Quality: Measuring & Improving Quality
Call for Papers
Quality: Ambulatory Care
Chair: David Hopkins, Pacific Business Group on Health
Sunday, June 25 • 5:45 pm – 7:15 pm
●Evaluating Sustained use of Quality Improvement
Interventions in VHA HIV Care
Candice Bowman, Ph.D., RN, Tuyen Hoang, Ph.D., Jason
Saleem, Ph.D., Allen Gifford, M.D., Matthew Goetz, M.D.,
Steven Asch, M.D., M.P.H.
Presented By: Candice Bowman, Ph.D., RN, QUERI-HIV
Implementation Research Coordinator, Health Services
Research & Development, VA San Diego Healthcare System,
3350 La Jolla Village Dr (111N-1), San Diego, CA 92130; Tel:
(858)552-8585 x5967; Fax: (858)552-4321;
Email: candice.bowman@va.gov
Research Objective: To confirm that quality improvement
(QI) interventions are institutionalized, it is important to
measure whether positive effects from an implementation
initiative persist after the study’s end. In our original study,
we analyzed the effects of two QI interventions targeting
provider performance on guideline-based care for treating HIV
disease (HIVGBC). To ascertain whether the implemented
interventions were sustained and became part of routine care,
we measured the original outcomes for one additional year
and evaluated continued intervention use at selected sites.
Study Design: In the main study, we implemented either
clinical reminders (R), a modified Breakthrough Series
collaborative (C), or both (R+C) at 16 VHA facilities in a 4-arm
quasi-experiment. We examined the odds of patients in each
intervention arm receiving any of 10 indicators of HIVGBC
(screening for Hepatitis A and C, toxoplasma, syphilis and
monitoring of lipids and CD4 cell counts and viral loads if on
antiretrovirals) compared to controls at 12 months past
baseline. Interviews with key informants selected from six of
the study sites exposed to either one or both of the
interventions about continued use at their facilities in the
follow-up year revealed that some study sites had ceased
using the interventions and some control sites had adopted
them; analyzing odds of patients receiving HIVGBC compared
to controls no longer made sense. Thus, we evaluated
sustained use as follows: At the facility- rather than the armlevel, we examined raw rates of patients receiving HIVGBC at
only those facilities in the intervention arms that had
significant effects in the study year to find whether they
continued to show a significant increase in these rates in the
following year compared to their raw rate at baseline.
Population Studied: HIV-infected patients in care at 16 VHA
facilities during 2002-4.
Principal Findings: Original study findings showed that Rs
significantly improved 3/10 care indicators, Cs improved 2/10,
and R+C improved 4/10 (totaling 6 HIVGBC indicators with
significant positive effects). For Hepatitis A screening, we
found that 4 out of the 5 sites that showed a significant
increase (p=.05) in their raw rate at 12 months also showed a
significant increase in their raw rate at 24 months compared
to baseline. For the other five significant indicators of
HIVGBC, all sites that showed significant increases in their
raw rates at 12 months also showed a significant increase in
their raw rates at 24 months compared to baseline.
Conclusions: These findings suggest that intervention effects
were sustained for one year at nearly all the sites that showed
significant increases in performance during the study period.
Qualitative results indicated that nearly all sites were using at
least some of reminders in the year after the study period.
Collaborative methods were still being used but only at the
most activated of the original study sites.
Implications for Policy, Delivery, or Practice: These findings
confirm that implementation projects can be important
mechanisms for ensuring that QI interventions improve care
delivery and actually become integrated into routine use.
However, exposure to the interventions in control sites in a
follow-up analysis can make this process difficult to assess.
Primary Funding Source: VA,
●Delayed Access to Health Care and Mortality
Julia Prentice, Ph.D., Steven D. Pizer, Ph.D.
Presented By: Julia Prentice, Ph.D., Health Services Research
Fellow, Center for Health Quality, Outcomes and Economic
Research, Department of Veterans Affairs, 200 Springs Road
(152), Bedford, MA 01730; Tel: 781-687-2882;
Email: jprentic@bu.edu
Research Objective: Long waits for healthcare have been
found to have a negative impact on patients and the
healthcare system. For example, long waits decrease patient
satisfaction and increase the use of the emergency room for
non-urgent conditions. Most importantly, policymakers argue
long waits will result in delays in diagnosis and treatment, and
these delays are hypothesized to negatively affect individual
health. Despite the assumed importance of ensuring timely
access to care to improve health outcomes, little research has
actually examined the empirical association between waiting
for outpatient care and health outcomes. This is likely due to
lack of data. Estimates of how long individuals wait for
healthcare are largely based on self-reported data. Data
collected by the Department of Veterans Affairs (VA) is an
important exception because it is one of the only health care
systems in the U.S. that automatically collects data on how
long patients wait for healthcare through its outpatient
scheduling system. Using these unique data, this study
examines the relationship between waiting for VA outpatient
health care and mortality.
Study Design: This was a retrospective observational study
using secondary data from VA administrative sources. The
main explanatory variable of interest was VA facility-level wait
times for outpatient visits. Waiting for outpatient care was
defined as the number of days between the appointment
request and the day the next available appointment was
scheduled. Facility-level data on the average wait by clinic
stop were extracted for 89 VA medical centers in 2001.
Individual-level logistic regression models were estimated that
included 1) facility-level wait times for outpatient care, 2)
standard risk-adjustors for prior individual health status and 3)
facility-level differences in case-mix. These models predicted
the odds of dying within a six month follow-up period.
Population Studied: Veterans over 65 who had at least one
geriatric outpatient appointment between October 2000 and
June 2001 were included. This sample was older and more
frail than the general population and ought to be particularly
sensitive to variation in the timeliness of access to medical
care.
Principal Findings: Veterans who visited a VA medical center
with facility-level wait times of 32 days or more had
significantly higher odds of mortality (odds ratio=1.20,
p=0.019) compared to veterans who visited a VA medical
center with facility-level wait times of less than 32 days.
Conclusions: Our findings support the widely assumed
association between long wait times for outpatient health care
and negative health outcomes, such as mortality. Future
research should focus on the causes of long waits for health
care, the consequences of these long waits on other health
outcomes (e.g. preventable hospitalizations) and effective
policies to decrease long waits for health care services.
Implications for Policy, Delivery, or Practice: Long waits for
healthcare can have a negative health impact, and
policymakers ought to focus on monitoring wait times and
reducing them if necessary. Healthcare systems should follow
the VA’s example of systematically monitoring wait times and
implementing and evaluating policies to decrease wait times.
Primary Funding Source: VA
●Developing a Physician-Level Measurement set for Back
Pain: Challenges and Opportunities
Sarah Sampsel, M.P.H., Philip Renner, M.B.A.
Presented By: Sarah Sampsel, M.P.H., Senior Health Care
Analyst, Quality Measurement, NCQA, 2000 L St., NW #500,
Washington, DC 20036; Tel: 505-986-9848; Fax: 202-955-3599;
Email: sampsel@ncqa.org
Research Objective: To evaluate the feasibility of the
development of performance measures to identify and
recognize physicians who provide high-value, patient-centered
care for patients with back pain.
Study Design: Structured literature review and clinical expert
panel consensus processes to identify and develop physician
level measures to assess care provided to adults experiencing
episodes of back pain lasting at least six weeks. Based on
published clinical guidelines, peer-reviewed publications and
expert panel advice, a clinical logic for back pain treatment
components was developed and utilized to identify
measurement opportunities.
Population Studied: Adults with a diagnosis of back pain and
documented physician visits with treatment duration of at
least 6 weeks. Back pain patients are identified utilizing ICD9-CM codes, physician registries or medical records. This
program was developed to be applicable across back pain
diagnoses, but may have particular importance to physicians
treating disc herniation, spinal stenosis and/or
spondylolisthesis.
Principal Findings: A total of 8 measures were identified for
inclusion in this newly developed spine care recognition
program (SCRP); a voluntary program for individual
physicians or physician practice sites providing care to people
experiencing subacute and chronic back pain. The measures
that will provide performance data include: Appropriate
Treatment – 1) Patient Assessment: Rate of medical records
with complete documentation of pain and functional status,
psychosocial/mental health status and employment status, 2)
Initial Visit: Rate of medical records with complete
documentation of physical examination, patient history with
assessment of “red flags” and assessment of prior treatment
and response, 3) Patient Education: Rate of medical records
with complete documentation that the following was provided:
review of treatment options, review of the risks and benefits of
each treatment option and smoking cessation, 4)Physical
Activity/Exercise: Rate of medical records with documentation
that patients were advised or recommended physical
activity/exercise; Inappropriate Treatment – 5) Imaging
Studies: Rates of imaging in acute low back pain in the
absence of “red flags” and repeated studies within 12
months, 6) Epidural Steroid Injections: Rate of epidural
steroid injections in patients without evidence of
radiculopathy/sciatica, 7) Post-surgical Complications: Rates
of re-hospitalization post-surgical procedure and repeated
surgical interventions; and Outcomes - Treatment Reevaluation: Rate of medical records with documentation of
treatment re-evaluation to assess pain and functional status.
Conclusions: Based on available evidence, published clinical
guidelines and expert consensus an initial set of measures to
recognize high quality back pain care has been identified and
specified for deployment. Measures were selected for
inclusion based on their potential to impact the quality of care,
face and content validity, and importance and meaningfulness
related to back pain care. A full pilot study on the feasibility,
reliability and validity at the physician level will all be assessed
during pilot test and public comment periods to occur in
early 2006.
Implications for Policy, Delivery, or Practice: Opportunities
to measure and improve the care of patients with back pain
exist but have not been implemented on a national basis.
This new physician recognition program is a first step in
improving the quality of care back pain patients receive and
will potentially provide a return on investment upon
implementation.
Primary Funding Source: No Funding
●Ambulatory Care Adverse Events and Preventable
Edverse Events Leading to Hospital Admission
Donna Woods, EdM, Eric Thomas, M.D., M.P.H., Jane Holl,
M.D. M.P.H., Kevin Weiss, M.D. M.P.H.
Presented By: Donna Woods, EdM, Research Assistant
Professor, Institute forHealthcare Studies Feinberg School of
Medicine, Northwestern University, 339 E Chicago Ave 7th Fl,
Chicago, IL 60611; Tel: (847) 571-2593; Fax: (312) 503-2936;
Email: woods@northwestern.edu
Research Objective: Most health care in the United States is
delivered in the ambulatory setting, but the epidemiology of
errors and adverse events in ambulatory care is understudied.
Study Design: We selected a representative sample of
hospitals from Utah and Colorado, and then randomly
sampled 15,000 non-psychiatric discharges from 1992. Each
record was screened by trained nurse reviewers for one of 18
criteria associated with adverse events. If one or more criteria
were present, the record was reviewed by a trained physician
to determine if an adverse event occurred. Adverse events
were defined as an injury caused by medical management
rather than disease processes that resulted in hospitalization
or disability at discharge. Ambulatory adverse events (AAEs)
were adverse events that took place in an ambulatory care
setting (physician’s office, day surgery center, emergency
department, hospital clinics, home) that directly resulted in
hospitalization. Two investigators judged preventability to
identify ambulatory preventable adverse events (APAEs). We
report percentages and 95% confidence intervals.
Population Studied: A population-based study of patients
who experiences an ambulatory care related adverse event
that led to a hospital admission in the states of Colorado and
Utah.
Principal Findings: We reviewed 14,700 hospital discharge
records and found 587 adverse events of which 70 were AAEs
and 31 were APAEs. When weighted to the general
population, there were 2,608 AAEs and 1,296 (44.3 %) APAEs
in Colorado and Utah in 1992. APAEs occurred most
commonly in physicians’ offices (43.1%, 46.8-27.8), the
emergency department (32.3%, 46.1-18.5), and at home
(13.1%, 23.1-3.1). APAEs in day surgery were less common
(7.1%, 13.6-0.6), but caused the greatest harm to patients.
The types of APAEs were broadly distributed among missed or
delayed diagnoses (36%, 50.2-21.8), surgery (24.1%, 36.7-11.5),
non-surgical procedures (14.6%, 25.0-4.2), medications
(13.1%, 23.1-3.1), and therapeutic events (12.3%, 22.0-2.6).
Provider types involved in the APAEs included primary care
(31.4%, 33.5-29.3), surgical specialties (22.6%, 24.5-20.7),
medical specialties (21.8%, 23.7-19.9), and emergency
medicine (18.5%, 20.3-16.7). Most APAEs occurred in adults
(45.5% in 21-64 year olds; 38.1% in patients 65 or older.
Overall, 10% of ambulatory preventable adverse events
resulted in permanent injury or death. The proportion of
APAEs that resulted in death was 31.8% for general internal
medicine, 22.5% for family practice, and 16.7% for emergency
medicine.
Conclusions: Although dated, these are the only populationbased epidemiological data that describe APAEs and related
harm. Nationally, over 75,000 hospitalizations per year are
due to preventable errors in the outpatient setting.
Implications for Policy, Delivery, or Practice: Broad-based
research and prevention efforts will be required due to the
diverse locations and providers involved and due to the
varying types of APAEs.
Primary Funding Source: No Funding
●The Quality of Diabetes Care by Insurance Status
James Zhang, Ph.D., Anne C. Kirchhoff, M.P.H., Jennifer W.
Walk, BA, Cynthia T. Schaefer, APRN,BC, Loretta J. Heuer,
Ph.D., R.N., Marshall H. Chin, M.D. M.P.H.
Presented By: James Zhang, Ph.D., Director of Health
Econometrics, Medicine, The University of Chicago, 5841 S.
Maryland Ave (MC 2007), Chicago, IL 60637; Tel: (773)8341956; Fax: (773)834-2238;
Email: xzhang@medicine.bsd.uchicago.edu
Research Objective: Access to quality care is important to
eliminate health disparities and increase the quality and years
of healthy life for all persons in the United States. Community
health centers (CHCs) provide primary health care to 12
million Americans with or without health insurance in
medically underserved areas. Over forty percent of CHC
patients are uninsured and over one-third are on Medicaid.
Currently the CHC program is in the process of expanding its
service capacity by 40% as part of a five-year initiative of
President Bush’s administration. Understanding whether
there are differences in quality of care among patients
according to insurance status can lead to the understanding
of the limits and potential additional gains of such an
initiative. The objective of this study is thus to compare the
quality of care for diabetes patients by insurance status at
CHCs.
Study Design: Twenty-seven CHCs in 17 West Central and
Midwest states were sampled in the year 2002. A total of
2,052 diabetes patients were enrolled in the study. An
algorithm to categorize diabetes patients into six mutually
exclusive groups was developed: no insurance,
Medicare/Medicaid dual eligible (DE) group, Medicare
without Medicaid, Medicaid without Medicare, private
insurance, and other. A set of six quality of care indicators
developed by the National Committee for Quality Assurance
(NCQA) were used. Multivariate regression analysis
technique was applied to analyze the association between the
insurance coverage and quality of care, adjusting for age,
gender, race, one dummy variable for urban location of
services, seven dummy variables for medical
comorbidity/complications, and CHC site fixed-effects.
Population Studied: 2,052 diabetes patients cared for in 27
CHCs in 17 states in the year of 2002.
Principal Findings: The mean age of patients was 54 years
old, and 60% of the patients were female. Twenty percent of
the patients were African American, 29% were Hispanic, and
41% of them were cared for in urban areas. Compared to the
no insurance group, those with Medicare without Medicaid,
DEs, private insurance, and others were three to seven
percentage point higher in having HbA1C testing (p=0.05,
0.004, 0.004, 0.05, respectively) Also, DEs and private
insurance groups were twelve to twenty-one percentage point
lower in having poor HbA1C control (p=0.07, 0.06,
respectively) than the no insurance group. The private
insurance group was also more likely to have eye examinations
and lipid profiles (p=0.009, <0.001, respectively).
Conclusions: Improved insurance coverage is associated with
higher quality of care within CHCs.
Implications for Policy, Delivery, or Practice: This study
argues for coupling expanding CHC service locations with
increasing insurance coverage to achieve the best health
outcomes for the thousands of indigent diabetes patients who
rely on safety-net providers for their primary care, as improved
insurance coverage is associated with higher quality of care.
Primary Funding Source: AHRQ
Call for Papers
Hospital Quality
Chair: Sheldon Greenfield, University of California, Irvine
Monday, June 26 • 8:30 am – 10:00 am
●Volume-Outcome Relationships: An Econometric
Approach to CABG Surgery
Hsueh-Fen Chen, M.S., Gloria Bazzoli, Ph.D., Askar
Chukmaitov, Ph. D.
Presented By: Hsueh-Fen Chen, M.S., Research Associate,
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: chenh@vcu.edu
Research Objective: To examine the effect of hospital volume
on risk-adjusted in-hospital mortality rate for coronary artery
bypass graft (CABG) surgery, and to demonstrate the
coefficient estimates change when the potential biases were
gradually controlled in the model.
Study Design: A longitudinal study design across six years
was applied to examine the volume-quality relationship. The
unit of analysis is the hospital. The dependent variable is the
in-hospital mortality rate, which was risk adjusted to control
for differences across hospitals in patient age, gender, the
interaction of age and gender, hospital, and APR-DRG. The
primary independent variable of interest is the total number of
CABG surgeries performed at a hospital in a year. Because a
curvilinear relationship may exist between volume and quality,
the logarithm of volume was applied in the model.
The dependent variable, independent variable, and the other
control variables were aggregated to the hospital level. A
series of specification tests was conducted to examine
endogeneity and heterogeneity. The tests indicated that
endogeneity and unobserved heterogeneity existed, and that
fixed or random effects estimation with instrumental variables
was needed. Bed size, tertiary capacity, and the level of
competition were used as instruments to predict long-run
CABG volume for hospitals. These instruments passed
specifications tests that they were related to CABG volume but
not related to the error term of the quality model. The
Hausman test indicated that random effects estimation was
consistent and thus preferred to fixed effects estimation.
Population Studied: All nonfederal, general short-term
hospitals included in the Health Care Cost and Utilization
Project State Inpatient Data (HCUP-SID) data from 1995-2000
for the states of AZ, CA, CO, FL, IA, MA, MD, NJ, NY, WA,
and WI. The sample was further restricted to hospitals that
had at least 6 CABG surgeries in these years, which led to a
hospital sample of 1,881 for the six study years. Due to
missing values for other control variables, the total hospitals
in the random effects models were 1,321. Dependent variables,
independent variable, and control variables were constructed
with HCUP-SID data. AHA Annual Survey provides hospital
characteristics and market variables. CMS provides DRG
weight. HMO penetration at MSA level was drawn from HMO
InterStudy.
