-patient mortality, a widely used indicator of quality, was used as the outcome measure. Demographics, LAB, vital signs, and comorbidities (identified by 6th digit ICD9 coding) were covariates. Multiple levels of LAB and vital signs were crafted per literature and empirical examinations. Five logistic regressions models, one for each disease group, were derived from 170,825 (11,129 deaths) admissions in 2000-2003 across 39 teaching and 39 non-teaching hospitals that electronically imported LAB data to AtlasTM (MediQual). C-statistic assessed model fit. Bootstrapping validated each model internally. Manually abstracted data (n=569,708; 37,573 deaths) from 80 teaching and 124 non-teaching hospitals validated the models externally. Population Studied: A total of 740,533 adult admissions in 2000-2003 for heart failure (n=261,349; 10,088 deaths), pneumonia (n=208,237; 9,354 deaths), ischemic stroke (n=98,361; 6,221 deaths), acute myocardial infarction (n=107,280; 11,334 deaths), and septicemia (n=65,306; 11,705 deaths) were analyzed. Principal Findings: The median age ranged from 75 to 77. The crude mortality was 3.9% (HF), 4.5% (pneumonia), 6.3% (stroke), 10.6% (AMI) and 17.9% (septicemia). The most significant predictors (p<.0001) entered into all five disease groups included: age, albumin < 2.5 g/dl, (odds ratio range: 1.8-2.1), BUN > 40 mg/dl or creatinine > 3.0 mg/dl (OR range: 1.6-3.6), pH arterial < 7.3 or > 7.48 (OR range: 1.7-3.0), systolic BP < 90 mm Hg (OR range 1.6-2.9), moderate/severe altered mental status (OR range:2.3-11.3), and metastatic cancer (OR range 2.0-3.4). The 95% confidence intervals for the predictors were narrow. Other significant predictors included hypo/hyperglycemia, hypo/hyperthermia, and other abnormal hematologic parameters and electrolytes. The c-statistics for all the derivative and validation models ranged from .82 to .85 with average of .835. Conclusions: Laboratory findings provide objective and precise measures of acute pathophysiologic conditions. Admission LABs indicating acidosis/alkalosis, malnutrition, hepatic or renal dysfunctions predict mortality for all five clinical conditions. Vital signs and altered mental status are highly significant and should be automated. Pathophysiologic variables commonly measured on admission can generate parsimonious and clinically plausible models with excellent accuracy. Implications for Policy, Delivery, or Practice: This study evaluated five major diseases across a large number of patients and hospitals. Based mainly on automated data, these models are cost effective to implement as a risk adjustment method for pay-for-performance. With increasing automation of clinical data, a valid and cost-effective risk adjustment system is feasible. Primary Funding Source: No Funding Source ●Medicare Physician Group Practice Demonstration: Design of Quality Incentives Michael Trisolini, Ph.D., MBA, John Pilotte, MHS, Gregory C. Pope, MS, John Kautter, Ph.D., Bela Bapat, MS, Heather Grimsley, MBA Presented By: Michael Trisolini, Ph.D., MBA, Senior Health Services Researcher, Program on Health Care Quality and Outcomes, RTI International, 411 Waverley Oaks Road, Suite 330, Waltham, MA 02452; Tel: (781) 788-8100; Fax: (781) 7888101; Email: mtrisolini@rti.org Research Objective: To design and implement financial incentives for quality improvement for physician group practices (PGPs) serving Medicare fee-for-service (FFS) beneficiaries. Study Design: The Medicare PGP Demonstration Project provides financial incentives for quality improvement if a PGP can also demonstrate cost savings. Cost savings are generated if a participating PGP holds its rate of growth in average Medicare cost per beneficiary below the rate in its surrounding service area. Those cost savings are then allocated to two bonus pools, one for quality improvement and one for cost containment. A goal of the demonstration is to promote broader use of disease management and coordination of care, as these efforts may provide for both quality improvement and cost containment by reducing complications and progression of chronic diseases that are suffered by many Medicare beneficiaries. The quality portion of the bonus pool is earned by meeting targets for quality indicators. Quality indicators include 32 measures in four modules: 1) diabetes measures; 2) congestive heart failure measures; 3) coronary artery disease measures; and 4) hypertension and cancer screening measures. Seven of them are measured using