Principal Findings: After controlling for endogeneity,
heterogeneity, and other factors, we found no relationship
between hospital volume of CABG surgeries and its riskadjusted in-hospital mortality rate.
Conclusions: Endogeneity and heterogeneity exist between
volume and quality. In order to reduce biases from these,
longitudinal data and instrument variable estimation are
recommended. Even though volume is easy to measure, it is
not a good proxy variable of quality of care. Volume is not a
major factor affecting quality of care for CABG surgery.
Implications for Policy, Delivery, or Practice: Current policy
in some states and recommendations from the employersponsored LeapFrog initiative recommend using volume
threshold for referral of CABG patients. Because we found that
volume is not related to quality of care, this policy needs to be
re-considered.
Primary Funding Source: No Funding
●A Comparison of Quality of Care in General Hospitals,
Specialty Hospitals, and Ambulatory Surgery Centers
Cheryl Fahlman, Ph.D., Phil Kletke, Ph.D., Jon Gabel, M.S., Joel
Hay, Ph.D., Chuck Wentworth, M.S.
Presented By: Cheryl Fahlman, Ph.D., Health Researcher II,
Center for Studying Health System Change, 600 Maryland
Avenue S.W., Suite 550, Washington, DC 20024;
Tel: (202)484-3094; Fax: (202)484-9258;
Email: CFahlman@hschange.org
Research Objective: To determine whether quality of care
differs among specialty hospitals, ambulatory surgical centers,
and general hospitals for similar procedures and diagnoses.
Study Design: We used claims data from 2002 and 2003 for a
large national group health plan to compare the quality of
inpatient care between specialty hospitals and general
hospitals and the quality of ambulatory care among ASCs,
specialty hospitals, and general hospitals. The plan enrolls
more than three million active workers, early retirees, and
dependents. The analysis is limited to beneficiaries, ages 18
to 64, residing in six states that have relatively large numbers
of specialty hospital or ASCs. The unit of analysis is the
episode of care, as defined by software developed by
Symmetry Health Data Systems. Inpatient care procedures
included: knee replacements (n=556); hip replacements
(n=1,005), coronary artery bypass grafts (n=843),
percutaneous coronary interventions (n=1,824), and heart
stents (n=135). Ambulatory care procedures included:
gastrointestinal procedures (n=49,914), back procedures
(n=2,652), and finger/hand/wrist procedures (n=4,009).
Quality of care was defined as the relative infrequency of
potentially adverse outcomes —- readmission rates, additional
surgery, surgical complications, and mortality rates. We used
logistic regression analyses to examine a dichotomous
dependent variable indicating whether a potentially adverse
outcome occurred during the episode of care. The regression
analyses controlled for severity of illness (using a risk adjuster
based on services and expenditures in the previous year),
demographic characteristics of the patient, and socioeconomic characteristics of the patient’s county of residence.
Population Studied: Members of a large national group
health plan, ages 18 to 64, residing in states with relatively
large numbers of specialty hospitals or ASCs.
Principal Findings: We found quality of care in specialty
hospitals and ASCs to be equivalent or superior to care in
general hospitals. The inpatient analysis found specialty
hospitals had lower rates of adverse outcomes. These
differences were generally not statistically significant when the
procedures were analyzed individually, due to the small
number of observations for specialty hospitals. However,
specialty hospitals had a significantly lower rate of adverse
outcomes when the three cardiac procedures were combined
into one analysis. The analysis of ambulatory care found that,
ASCs had 28 percent fewer adverse outcomes than general
hospitals for GI procedures; that differences between ASCs,
general hospitals, and specialty hospitals were not significant
for back procedures; and that findings were mixed for
finger/hand/wrist procedures.
Conclusions: We found that the quality of care in ASCs and
specialty hospitals is equal to or better than care in general
hospitals.
Implications for Policy, Delivery, or Practice: Our findings
support the proposition that specialization leads to better
outcomes. Policy makers need to weigh the advantages of
specialty hospitals and ASCs against the perceived
disadvantages.
Primary Funding Source: No Funding
●Development and Evaluation of Quality Indicators in the
Intensive Care Unit: Preliminary Results
Wilco Graafmans, Ph.D., Maartje de Vos, MSc, Gert Westert,
Ph.D., Peter van der Voort, M.D., Ph.D.
Presented By: Wilco Graafmans, Ph.D., Epidemiologist,
Centre for Prevention and Health Services Research, National
Institute for Public Health and the Environment, P.O. Box 1,
Bilthoven, 3720 BA; Tel: +31 30 274 2595; Fax: +31 30 274
4407; Email: wilco.graafmans@rivm.nl
Research Objective: The objective of this study was to
develop and implement a set indicators for quality of care in
intensive care units (ICU) in Dutch hospitals. These indicators
will serve to monitor and improve quality of care.
Study Design: To obtain a set of indicators for quality of care
in ICUs, we reviewed the literature and discussed the results
in expert meetings. A literature search was carried out to
obtain articles from the years 2000 to 2005. Publications were
included describing process or structure indicators in care,
which are associated with improved patient outcome. The
final selection of indicators was made by a consensus
procedure among members of the Dutch Society of Intensive
Care Medicine (NVIC). To evaluate the feasibility of the
registration of the identified indicators, a pilot study was
carried out during six months in 2005. Registration was
electronically, using an application on the personal computer.
The use of the indicators was evaluated by means of
questionnaires and interviews with two representatives of the
participating ICUs. Among other issues, the information
focused on workload and perceived validity and reliability.
Population Studied: ICUs of 18 hospitals in the Netherlands:
one categorical hospital, nine teaching hospitals and eight
non-teaching hospitals.
Principal Findings: Twelve indicators were selected to
measure the quality of care regarding structure, process, and
outcome. The following indicators were selected for the
structure: 1) intensivist availability, 2) nurse-patient ratio, 3)
medication error prevention policy, and 4) registration of
patient/family satisfaction. Selected process indicators were:
5) ICU length of stay, 6) duration of mechanical ventilation, 7)
inter-clinical transport, 8) days of 100% bed occupation, and
9) glucose regulation. The outcome indicators selected were:
10) mortality, 11) severe decubitus and 12) unplanned
extubation. Investment of time for registration of the
indicators varied from less than thirty minutes a day (46% of
the respondents) to more than sixty minutes (17%). This
workload was considered acceptable by 86% of the
respondents. The indicators ‘ICU length of stay’,‘unplanned
extubation’, and 'mortality' were not considered to strongly
represent the quality of care by at least 30% of respondents.
Reliability of registration of ‘unplanned extubation’ was
considered to be inadequate. More than 80% of respondents
supported further implementation of 9 indicators of the initial
set.
Conclusions: A set of 12 quality indicators was defined for the
ICU. In general, the registration was considered feasible and
desirable. Nine indicators of this set were considered
applicable for future implementation.
Implications for Policy, Delivery, or Practice: Based on
these results and future analyses of validity and reliability, a
final set of quality indicators in ICUs in the Netherlands will be
determined. The Dutch Society of Intensive Care Medicine
aims to implement this set of indicators in all Dutch hospitals
in the coming years.
Primary Funding Source: Dutch Health Care Inspectorate
and the Dutch Order of Medical Specialists
●Surgery Volume and Mortality: A Re-Examination Using
Fixed-Effects Regression
Amresh Hanchate, Ph.D., Arlene S. Ash, Ph.D.
Presented By: Amresh Hanchate, Ph.D., Research Associate,
General Internal Medicine, Boston University School of
Medicine, 720, Harrison Ave, DOB, Suite 1108, Boston, MA
02118; Tel: (617) 638 8889; Fax: (617) 638 8026;
Email: hanchate@bu.edu
Research Objective: To estimate the independent effects of
surgeon and hospital volumes on operative mortality using
less restrictive statistical methods than primarily used in the
literature.
Study Design: The voluminous literature on this issue
overwhelmingly finds that higher surgeon and hospital
volumes are associated with lower operative mortality.
However, methodologically, this research strand rests on
assumptions that may be untenable. Given nested data
(patients, surgeons, hospitals), while multilevel regression
methods are appropriate, most of existing studies are based
on the unrealistic assumption of randomness of unobserved
provider effects (random effects regression). This arises from
the fact that if there is indeed a protective volume effect, then
a strategic response from decision makers at low volume
hospitals would be to respond aggressively with other counter
measures to keep mortality low. Our objective is to re-estimate
this relationship using a less restrictive multilevel model (fixed
effects regression). The surgical procedures studied are
abdominal aortic aneurysm (AAA) and coronary artery bypass
grafting (CABG). Using fixed effects regression the effects of
the two volume measures (surgeon and hospital) are
estimated separately. Almost all CABG and AAA surgeries
were performed in hospitals with two or more surgeons –
consequently, fixed hospital effect regression enables
unbiased estimation of surgeon volume effect even when
unobserved hospital effects are correlated with regression
covariates. Estimating the hospital volume effect pursues an
analogous approach by limiting analysis to surgeons who
practice in two or more hospitals. Operative mortality rate per
100 surgeries for each surgeon in each hospital is the
outcome measure. Covariates are patient demographics,
socioeconomic and illness profile, and hospital characteristics
(ownership, teaching status).
Population Studied: Our data is based on 100 percent of
inpatient Medicare Fee for Service data for 1998 and 1999. A
total of 220,592 CABG surgeries are aggregated at the surgeon
level (N=2,772) within each of 958 hospitals. For AAA, 39,794
surgeries were grouped into 6,276 surgeon records across
2,202 hospitals.
Principal Findings: For CABG surgery, the effects of surgeon
and hospital volumes from fixed effects (FE) regression are
different from those from random effects (RE) regression both
qualitatively as well as in magnitude. While FE indicates
protective effects of both volumes, RE estimates do not find
hospital volume to be protective. Also the protective effects
are steeper from FE than from RE. For AAA, both models find
protective effect of only surgeon volume, and again the effect
is steeper from FE than from RE. Estimates of protective
effect from unobserved hospital effects are inversely correlated
with hospital volume.
Conclusions: While higher volume is one source of reducing
operative mortality, our study finds evidence of other
protective measures. Actual mortality in low volume hospitals
were systematically lower than what would be expected based
on their volumes alone.
Implications for Policy, Delivery, or Practice: There has
been a concerted move from consumer advocacy groups,
health insurance coalitions (Leapfrog) and state agencies, all
aimed at transferring patients away from low-volume
providers. Our findings suggest a tempering of the singular
focus on volumes, and expanding the search for surgeryrelated patient safety interventions.
Primary Funding Source: AHRQ
●Therapy Disruptions in ICU Settings: Implications for
Quality Initiatives
Lorraine Mion, Ph.D., RN, Ann F. Minnick, Ph.D., RN, Cathy
Catrambone, DNSc, RN, Mary Johnson, Ph.D., RN, Rosanne
Leipzig, M.D., Ph.D.
Presented By: Lorraine Mion, Ph.D., RN, Director, Nursing
Research and Geriatric Nursing, Nursing, MetroHealth
Medical Center, 2500 MetroHealth Drive, Cleveland, OH
44109; Tel: (216) 778-4412; Fax: (216) 778-3939; Email:
lmion@metrohealth.org
Research Objective: Little is known about the rate, contexts
or consequences of patient-initiated therapy disruption in
ICUs; most studies have focused only on self-extubation.
Although clinicians typically use physical restraint (PR) to
prevent premature therapy disruption, the relationship of PR
use to various types of therapy disruptions is unknown.
Information is needed to help guide quality initiatives in ICUs
addressing the federal regulations aimed at minimizing PR.
The research objective was to determine the scope and
consequences of patient-initiated therapy disruption in ICUs
and relation to PR use.
Study Design: A prospective prevalence study was used in 49
ICUs from a random sample of 40 hospitals from 5 states.
Data on PR use and therapy disruptions were collected daily
(range 30 – 90 days). All patients were directly observed by
trained RN data collectors for presence of PR. Therapy
disruption was ascertained in three ways: nurse-reports on
specially designed forms, daily chart audits, and daily nurse
interviews. A positive response to any of the three methods
constituted a therapy disruption event. For each reported
event, data were collected on time of event, age, gender, use
of PR, sedation, harm as a direct consequence of the
premature disruption, whether the device was reinstated, and
whether additional treatments or procedures were required.
Population Studied: All patients in the study ICUs during the
data collection period.
Principal Findings: One or more devices were disrupted on
1165 occasions out of the 49,002 patient days (overall rate
23.8/1000 patient-days; ICU range: 0 – 102.4/1000 patientdays). At the time of the event, 44% were in PR. Therapy
disruption events were weakly correlated with unit census (r =
0.15) and number of patients in PR (r = 0.10). Harm occurred
in 263 (23%) of therapy disruption events: 185 (16%) minor
harm (e.g., ecchymosis), 66 (6%) moderate harm (e.g.,
sutures), and 12 (1%) serious harm (e.g., fractures). No
deaths resulted. Older adults (> 64 years) accounted for 63%
of the disruption events; men accounted for 57%; half the
occurrences were on day shift. There were 1620 disrupted
devices/49,002 patient-days (33.1 devices /1000 patient-days).
Most frequently disrupted devices were nasogastric tubes
(27%), oxygen (22%), peripheral intravaneous lines (IV)
(20%), monitor leads (19%), and endotracheal tubes (ETT)
(15%): Reinstating the device varied by type of device: monitor
leads (90%), oxygen (86%), IVs (82%), bladder catheters
(77%), CPAP (76%), ETT (49%), all others less than 40%.
Additional resources or treatments (e.g., x-rays, surgical
procedures) were done in 58%.
Conclusions: Patient-initiated therapy disruptions are
relatively common, but rates vary substantially. Use of PR
does not seem to impact the occurrence of therapy
disruptions. Significant hospital resources are expended when
patients prematurely disrupt therapy and a portion of patients
suffer some degree of harm.
Implications for Policy, Delivery, or Practice: Given the
frequency, the potential harm, and need for additional
treatments, further studies are needed to quantify the costs
related to therapy disruptions and to guide quality initiatives
implementing safe restraint reduction protocols in ICUs.
Primary Funding Source: NIA
Call for Papers
Quality: Coordination & Transitions
Chair: Gregory Pawlson, National Committee for
Quality Assurance
Monday, June 26 • 2:00 pm – 3:30 pm
●Measuring Coordination of Care for Children with Special
Health Care Needs: Alternative Methods and Findings in
National and State Level Surveys
Christina Bethell, Ph.D., MBA, M.P.H., Debra Read, M.P.H.
Presented By: Christina Bethell, Ph.D., MBA, M.P.H.,
Associate Professor, Pediatrics, Oregon Health and Science
University, Mailcode CDRCP, 707 SW Gaines Street, Portland,
97239-3098; Tel: 503-494-1892; Fax: 503-494-2475;
Email: bethellc@ohsu.edu
Research Objective: To compare care coordination (CC)
measurement methods and findings for children with special
health care needs (CSHCN) in national and state level
surveys.
Study Design: CC survey items were identified and measures
constructed using data from the National Survey of Children
with Special Health Care Needs (NS-CSHCN), the National
Survey of Children's Health (NSCH) and the Consumer
Assessment of Health Plans Survey-Children with Chronic
Conditions (CCC). Methods and findings were compared both
conceptually and statistically for all and subgroups of CSHCN,
who were identified using the same CSHCN Screener across
all data sets.
Population Studied: Nationally representative samples of
children from the NS-CSHCN (n=372,174) and NSCH (n =
102,353) and children represented in CCC data in one state
Medicaid program (n=10,792)
Principal Findings: The NS-CSHCN includes 8 items relevant
to CC focused on parent perceived need for and provision of
professional CC. The NSCH includes 9 items focused on
whether children get help accessing needed care from
specialists and/or other special services when problems arise
and also receive follow-up from their personal doctor or nurse
(PDN). The CCC focuses on whether children receiving care
from more than one provider get CC assistance, a method
used to assess CC in the upcoming 2005-2006 NS-CSHCN.
In the NSCH over 60% of CSHCN with a PDN needed
specialist and/or other special care in the past year.
Depending on the specialized service needed, 16%-21% had
problems accessing care and 52%-70% of these got help from
their PDN to access care and 50%-61%% received follow-up
from their PDN after specialized services. In the CCC, 56% of
CSHCN needed care from more than one provider and 55.672.5% of CSHCN across 14 health plans received CC help.
CSHCN with a PDN were more likely to have received CC (OR
1.89). In the NS-CSHCN only 11% of CSHCN had parents
reporting needing professional CC. CSHCN experiencing
emotional or behavioral problems were more likely to report
such a need (20.5%) as were children with five or more service
needs (18.1%). About 20% did not receive needed
professional CC and 30% needing a professional CC said they
usually or always received it. Nearly 50% said the professional
care coordinator operated out of their child's primary care
provider's office. Wide, significant variations exist across
states and sociodemographic, insurance and health status
subgroups of CSHCN for NSCH and NS-CSHCN findings.
Conclusions: Measurement of CC for CSHCN varies by
whether professional or primary-care based CC is evaluated
and whether objective vs. subjective screening criteria
determine the need for care coordination. Methods also vary
in terms of whether continuity of care and having a personal
doctor or nurse are incorporated into the concept of CC as
well as whether both help accessing care and integrating care
are included. Regardless of methods used, a large proportion
of CSHCN who need CC appear to have some unmet needs
in this area.
Implications for Policy, Delivery, or Practice: Given the
importance of CC to the quality of care for CSHCN, advances
can and should be made to further validate and align
measurement methods.
Primary Funding Source: HRSA
●Cost Effectiveness of Hernia Surgery:Implications for
Practice
Denise Hynes, Ph.D., M.P.H., RN, Kevin T Stroupe, Ph.D.,
Ping Luo, Ph.D., Anita Giobbie-Hurder, MS, Domenic Reda,
Ph.D., Leigh Neumayer, M.D.
Presented By: Denise Hynes, Ph.D., M.P.H., RN, Research
Health Scientist, VA Information Resource Center,
Department of Veterans Affairs, Edward Hines Jr VA Hospital,
5th and Roosevelt Roads, PO Box 5000 (151V), Hines, IL
60141; Tel: (708) 202-2413; Fax: (708) 202-2415;
Email: denise.hynes@va.gov
Research Objective: To estimate two-year costs, benefits, and
cost effectiveness of two types of inguinal hernia repair in an
outpatient setting to further assist physicians, patients and
insurers in making informed risk/benefit decisions.
Study Design: Randomized controlled trial of open (OPEN)
versus laparoscopic (LAP) hernia repair using mesh
conducted at fourteen Department of Veterans Affairs
medical centers including economic and quality of life data
using an intent-to-treat analysis with two-year follow-up.
Specific outcome measures included surgical and
postoperative costs, quality adjusted life years (QALY), and
incremental cost per QALY gained or the incremental cost
effectiveness ratios (ICER)
Population Studied: 2,164 men with inguinal hernia were
randomized, 1,395 patients (708 OPEN and 687 LAP) with
outpatient hernia operations were included in the costeffectiveness analysis.
Principal Findings: Over two years, LAP costs an average of
$638 more than OPEN. QALYs at two years were similar,
resulting in $45,899 per QALY gained (95% CI: -$669,045,
$722,457). The probability that LAP is cost effective at the
$50,000 per QALY level (slightly more costly but more
effective), was 51%. For unilateral primary and unilateral
recurrent hernia repair, the probability that LAP was cost
effective at the $50,000 per QALY level was 64% and 81%,
respectively. For bilateral hernia repair OPEN was less costly
and more effective.
Conclusions: Overall, laparoscopic hernia repair is not cost
effective compared with open repair. However, for patients
with unilateral (primary or recurrent) hernia, laparoscopic
repair is a cost effective treatment option.
Implications for Policy, Delivery, or Practice: Costeffectiveness analysis examines a different dimension than the
traditional surgical outcomes measures of morbidity and
mortality, and in the case of hernia repair, the rate of
recurrence. Cost-effectiveness analysis highlights the
importance of patient-reported outcomes and costs when
evaluating surgical procedures. Because laparoscopic repair is
not cost effective for all patients, surgeons and patients
should carefully consider the patient-specific benefits and risks
of open versus laparoscopic hernia repair. While we found
that laparoscopic repair is cost-effective for unilateral recurrent
hernias, we were unable to support a recommendation for
laparoscopic repair of bilateral hernias on the basis of costeffectiveness. Consideration of patient centered outcomes
and cost outcomes rather than traditional morbidity and
mortality measures alone, is important.
Primary Funding Source: VA
●Primary Care Teams: Effects on the Quality of ClinicianPatient Interactions and Patients’ Primary Care
Experiences
Hector Rodriguez, M.P.H., William H. Rogers, Ph.D., Richard
E. Marshall, M.D., Dana Gelb Safran, Sc.D.
Presented By: Hector Rodriguez, M.P.H., Pre-Doctoral Fellow,
Institute for Clinical Research and Health Policy Studies, TuftsNew England Medical Center, 750 Washington Street, Box 345,
Boston, MA 02111; Tel: (617) 636-5751; Fax: (617) 636-5000;
Email: hrodrig@fas.harvard.edu
Research Objective: Multidisciplinary care teams are
increasingly seen as important for advancing the quality of
chronic disease management and primary care practice
generally. This study examines the influence of primary care
teams on patients’ reported experiences in their primary care
practices, including the quality of clinician-patient interactions.
Study Design: Patients’ primary care experiences were
assessed using the Ambulatory Care Experiences Survey
(ACES), a well-validated survey comprised of 9 summary
measures across two domains: clinician-patient interaction
quality and organizational features of care. Using
administrative data, standardized indices of visit continuity
were calculated, indicating the type and extent of each
patient’s continuity over 6 months preceding the survey.
Information on team composition (primary care physician
[PCP] matched with specific physician assistants, nurse
practitioners, and registered nurses) was used to classify each
primary care visit as either PCP, on-team, or off-team. For
each ACES measure, regression models controlling for patient
characteristics and utilization, evaluated the effects of
continuity on patients’ reported primary care experiences,
including the quality of interactions with their PCP and other
team members.
Population Studied: From March 2004 through March 2005,
a large multispecialty practice in Massachusetts administered
surveys monthly to a random sample of patients visiting each
of 145 primary care physicians. Eligible patients were those
with at least one visit to their PCP the month prior. Our
analytic sample includes 14,835 patients (average per
physician=102) with 2 or more primary care visits over the
prior 6 months.
Principal Findings: Among patients with 2 or more visits,
35% saw only their PCP, 15% had only “on-team” visits (PCP
and team members), 9% had both on- and off-team visits,
and the remainder (41%) had only “off-team” visits when not
seeing their PCP. Higher PCP continuity was associated with
more favorable assessment of physician-patient interactions,
including communication, knowledge of the patient, health
promotion, and patient willingness to recommend the
physician. Effects ranged from 0.6-1.9 points for every
standard deviation increase in PCP continuity (p<.001 for all
measures). Patients’ assessments of the clinical team were
significantly better for those with “on-team” vs. “off-team”
visits (1.5 points, p<.01). However, for all other ACES
measures, the effect of PCP discontinuity was the same for
patients with on-team vs. off-team visits.
Conclusions: Our findings suggest that visit discontinuities
between patients and their PCPs are associated with a
decrement in patients’ assessments of their care, irrespective
of whether those discontinuities involve visits to clinicians
who are formally part of the team vs. others in the practice.
The sole exception to this involved patients’ assessments of
their team interactions, where on-team visits were associated
with more favorable assessments than off-team visits.
Implications for Policy, Delivery, or Practice: The findings
highlight the challenges of incorporating teams into primary
care practice in ways that positively affect patients’ overall
experiences and don’t impede strong PCP-patient
relationships. The finding that on-team vs. off-team visits
were equally negative in their effects on most aspects of
patients’ experiences suggests the need for improving the
coherence and value of team approaches from patients’
perspectives.
Primary Funding Source: No Funding
●Nursing Characteristics and Patient Outcomes: A Multilevel Model
Jean Ann Seago, Ph.D.
Presented By: Jean Ann Seago, Ph.D., Associate Professor,
Community Health Systems, University of California, San
Francisco, 2 Koret Way, Rm N505, San Francisco, CA 941430608; Tel: 415-502-6340; Fax: 415-476-6042;
Email: jean.ann.seago@nursing.ucsf.edu
Research Objective: Although many studies have
demonstrated a relationship between nurse staffing and
patient outcomes, there is a need for more complete models
of prediction.The purpose of this study was to investigate the
relationship between nurse perception of autonomy, control
over practice, and relationships with physicians with patient
satisfaction with pain management, teaching, and physical
care on medical surgical units, after controlling for potential
patient, nurse, unit, and hospital confounders.
Study Design: The study design was correlational, descriptive,
and cross sectional. The sample was a convenience sample of
acute care hospitals in California; however every effort was
made to include both rural and urban hospitals and teaching
and community hospitals. Data were collected by patient
structured interview, nurse survey, chart review, and using
administrative databases. The study questions were addressed
using multiple linear regression adjusted for robust standard
errors and clustering and each outcome variable was assessed
using four models. Unit level nurse staffing was a control
variable and can be measured in several ways. Two of the
most common methods, skill mix with total nursing hours per
patient day (HPPD) and RN HPPD with other HPPD, were
used to estimate models for each outcome. Model one used
skill mix with total hours per patient day and model two used
RN and other hours per patient day. Additionally, it was
anticipated that there would likely be a hospital effect, so
models were estimated both without (model three) and with
(model four) dummy variables for the 21 hospitals in the
study.
Population Studied: medical surgical hospital inpatients
Principal Findings: For the estimate of satisfaction with pain
management, without hospital dummies the only significant
predictor was higher patient functional status. For patient
satisfaction with teaching, lower nurse perception of
autonomy is significantly associated with higher satisfaction
with teaching in all the models, and lower scores of
collaboration with physicians was predictive of higher
satisfaction with teaching in one model. The satisfaction with
physical care none of the predictors of interest were
significant.
Conclusions: One of the most interesting findings of this
study is what is not significant in the models, that is skill mix,
RN HPPD, and total HPPD are not significant predictors.
Another finding that seems counterintuitive and frustrating is
that lower autonomy scores and, in one case, lower
collaboration scores predict higher satisfaction with patient
teaching. These are the only significant findings in the models
for the variables of interest, control of practice, autonomy, and
relationships/collaboration with physicians. It is notable that
for satisfaction with pain management and teaching, there are
several large hospital effects. Therefore, it was concluded that
there are important predictors not in the model that are
specific hospital characteristics that would explain these
hospital effects. The results were disappointing but may prove
useful in stimulating a dialogue about the work of nurses in
hospitals.
Implications for Policy, Delivery, or Practice: It is imperative
that further research be done to explore the mechanism by
which nursing care is related to positive hospital patient
outcomes.
Primary Funding Source: University of California, San
Francisco Academic Senate and School of Nursing Research
Committee
●Successful Governing Models for Physician Groups
Leading to Improved Quality Performance
Amy Smalarz, Ph.D.
Presented By: Amy Smalarz, Ph.D., Health Science Specialist,
Center for Organization, Leadership & Management Research,
VA, 13 Edith Rd, Framingham, MA 01701; Tel: 857-364-2625;
Email: amy.smalarz@gmail.com
Research Objective: This study examined 50 physician
groups and their office managers/administrators to answer
the following research questions: What is the effect of
physician groups’ cultural dimensions and structural
characteristics on specific quality of care outcomes, as
specified by HEDIS measures? Are there best performers
among the physician groups? Do commonalities exist among
the best performers from which other physician groups can
learn from and emulate?
Study Design: The first step involved 1) surveying the
physicians using a cultural dimension survey instrument
(Kralewski et al., 2005) and 2) surveying the office
managers/administrators about the structural characteristics
of their group practice, using an instrument which was
developed for this project. The second step included analyzing
physician group performance based on eight specified quality
of care performance measures from an insurer’s claims
database. The third step linked the performance results to the
cultural dimension and structural characteristic surveys to
determine the extent to which cultural dimensions and
structural characteristics affect specified patient outcomes.
Both linear regression (OLS) and Data Envelopment Analysis
were performed. Unlike OLS, which measures average
performance, DEA identifies best performance, in essence
creating a “frontier” representing the best performers.
Population Studied: The population for this study consisted
of 1,236 physicians composing 57 physician group practices in
the state of Massachusetts, as well as their respective office
managers/administrators. The final sample size was 734
physicians from physician group practices and their
corresponding office managers/administrators. Of the 50
physician group studied, 15 were members of a larger health
care organization, with independent practice sites and the
remaining 35 were independent group practices. Quality
performance data for all eight quality measures was provided
for each of the 50 physician groups.
Principal Findings: Overall, based on the linear regression
results, the cultural dimensions and structural characteristics
included in the analyses explained 17% of the variation in
diabetic eye exam rates, 32% of the variation in HbA1c control
rates, 42% of variation of cholesterol management rates, 26%
of the variation of Chlamydia screening, 10% of Diabetes
Nephropathy monitoring and 15% of variation of adolescent
well visits. The DEA analysis revealed that there were nine
physician groups, out of the 50 studied, that were members of
the “frontier,” i.e. best performers. The nine best performers
were statistically significantly different from the rest for each
cultural dimension. There were no differences among
structural characteristics between the physician groups on the
frontier and the remaining physician groups. Two different
governing models of physician groups, professional and
administrative, were found upon further examination. Four of
the nine physician groups, which were independent practices,
represent the professional model. They emphasize
organizational trust/identity, business, innovativeness and
autonomy. The other five physician groups, which are
members of a larger healthcare organization, represent the
administrative model. They emphasize information,
standardization and formal structures.
Conclusions: Whether analyzing performance reporting
results for average performers or for best performers, the
cultural dimensions present in physician groups are
important. Both analyses demonstrate that cultural
dimensions affect physician groups’ quality performance. In
addition, with the existence of two successful governing
models, these results demonstrate that there is more than one
model that can lead to successful quality performance by
practicing physician groups.
Implications for Policy, Delivery, or Practice: This study
provides empirical evidence that there is more than one
model for a physician group to emulate in order to improve its
quality performance. Further study is warranted into whether
the governing models are related to the types of ownership
among physician organizations. Evidence from this study
demonstrates that DEA analysis may be a better tool for
identifying best performance than OLS analyses. Using a tool
that is meant to measure best performers and that can create
a “frontier” of quality may be more useful than using a tool
that is meant to find variables associated with average
performers, especially when it comes to measuring physician
performance with the various pay-for-performance models
being used today.
Primary Funding Source: AHRQ
Call for Papers
Patient Safety & Methodological Issues
Chair: Dana Gelb Safran, Tufts- New England Medical Center
Tuesday, June 27 • 10:30 am – 12:00 pm
●Validating and Reporting Quality Performance
Measurement Models with Multilevel Analysis
Ronald Fisher, Ph.D. (candidate)
Presented By: Ronald Fisher, Ph.D. (candidate), Research
associate, Health Administration, Virginia Commonwealth
University, P. O. Box 980203, Richmond, VA 23298-0203; Tel:
804-873-7844; Fax: 804-828-1894; Email: rlfisher@vcu.edu
Research Objective: Health care research is an applied field
challenged by complex and nested sources of variability.
Measurement observations often need to be connected across
unit of analysis boundaries: e.g., patients under the treatment
of clinicians, patient events within hospitals or other facilities,
clinician performance within (and across) facilities,
longitudinal measurements of subjects, etc. There is a need to
model relevant coefficients, such as risk-adjustment at the
patient-level, and variance components at each level in order
to achieve unbiased assessment at the aggregate performance
level. The objective of this methodological study is to
demonstrate the advantages in applying multilevel analysis
when compared to more traditional fixed-effected regression
techniques.
Study Design: A four year longitudinal multilevel design was
employed to model relationships between hospital quality and
economic performance. A first stage analysis was conducted
to develop a quality performance measurement model of riskadjusted patient outcomes with patients nested within APRDRG diagnostic group severity categories and hospitals. The
“goodness” of the risk-adjustment methodology was assessed
by analysis of the variance accounted for by the patient-level
risk factors, and whether the remaining between hospital
variance component was significant (e.g., that the hospital
specific effect remained significant after the risk-adjustment
factors were accounted for). The residual log-odds of the
hospital specific effects produced in this measurement model
stage were then used in longitudinal analysis to model
hospital-level economic performance and market-area factors
on hospital quality performance. Economic performance
measures and organizational characteristics were used to
model hospital-level effects on quality outputs. Local market
structure factors of competition, health plan purchasing
leverage, and market wage index where also modeled within
the hierarchical analysis.
Population Studied: Acute hospitals operating in Virginia
during 1998 and 2001.
Principal Findings: Only five of the seventeen risk-adjusted
patent-level indicators where found acceptable in the stageone analysis. The five included four AHRQ patient safety
measures (decubitus ulcer, failure to rescue, infection due to
medical care, accidental puncture or laceration) and a rather
global, if qualified, mortality rate. At stage-two, significant
between hospital variance was demonstrated, with hospital
specific effects accounting for approximately 50% of the total
variance of the quality performance measures. No significant
trend of quality improvement was found, though the trajectory
of two error rates indicated a general lowering over time.
Comparisons between economic and market dynamic models
and “empty models” did not reveal any expected market effect
on the variance component of the growth curve. Some
hospital-level fixed-effects were found to be positively
associated with adverse events, though none were predictive
of “expected” pricing mechanisms. A market-level effect for
competition was found to be associated with better quality
performance in the two “error” rate PSIs. Comparative
differences between the results obtained from traditional
regression techniques and the multilevel analysis were
substantive.
Conclusions: Discriminate validity of the measurement
model for hospital quality was provided via the multilevel
technique, and for which there is no comparable inference
from traditional analysis. Multilevel analysis allows for an
unbalanced design, providing for more complete use of
available information. Estimation of the multilevel empirical
Bayes coefficients is comparatively different from traditional
analytic techniques, and represents a more realistic and
accurate approach to model fitting.
Primary Funding Source: No Funding
●Consequences and Costs of Medical Injuries in Medicare
Inpatients
David A. Foster, Ph.D., M.P.H.
Presented By: David A. Foster, Ph.D., M.P.H., Chief Scientist,
Center for Healthcare Improvement, Solucient, LLC, 5400
Data Ct, Ann Arbor, MI 48118; Tel: (734) 669-7982; Fax: (734)
930-7611; Email: dfoster@solucient.com
Research Objective: To evaluate the incremental
consequences of specific inpatient medical injuries in terms of
mortality, length of stay, and cost per case among Medicare
beneficiaries, and to estimate the national impact of such
injuries during hospitalization.
Study Design: The Agency for Healthcare Research and
Quality (AHRQ) Patient Safety Indicators (PSIs) were used to
identify selected medical and surgical injuries in about 20
million hospital discharge records from over 5,000 acute care
hospitals in the CMS Medicare Provider Analysis and Review
(MedPAR) data files for federal fiscal years 2003 and 2004.
Propensity score analyses were conducted to obtain riskadjusted results. Unadjusted results were also reported.
Population Studied: Medicare beneficiaries who were 65 or
more years of age and were medical or surgical inpatients in
short-term, acute-care hospitals in federal fiscal years 2003 or
2004 were included.
Principal Findings: The consequences of potentially
preventable adverse outcomes in Medicare beneficiaries can
be very severe. In all PSIs studied, risk of death, LOS, and
estimated hospital cost per case was significantly increased
among patients with the PSI adverse outcome over similar
patients who did not experience the outcome. For several
PSIs the risk of death among patients with an adverse
outcome was increased by more than 20-fold, length of stay
was increased by as much as two weeks, and increases in
average hospital cost per case were as high as $30,000 over
similar patients who did experience the outcome. Unadjusted
and risk-adjusted results were alike in direction and
magnitude. In just one of the most serious PSIs studied,
postoperative respiratory failure, it was estimated that 4,122
deaths, more than 212,000 days of stay, and almost $444
million in total estimated hospital costs were attributable to
the outcome over the two-year study period. Risk-adjusted
rates of the PSIs varied considerably across hospitals and
geographic regions.
Conclusions: Adverse outcomes due to medical injuries that
are potentially preventable are responsible for substantial
amounts of excess death, days of stay, and hospital costs.
High levels of variability exists across hospitals and
geographic regions in the risk-adjusted rates of these patient
safety outcomes, therefore it is reasonable to believe that
considerable room for improvement exists.
Implications for Policy, Delivery, or Practice: Potentially
preventable medical injuries contribute substantially to excess
mortality and cost of care of Medicare inpatients. It is likely
that medical injuries impact the care of non-Medicare
inpatients to a similar degree. More and better research is
needed to identify endogenous factors which hospitals can act
upon to reduce the incidence of preventable medical injuries.
Primary Funding Source: Solucient, LLC
●Socioeconomic Status and the Prevention Quality
Indicators
Jeffrey Geppert, JD EdM, Sheryl Davies, MA, Corinna
Haberland, M.D., Amy Ku, MHSA, Kathryn McDonald, MM,
Patrick Romano, M.D.