Medicare claims data; 25 are measured using medical record abstracts. For example, one claimsbased quality measure is the rate of HbA1c testing done annually for diabetics. A medical-records based measure is the level of HbA1c control achieved for diabetics. Two types of targets can be met to earn the quality bonuses for each measure. Threshold targets can be met if the PGP achieves a higher rate of performance than a pre-determined threshold. An improvement target is satisfied by demonstrating a 10% improvement. Meeting either the improvement or the threshold target means the PGP will earn the bonus payment for that quality measure. Total quality bonus payments earned each year will depend on the number of quality measures for which either improvement or threshold targets were reached. Population Studied: Medicare beneficiaries Principal Findings: This presentation reports on the design of quality incentives for this Medicare pay for performance demonstration project. The demonstration is slated to begin on April 1, 2005 and run through March 30, 2008. This presentation also includes comparisons with the design of other quality improvement incentive programs for physicians. Conclusions: A quality incentive program that has the potential to improve the quality of care for Medicare beneficiaries can be designed for physician group practices treating fee-for-service Medicare beneficiaries. Implications for Policy, Delivery, or Practice: The PGP Demonstration is a new reimbursement mechanism for Medicare, and may be expanded beyond the PGPs participating in this three-year demonstration if it proves successful. Primary Funding Source: CMS ●Overcoming Bias in Estimating the Volume-Outcome Relationship Alexander Tsai, Mark Votruba, Ph.D., John F.P. Bridges, Ph.D., Randall D. Cebul, M.D. Presented By: Alexander Tsai, Department of Epidemiology and Biostatistics, Case Western Reserve University School of Medicine, 2511 Overlook Road, Suite 6, Cleveland Heights, OH 44106-2459; Tel: (216) 321-5496; Email: act2@case.edu Research Objective: To examine the effect of hospital volume on 30-day mortality for patients with congestive heart failure (CHF) using administrative and clinical data in conventional regression and instrumental variables (IV) estimation models. Study Design: The patient was the primary unit of analysis. We fit a linear probability model to the data to assess the effects of hospital volume on patient mortality within 30 days of admission. Both administrative and clinical data elements were included as risk adjusters to illustrate the importance of appropriate risk adjustment. Linear distances between patients and hospitals were used to construct the instrument, which was then used to assess the endogeneity of hospital volume. Population Studied: The primary data consisted of longitudinal information on comorbid conditions, vital signs, clinical status, and laboratory test results for 21,439 Medicareinsured patients aged 65 years and older hospitalized for CHF in northeast Ohio in 1991-1997. Principal Findings: When only administrative data elements were included in the risk adjustment model, a statistically significant inverse volume-outcome relationship was identified (P=0.045). The estimate was markedly attenuated when clinical data were added to the model as risk adjusters (P=0.28). IV estimation shifted the estimate in a direction consistent with selective referral, but we were unable to reject the consistency of the linear probability estimates. Conclusions: Use of only administrative data for volumeoutcomes research may generate spurious findings. The IV analysis further suggests that conventional estimates of the volume-outcome relationship may be contaminated by selective referral effects. Implications for Policy, Delivery, or Practice: Taken together, our results suggest that efforts to concentrate hospital-based CHF care in high-volume hospitals may not reduce mortality among elderly patients. Primary Funding Source: AHRQ ●Adverse Events in Acute Care: Pressure Ulcer Risk & Occurrence Over 10 Years Elizabeth VanDenKerkhof, RN, DrPH, Margaret B. Harrison, RN, Ph.D., Elaine Friedberg, RN, MHA Presented By: Elizabeth VanDenKerkhof, RN, DrPH, Assistant Professor, Faculty of Health Science, School of Nursing & Community Health, Queen's University, 92 Barrie Street, Kingston, Ontario, K7L 3N6; Email: vandenke@kgh.kari.net Research Objective: To analyze the risk and occurence of pressure ulcers in a large acute