Presented By: Jeffrey Geppert, JD EdM, Senior Health
Reseach Scientist, Centers for Public Health Research and
Evaluation, Battelle Memorial Institute, 8142 Ardenness Drive,
Sacramento, CA 95829-6504; Tel: (916) 682-9965; Fax: (703)
527-5640; Email: geppertj@battelle.org
Research Objective: The AHRQ Prevention Quality Indicators
(PQI) measure hospitalization rates for ambulatory care
sensitive conditions such as asthma, congestive heart failure
and diabetes. The PQI are indices of potential problems in
access to high quality ambulatory care. Socioeconomic status
has been found in many research studies to be a confounding
factor independent of access. However, the lack of a widely
available measure of socioeconomic status hampers our
ability to account for this factor. This study reports on a
potential measure that could be used to adjust PQI rates for
differences in socioeconomic status across all counties in the
U.S.
Study Design: We used findings from the Public Health
Disparities Geo-coding Project to construct a measure of
socioeconomic status based on the percent of persons living
below the U.S. federal poverty level. Using county level
poverty estimates and population counts from the U.S Census
Bureau, we assigned counties to deciles with poverty rates
that varied from 5.7% in the lowest poverty areas to 23.6% in
the highest poverty areas. We then computed county level
PQI rates based on the location of the patient residence, and
compared residual (after age and sex adjustment) rates across
the poverty level deciles.
Population Studied: We used all-payer state hospital
discharge data from the AHRQ Healthcare Cost and
Utilization Project (HCUP) from 38 states and population data
from the U.S. Census Bureau for 2,500 counties in the U.S. for
calendar year 2003. These data include approximately 90% of
all hospital discharges in the nation.
Principal Findings: Of the 16 PQI included in this study, 15
were significantly and monotonically related to the poverty
level of the county, with higher hospitalization rates in
counties with higher poverty rates. On average, rates were
90% greater in the highest poverty areas than in the lowest
poverty areas. The increase in the hospitalization rate was
greatest for pediatric asthma (105 to 352 per 100,000),
pediatric gastroenteritis (72 to 240) and hypertension (27 to
89). The one exception was perforated appendix, which is the
rate of ruptured appendices among those hospitalized with
acute appendicitis. Recent literature suggests a plausible
rationale for the lack of association, given the importance of
timely access to surgery and the tendency of lower income
patients to seek primary treatment in hospital Emergency
Departments.
Conclusions: The county level percent of persons living below
the federal poverty line is strongly associated with rates of
potentially avoidable hospitalizations, which could either
mean that poverty status by itself is an important factor in
access to high quality ambulatory care, or that poverty status
is a potential proxy for a broader range of socioeconomic
characteristics.
Implications for Policy, Delivery, or Practice: Accounting for
area level differences in socioeconomic status as a
confounding factor can help to more accurate identify the
determinants of inequalities in access to high quality
ambulatory care. The relationship between poverty and health
status also has intrinsic interest to public health. A simple
measure of socioeconomic status based on poverty level
discriminated well and has direct policy relevance as a
criterion of eligibility for federal and state anti-poverty
programs.
Primary Funding Source: AHRQ,
●Toward a Nationally Standardized Methodology
Joachim Roski, Ph.D. M.P.H., Dan Dunn, Ph.D., Sally
Turbyville, MA, Kim Sanborn, MA, David Knutson, MS
Presented By: Joachim Roski, Ph.D. M.P.H., Vice-President,
Performance Msmt, Research & Contracting, NcQA, 2000 L
St NW, Ste 500, Washington, DC 20036; Tel: 202.955.5139;
Fax: 202.955.3599; Email: roski@ncqa.org
Research Objective: To develop and initially test a
transparent, standardized, meaningful, methodologically
sound, and feasible approach to measuring resources
expended to achieve quality outputs by managed care
organizations (MCOs).
Study Design: This was a cross-sectional study.Resource use
was calculated using medical and pharmacy claims and
enrollment data for members with cardiovascular disease,
diabetes, asthma/chronic obstructive pulmonary disease
(COPD), or arthritis/low back pain (LBP). Relative resource
use was captured using a standardized costing methodology
that included inpatient, pharmacy, evaluation and
management, and procedural services. Results were
calculated overall, and by clinical condition comparing two
risk-adjustment approaches. Morbidity adjustment was
applied to ensure valid comparisons across MCOs.
Population Studied: Twelve commercial populations,
representing a mix of HMO, PPO, and POS products from
accross the US. MCO enrollment for these populations
ranged from less than 250,000 to over 500,000 members.
Total study population exceeded 7 million individuals.
Principal Findings: Total resource varied substantially
between MCOs. Prevalence of clinical conditions studied were
similar across populations. MCO resource use compared
similarly to other MCOs accross clinical conditions. Resource
use between MCOs differed by up to 30% for all clinical
conditions under study as well as overall. Risk adjustment
based on an age-gender morbidity model and a proprietary
population risk adjustment system yielded similar findings.
Conclusions: A methodologically robust, standardized
approach to measure MCO relative resource use was
developed. When linked to measures of quality of care, results
of this measurement approach can provide insights into the
efficiency or value of care rendered by MCOs. Identified
measures and methods may lend themselves to be integrated
into public performance reporting efforts of health plans.
Implications for Policy, Delivery, or Practice: As the delivery
of high-value care (quality/cost) becomes a critical policy
concern, these methods may provide a critical instrument to
judge the effectiveness of quality and resource use
management in managed care organizations.
Primary Funding Source: No Funding
●Assessing Patient Safety: Potentially Harmful DrugDisease Interactions in the Elderly
Sarah Sampsel, M.P.H., Russell E. Mardon, Ph.D., Philip
Renner, MBA, Lok Wong, MHS
Presented By: Sarah Sampsel, M.P.H., Senior Health Care
Analyst, Quality Measurement, NCQA, 2000 L St., NW #500,
Washington, DC 20036; Tel: 505-986-9848; Fax: 202-955-3599;
Email: sampsel@ncqa.org
Research Objective: To evaluate whether elderly patients are
prescribed medication(s) contraindicated for use in people
with the specified condition(s). Many older persons take
multiple drugs for the treatment of several conditions
increasing the chance of adverse reactions including drug–
disease interactions.
Study Design: Retrospective analysis of claims/encounter and
pharmacy data from 4 Medicare Advantage Health Plans.
Drug-disease interactions included in this study were
identified using the Beers Criteria for Potentially Inappropriate
Medication Use in Older Adults and the Canadian Criteria
Defining Inappropriate Practices in Prescribing for Elderly
People. Percentages of enrollees with at least one of the
specified conditions and use of at least one contraindicated
medication were calculated and reported by plan, age, gender,
prescribing provider specialty and condition.
Population Studied: The field test sites submitted data from
a membership total of 45,461 Medicare Advantage enrollees
65 years and older. Health plans submitted data for Medicare
enrollees with evidence of any of the following conditions:
benign prostatic hypertrophy, history of falls, heart block, heart
failure, dementia, peptic ulcer disease and chronic renal
failure.
Principal Findings: This study identified a significant number
of Medicare enrollees with evidence of conditions that are
more susceptible to adverse drug interactions. In the total
population studied, 10.8% had a potentially harmful drugdisease interaction, with the rates varying from 0.8% to 16.6%
per condition. Patients with a history of a fall were mostly
likely to get a potentially harmful drug, with 16.6% receiving an
anticholinergic agent. Other common potentially harmful
drug-disease interactions included the use of antichoniergic
agents among patients with dementia (8.6%) and the use of
non-steroidal anti-inflammatory drugs among patients with
chronic renal failure (8.3%).
Drug-disease interaction rates varied by gender and age; with
more women then men identified as having a potential
harmful interaction. The rate in the female population was
6.5% and in the male population 4.9%. Rates were also higher
as age increased: 5.3% in enrollees aged 67-74 years, 5.7% for
those 75-84 years and 6.6% in the 85 and older range. For any
condition, primary care providers prescribed the greatest
number of prescriptions that were identified as
contraindicated in the study population (49.5%), however; this
varied across the conditions. Medical records were reviewed
to validate the administrative findings; disease and medication
confirmation rates were very high (87.5% - 89.6%).
Conclusions: Adverse drug events have been linked to
preventable problems such as depression, constipation, falls,
immobility, confusion and hip fractures. Use of medications
that pose high risks to the elderly is likely to increase the cost
of care while decreasing the quality of care. The results from
this study indicate the potential for drug-disease interactions
may be more problematic for some diagnoses versus others,
however, due to the potential size of the eligible population,
even small improvements in recognizing and avoiding
potentially harmful drug-disease interactions will help in the
prevention of drug-related morbidities.
Implications for Policy, Delivery, or Practice: This study
highlights the need to continue 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
Quality: Measuring & Improving Quality
Poster Session B
Monday, June 26 • 5:30 pm – 7:00 pm
●The Accuracy of Electronic Billing Data in the
Identification of Bone Mineral Density Screening and
Osteoporosis Risk Factors
Shilpa H. Amin, MBsc, M.D., FAAFP, James Naessens,
M.P.H., Erin McMurtry, MS, Jane Boots, ADN, RN, Rosa
Cabanela, Ph.D., Sidna Scheitel, M.D., M.P.H.
Presented By: Shilpa H. Amin, MBsc, M.D., FAAFP, Staff
Physician, Geriatrics and Women's Health, Division of Health
Care Policy and Research; Division of Primary Care, Mayo
Clinic Rochester; Medical Department, The Hebrew Home of
Greater Washington, Mayo Clinic Rochester and The Hebrew
Home of Greater Washington and Hirsh Health Center, 6121
Montrose Road, Rockville, MD 20852; Tel: 301-816-5056; Fax: ;
Email: shilps716@aol.com
Research Objective: To assess the accuracy of using
electronic administrative data sources for identifying the
occurrence of bone mineral density screening and risk factors
for osteoporosis.
Study Design: A retrospective review comparing the last 7
years of administrative billing data with the corresponding
length of time in the electronic medical record. Bone mineral
density (BMD) screening status and fourteen risk factors for
osteoporosis were patient variables assessed. Patients were
stratified by age greater than or less than 65, Those under 65
were further stratified by presence of an osteoporosis risk
factor in the billing data.
Population Studied: Random samples of 200 women, ages
50 to 80 empanelled during December 2002 in either primary
care internal medicine or family medicine panels at the Mayo
Clinic, Rochester.
Principal Findings: A positive predictive value (the percent of
women with a BMD noted in the billing data to have BMD
results documented in the medical record) of 99% (197/199
women) was was found. The negative predictive value (the
percent of women without a BMD noted in the billing data
who were also noted to have no BMD noted in the medical
record) was 174/179 (97.2%). Four of five women with a BMD
in the medical record had the screening performed outside of
the Mayo Clinic. Four women with documented osteoporosis
had no BMD identified in either source. Risk factor data were
not accurately identified by the administrative billing data.
Only 61% of patients with osteoporosis risk factors were
identified in seven years of billing data. It was shown that 182
of 190 patients who had osteoporosis risk factors identified in
the billing data also had those risk factors identified in the
medical record.
Conclusions: Electronic Billing data is a valid tool for
identification of BMD screening in primary care patients.
Osteoporosis risk factor identification is not accurately
assessed by current electronic billing data sources in
comparison to the medical record.
Implications for Policy, Delivery, or Practice: Based on high
positive and negative predictive values, electronic billing data
can be used to assess primary care BMD screening practices
among physicians. However, in the future, risk. Risk factors for
chronic medical diseases should be adopted to electronic
medical record systems to enhance identification of patient’s
needing specific primary care screens and those set by the US
Preventive Task Force. The study method was a blinded,
retrospective randomized sampling of patients from either
internal medicine or family medicine primary care panels..
The consistency of results of a high positive predictive value of
the identification of BMD’s in the administrative billing data
compared to BMD documentation in the medical record
across all strata suggests these results are generalizable to the
target population of all women ages 50 to 80. This has greater
implications for designing EMR software or implementing
mechanisms in current systems to identify and monitor the
efficacy of other primary care screening services in adults and
children.
Primary Funding Source: No Funding
●Mammography Rescreening Differences Between
Beneficiaries Enrolled in Medicare Managed Care and Feefor-service: Effects of Medicare Coverage Type,
Race/Ethnicity, and Socio-economic Status
Ashley Antler, B.A., Kelle Eason, M.P.H., Michelle Fernandez,
M.P.H., Susan Merrill, Ph.D.
Presented By: Ashley Antler, B.A., Healthcare Information
Specialist, Scientific Affairs, Lumetra, One Sansome Street,
San Francisco, CA 94114; Tel: (415) 677-2046;
Email: aantler@caqio.sdps.org
Research Objective: Evidence suggests that a large number
of women are not regularly screened (rescreened) for breast
cancer, and that women in vulnerable subpopulations- nonwhite, low socio-economic strata- are even less likely to obtain
routine mammograms than their less vulnerable counterparts.
Some studies also report that Medicare Advantage (MA; the
Medicare managed care program) beneficiaries are more likely
to receive mammograms than their Fee-for-Service (FFS)
counterparts. The primary objective of this study is to identify
mammography rescreening patterns within and across diverse
Medicare populations in both MA and FFS settings. Key
demographic and socio-economic factors are evaluated to
determine their influence on rescreening patterns in both
settings.
Study Design: Three years of Health Plan Employer Data and
Information Set (HEDIS) mammography population data
were obtained from one MA plan serving California
beneficiaries aged 65 years and older. Female MA beneficiaries
were matched to FFS enrollees by age and baseline county of
residence. Medicare data were used to assess screening status
of FFS enrollees and demographic variables for both
populations, and 2000 Census data were used to estimate
socio-economic (SES) factors.
Analyses examined the combined and interactive effects of
individual demographic characteristics that predicted, or had
an inverse effect on the receipt of mammography rescreening,
defined as a woman having a mammogram during the first
measurement period and at least one more mammogram
during the next two measurement periods. Differences and
similarities in rescreening patterns and predictors of these
patterns in and across settings were identified.
Population Studied: Study population includes women aged
65-67 as of January 2001 and their mammography screening
status during three measurement periods within the
timeframe of 1/01-12/04.
Principal Findings: Those MA members who were initially
screened were more likely to be rescreened (have a
mammogram in the succeeding periods) than women not
initially screened (74% and 26%, respectively). This pattern
persisted among each race/ethnic group except Black women,
among whom sharply contrasting patterns emerged.
Specifically, initially non-screened Black women were more
likely to be non-screened in the follow-up periods, and those
initially screened were less likely to be rescreened. While all
women in FFS were less likely to be screened than their
counterparts in MA, this gap in care by setting is most
pronounced for Black women.
Conclusions: Overall, women in one MA plan who received
an initial mammogram were more likely to be rescreened than
women not initially screened. However, Black women in this
MA plan are disproportionately affected, as this group
received initial and follow-up mammography screening at
lower rates compared to other race/ethnic groups. Similar
patterns were observed in FFS: Black women were least likely
to obtain initial and/or repeat screening. Dual eligibility and
SES characteristics also influenced mammography
rescreening patterns in both the MA and FFS settings.
Implications for Policy, Delivery, or Practice: To make care
more equitable, patient-centered, and effective, disparities in
Mammography rates by race need to be addressed. These
results, in conjunction with previous research findings that
show Black women have lower mammography screening rates
and higher breast cancer mortality than White women,
support the need to target mammography screening and
rescreening efforts toward this particularly underserved group.
Primary Funding Source: CMS
●Case-mix Adjustment Strategy for the Dutch Version of
the Hospital CAHPS® Instrument
Onyebuchi A. Arah, M.D., M.P.H., Ph.D., Peter
Spreeuwenberg, M.Sc., Diana M. J. Delnoij, Ph.D., Piet J. A.
Stam, M.Sc., Aldien Poll, Niek S. Klazinga, M.D., Ph.D.
Presented By: Onyebuchi A. Arah, M.D., M.P.H., Ph.D.,
Assistant Professor, Department of Social Medicine,
Academic Medical Center of the University of Amsterdam,
Meibergdreef 9, PO Box 22700, Amsterdam, 1100 DE;
Tel: +31205665049; Fax: +31206972316;
Email: o.a.arah@amc.uva.nl
Research Objective: Beginning January 1, 2006, the
Netherlands has a new national health insurance system with
open enrolment, where every citizen gets a mandatory, basic
health services basket without premium differentiation within
insurers, and insurers are expected to compete on quality and
costs within a regulated market. This new health insurance
system engenders performance evaluation, disclosure and
use, and patients are encouraged to become informed
consumers. To aid these developments, we have been
developing patient experience survey instruments for
collecting performance information on insurers, providers and
hospitals. A new Dutch hospital experience survey instrument
was crafted after the U.S. Consumer Assessment of
Healthcare Providers and Systems hospital survey (HCAHPS®). During phase one of the survey, we found the
Dutch version of the H-CAHPS® to be reliable and valid.
However, to aid cross-hospital comparisons the instrument
needs to have an adequate case-mix adjustment. This phase
two study develops a case-mix adjustment model for the new
instrument.
Study Design: We used the 2004 response file from the
second phase of the Dutch patient experience survey based on
the H-CAHPS® instrument. We estimated the effect
(predictive power) of patient characteristics on three global
ratings of nurse, doctor and hospital, by quantifying the
variable-specific improvement in model fit. We set the variable
inclusion p-value at 0.005 to ensure parsimony. We also
calculated the heterogeneity factor to account for differential
distribution of each patient characteristic across hospitals. The
predictive power and heterogeneity factor were then combined
to quantify the overall impact factor of each patient
characteristic. We subsequently selected characteristics with
high impact on any of the three global ratings. We also fit
hierarchical models to check the effect of patient
characteristics on the global ratings when survey items were
nested within patients and patients within hospitals.
Population Studied: The responding 7209 adult patients
aged 18 years or older who were discharged between April and
June 2004 from any of seventeen hospitals in the
Netherlands.
Principal Findings: We found that the most important casemix variables were: age, general health status, education, not
speaking Dutch at home, hospital service (medical, surgical,
obstetric), and interactions of age with the hospital service
variables. Age, general health status, and education affected
all three global ratings, and all but obstetric services. Not
speaking Dutch at home affected hospital and doctor global
ratings for obstetric service only. Overall, the case-mix
variables had only modest influence on hospital rankings on
all domains of patient experience and global ratings.
Conclusions: Although case-mix adjustment has a small
impact on hospital performance ratings, they can contribute
to non-negligible reductions in bias when hospitals are
compared. Furthermore, case-mix adjustement ensures that
healthcare stakeholders are more receptive to comparative
analysis.
Implications for Policy, Delivery, or Practice: Patient
experience data are used by patients, providers, insurers and
policymakers to make important decisions that are related to
quality improvement, provider and institutional selection, and
healthcare purchasing. In order to ensure fairness, accuracy,
and continued validity of comparative hospital performance
information, this follow-up study was designed to develop a
case-mix adjustment model for unbiased performance
evaluation.
Primary Funding Source: Agis Zorgverzekeringen
●Randomized Trial of Transitional Care Pharmacy Services
K. Bruce Bayley, Ph.D., Lucy Savitz, Ph.D., M.B.A.
Presented By: K. Bruce Bayley, Ph.D., Regional Director,
Center for Outcomes Research and Education, Providence
Health System, 5211 NE Glisan, Portland, OR 97213; Tel: (503)
215-7188; Fax: (503) 215-7178;
Email: bruce.bayley@providence.org
Research Objective: Transitions of care within (e.g.,
inpatient) and between (e.g., ambulatory to inpatient care)
components of the health care continuum present risks and
hazards to patients in the form of adverse events, early returns
to the hospital, and mortality. The objective of this study was
to determine whether transitional care pharmacy services
offered in the hospital were effective in reducing hospital
readmission rates and mortality in the Medicare population.
Study Design: Patients admitted to the general medical units
of a community hospital were randomized to usual care vs
additional care by a transitional care pharmacist. The
pharmacist was responsible for medication reconciliation
throughout the stay, consultation with the hospitalist on
medication choice, dosing and route, patient education, and
transfer of a complete pharmaceutical care plan to the primary
care physician at the time of hospital discharge. The number
and nature of pharmacist interventions was tracked for each
patient, as well as their demographics, diagnoses,
medications, and prior admission history. The outcome
measures of 30-day and 180-day non-elective readmissions to
any hospital or emergency department, as well as patient
mortality, were measured using hospital and primary care
record systems. Data were analyzed using chi-square for
raw readmission and mortality rates, logistic regression for
consideration of potential group differences despite
randomization, and Cox Regression to test group differences
in time to readmission or death.
Population Studied: The study population was comprised of
HMO Medicare members admitted to a community hospital
and cared for by employed hospitalists. Two-hundred and ten
patients were randomized to treatment vs usual care, over a
nine-month period in 2004. Patients with dementia, on
hospice, or staying less than 1 day were excluded. Additional
data were collected on all HMO Medicare members admitted
to the same hospital during 2003 and 2004 (n= 43178) and to
a similar local hospital (n=70471), to determine the
representativeness of the sample and to identify underlying
readmission trends.
Principal Findings: Treatment and usual care groups did not
differ in 30-day readmissions (20/105 vs 27/105, respectively,
p=.25). Nor were they statistically different in six-month
readmissions (56/105 vs 63/105 respectively, p=.33). Mortality
rates were higher in the usual care group, but not significantly
so (12/105 vs 8/105, p=.35). Important factors increasing the
probability of 30-day readmission rates were a) age, b) more
previous hospital admissions, and c) discharge to a skilled
nursing facility. Treatment arm did not reach significance
(p=.11) in logistic regression analysis. Time to readmission or
death was slightly longer in the treatment group vs usual care
(88.9 vs 67.2, p = 0.12), and Cox Regression analysis showed
slightly better survival (to readmission or death) in the
treatment group (p=.07) once age, discharge disposition, and
previous admissions were taken into consideration.
Conclusions: The services of a transitional care pharmacist
have a favorable but non-significant effect on hospital
readmission and mortality. Limitations of the study included
a relatively small sample size and the possibility that the
pharmacist had a spill-over effect on the hospitalists providing
usual care.
Implications for Policy, Delivery, or Practice: Current efforts
to institute medication reconciliation within and between
settings are an important step toward quality and patient
safety, and may decrease the chance of short-term return to
the hospital or emergency department. However, additional
resources beyond medication reconciliation are necessary to
prevent readmissions and mortality, possibly including more
timely post-hospital follow-up care, patient adherence to the
improved home medication regimen, and long-term disease
management.
Primary Funding Source: AHRQ
●Adverse Events and Need for Additional Care Following
Hospitalization
K. Bruce Bayley, Ph.D., Carol Baird, M.D.
Presented By: K. Bruce Bayley, Ph.D., Director, Center for
Outcomes Research and Education, Providence Health
System, 5211 NE Glisan, Portland, OR 97213; Tel: (503) 2157188; Fax: (503) 215-7178; Email: bruce.bayley@providence.org
Research Objective: Patients discharged from the hospital
have been shown to be at risk for adverse events, often rooted
in the changes in medications and care plans across settings.
The objective of this study was to determine the extent of
post-hospital adverse events at 30-days after acute care
hospitalization, the nature of the medications and symptoms
involved, and the impact of these events on care seeking and
non-elective ED or hospital return.
Study Design: Patients were contacted via telephone by a
clinical pharmacist at 30-days post discharge, and interviewed
using a pre-tested questionnaire addressing medication
compliance, symptoms, health status, and efforts to seek
additional care or advice. Hospital and clinic records were
used to determine actual visits to hospitals or clinics. Trained
reviewers coded each questionnaire to determine the
likelihood of an adverse drug event, based on Naranjo criteria.
Frequencies and means for all questions were analyzed using
SPSS. In addition, logistic regression was used to identify
factors related to an adverse event.
Population Studied: Potential respondents were selected
from the daily census of HMO Medicare patients cared for by
hospitalists at a 420-bed community hospital. These patients
were approached for participation in a randomized trial of
clinical pharmacy services, and after exclusions for possible
dementia, hospice status, or extremely short (1-day) stays, 210
patients consented to participate. Results are based on the
128 patients successfully contacted at 30 days after discharge.
Caregivers (principally for patients residing in nursing
facilities) provided responses for 13 patients. Differences
between the initial stuffy population and the survey
respondents were minimal, with the exception of a smaller
percentage of respondents among discharges to nursing
facilities.
Principal Findings: More than 80% of respondents reported
at least one symptom, and nearly 70% reported a symptom
that lasted more than one day. The 128 respondents
mentioned a total of 339 symptoms, an average of 2.6 per
patient. The most frequently reported symptoms were
weakness/fatigue, headache/dizziness, cough, and swelling in
the legs or ankles. More than one-quarter (28%) of patients
rated the discomfort from theses symptoms as “quite a bit” or
“extreme”, and 38% percent rated the pain from these
symptoms as 5 or more on a 10-point scale.
Fifty-five percent of patients reporting a symptom believed
that a medication had caused it. Trained physician reviewers
identified 31% of patients as having an adverse event due to
medical management. Cardiac and pain medications were the
most frequent contributors to adverse drug events. The single
best predictor of an adverse medical event was the number of
medications taken by the patient (p= .001). Respondents had
uniformly high ratings of understanding of their medications
and satisfaction with information they had been given in the
hospital (all rated an average of 4.4 on a 1-5 scale). Less than
7% of patients reported any problems obtaining or taking their
medications, even though 55% of patients reported taking
more than ten. Forty-three percent of patients sought care or
advice from others about these symptoms, 32% from their
doctor. Of those contacting their physician, 62% had changes
made in their medications. Twenty-three percent of
respondents visited the ED or had a non-elective hospital stay
in the 30-days post discharge.
Conclusions: Despite efforts to reconcile medications at
hospital discharge and educate patients as to any changes,
quality and patient safety can be compromised at transitions
in care. Adverse medical events occur at a high rate, even
though patients profess to understand and be satisfied with
information about their medication regimen. Among the
factors contributing to high rates of adverse events is the
sheer number of medications taken by the hospitalized
Medicare population.
Implications for Policy, Delivery, or Practice: Transitions in
care continue to present quality and safety problems for
hospitalized elderly. Attention must be paid not only to better
patient education and medication reconciliation, but also to
simplifying medication regimens and reducing the number of
medicines. The push to implement evidence-based guidelines
to prescribe additional chronic disease medications at hospital
discharge should be reviewed for its impact on the frail elderly.
Primary Funding Source: AHRQ
●Patient Complaints offer Window on Serious Threats to
Patient Safety
Marie Bismark, MBChB LLB MBHL, Troy Brennan, M.D. JD
M.P.H., Ron Paterson, LLB (Hons) BCL, Peter Davis, Ph.D.,
David Studdert, LLB ScD M.P.H.
Presented By: Marie Bismark, MBChB LLB MBHL, Senior
Solicitor, Health Law, Buddle Findlay, 440 The Esplanade,
Island Bay, Wellington, 6015; Tel: +64 4 3838066;
Email: mariebismark@gmail.com
Research Objective: There is growing international interest in
harnassing patient dissatisfaction and complaints to address
problems with quality in healthcare. The objectives of this
study were to estimate the proportion and characteristics of
medically injured patients who complain to a national health
ombudsman.
Study Design: This study is the first to match epidemiological
data on medical injuries to complaints about quality of care
lodged with a national health ombudsman. The percentage of
injured patients who lodge complaints was estimated by
linking the New Zealand Health and Disability
Commissioner's national complaints database to records
reviewed in the New Zealand Quality of Healthcare Study.
Bivariate and multivariate analyses investigated differences
between complainants and injured non-complainants.
Population Studied: The study population included (1)
patients who lodged complaints with the New Zealand Health
and Disability Commissioner (n=398); and (2) patients
identified by the New Zealand Quality of Healthcare Study as
having suffered an adverse event who did not lodge a
complaint with the Commissioner (n=847). All iatrogenic
injuries captured in the study occured in 1998, and
subsequent complaints were sampled through to 30 June
2004.
Principal Findings: The vast majority of complaints were filed
within 2 years of the date of injury (mean 10 months, median
5 months). Among adverse events identified by the New
Zealand Quality of Healthcare study, 0.4% resulted in
complaints; among serious, preventable adverse complaints,
4% resulted in complaints. Injury severity was a strong
predictor of complaining: odds of complaint were 11 times
greater after serious permanent injuries than after temporary
injuries and 18 times greater after deaths. Preventable events
were significantly more likely to result in a complaint than
unpreventable ones. We found no sex differences in complaint
behaviour. Odds of complaining were significantly lower
among patients who were elderly (odds ratio 0.2, 95%CI 0.1 to
0.4), of Pacific ethnicity (odds ratio 0.3, 95%CI 0.1 to 0.9), or
lived in the most deprived areas (odds ratio 0.3, 95%CI 0.2 to
0.6).
Conclusions: Despite the availability of a free independent
complaints mechanism, most adverse events in New Zealand
never trigger a patient complaint. Underclaiming is not spread
uniformly across the patient population: elderly patients and
socioeconomically disadvantaged patients are especially
unlikely to complain despite having suffered an adverse event.
The probability of complaint increases sharply with severity of
injury, and preventable complaints are much more likely to
result in complaints than unpreventable ones.
Implications for Policy, Delivery, or Practice: Some
physicians feel under seige from complaints processes. Yet,
when complaints are set against the underlying rate of injury,
it becomes apparent that they represent only the tip of the
iceberg. Troubling disparities in access to and utilisation of
complaints processes echo previous studies from the United
States in which old age and lower socioeconomic status were
associated with a lower propensity to sue. Further work is
required to understand and address these disparities. There is
also a clear bias in the severity and types of injuries that give
rise to complaints. Complaints data should not be construed
as representative of general patterns of medical injury. On the
other hand, the observed skew towards serious and
preventable events is exactly what policymakers might hope
for from a system whose goals are to identify the most serious
safety hazards and identify opportunities for quality
improvement. Complaints may thus offer a valuable portal for
observing serious threats to patient safety, and increased
attention to the patient voice may facilitate efforts to improve
patient care.
Primary Funding Source: CWF
●Predicting Adverse Drug Events Associated with Patient
Transfer
Kenneth Boockvar, M.D., MS, Bianca Lee, Pharm.D., Nathan
Goldstein, M.D., Cornelia Dellenbaugh, MS, Terri Fried, M.D.
Presented By: Kenneth Boockvar, M.D., MS, Assistant
Professor, GRECC, Bronx VA Medical Center, 130 West
Kingsbridge Rd, Bronx, NY 10468; Tel: 718-584-9000 x3807;
Fax: 718-741-4211; Email: kenneth.boockvar@mssm.edu
Research Objective: Patient transfer between sites is regular
practice during an episode of care, but is associated with poor
communication of health information, drug prescribing errors,
and adverse drug events (ADEs). The objective of this study
was to test the reliability and validity of a new measure
designed to identify patients at higher risk of experiencing
transfer-related ADEs.
Study Design: Observational study of hospitalized nursing
home residents. A research assistant abstracted medication
data retrospectively from nursing home and hospital records.
Medication alterations and discrepancies associated with
transfer to and from the hospital were rated by two clinician
raters on a 4-point Likert scale indicating ADE risk (1=none,
2=small, 3=moderate, 4=great) and averaged. A sum of
averaged risk ratings for all medication changes associated
with an episode of transfer resulted in a transition drug risk
score. Clinician raters then performed structured review of
nursing home and hospital records, identifying incidents that
were possible ADEs according to a priori criteria and rating
the certainty that an incident was caused by a transfer-related
medication change on a 6-point scale. An ADE was defined as
any incident for which both raters’ certainty scores, after
consensus discussion, was at least 4. Data were also
collected on patient gender, age, Charlson comorbidity,
APACHE illness severity, and nursing home and hospital
length of stay.
Population Studied: Sixty residents of 2 nursing homes in
New York City (1 VA and 1 non-VA) who had been admitted to
the hospital. Subjects were 53% female, 80.6 (s.d. 9.1) years
old on average, and had been residing in the nursing home for
a median of 5.3 (range 0-156) months. Twenty (33%) cases
were in the VA setting. The mean number of medication
changes associated with transfer was 4.1 (s.d.2.1), not
including vitamins, minerals, topicals, and as needed
medications.
Principal Findings: Mean transition drug risk was 9.0 (s.d.
5.1), median 8.25 (range 0-25.5), and interquartile range 5.512.5. Inter-observer reliabilities between 2 physician raters and
between a physician and pharmacist rater were good
(weighted kappa= .74 and .57, respectively). In a multivariate
linear regression model, higher transition drug risk was
associated with greater number of medication changes and
worse APACHE illness severity. Eleven patients (18%)
experienced 1 transfer-related ADE and 7 (12%) more than
one. In a multinomial regression model, occurrence and
number of transfer-related ADEs was associated (p=0.025)
with higher transition drug risk but no other variable,
including number of medication changes. Fifty percent of
patients in the highest (4th) quartile of transition drug risk
experienced a transfer-related ADE, 36% in the 3rd quartile,
and 17% in the bottom 2 quartiles. Seventy-five percent
(21/28) of transfer-related ADEs occurred in patients at
median or greater transition drug risk.
Conclusions: A new measure predicts ADEs associated with
patient transfer.
Implications for Policy, Delivery, or Practice: This measure
could be used to evaluate the effectiveness of interventions
designed to improve drug prescribing safety during transfer,
or to identify patients most likely to benefit from such
interventions.
Primary Funding Source: VA
●Evidence-based Education for Heart Failure
Suzanne Boren, Ph.D., M.H.A., Teira Gunlock, BA
Presented By: Suzanne Boren, Ph.D., M.H.A., Assistant
Professor, Department of Health Management & Informatics,
University of Missouri, 324 Clark Hall, Columbia, MO ; Tel:
573-882-1492; Fax: 573-882-6158;
Email: BorenS@health.missouri.edu
Research Objective: To advance effective patient education
programs in congestive heart failure based on the results of
randomized controlled trials testing educational techniques.
Study Design: We systematically reviewed the literature and
applied a quantitative meta-analysis to identify the specific
education content most crucial for improving clinical practice.
Eligible studies were randomized controlled trials evaluating
congestive heart failure self-management education programs
with assessment measured on outcome of patient care.
Studies were selected and educational content and outcomes
data were extracted independently by two investigators. The
educational content items from all studies were clustered to
create an education topic list. The outcomes of the studies
were clustered to create a comprehensive description of
outcomes that were measured and if the outcome was
significant or non-significant as indicated in the original
article.
Population Studied: Persons with congestive heart failure.
Principal Findings: A total of 5589 patients participated in the
27 congestive heart failure studies. Twenty education topics
were identified. Univariate analysis revealed that educational
interventions containing five of the education topics were
significantly more likely to improve patient social functioning
than education interventions lacking the topics: diagnosis and
prognosis, aims of treatment, medication review and
discussion of side effects, social interaction and support, and
dietary assessment and instruction.
Conclusions: Several topics were correlated with improved
patient social functioning. Clinicians and health educators
should include theses topics when educating patients with
congestive heart failure.
Implications for Policy, Delivery, or Practice: Educating
patients about good chronic care needs to be based on
scientifically sound evidence. Most patient education is based
on an ad hoc set of messages and skills that clinicians believe
patients need to acquire.
Primary Funding Source: No Funding
●Patient Safety and Nursing Workload
James Bramble, Ph.D., Bartholomew Clark, Ph.D., Susan
Draftz
Presented By: James Bramble, Ph.D., Associate Professor,
Creighton Health Services Researh Program (CHRP),
Crieghton University Medical Center -- School of Pharmacy
and Health Professions, 2500 California Plaza, Boyne 143,
Omaha, NE 68178; Tel: (402)280-4129; Fax: (402)280-4809;
Email: jbramble@creighton.edu
Research Objective: This study examined the association of
organizational factors, such as, hospital workload, and the
quality of patient care. With the consistent restructuring of
health care organizations, this study explores, from a macro
perspective, how these changes may impact patient care by
examining the incidences of adverse events utilizing the
Agency for Healthcare Research and Quality (AHRQ) Patient
Safety Indicators (PSI). This report specifically addressed the
following question: Is the incidence of adverse events related
to the hospital’s nurse-to-patient ratio?
Study Design: We conducted a quasi-experimental crosssectional study using administrative data to examine the
proposed hypothesis. We merged patient safety measures and
patient characteristic data from the 2002 HCUP National
Inpatient Sample with hospital and market data from the
American Hospital Association and the Area Resource File,
respectively. The dependent variable of interest in this study is
the incidence of adverse events. To measure adverse events,
we used AHRQ’s PSI, a measure that uses patient discharge
data to help identify preventable adverse events. The
independent variable was a measure of hospital nurse
workload and was measured as the number of registered
nurses per patient day and the number of licensed practical or
vocational nurses per patient day. We controlled for patient,
hospital, and market characteristics.
Population Studied: We calculated the PSIs using the 2002
HCUP National Inpatient Sample and linked these indicators
to their respective hospitals. After merging the specific data
needed, the unique dataset consisted of 358 hospitals in 25
states.
Principal Findings: Using hierarchical linear modeling, we
found a significant inverse relationship between the number of
registered nurses per patient day and the PSI for postoperative
sepsis. Specifically, higher levels of RNs per patient day were
found to be associated with fewer occurrences of
postoperative sepsis (p = 0.0013).