care facility over 10 years. Study Design: Yearly risk and point prevalence surveying and population profiling was conducted between 1994 and 2003. The data was collected prospectively with a dedicated RN team in one large teaching hospital. Two phases of analysis were undertaken: Descriptive, observational Population Risk and prevalence Trends over 10 years Multivariable modelling Identification of factors associated with pressure ulcers (stage II or greater) Population Studied: From 0600-1800 all patients on service or admitted during the day to medical, surgical or critical care units were included in the yearly studies. In total 95% of eligible patients (8143/8562) particapted in the surveys. Principal Findings: Of the 8,143 patients surveyed over the 10-year period, 885 were found to have stage II or greater ulcers. Severity of pressure ulcers was variable over the period in terms of staging and proportions with multiple ulcers. In multivariable analysis increasing age and being male sex, length of time in hospital, service, and Braden activity and friction/shear deficits were associated with the presence of pressure ulcer of stage II or greater. After controlling for age, sex, time in hospital, service, and year, the odds of pressure ulcer in patients with a friction/shear problem dropped from 14 (95% CI 11, 18) to 4.6 (95% CI 3.1, 6.9), and in bedfast patients from 12 (95% CI 8.6, 17) to 3.5 (95% CI 2.2, 5.6). Conclusions: Over a 10 year period pressure ulcer prevalence estimates in one large hospital remained fairly constant in spite of the risk profile increasing. Activity and friction/shear deficits are associated with pressure ulcers of stage II or greater, and indicate continuing risk with these patients. However, demographic and clinical characteristics also play an important role in describing the characteristics of patients with pressure ulcers. Implications for Policy, Delivery, or Practice: Pressure ulcer occurrence continues to be a significant problem in acute care. Current emphasis on adverse events and patient safety should bring a renewed focus on the personal, environmental and system factors related to this longstanding problem. Primary Funding Source: Canadian Institutes of Health Research Personnel Award ●Strategies for Assessing Health Plan Performance on Chronic Diseases: Selecting Performance Indicators and the Application of Health-Based Risk Adjustment Ann Volpel, MPA, Jonathan Weiner, DrPH, Steven Johnson, Ph.D., John O'Brien, MPA, David Idala, MA, Hoon Byun, MA Presented By: Ann Volpel, MPA, Senior Research Analyst, Center for Health Program Development and Management, University of Maryland, Baltimore County, 1000 Hilltop Circle, Social Sciences 3rd Floor, Baltimore, MD 21250; Tel: (410)4556518; Fax: (410)455-1309; Email: avolpel@chpdm.umbc.edu Research Objective: As Medicaid managed care programs mature, states are looking to refine their methods of measuring and improving the performance of health plans. Along with other private and public health insurers, states are developing strategies to pay for performance and implement value-based purchasing. Many states are also increasing their scrutiny of the care provided to enrollees with chronic diseases. This research, funded by the Center for Health Care Strategies, piloted an approach to measure health plan performance using administrative data to evaluate the care provided to enrollees with chronic diseases. Study Design: Using encounter data from Maryland’s Medicaid program, a series of performance measures were developed and calculated for enrollees with four chronic diseases: asthma, diabetes, HIV/AIDS, and schizophrenia. The focus on chronic diseases sets the research apart from many performance measurement approaches that concentrate on preventive services such as well child visits. A unique feature of our approach is the application of health-based risk adjustment to performance measurement, a relatively new field in Medicaid. Health-based risk adjustment provides the ability to reduce the confounding effects of case-mix differences between health plans. The results of our analysis were used to profile health plans to identify differences in performance. Selected performance indicators allowed us to evaluate patterns of service utilization across all four diseases, as well as to measure disease-specific performance for each plan. This study also identified performance indicators that are associated with desired outcomes. Logistic regression models were used to identify factors associated with a decreased risk for hospitalization for each disease studied. The regression model controlled for health status, eligibility category, and demographic factors including sex, age, race, and geographic region. Our analysis provides a foundation identifying performance indicators that a state may consider including in a performance measurement program. Population Studied: We analyzed administrative data from the Maryland Medicaid program in CY2002. We selected individuals with at least 320 days of enrollment in one of six health plans who met criteria for one of the following four conditions: asthma n=16,836, diabetes n=7,121, HIV/AIDS n=2,162, and schizophrenia n=5,038. Principal Findings: The research yielded three important findings: 1-Outcome measures such as inpatient admissions are highly sensitive to risk and should be risk-adjusted if used to assess performance; 2- Process measures, such as preventive services, are relatively insensitive to risk and do not need to be risk-adjusted to be considered valid performance indicators; 3-Regular access to ambulatory care services significantly reduces the likelihood of a hospital admission. These findings held across all four chronic conditions examined. Conclusions: Administrative data can be a rich source of information for states to assess health plan performance for chronically ill populations across a number of parameters. Administrative data, when of high quality, can also support sophisticated risk adjustment techniques. Implications for Policy, Delivery, or Practice: State policy makers may use our techniques to evaluate potential or existing pay for performance and value based purchasing strategies. This paper provides a detailed template and specific examples regarding how to design and execute such evaluations. Primary Funding Source: RWJF, Center for Health Care Strategies ●Impact of State Tort Reforms on Medical Malpractice Payments Teresa Waters, Ph.D., Peter P. Budetti, M.D., JD Presented By: Teresa Waters, Ph.D., Associate Professor, Preventive Medicine, Center for Health Services Research, University of Tennessee Health Science Center, 66 North Pauline Street, Suite 463, Memphis, TN 38163; Tel: (901)448.5826; Email: twaters@utmem.edu Research Objective: State tort reform is a high priority for many states as they struggle to respond to rising medical malpractice premiums. Very little empirical evidence exists on the impact of various state-level reforms. The purpose of this study is to investigate the impact of state-level tort reforms on the number and size of medical malpractice payments. Study Design: Cross-sectional, time series data on the number of physician paid claims and average physician claim payment were constructed for each state for the years 19912003. A comprehensive review of state laws was conducted to create Likert scales rating state tort laws in 13 separate areas. Multivariate fixed effects models were used to explore the impact of the 13 tort area scales on number and size of claims payments. Population Studied: Medical malpractice payments for US physicians were identified using the National Practioner Data Bank public use file. Number of practicing physicians were identified using AMA data. Principal Findings: Total number of paid claims and number of paid claims per 1,000 practicing physicians (per state, per year) were negatively related to strong expert witness requirements (p<0.01). These measures were also negatively related to stronger frivolous suit penalties (p<0.05), caps on damages (p<0.10), and ad damnum clauses (p<0.10). Average paid amounts for physician claims and physician medical malpractice payment dollars per 1,000 practicing physicians (per state, per year) were also negatively related to expert witness requirements (p<0.01). Dollars per 1,000 physicians were also negatively related to caps on damages (p<0.10) and periodic payment of damage awards (p<0.10). Conclusions: Our results suggest that expert witness requirements and frivilous suit penalties may have a negative effect on number and size of paid physician medical malpractice claims. We find weaker evidence that stronger caps on damages, ad damnum clauses and periodic payment of damage awards may also have a negative impact. A number of other tort reforms did not appear to have an effect, including: caps on attorney's fees, immunity provisions, pretrial screening, shorter statute of limitations, and joint and