Conclusions: This preliminary study demonstrated that the
manner in which hospitals structure their staff may impact
patient safety. Previous research has shown that there are a
number of nurse sensitive events where the actions of nurses
during the course of patient care may make a difference in the
quality of care that is experienced by patients. In addition this
research demonstrated how administrative discharge data and
AHRQ’s PSI can identify patient safety concerns within the
hospital.
Implications for Policy, Delivery, or Practice: Driven by
various stakeholders, such as consumers, providers, and
health care organizations, there is a widespread and
noticeable movement to improve the quality of patient care.
Information from this study will help researchers, policy
makers, and administrators better understand the sources of
medical errors. This understanding is critical so systems can
be put in place to reduce errors from occurring and mitigate
its impact when errors occur.
Primary Funding Source: SPAHP Faculty Development
Research Program
●The Dissemination of CT Screening Services into
Practice: What do Radiologists Think?
Ingrid Burger, BS, Jonathan Sunshine, Ph.D., Nancy Kass, ScD
Presented By: Ingrid Burger, BS, Graduate Student, Health
Policy and Management, Johns Hopkins Bloomburg School of
Public Health, 326 S. Chapel St, Baltimore, MD 21231; Tel: 443750-9024; Email: iburger@jhsph.edu
Research Objective: Several different screening exams using
computed tomography (CT) have diffused into practice prior
to empirical evidence and professional agreement about their
medical efficacy and cost-effectiveness. These include CT
screening for lung cancer, coronary artery disease, and wholebody screening in asymptomatic patients. The possibility of
patient self-referral for these exams, along with direct-toconsumer marketing activities of some screening centers
raises policy and ethics questions about the appropriate scope
of patient access to unproven medical technology as well as
physicians’ professional roles and responsibilities within the
setting of medical innovation. The aims of this study are to
describe the prevalence of radiologists reading CT screening
tests; to describe attitudes about the appropriateness of CT
screening in high-risk patients, direct-to-consumer marketing,
patient self-referral, decision-making for screening; and to
compare these beliefs and attitudes among radiologists who
do and do not read CT screening tests.
Study Design: A cross-sectional, self-administered mailed
survey. Questionnaires containing questions exploring several
domains of interest were sent to a national sample of
radiologists. Questionnaires were developed and tested for
content validity.
Population Studied: Diagnostic radiologists currently
practicing in the United States. A random sample of
radiologists was drawn from the AMA’s Masterfile Physician
Database.
Principal Findings: A total of 395 radiologists participated in
this survey. Of the sample, 30% reported that they read CT
screening exams; 73% of these read coronary artery calcium
scoring, 53% read lung cancer screening, and 27% read wholebody CT. For those reading screening exams, 83% are in
private practice and 15% are academic. Most radiologists
believed that lung and coronary screening should be provided
to all high-risk patients (33% and 37%) or patients who really
want them (42% and 46%). Most radiologists (66%) believed
whole-body CT screening is inappropriate for high-risk
patients, although 30% believed that patients who really want
one should have access. As a group, 56% radiologists favored
informed consent for screening exams and 51% percent
believed radiologists should be involved in decision-making
about screening. Fewer radiologists supported patient selfreferral (30%) than with a physicians’ referral (45%), although
radiologists reading screening were more likely to support selfreferral. While only 29% believed that direct-to-consumer
marketing of screening is appropriate, radiologists who read
screening exams were significantly more likely to find it
appropriate. Further analysis will be performed before June
involving multivariate modeling to predict which attitudes and
demographic characteristics are associated with a radiologist’s
involvement with screening.
Conclusions: A large minority of radiologists read CT
screening exams today despite lack of evidence of benefit and
professional endorsement of these tests. Pro-screening
attitudes exist among a majority of radiologists while there is a
deep divide in beliefs about the need for informed consent or
radiologists’ involvement in screening decisions. The major
differences in attitudes among screeners and non-screeners
concern the appropriateness of marketing and patient selfreferral.
Implications for Policy, Delivery, or Practice: The extent of
pro-screening attitudes among radiologists surveyed in this
study may indicate a general climate that aids the premature
diffusion of screening tests. Disagreement about certain
details of screening provision signals the need for professional
dialogue and education about radiologists’ professional roles
and responsibilities regarding the appropriate use and
conduct of screening exams.
Primary Funding Source: Departmental funds
●Effect of E-mail Versus Postal Reminders for
Mammogram Screening
Rosa Cabanela, Ph D, Rajeev Chaudhry, MBBS, M.P.H., Erin
McMurtry, MS, Ahmed Rahman, BS, Dorinda Leutink, RN,
Sidna Scheitel, M.D., M.P.H., James M Naessens, M.P.H.
Presented By: Rosa Cabanela, Ph D, Health Services Analyst,
Health Care Policy & Research, Mayo Clinic, Pavilion 3,
Rochester, MN 55905; Tel: (507)538-0220; Fax: (507)284-1731;
Email: cabanela.rosa@mayo.edu
Research Objective: To assess the relative effectiveness of email versus letter reminders to women due for mammogram
screening
Study Design: Women aged 40-75 who had not had
mammography screening in the previous 9 months were
randomly assigned to either a) be sent a reminder about
breast cancer screening with a phone number to schedule an
appointment or b) receive no reminder. Those women who
were employees of the medical center randomized to receive a
reminder were randomized to receive the reminder a) by mail
or b) by e-mail. The effectiveness of the trial was assessed by
the percent of women in each group who were up-to-date on
mammography screening at the end of the trial
Population Studied: Women eligible for screening were
identified from patients in primary care panels who had not
had mammograms performed within the system in the prior
nine months. Reminders were sent out to those in the
intervention arm. Women who did not schedule a
mammogram were reminded again in 4-6 weeks.
Administrative data covering from 1/2002 to 10/2004 was
processed to determine whether mammograms had been
performed.
Principal Findings: Women who received reminders had
screening rates significantly higher than those who did not
receive reminders (64.3% vs. 55.3%; p<0.001). Among the
subset of employees, there was no significant difference
between those with postal versus e-mail reminders (postal:
68.1%, 95% CI: 63.7 – 72.4%; e-mail: 72.2%, 95% CI: 67.876.6%.
Conclusions: Reminding women to receive timely preventive
services increases the rate at which they have them
completed, particularly for mammography screening. Among
those with e-mail availability at work, there appears to be no
difference if the reminder is mailed to their home or e-mailed
to their work address.
Implications for Policy, Delivery, or Practice: Reminders
work. As more and more patients have access to e-mail
accounts, health care providers need to incorporate this mode
of communications with their patients.
Primary Funding Source: Mayo Foundation
●Establishing Reliability and Validity of the SF-12v2 in the
MEPS
Nancy Cheak, MA, Kathleen Wyrwich, Ph.D., Timothy
McBride, Ph.D.
Presented By: Nancy Cheak, MA, Graduate Research
Assistant, School of Public Health- Health Policy and
Management, Saint Louis University, 3545 Lafayette Ave., St.
Louis, MO 63104; Tel: 314-977-8128; Fax: 314-977-1674; Email:
cheaknc@slu.edu
Research Objective: This study was designed to evaluate the
reliability and validity of the SF-12v2 in the 2003 Medical
Expenditure Panel Survey (MEPS).
Study Design: We analyzed the reliability and validity of the
SF-12v2 in the household component of the 2003 MEPS. The
2003 MEPS is the most recently released version of this health
services and expenditure survey cosponsored by the Agency
for Healthcare Research and Quality and the National Center
of Health Statistics. Internal consistency reliability was
assessed by calculating a Cronbach’s alpha for the SF-12v2
components items. The EQ-5D, perceived health, and mental
health questions were used to test construct and discriminate
validity. ANOVA tests were used to determine the predictive
and concurrent validity of the SF-12v2 in this dataset. SF-12v2
Physical Component Summary (PCS) and Mental Component
Summary (MCS) scores were compared for participants
reporting either limited ability to work, physical functioning
limits, or cognitive limitations to participants that did not
report such limits. Scores were also compared for
respondents based on the number of reported chronic
conditions.
Population Studied: The MEPS sample frame is drawn from
the National Health Interview Survey representing the general
civilian non-institutionalized public with over sampling of
Hispanic, African American, Asians, and families expected to
have incomes below 200% of the poverty line. The selfadministered questionnaire of the 2003 MEPS household
component was used in this analysis, and 20,661 adult United
States residents completed the SF12v2 items.
Principal Findings: PCS and MCS scores were shown to have
high internal consistency reliability within the 2003 MEPS
(PCS, a = .91; MCS, a = .86). PCS scores demonstrated high
convergent validity for EQ-5D items (except self-care) and
perceived health status (r >.56). MCS scores demonstrated
moderate convergent validity on EQ-5D items and perceived
mental health (.61 > r > .38). PCS scores were able to
distinguish between groups with different work and physical
functioning limitations. Similarly, MCS scores were able to
distinguish between groups with and without cognitive
limitations. Both component scores showed strong
concurrent validity and perfect dose-response change when
the mean group scores were compared using participants’
chronic condition status.
Conclusions: The Physical and Mental Component Scores of
the SF-12v2 showed adequate reliability and validity with the
2003 MEPS and should be suitable for use in a variety of
relevant group comparison proposes within this database.
Implications for Policy, Delivery, or Practice: The MEPS is a
resource used by countless policymakers, healthcare
administrators and professionals within academic research
and health care practice. The results from this study will allow
this diverse group to confidently use the SF-12v2 component
scores as physical and mental health status measures within
the 2003 MEPS database.
Primary Funding Source: No Funding
●The Influence of System Change on Improving Quality of
Care Transitions
Eric Coleman, M.D., M.P.H., Carla Parry, Ph.D., MSW, Sandra
Chalmers, M.P.H., Eldon Mahoney, Ph.D.
Presented By: Eric Coleman, M.D., M.P.H., Associate
Professor, Health Care Policy and Research, University of
Colorado Health Sciences Center, 13611 E. Colfax Ave., Suite
100, Aurora, CO 80011; Tel: (303) 724-2456; Fax: (303) 7242486; Email: Eric.Coleman@uchsc.edu
Research Objective: Attempts to use performance
measurement as a tool to drive quality improvement have
largely been focused at the level of individual practitioners.
This study examines the role of system-level influence on
improving the quality of care transitions.
Study Design: Monthly performance evaluation of hospital
wards that were in the process of implementing a patientcentered intervention. Individualized self-management
training that emphasizes the skills needed after discharge,
patient education materials that address warning symptoms
and steps to take if these occur, and mechanisms to involve
patients and caregivers in the development of discharge plans.
The primary measurement tool was the Care Transition
Measure (CTM), a validated self-report measure of care
transition quality. Potential internal and external influences on
performance were monitored.
Population Studied: Adults 65 years and older hospitalized
with CHF or diabetes in a moderate sized community hospital
in the Pacific Northwest.
Principal Findings: Over the 12-month study period, CTM
scores significantly (p<0.05) increased over time. However,
quality gains were counteracted by discreet events that
disrupted stability. Examples included a transient financial
downturn and a change in policy that affected hospital staffing
patterns. Improvements in CTM scores were not sustained
after support for the intervention ended.
Conclusions: System-level factors appeared to significantly
influence attempts at quality improvement efforts and were
largely outside of the control of practitioners.
Implications for Policy, Delivery, or Practice: System
influences on the relationship between performance
measurement and quality need to be explicitly measured and
addressed to effectively improve the quality of care transitions.
Primary Funding Source: CWF
●Nursing Home Acquired Pneumonia (NHAP): Does Care
Consistent with Consensus- and Evidence-based
Treatment Guidelines Impact 30-day Mortality?
Laura Eaton, M.D., M.P.H., David Mehr, M.D., MS, Robin
Kruse, Ph.D.
Presented By: Laura Eaton, M.D., M.P.H., Postdoctoral
Fellow, Family and Community Medicine & The Institute for
Health Policy Studies, UCSF, Box 0936, San Francisco, CA
94143-0936; Tel: (415) 661-3538; Fax: (415) 476-0705; Email:
leaton@fcm.ucsf.edu
Research Objective: Pneumonia is a major cause of
mortality, hospitalization and functional decline in nursing
home patients. Mortality from nursing home acquired
pneumonia (NHAP) is as high as 44%, with hospitalization
rates varying between 21 and 30%. Studies have shown that
high quality care, including appropriate antibiotic use,
hospitalization when indicated, and rapid identification of and
response to respiratory symptoms improve survival, however,
many nursing homes still provide less than adequate care.
Recently, a national panel of experts from multiple disciplines
and specialties convened to develop a comprehensive,
evidence and consensus-based guideline for the management
of NHAP with the hope that a standardized NHAP care
pathway would improve survival of nursing home residents
who acquire pneumonia. (Hutt E, Kramer A. J Fam Pract
2002;51:709-716) Assessing the potential impact of care
consistent with the Hutt guideline for the management of
NHAP on 30-day all-cause mortality in nursing home residents
with probable pneumonia.
Study Design: Secondary data analyses of the Missouri
Lower Respiratory Infection (LRI) Study, a prospective cohort
study. 36 nursing homes in central Missouri and the St.
Louis, MO, area. Nine recommendations adapted from the
new Hutt evidence- and consensus-based guideline for
evaluating and treating NHAP. Thirty-day all-cause mortality
Population Studied: 781 residents with probable pneumonia
(according to the definition from the Hutt guideline for
NHAP) who where initially treated with antibiotics from the
Missouri LRI Study between August 15, 1995 and September
30, 1998.
Principal Findings: Most Missouri LRI subjects met the Hutt
guidelines’ definition of probable pneumonia; however, only
about 2/3 had either possible or probable pneumonia
according to a rating of radiology reports. Treatment
consistent with some guideline recommendations is
associated with lower 30-day mortality in bivariate and
multivariate analyses. Nursing home residents who were
initially treated with any type of antibiotic for at least 10 days
were 70% less likely to be dead at 30 days than those who
received antibiotics for less than a total of 10 days (adjusted
OR=0.3, 95% CI= 0.19-0.47).
Conclusions: Observational data from the Missouri Lower
Respiratory Infection (LRI) only partially support specific
recommendations from a recently published evidence- and
consensus-based guideline for the management of NHAP.
Our findings suggest that using antibiotics for at least 10 days
may reduce 30-day mortality which is consistent with the Hutt
guideline recommendations. In contrast, we found that the
choice of initial empiric antibiotic had no impact on 30 days
mortality. This result is most likely due to the problem that
sicker patients tend to receive more aggressive antibiotic
treatment (confounding by indication). It is difficult to
demonstrate the benefit of specific antibiotic regimens in this
or other observational studies.
Implications for Policy, Delivery, or Practice: Implementing
standardized comprehensive and evidenced-based NHAP
treatment guidelines has the potential to improve (1) the
survival of nursing home residents who acquire pneumonia
and (2) the overall quality of care they receive. We
recommend that the efficacy and effectiveness of these new
NHAP treatment guidelines should be carefully and fully
examined in a large prospective study across multiple centers.
Primary Funding Source: AHRQ
●Coordinated Care for Homeless Youth: An Inside View
Josephine Ensign, M.P.H., Dr.P.H.
Presented By: Josephine Ensign, M.P.H., Dr.P.H., Associate
Professor, Psychosocial and Community Health, University of
Washington, UW Box 357263, Seattle, WA 98195; Tel: 206-6850816; Fax: 206-685-9551; Email: bjensign@u.washington.edu
Research Objective: In order to develop and improve
effective comprehensive, coordinated health services for
homeless young people, it is important to seek their advice as
to what works. The objective of this study was to document
the perspectives of homeless youth on what works and what
does not work in terms of coordinated health care services.
Study Design: The research was qualitative, using a series of
four focus groups with a total of 30 homeless young people.
Two of the focus groups were conducted at a clinic serving
homeless young people. The other two focus groups were
done street-side in coffee shops close to major homeless
youth ‘hang outs.” All focus groups were recorded, coded,
and analyzed using the Atlas-ti qualitative software program.
Population Studied: The research was conducted with
homeless youth ages 16- 24 years (average age 21 years, equal
numbers male and female) in a large West-coast city.
Principal Findings: The most common recommendation for
improving health care for homeless youth was increasing the
They also related an increase in access to primary health care
with a decrease in homeless youth reliance on emergency
departments (EDs). They recommended having a health care
hot-line, so that homeless youth could call to get advice on
health issues and situations, and know where to go for health
care. Related to this, they wanted a place where they could go
to rest and recuperate from illnesses or when they were
“burned out” on either drugs or living on the streets. They
pointed out that such a place could help them connect with
other services that could help them transition off the streets.
Many youth said that the main recommendation they had for
improving health care services for homeless youth was having
a full range of alcohol and drug services. This included
chemical dependency counselors available to meet with youth
in primary care clinics, as well as in-patient drug rehab
programs that were youth-specific. Youth often linked the
need for substance abuse treatment with the need for mental
health treatment. Youth wanted greater coordination of health
care between different providers in the same city, especially
when it came to paperwork. Many young people stated that
changes in the entire US health care system were necessary
for improving health care for homeless youth. While no
differences were found in key themes by gender, differences
were found between the street-based and service-based young
people, mainly in that the street-based youth were more likely
to indicate a need for chemical dependency services.
Conclusions: The young people identified a series of
recommendations for improving comprehensive, coordinated
health care for homeless youth. One of their main
recommendations was increasing access to appropriate
chemical dependency services as well as mental health
treatment. Important differences were noted between
responses of youth from the street-based versus service-based
venues.
Implications for Policy, Delivery, or Practice: The youthidentified recommendations can be used in improving
comprehensive, coordinated health care for homeless young
people. Greater emphasis should be placed on including
street-based youth in future research.
Primary Funding Source: AHRQ
●Relationship of the Patient Activation Measure (PAM) to
Employee Characteristics
Jinnet Fowles, Ph.D., Paul Terry, Ph.D., Susan Adlis, MS,
Christine Taddy Bloom, M.P.H., Ted Wegleitner, MBA, Lisa
Harvey, MS
Presented By: Jinnet Fowles, Ph.D., Sr. Vice President,
Research, Health Research Center, Park Nicollet Institute,
3800 Park Nicollet Blvd., Minneapolis, MN 55416; Tel: (952)
993-1949; Fax: (952) 993-3741; Email: fowlej@parknicollet.com
Research Objective: To assess the relationship of a new
patient activation measure PAM to employee characteristics
including gender, age, education, income, marital status, body
mass index, smoking status, health risk status, and health
status
Study Design: cross-sectional survey, Spring 2005
Population Studied: Two predominately white-collar
employee groups in the Midwest (health care workers, airline
reservationists)(n= 634)
Principal Findings: At the bivariate level, PAM scores were
significantly related to education, income, marital status, body
mass index, smoking status, health risk status and health
status. In a regression model that included all these variables
and controlled for age and gender, lower PAM scores were
predicted by the lowest educational level (ref: >college
grad)(beta = -5.8, p = 0.04), the highest risk category (ref:
excellent) (beta = -14.28, p = .0001), the second highest risk
category (beta = -8.96, p = .003), as well as for those with very
good health status (ref: excellent)(beta = -5.53, p = .001), good
heath status (beta = -12.53, p <.0001), and fair or poor health
status (beta = -11.69, p < .0001). Variables that were not
significant included: marital status, income, body mass index,
and smoking status. The adjusted R-sq for the model was
0.20.
Conclusions: Patient activation skills are critical for effective
self management of chronic illness, but have antecedent
expression in health risk status.
Implications for Policy, Delivery, or Practice: Activation
status should be addressed for people with higher health risks
as well as for people with chronic illness.
Primary Funding Source: CDC
●Healthcare Expenditure and Patient Satisfaction: Does
Better Healthcare Quality Cost More?
Alex Z. Fu, Ph.D., Nan Wang, M.S.I.S.
Presented By: Alex Z. Fu, Ph.D., Assistant Staff, Quantitative
Health Sciences, Cleveland Clinic Foundation, 9500 Euclid Ave
/ Wb-4, Cleveland, OH 44195; Tel: 216-445-7745; Fax: 216-4452781; Email: fuz@ccf.org
Research Objective: Common sense makes us believe that in
order to get better healthcare quality, patients have to pay
more. The objective of this study is to test such a hypothesis
and provide the empirical evidence for the U.S. national
population.
Study Design: Retrospective database analysis with crosssectional study design was conducted using the most recently
available 2003 Medical Expenditure Panel Survey (MEPS).
Given the heavily right-skewed distribution of the cost data, a
generalized linear model with log-link function was employed
to identify the relationship between patients’ self-rating of
healthcare quality (from 0 (worst healthcare possible) to 10
(best healthcare possible)) and their healthcare expenditures,
after controlling for individual demographic covariates,
comorbidity profile (AHRQ comorbidity software), and
functional and activity limitations. In order to generalize the
results to the whole U.S. population, the complex survey
sampling design of the MEPS was taken into account using
the specified sampling weight, variance estimation stratum
and primary sampling unit.
Population Studied: A nationally representative noninstitutionalized sample of 13,980 adults (age >= 18) with their
overall self-rating of healthcare quality (patient satisfaction)
reported for the 12 months before the survey. The annual
individual healthcare expenditure was derived from the same
period.
Principal Findings: The average annual healthcare
expenditure ranged from $4,000 to $6,000 with the mean
value $4,842 for all patients rating their received healthcare
quality from 0 to 10. Patients with higher ratings of their
healthcare quality did not spend significantly more compared
to patients with lower ratings (p = 0.92). Within expectation,
comorbidities and functional and activity limitations were
significant predictors of the total healthcare expenditure.
Conclusions: This study adds to the literature of healthcare
quality improvement by providing the empirical evidence at
the national level. Patients do not need to spend more to
achieve higher satisfaction for their healthcare. More research
is needed to establish the relationship between healthcare
quality and the expense on the healthcare provider side.
Implications for Policy, Delivery, or Practice: Healthcare
quality improvement may not require additional healthcare
costs, at least from the patient perspective.
Primary Funding Source: No Funding
●How Does the Stability of Health Plan Membership
Affect the Assessment of its Quality?
Lauren Goldman, M.D., Arpita Chattopadhyay, Ph.D., Andrew
B. Bindman, M.D.
Presented By: Lauren Goldman, M.D., General Medicine
Fellow, Medicine, U.C.S.F., 1001 Portrero Ave., Box 1364, San
Francisco, CA 94110; Tel: (415) 476-1109; Fax: (415) 206-5586;
Email: legoldman@medsfgh.ucsf.edu
Research Objective: Medicaid health plan membership is
often dynamic—with fluctuations in enrollment throughout
the year. In applying the HEDIS indicators to evaluate health
plan performance, NCQA limits its assessment to plan
members who are continuously enrolled throughout the year.
It is unclear whether the stability of a plan’s membership
affects the assessment of its performance and whether health
plan rank changes if quality indicators are applied to all of its
members during a year or only to those with at least a year of
continuous enrollment.
Study Design: We conducted a cross-sectional analysis of the
1999-2001 California Medicaid managed care plans. Our
independent variable, the “percent of members continuously
enrolled”, was defined as the number of plan members who
were continuously enrolled for 12 months of a calendar year
divided by all members enrolled in the plan for part or all of
that year. We used the hospitalization rate for ambulatory care
sensitive conditions (ACSC), as the health plan quality
indicator. We applied the Agency for Health Research and
Quality’s definition of ACSC hospitalizations to calculate a
health plan’s annual rate (1) among all beneficiaries during
their time in the plan regardless of the duration of their
enrollment and (2) among only beneficiaries continuously
enrolled throughout the year. We conducted a repeated
measures (clustered by year) linear regression to assess
whether the percent of patients continuously enrolled
predicted a health plan’s ACSC hospitalization rate defined
both ways. We also calculated the Spearman correlation of
health plan ranks based on the two calculated ACSC
hospitalization rates.
Population Studied: We included plans serving the
Temporary Assistance to Needy Family (TANF) eligibility
group with membership populations greater than 1,000.
Principal Findings: Forty-three Medicaid plans representing
over 2 million TANF members were included. On average,
50% (range 15% to 79%) of the beneficiaries were enrolled for
12 months continuously over a given year. For every 1%
percent increase in the percentage of continuously enrolled
members, the ACSC hospitalization rate among continuously
enrolled members decreased by 4.4% (p =0.003). A similar
negative association between the percentage of continuously
enrolled members and ACSC hospitalization rates was seen
among all members enrolled in the plan (4.1%, p = 0.006).
Plan rankings based on the ACSC hospitalization rate for all
members versus those enrolled throughout the year were
highly correlated (Spearman 0.91, p < 0.0001). However, two
health plans whose ACSC hospitalization rates were greater
than 2 standard deviations from the mean when calculated on
the continuously enrolled population were within the 2
standard deviation range when calculated on the entire
member population.
Conclusions: Scores on Medicaid health plan performance
indicators can be affected by the stability of a plan’s patient
population. Though highly correlated, a Medicaid health
plan’s performance ranking relative to other Medicaid health
plans may be influenced by whether or not the evaluated
population is restricted to those who are continuously enrolled
throughout the year.
Implications for Policy, Delivery, or Practice: Future studies
should determine the best sampling strategy to evaluate
health plan performance without introducing systematic bias.
Primary Funding Source: NRSA 1 T32 HP19025
●Measuring Healthcare Quality Across Countries: Using
the AHRQ QIs with Italian Discharge Data
Paul Gorrell, Ph.D., Cinzia Marano, Ph.D., Edward Kelley,
Ph.D., Andrew Mosso, MS
Presented By: Paul Gorrell, Ph.D., Programming Manager,
Social & Scientific Systems, Inc., 8757 Georgia Avenue, 12th
Floor, Silver Spring, MD 20910; Tel: (301) 628-3237;
Email: pgorrell@s-3.com
Research Objective: Measuring quality of care and health
system performance is on the national agenda in an
increasing number of countries. Sharing lessons learned has
the potential to increase the applicability of innovative
solutions, as well as yield maximum benefit for limited
resources. But such sharing, to be meaningful, requires a
methodological common ground. The aim of this ongoing
study is to investigate methodological and analytic issues in
using a set of quality indicators developed for use with U.S.
data (the AHRQ QIs) to analyze data from the Italian
Administrative Database.
Study Design: Discharge data from the 2003 National
Hospital Administrative Database were formatted for use with
the AHRQ Quality Indicators (the Patient Safety Indicators
and the Inpatient Quality Indicators). Output was then
compared with data from the 2003 Nationwide Inpatient
Sample. Specific methodological issues involved aligning
specifications for admission type, length of stay,
diagnosis/procedure counts, as well as the definition of
principal diagnosis, for the quality indicator software.
Population Studied: The 2003 Italian National Hospital
Administrative Database includes approximately 12.5 million
community hospital records (approximately 1,400 hospitals).
The 2003 U.S. Nationwide Inpatient Sample contains records
for approximately 8 million community hospital discharges
(approximately 1,000 hospitals), weighted for national
estimates.
Principal Findings: Variations among quality indicator rates
were similar between the two populations. Differences in the
structure of the administrative data likely accounted for a
considerable proportion of the disparity. Evidence of
established differences in health utilization and health status
between the two nations (e.g., birth rate, admission structure)
were reflected in the results. Differences in rates between
Italian and American populations were also observed and
provide a fruitful basis for further investigation. For example,
rates for hypertension admission and cesarean section are
higher in Italy. Possible explanations for these and other
differences will be discussed. A significant methodological
finding is that it is possible, with reasonable effort, to format
data from the Italian Hospital Administrative Database to
meet the requirements of the AHRQ Quality Indicators
software.
Conclusions: Using the AHRQ Quality Indicators with Italian
data, and comparing indicator rates with those of a
comparable U.S. database, yields promising results, both for
the specific case of investigating differences in type and quality
of care between the U.S. and Italy, and more generally for
establishing a basis for facilitating comparisons and quality
measurement across countries.
Implications for Policy, Delivery, or Practice: The specific
methodology derived from this work will be applicable to other
European countries, as well as all countries which use similar
data collection and database procedures. The use of a
common set of indicators across different countries shows the
potential for sharing quality measurement methodologies and
approaches to heath care quality and patient safety issues.
Although significant issues remain to be addressed, e.g. the
current lack of international benchmarking, the present study
increases the potential for both a methodological and an
analytic common ground in international studies of quality of
care.
Primary Funding Source: AHRQ
●Treatment Outcomes in a Secondary Prevention Lipid
Clinic
Anthony Greisinger, Ph.D., Brenda Brehm, MA, Oscar
Wehmanen, MS, Hope Nora, Ph.D., Ali Mortazavi, M.D., Kim
Birtcher, PharmD
Presented By: Anthony Greisinger, Ph.D., Vice President for
Research and Development, Kelsey Research Foundation, 5615
Kirby Drive, Suite 660, Houston, TX 77005; Tel: (713) 442-1214;
Fax: (713) 442-1228; Email: AJGreisinger@kelsey-seybold.com
Research Objective: Previous analyses have demonstrated
that patients enrolled for 3 or more years in our
multidisciplinary Secondary Prevention Lipid Clinic (SPLC)
have shown significant improvement in achieving treatment
outcomes (total cholesterol <200, HDL >40, LDL <100, and
triglycerides <150). In this study, we evaluated these
treatment outcomes for SPLC patients (N = 360) compared to
non-SPLC patients (N = 360).
Study Design: The SPLC includes cardiologists, a pharmacist,
nurses, and a registered dietician. This approach uses a
protocol based on the National Cholesterol Education Project
(NCEP) guidelines that provides individualized patient and
family education to promote lifestyle changes (e.g., cholesterol
reduction, diet/weight management, exercise, medications,
and tobacco cessation). One half of the study sample
included patients enrolled in the SPLC (N=360) for three or
more years and the other half were non-SPLC patients
(N=360) under the care of a cardiologist during the same time
period. Non-SPLC patients received standard care from their
cardiologists. The two groups were closely matched on age,
gender and co-morbidities such as diabetes, cancer, and
cardiac conditions. To determine whether participation in the
SPLC resulted in greater improvement in treatment outcomes,
two-way between-groups analyses of variance (ANOVA) were
conducted. Change scores between Year 2001 levels of total
cholesterol, HDL, LDL, and triglycerides and Year 2004 levels
were used as dependent variables. Clinic participation and
gender were included as independent variables. Separate
ANOVAs were conducted on each of the independent
variables.
Population Studied: Patients with a history of CV disease
were identified and treated at a large multi-specialty medical
organization. The organization has 21 clinics and more than
300 physicians serving the greater Houston, Texas
metropolitan area.
Principal Findings: The ANOVAs showed a significant main
effect for SPLC participation [F(1,503)=19.27, p=.001], with
SPLC patients exhibiting a significantly greater decrease in
their total cholesterol levels after three years (M=-21.63,
SD=39) than non-SPLC patients (M=-4.83, SD=45.74). There
was also a significant main effect for SPLC patients
[F(1,436)=8.31, p=.01], who experienced a significantly greater
decrease in LDL levels (M=-17.82, SD=33.29) than non-SPLC
patients (M=-8.57, SD=35.93) over the same three-year period.
There were no significant main effects or interactions on HDL
and triglycerides.
Conclusions: Patients participating in our multidisciplinary
SPLC showed greater improvement in achieving total
cholesterol and LDL goals for cholesterol management than
patients receiving standard care by a cardiologist.
Implications for Policy, Delivery, or Practice: These findings
provide additional support for the effectiveness of
multidisciplinary lipid clinics in improving patient outcomes
and quality of care. Future studies are needed to examine
adherence to lipid clinics over time, the cost-effectiveness of
these programs, and whether lipid clinics reduce the risk of
future CV events.
Primary Funding Source: No Funding
●Patient Safety and Communication on a Labor and
Delivery Unit
Bill Grobman, M.D. MM, Donna M Woods, EdM, Ph.D.,
Marilyn Szekendi, RN, MSN, Betsy Wassilak, BSRN, MSN,
Kristine Gleason, RPh, Jane L Holl, M.D., M.P.H.
Presented By: Bill Grobman, M.D. MM, Faculty Maternal
Fetal Medicine, Maternal Fetal Medicine, Northwestern
Memorial Hospital and the Center for Pateint Safety
Northwestern University Feinberg School of Medicine,
Prentiss Hosptial 4th Fl, Chicago, IL 60611;
Email: w-grobman@northwestern.edu
Research Objective: The Institute of Medicine report, To Err
Is Human, has provided evidence of the widespread
prevalence of medical error and related patient harm. The
Joint Commission on Accreditation of Healthcare
Organizations has estimated that 65% of sentinel events are
associated with lapses in communication. The dynamics of
communication have been understudied in the obstetric
context. The aim of this study was to characterize the context
and modes of effective and problematic communication on a
labor and delivery unit.
Study Design: Focus groups were convened, comprised of
attending obstetricians, attending and resident
anesthesiologists, resident physicians, obstetrical nurses,
pharmacists, and unit secretaries. All participants provided
care on a single obstetrical unit at a large urban academic
medical center. To enhance participants’ comfort level and
willingness to speak freely, all groups were composed of
individuals of the same discipline and professional level. Each
focus group was facilitated by a trained leader using a
standardized protocol to elicit: 1) the normal means and
patterns of communication among clinicians; 2) instances
and patterns of effective and problematic communication; and
3) how communication among clinicians affects the delivery of
safe patient care. Each 90-minute focus group was audiotaped and transcribed such that no identifying information
was included in the transcripts. A minimum of three
investigators reviewed and independently analyzed each
transcript using an emerging themes approach and achieved
consensus on final theme assignment when discrepancy
occurred.
Population Studied: Obstetrical care providers and staff at a
hospital with a large obstetrical intrapartum service
Principal Findings: Eighteen focus groups of 3 to 8
individuals were convened, with a total of 92 participants. A
range of contexts and modes for effective and problematic
communication was identified. Participants identified face-toface interaction and interdisciplinary teamwork as primary
features of effective communication. Of note, participants
described enhanced communication occurring during high
risk or emergency situations, comparing it favorably to that
occurring during the course of routine care. Devices such as
web pagers and portable phones were recognized as being
crucial to effective communication.
The two factors most frequently associated with problematic
communication were difficulties reaching providers in a timely
fashion and multiple, non-standard sources of clinical
information (including paper, wall-boards and electronic
sources). Also noted were continuity of care issues related to
the existence of multiple providers and unprofessional,
disrespectful behavior.
Conclusions: Problematic communication on an inpatient
intrapartum obstetric service often occurs at the nexus of
interaction between the many types of care providers. Several
strategies identified by caregivers may ameliorate some of the
impediments to effective communication, and further
interventions may continue to improve communication.
Implications for Policy, Delivery, or Practice: Qualitative
analysis is a useful tool to elucidate impediments to effective
communication and strategies that can be used to improve
communication practices.
Primary Funding Source: Excellence in Academic Medicine
grant from the State of Illinois
●Concentration of Medical Care Utilization at End-of-Life
for Ovarian Cancer Patients
Mark C. Hornbrook, Ph.D., Donald J. Bachman, MS, Lisa
Herrinton, Ph.D., Christine Neslund Dudas, Ph.D., Sheron J.
Rolnick, Ph.D.
Presented By: Mark C. Hornbrook, Ph.D., Chief 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: Prior studies of Medicare beneficiaries
indicate that health care utilization increases exponentially as
aged persons approach end of life, both overall and for cancer
patients. Other studies indicate significant underuse of
hospice care, although less so for cancer patients. Our
primary aim is to examine the rate at which women dying of
ovarian cancer were referred to hospice and the timing of that
referral. A secondary aim is to examine the degree of
concentration of medical care use towards the end of the last
six months of life by type of service and patient demographics
and disease severity.
Study Design: This is a retrospective cohort descriptive study
using data extracted from automated medical records,
hospital discharge abstracts, claims, and administrative
information systems of the participating health plans.
Enrollees who died from ovarian cancer were identified by
matching health plan enrollment databases to cancer registry
data (via health record numbers) and State vital statistics
information on deaths (via Social Security numbers or names,
dates of birth, and addresses). Expert reviewers confirmed
each case’s cause of death through chart abstraction, so that
the study cohort included only women whose primary cause of
death was ovarian cancer or a complication thereof.
Population Studied: We included all women aged 18 years
and older who died at any age of epithelial ovarian cancer
during the six-year period 1995-2000. The number of women
eligible for the study was 421. All study cases were
continuously enrolled in one of three integrated healthcare
delivery systems during their last 18 months of life.
Principal Findings: Use of ambulatory visits was evenly
distributed across the last six months of life for the sample
overall and by selected patient characteristics. Persons
diagnosed in the same year as they died had the highest
concentration in all services. Hospitalizations and
dispensings showed modest concentration towards the end of
life. Referral to home health and hospice were the most
concentrated services, with about 70% of patients referred to
hospice less than 60 days before death. Patients with the
highest comorbidity scores were referred to home health
much earlier than the remaining cases. Patients who never
had an escort to a clinical encounter have a higher
concentration of inpatient stays at the end of life.
Conclusions: Virtually all patients studied had continuous
access to medical care during the last six months of life, as
revealed by their office visit and medication use patterns.
Access to hospice was the most significantly rationed
resource, followed by home health and hospital care. Patients
whose ovarian cancer was diagnosed late, and had fewer
comorbidities represented a group whose cancer episodes
were highly compressed, thereby leaving much less time to
prepare for death.