several liability. Implications for Policy, Delivery, or Practice: Our study provides some of the first comprehensive evidence on the relative impact of various state-level reforms using a national database of physician medical malpractice payments. When seeking to identify tort reforms that might lower the number and level of physician paid claims, state and national lawmakers should consider the evidence on the relative 'efficacy' of the available tort reform measures. Primary Funding Source: Foundation, Kaiser Family Foundation ●Practice Pattern Analysis: Improving the Efficiency of Care William Westerfield, MA, Ray Phillippi, Ph.D., Ken Patric, M.D., Judy Slagle, RN, MPA, Steven Coulter, M.D. Presented By: William Westerfield, MA, Research Scientist, Health Services Research, BlueCross BlueShield of Tennessee, 801 Pine Street 3E, Chattanooga, TN 37402; Tel: (423) 7556260; Email: William_Westerfield@bcbst.com Research Objective: The primary objective of this study was to describe the Practice Pattern Analysis (PPA) methodology of a single-state health plan in the Southeastern United States. The secondary objective of this study was to measure the impact of the PPA on the efficiency of network practitioners. Study Design: The PPA employed by the health plan was designed to reveal differences in the efficiency of care of an individual practitioner compared to his/her specialty. The main measures reported on the PPA were based on an episode of care model. An episode of care is a unit of analysis that includes all the services a patient received to treat a particular illness. Each episode of care was assigned to one of nearly 600 Episode Treatment Groups (ETGs). Some of the measures reported included the case mix index, efficiency ratio, episode volume, patient satisfaction, and efficiency by service category (ancillary, facility, management, pharmacy and surgery). The PPA includes a detailed report of the practitioners Top 5 ETGs that compares total and service category costs per episode. A pretest-posttest design was used to measure the impact of the PPA on the efficiency of network practitioners. The efficiency of practitioners was measured using the PPAs produced in the 2nd quarter of 2003 and the 1st quarter of 2004. Efficiency Ratios from 2003 and 2004 were compared for all practitioners and practitioners identified as outliers. Expected costs for each ETG were calculated and compared to a practitioner’s actual costs to calculate the Efficiency Index. The Efficiency Index was adjusted for specialty differences by dividing the Practitioner Efficiency Index by the Specialty Efficiency Index to create the Efficiency Ratio. The 2003 Efficiency Ratios were adjusted for medical inflation using the Consumer Price Index for the Southeast. Population Studied: The study population included 8,902 practitioners who received a PPA in 2003 and 8,857 practitioners who received a PPA in 2004. The study population also included a subset of 573 outlier practitioners who received a PPA in 2003. Outliers were practitioners whose performance was in the lowest 10 percentiles in 2003 for all practitioners. Principal Findings: The methodology employed by the health plan significantly impacted the efficiency of the network practitioners. Preliminary findings indicate the mean efficiency score of all practitioners improved 8% between 2003 and 2004 (difference of means p<.05). Only 26% of practitioners who were outliers in 2003 remained outliers in 2004 (117 out of 573). The mean efficiency score of practitioners who were outliers in 2003, but not in 2004 improved an astounding 80% in single year (difference of means p<.05). Conclusions: The PPA methodology employed by the health plan had a positive effect on the efficiency of the all practitioners. More importantly, PPA methodology had a major impact on the efficiency of outlier practitioners helping these practitioners substantially improve their efficiency. Implications for Policy, Delivery, or Practice: : Efficiency based PPAs are an effective method to improve the efficiency of care delivered by practitioners. The PPA employed by this health plan is portable and could easily be employed by other health plans. Primary Funding Source: BlueCross BlueShield of Tennessee ●Culture Schlock: Is Safety Culture Measurable or Relevant? Arthur Williams, Ph.D., Todd Huschka, BS BA, Timothy Beebe, Ph.D., Rosa Cabanela, Ph.D., Erin McMurtry, BS, Sandra Bruggeman, BA Presented By: Arthur Williams, Ph.D., Chair, Division of Health Care Policy & Research, Health Care Policy & Research, Mayo Clinic, 200 1st Street SW, Rochester, MN 55905; Tel: 507-284-3356; Email: williams.arthur@mayo.edu Research Objective: Evaluate the IHI Safety Climate Survey to determine whether it provides an accurate descriptive measure of “safety culture.” Use findings to comment upon the “safety culture” concept in terms of interpretability and administrative relevance. Comment upon a review of the “safety culture” literature. Study Design: Survey questionnaires containing the 19 IHI Safety Climate Survey questions and a Single Safety Climate Question (SSCQ) were distributed via intranet to more than 3000 nurses, physicians, and residents. The overall response rate was 45.4%, resulting in 1,451 useable survey responses. Exploratory (EFA) and confirmatory factor analyses (CFA) were done using Stata version 8.2 and AMOS version 5.0. The CFA were fit with different sets of items and factors to model underlying relationships. Standard criteria for EFA (loadings >.40, eigenvalues > 1.00, no cross-loading, etc.) and CFA (CFI >.900, RMSEA < 0.05, CMIN/DF 1.00 to 3.00, etc) were used. Item Response Theory (IRT) was applied to assess survey questions against several constructs. Cronbach’s alphas were calculated, and correlation analyses were done among latent constructs and items. Population Studied: 1,451 clinical survey respondents at a large mid-western hospital. Principal Findings: The underlying correlations between the IHI Safety Climate Survey score and the SSCQ was 0.60. Several models were generated using EFA and CFA with anywhere from 4 to 10 items from the IHI survey loading on factors. In no case was more than 30.7% of the sample variance accounted for on a factor. CFA indicated that the 19 IHI questions are not consistent with a “safety culture” construct. Correlations from many combinations of items with the SSCQ ranged from 0.51 to 0.61. Combinations of items did not have distinct relationships with the SSCQ; almost any combination could be selected to obtain a similar correlation. The simplest model of 4 items (Cronbach’s alpha = 0.70) had a correlation of 0.51 with SSCQ and 0.91 with the IHI 7 question scores. IRT indicated that 9 of the 19 IHI questions had an alpha <0.90 implying low discrimination. Deleting these showed minimal information loss. No total information was provided at the upper ends of trait distribution (>1std). Conclusions: There are a multitude of seemingly random collections of items that could be used to model factors and correlations with the IHI scores. Many of the questions included in the various models did not include the word safety, nor did they seem to be related to safety. In general, one could randomly select between 4 and 10 questions and create his or her own purported measure of “safety culture.” Implications for Policy, Delivery, or Practice: Increasingly healthcare organizations are becoming aware of the importance of transforming organizational culture in order to improve patient safety. This growing awareness has created an interest in developing assessment tools focused on the culture of safety. However, in the desire to create bigger or better tools researchers run the risk of measuring that which cannot be easily measured, drawing conclusions unrelated to the question at hand, removing resources from direct patient care (opportunity costs), and making dysfunctional administrative prescriptions. Primary Funding Source: Internal ●Extent of Potentially Inappropriate Antibiotic Prescribing for Healthy Adults with Acute Bronchitis in Managed Care Lok Wong, MHS, Jennifer Lis, MBA, MHSA, Barbara Souder, Ph.D., MPH, RN, Mahil Senathirajah, MBA Presented By: Lok Wong, MHS, Senior Health Care Analyst, Quality Measurement, National Committee for Quality Assurance, 2000 L Street, NW, Suite 500, Washington, DC 20036; Tel: (202)955-1784; Fax: (202)955-3599; Email: wong@ncqa.org Research Objective: To investigate the extent to which antibiotics are inappropriately prescribed to healthy adults with acute bronchitis in managed care. Study Design: Retrospective claims analysis of data from four commercial and one Medicaid managed care plan; percentages of healthy enrollees with acute bronchitis who received an antibiotic prescription were calculated. A validation study was also conducted with a sample of 150 patient records per plan to determine accuracy of administrative data. Population Studied: 2.6 million managed care enrollees ages 18 to 64, of which 1.6 million were identified with an acute bronchitis episode in 2003, excluding unhealthy enrollees with comorbid conditions or acute competing diagnoses, i.e. bacterial where antibiotics may be appropriate, or prior treatment with antibiotics. Principal Findings: Consistently high rates of inappropriate antibiotic prescribing were found across the five plans, average 70%; plan range 64% to 74. These rates of prescribing were found regardless of age cohort - 18-49 and 50-64 years of age, location of treatment - ER vs. non-ER - and type of bronchitis diagnosis - acute bronchitis vs. bronchitis not otherwise specified. Similar but slightly lower rates of prescribing were found in patients with comorbidities, i.e. COPD or other competing diagnoses, where antibiotics may be appropriate. Conclusions: Overall high rates of antibiotic prescribing in patients without clinical characteristics that warrant antibiotic treatment found in the study, i.e. comorbid condition or competing diagnosis highlights the extent to which physicians in managed care are inappropriately prescribing antibiotics for healthy adults with acute bronchitis. Managed care plans should be encouraged to develop quality improvement activities to reduce inappropriate antibiotic prescribing in clinical practice through the introduction of a standardized clinical performance measure targeting inappropriate antibiotic prescribing in healthy adults with acute bronchitis. Restricting a measure to patients identified with acute bronchitis - ICD-9 diagnosis code 466.0 - will improve the clinical validity of the measure denominator, even though results indicate all patients regardless of the type of bronchitis diagnosis - acute vs. not otherwise specified - are receiving antibiotics at similar rates. Patients with chronic bronchitis were considered to have a comorbid condition and excluded from the study. Implications for Policy, Delivery, or Practice: A HEDIS Effectiveness of Care performance measure highlighting the extent of inappropriate antibiotic prescribing in adults with acute bronchitis will help to drive quality improvement activities in managed care organizations and improve overall awareness of the widespread overuse and misuse of antibiotics that contributes to antibiotic drug resistance. Similar measures addressing inappropriate antibiotic prescribing in pediatric populations have been introduced into pay for performance schemes to drive quality improvement in medical groups and among physicians. In addition, dissemination of these study findings will improve physicians’ awareness of poor compliance with treatment guidelines for managing otherwise healthy patients with acute respiratory conditions such as acute bronchitis. Further research is needed to understand therapeutic decision-making by physicians for acute bronchitis patients with comorbid conditions and why antibiotic prescribing rates may be lower than in healthy adults. Primary Funding Source: CDC, Council for Affordable Quality Healthcare, National Committee for Quality Assurance ●Presence and Correction of Regression to the Mean When Providers’ Quality of Care is Evaluated by Changes in Beneficiaries’ Health Status Over Time Ning Wu, Ph.D., M.D., Jason Roy, Ph.D., Vincent Mor, Ph.D., Susan C Miller, Ph.D., Kate Lapane, Ph.D. Presented By: Ning Wu, Ph.D., M.D., Abt Associates, Inc, 55 Wheeler Street, Cambridge, MA 02138; Tel: (617)520-2588; Fax: (617)349-2675; Email: Ning_Wu@abtassoc.com Research Objective: To evaluate the probable impact of regression to the mean on the evaluation of quality of care provided by Medicare + Choice Organizations (M+CO's). Study Design: Simulation Population Studied: Medicare beneficiaries who enrolled in Medicare + Choice Organizations (M+CO's). Principal Findings: We used the data from the first cohort of Health Outcome Survey (HOS), a study evaluating the health care quality of Medicare + Choice Organizations (M+COs). In the HOS, Medicare beneficiaries’ health status is measured by SF-36 at baseline and two years later. Beneficiaries’ health transition status, e.g., mental health “being the same or better” or otherwise, is evaluated by comparing the changes in the SF-36 scores over a two-year period to a pre-defined cutoff. The health transition status is then aggregated at the provider level to evaluate M+COs’ performance. In simulation, we demonstrate that due to RTM, beneficiaries with low baseline scores are more likely to be misclassified as mental health being the same or better although they experience deterioration, and vice versa for beneficiaries with high baseline scores. As a result, health transition rates aggregated at the plan level may be based either downwards or upwards conditional on the distribution of baseline scores in the M+CO. That is M+CO’s with high baseline scores may appear to perform worse than M+CO’s with low baseline scores because of RTM. We also propose a method to correct the bias introduced by RTM and demonstrate via simulation that estimations of transition rate has been improved by the correction method, so has the provider profiling. Implications for Policy, Delivery, or Practice: Because improving individuals’ health outcomes is the ultimate goal of health care, we expect more and more health services researchers evaluate provider performance by measuring the changes in patients’ health status. The problem of RTM has not been discussed or examined in such longitudinal studies. Using simulation, we demonstrated that RTM may bias the comparison of provider performance and lead to unwarranted reward or sanction of providers. Further research is needed to examine RTM using real data. The knowledge gained from further research on RTM will improve the validity of HOS study results, and provide insight or even solutions to other health services researchers facing similar problems. Primary Funding Source: No Funding Source ●How Much Does Medicare Pay Hospitals for Medical Injuries? Chunliu Zhan, M.D., Ph.D., Bernard Friedman, Ph.D., Andrew Mosso, MS, Peter Pronovost, M.D. Presented By: Chunliu Zhan, M.D., Ph.D., Service Fellow, AHRQ, 540 Gaither Road, Rockville, MD 20850; Tel: (301)4271225; Fax: (301)427-1341; Email: czhan@ahrq.gov Research Objective: Previous studies indicate that medical injuries during hospitalization significantly extend length of stay and incur significant cost, in addition to increasing a patient’s risk of dying in the hospital. But the extent to which hospitals are reimbursed by payers for treating medical injuries is not clear. This study assesses how much Medicare, the biggest payer in the US, is paying hospitals for medical injuries and explores the implications. Study Design: The Agency for Healthcare Research and Quality Patient Safety Indicators (PSIs) were used to screen selected types of medical injuries, based on ICD-9-CM codes in discharge summaries for Medicare beneficiaries in the 2002 Health Care Utilization Project’s Nationwide Inpatient Sample. Diagnosis-Related Groups (DRGs) were assigned to the claims with medical injuries based on the ICD-9-CM codes using 3M’s DRG Grouper Software, and payments were calculated based on published Medicare payment formulae and payment determination data such as DRG relative weights and other adjustment factors. Second, the ICD-9-CM codes indicating the medical injuries were removed from the claims, and DRGs were re-assigned and payments recalculated. The changes in DRGs and payments were examined. Population Studied: Medicare beneficiaries aged 65 or over. Principal Findings: Of the Medicare claims that had iatrogenic pneumothorax (IP, n=2,466), postoperative sepsis (PS, n=1,399) and decubitus ulcer (DU, n=34,115), the percentages of claims changed DRGs after removing PSIindicating ICD-9-CM codes are 2.11%, 0.79% and 2.38%. With few exceptions, these changes were from DRGs with complication to DRGs without complication. Medicare extra payments were at $1369 (Confidence Interval, CI=57) per IP event, $8807(CI=296) per PS event, and $735(9) per DU event, which account for 6.75%, 19.62% and 4.60% of the average Medicare payment. Of the extra payment, 92%, 99% and 91% respectively were due to changes in the outlier payment for cases with unusually high costs. National Medicare payment for IP, PS and DU events were estimated at $16.4 million, $59.2 million and $155.0 million respectively. Conclusions: The Medicare DRG-based inpatient payment system pays a substantial amount to hospitals for treating medical injuries, but the amount appears to be a small fraction to the cost incurred to hospitals. Implications for Policy, Delivery, or Practice: Medicare could achieve savings by investing in patient safety improvement. These improvements could also yield savings for hospitals that serve Medicare patients and for private payers by reducing payment for treating medical injuries and avoiding cost-shifting from Medicare. Primary Funding Source: No Funding Source