Implications for Policy, Delivery, or Practice: Much more
work is needed to improve timely referral to hospice. Patients
who are expected to live with their ovarian cancers for less
than a year should receive assistance in preparing for the
decisions and transitions associated with end of life no matter
their age or co-morbidity status
Primary Funding Source: CDC
●The Impact of a Nationwide Quality Improvement
Initiative to Increase Organ Donation Rates
David Howard, Ph.D.
Presented By: David Howard, Ph.D., Assistant Professor,
Department of Health Policy and Management, Emory
University, 1518 Clifton Road NE, Atlanta, GA 30322; Tel: 404727-3907; Fax: 404-727-9198; Email: david.howard@emory.edu
Research Objective: Continuous quality improvement is
widely used in health care, but it is often difficult to evaluate in
a rigorous manner. We examine the first phase of the Organ
Donation Breakthrough Collaborative, a nationwide quality
improvement initiative led by HRSA to increase organ
donation rates. The first phase involved 95 hospitals in 43
organ procurement organizations (OPOs), the regional
entities that oversee the request and removal process. The
Collaborative intervention consisted of a series of “Learning
Sessions”, attended by teams of hospital and OPO staff, and
dissemination of printed materials describing “best practices”
for increasing donation.
Study Design: We obtained data on actual eligible donors and
total eligible donors (decedents age<70 experiencing brain
death with no cancer or HIV) from the national transplant
registry. We computed the conversion rate -- the proportion of
eligible donors who became actual donors – by hospital group
(intervention versus control) and period (pre versus post).
The intervention group consists of the 95 hospitals that
participated in the first phase of the Breakthrough
Collaborative. The control group consists of 99 hospitals with
at least five eligible donors in the pre-period located in OPOs
that did not participate in the Collaborative. The pre-period is
the year prior to the start of the Breakthrough Collaborative
(August 2002-July 2003), the post-period is the final three
months of the first phase of the Collaborative (May 2004August 2004). To assess statistical significance, we estimated
a logistic regression with hospital random effects at the level
of the eligible donor. Period, group, and an interaction
between period and group were independent variables. Using
the estimated coefficients, we computed the “difference-indifference” estimator and used the delta method to compute
its standard error.
Principal Findings: The number of actual eligible donors per
month in Collaborative and control hospitals increased by
26.9% (from 101 to 128) and 4.2% (from 74 to 77),
respectively. The number of total eligible donors per month in
Collaborative hospitals increased by 5.6% but decreased by
1.3% in control hospitals. The pre-period conversion rate was
the same in Collaborative and control hospitals: 52.6%. In the
post-period, the conversion rate increased to 63.2% in
Collaborative hospitals and 55.6% in control hospitals. The
increase attributable to the Collaborative is 7.7 percentage
points [95% CI: 0.6 to 14.7]. Applying this estimate to the
number of eligible donors in Collaborative hospitals, the first
phase resulted in an increase of 187 actual donors annually.
Conclusions: The first phase of the Breakthrough
Collaborative was successful in increasing organ donation
rates in participating hospitals. Assuming that the solid
organs removed from a typical donor are associated with a
gain of 30.8 life years to recipients, the Collaborative led to a
gain of 5,767 life years at a cost of only $3 million per year.
Implications for Policy, Delivery, or Practice: Quality
improvement initiatives are a highly cost-effective mechanism
for increasing organ supply. Recent data suggest that “best
practices” have been adopted nationwide, leading to large,
unprecedented increases in the number of organ donors.
However, demand still outstrips supply.
Primary Funding Source: No Funding
●Healthcare Utilization Patterns of California Veterans
with Colon Cancer – VA Patients Rely on Medicare
Denise Hynes, Ph.D., M.P.H., RN, Ruth Perrin, MA, Kristin
Koelling, M.P.H., Elizabeth Tarlov, Ph.D., Todd A. Lee,
PharmD, Ph.D., Rosario Ferreira, M.D., MAPP
Presented By: Denise Hynes, Ph.D., M.P.H., RN, Research
Health Scientist, Midwest Center for Health Services and
Policy Research, Veterans Administration, Hines VA Hospital,
5th Ave & Roosevelt Rd, Bldg 1/Rm C305 (151V), Hines, IL
60141; Tel: 708-202-2413; Fax: 708-202-2316;
Email: Denise.Hynes@va.gov
Research Objective: To assess and compare colon cancer
care across Veterans Administration and Medicare systems of
healthcare
Study Design: Based on a quality of care theoretical model,
this study links clinical data from the California Cancer
Registry with VA and Medicare workload and claims data for a
retrospective cohort of colon cancer patients to characterize
and compare healthcare use in terms of patient
demographics, clinical characteristics and provider choice (VA
or Medicare) for surgery and adjuvant chemotherapy.
Healthcare use in 1999 through 2003 was examined.
Population Studied: A retrospective cohort of incident colon
cancer patients, identified from California Cancer Registry
data, who were at least 66 years old and eligible to use both
VA and Medicare healthcare between 1999 and 2001.
Principal Findings: The California Cancer Registry matched
cancer cases in their database with a list of potential subjects
qualified by age and VA and Medicare eligibility. Selecting only
colon cancer cases (ICD-O-3 = 18.0 – 18.9) diagnosed in 1999
– 2001 and eliminating Medicare HMO participants, for
whom we have incomplete Medicare records, produced our
analytic cohort of 976 cases. Of the 976, 93% were male.
African Americans composed 16% of the cohort. Surgery was
performed on 707 (72%). Of the 707 surgery cases, 544 (77%)
had their surgery in the Medicare healthcare system. Most
patients who received adjuvant chemotherapy received it in
the same healthcare system where they had their surgery. Only
5% of VA surgery patients and 1% of Medicare surgery
patients switched to the other system or used both systems
for chemotherapy. There was no significant difference in stage
at diagnosis or in number of comorbidities in patients
selecting VA care versus Medicare.
Conclusions: In California, patients eligible for both VA and
Medicare rely heavily on Medicare for cancer care. For patients
enrolled and actively engaged in VA healthcare, dual system
use raises questions of coordination and costs of care.
Although California’s cancer registry represents a large,
diverse population and reveals much about the treatment of
cancer, to increase the generalizability of our knowledge of the
quality and coordination of colon cancer care in both VA and
Medicare, this study will be expanded to include data from
nine additional cancer registries, including the VA Central
Cancer Registry.
Implications for Policy, Delivery, or Practice: As national
cancer organizations and policymakers consider ways to
improve colon cancer care in Medicare, and with particular
interest focused on the quality of colon cancer care in the VA
by the Government Performance and Results Act, this study
uses well-validated approaches to ascertain healthcare
resource use and quality of care in both systems to provide
timely and important information about how quality and
efficiency of colon cancer care can be improved.
Primary Funding Source: VA
●Assessing the Feasibility and Acceptability of Quality
Indicators for End-of-life Breast Cancer Care
Grace Johnston, Ph.D., Eva Grunfeld, M.D., DPhil, FCFP, Lisa
Cicchelli, BN, Lynn Lethbridge, MEcon, Eric Mykhalovskiy,
Ph.D., Frederick Burge, M.D., MSc
Presented By: Grace Johnston, Ph.D., Associate Professor,
School of Health Services Administration, Dalhousie
University, 5599 Fenwick Street, Halifax, Nova Scotia, B3H
1R2; Tel: (902)494-1309; Fax: (902)494-6849;
Email: Grace.Johnston@Dal.Ca
Research Objective: 1: to assess the feasibility of measuring
Quality Indicators (QIs) using routinely-collected
administrative data. 2: to assess agreement among
stakeholder groups on these measurable QIs.
Study Design: After an extensive literature review an expert
panel identified 19 QIs that were considered potentially
measurable using routinely collected administrative data.
Objective 1: Using population-based cancer registry and vital
statistics data, we selected all women in Nova Scotia who died
of breast cancer between 01/01/1998 and 31/12/2002. The
EOL study period was the last 6 months of life. Physician
billings, hospital admission/discharge, seniors pharmacare
data, and oncology clinic data were used to calculate statistics
to represent each indicator. Objective 2: Women with
metastatic breast cancer, bereaved caregivers of women who
died of breast cancer, and health care professionals were the 3
stakeholder groups identified. Focus groups with patients,
focus groups with caregivers, and modified Delphi panels with
health care professionals were conducted. Sessions were
audio-taped, transcribed verbatim, audited and a thematic
analysis was undertaken.
Population Studied: All women in Nova Scotia who died of
breast cancer between 01/01/1998 and 31/12/2002
Principal Findings: Objective 1: A cohort of 864 patients was
identified. Benchmark measures of care across the cohort
show 63.4% died in hospital, a mean continuity of care index
of 0.786, and the mean number of inpatient days in the last 30
days was 9.9. Indicators of aggressive care include 9.3% had
chemotherapy in the last 14 days, 5.6% had more than 1
emergency room visit in the last 30 days, and 29.1% had more
than 14 inpatient hospital days in the last 30 days. Objective 2:
A total of 16 patients, 7 bereaved caregivers and 23 health care
professionals participated in the study. There was good
agreement on QIs amongst patient and caregiver groups. The
need for effective communication was identified as an
important QI that was not measurable. The Delphi process
identified overall moderate agreement with QIs amongst
health care providers. 3/19 quality indicators showed poor
agreement amongst health care providers. Results from the
modified Delphi process indicate that patient preferences and
differences in health care delivery between different
jurisdictions must be considered.
Conclusions: QIs must be both measurable and acceptable
to stakeholder groups in order to be useful for improving
quality of care. We have tested both these characteristics in 19
QIs of EOL breast cancer care, yielding a useful set of
measurable indicators. However, many aspects of quality EOL
care considered important by stakeholder groups are not
measurable from routinely collected data.
Implications for Policy, Delivery, or Practice: Quality
indicators (QIs) are tools to measure quality of care and help
enhance quality through identifying areas needing
improvement. Breast cancer offers a disease model to
examine indicators of quality end-of-life (EOL) care.
Primary Funding Source: Canadian Institutes for Health
Research
●The Culture of Safety in 24 Critical Access Hospitals
Katherine Jones, Ph.D., PT, Gary Cochran, PharmD, SM, Anne
Skinner, BS, RHIA, Andrea Fellows, BS
Presented By: Katherine Jones, Ph.D., PT, Assistant Professor,
Preventive and Societal Medicine, University of Nebraska
Medical Center, 984350 Nebraska Medical Center, Omaha, NE
68198-4350; Tel: 402-559-8913; Fax: 402-559-7259;
Email: kjonesj@unmc.edu
Research Objective: To describe the culture of safety in 24
Critical Access Hospitals (CAHs) using the Agency for
Healthcare Research and Quality (AHRQ) Hospital Survey on
Patient Safety Culture, to compare aggregate results from the
12 survey dimensions to AHRQ’s preliminary benchmark, and
to compare an outcome variable from the survey to the
severity of voluntarily reported medication errors in each CAH.
Study Design: We used the Dillman four-contact method to
survey CAH employees within the four categories suggested
by the Survey User’s Guide. Participating CAHs provided a list
of eligible employees that we coded to control for duplicate
responses. A prenotification letter, surveys, personalized cover
letters, and postage paid envelopes addressed to the
Nebraska Center for Rural Health Research were mailed to
each hospital for internal distribution in three waves over two
months. Returns were scanned into a database and analyzed
using the Excel tool developed by Premier, Inc. We obtained
each CAH’s voluntarily reported medication errors for the
quarter corresponding to the survey period through the
MEDMARXTM national reporting database as part of our
AHRQ-funded grant.
Population Studied: We surveyed 2,331 eligible employees in
24 CAHs participating in our grant project, which includes 23
in Nebraska and 1 in Wyoming.
Principal Findings: The overall response rate was 71% (range
49% - 92%). Aggregate positive responses ranged from 82%
in the “teamwork within department” dimension to 53% in
“nonpunitive response to error.” The greatest range of
positive responses across the CAHs were within “staffing”
(45% - 93%), “hospital handoffs and transitions” (38% - 86%),
and “teamwork across departments” (47% - 92%). A strong
correlation in rank order of positive responses (r = 0.86) was
found between the CAHs and the AHRQ benchmark over all
dimensions. The aggregate percent of positive responses from
the CAHs were at least 8% higher than the AHRQ benchmark
for 9 of the 12 dimensions. We found weak correlation
between proportions of total errors that the CAHs reported as
“near misses” and the survey item asking how often near
misses were reported within an organization.
Conclusions: Aggregate comparisons between the CAHs and
the AHRQ benchmark indicate that teamwork and
organizational learning are likely to be perceived most
positively while nonpunitive response to error, coordinated
handoffs, and feedback about error reporting were least
positively perceived across organizations. The weak
correlation of survey responses with known error reporting
behavior supports the need for further research to understand
the relationship between measures of beliefs and reporting
behavior.
Implications for Policy, Delivery, or Practice: Any aggregate
use of these findings for benchmarking purposes should be
done with the knowledge that these 24 organizations were
participating in a structured patient safety program at the time
of the survey and may not be representative of CAHs in
general. Given the early stage of development of safety culture
assessments, findings from surveys should be combined with
additional measures of patient safety and used primarily to
identify opportunities for organizational change and
secondarily in comparisons to an aggregate benchmark.
Primary Funding Source: AHRQ
●Appropriateness of Tympanostomy Tube Placement in
Children
Salomeh Keyhani, MS, M.D., M.P.H., Rebecca Anderson,
M.P.H., Mike Rothchild, M.D., Melissa Simon, BS, Mark
Chassin, M.D., M.P.P., M.P.H.
Presented By: Salomeh Keyhani, MS, M.D., M.P.H., Assistant
Professor, Health Policy, Mount Sinai School of Medicine,
1 Gustave L. Levy Place, Box 1077, New York, NY 10029; Tel:
(212) 659-9563; Email: salomeh.keyhani@mountsinai.org
Research Objective: Tympanostomy tube placement, which
requires general anesthesia, is the most common surgical
procedure performed on children. We examined the
appropriateness of tympanostomy tube placements in 1046
children using data from medical records.
Study Design: We performed a retrospective cohort study of
all children who had tympanostomy tubes placed in 2002 in
five hospitals in the New York metropolitan area. Based on a
review of the research literature, we developed a list of 2228
mutually exclusive indications for tympanostomy tube
placement and asked a panel of nationally representative
expert physicians (4 otolaryngologists, 4 pediatricians and 1
family physician) to rate the appropriateness of each
indication using the RAND appropriateness method. One
indication was assigned to each case based on the criteria
developed by the panel.
Population Studied: Our cohort included a racially and
ethnically diverse population of children. After the
implementation of the Health Insurance Portability and
Accountability Act many community physicians refused access
to their medical records. We were able to obtain complete
information from 3 sources (hospital, pediatrician and
surgeon charts) for 682 (65%) of the cases. Detailed data on
all cases were abstracted from all three medical records for
one year prior to tube placement.
Principal Findings: Of the 682 patients, 48 (7%) of tubes
were inserted for appropriate, 159 (23%) for uncertain and 476
(70%) for inappropriate indications. The principal reasons
that explained inappropriate cases were short duration of
effusion (< 60 days) and low frequency of recurrent Otitis
Media (fewer than 4 infection in 6 months or fewer than 6
infections in 12 months). Changing some of the clinical
definitions used by the expert panel resulted in 45.4% cases
judged appropriate, 24.4% uncertain and 30.2%
inappropriate.
Conclusions: Half of tympanostomy tubes may have been
placed in children for inappropriate or uncertain indications in
a heterogeneous population in New York.
Implications for Policy, Delivery, or Practice: Primary care
physicians who refer patients and otolaryngologists who insert
tympanostomy tubes should consider performing a more
systematic assessment of candidates for this procedure.
Better communication on the extent of middle ear disease
between physicians may decrease inappropriate
tympanostomy tube insertions.
Primary Funding Source: AHRQ
●Adoption of Device-Related Infection Prevention
Practices by U.S. Hospitals
Sarah Krein, Ph.D., RN, Timothy P. Hofer, M.D., MSc,
Christine Kowalski, M.P.H., Jane Banaszak-Holl, Ph.D., Sanjay
Saint, M.D., M.P.H.
Presented By: Sarah Krein, Ph.D., RN, Health Research
Scientist, HSR&D (11H), VA Ann Arbor Healthcare System,
PO Box 130170, Ann Arbor, MI 48113; Tel: 734--769-7100 x
6224; Fax: 734-761-2939; Email: skrein@umich.edu
Research Objective: Device-related healthcare-associated
infection is a common and costly patient safety problem.
Unfortunately, there is no current information about which
practices are used in U.S. hospitals to prevent the most
common device-related infections. The purpose of this study
was to determine the degree to which evidence-based
practices to prevent device-related infections are adopted by
US hospitals and identify factors associated with their use.
We specifically focused on whether centralization, as
demonstrated by the Department of Veterans Affairs (VA)
healthcare system, might influence the adoption of evidencebased infection prevention practices.
Study Design: We conducted a written survey of Infection
Control Coordinators at 719 hospitals nationwide.
Respondents identified whether certain practices were used
for the prevention of the three most common device-related
infections: catheter-related urinary tract infection (UTI), central
venous catheter-related infection and ventilator-associated
pneumonia (VAP). For this analysis we focused on 7 key
prevention practices, 2 each related to UTI and VAP and 3
related to central venous catheters. We defined adoption as a
report of using a practice always or almost always. Data
analysis includes both descriptive and multivariable
techniques with population weights based on the respondent
sample.
Population Studied: A national random sample of nonfederal general medical hospitals with an ICU and more than
50 hospital beds, and all VA hospitals.
Principal Findings: Survey response rates were 80% for VA
and 70% for the non-VA sample. Both VA and non-VA
hospitals have adopted, on average, 3 of the 7 practices of
interest. The percent of hospitals, VA versus non-VA, adopting
at least one prevention practice within each domain were: for
UTI, 21% vs. 35%; for central venous catheters, 99% vs. 89%;
and, for VAP, 88% vs. 83%. VAs were significantly less likely to
use antimicrobial urinary catheters, compared to non-VA
hospitals (14% vs. 30%, p = .002), but were significantly more
likely to use maximum sterile barrier precautions (84% vs.
71%, p = .01) and chlorhexidine gluconate for antisepsis of the
central venous catheter insertion site (91% vs. 69%, p < .001).
Some of the factors associated with adoption were number of
intensive care beds, infection control practitioner certification,
residency training, and hospital safety culture; these factors
varied by type of practice.
Conclusions: The number of infection prevention practices
adopted is similar between VA and non-VA hospitals. There
are important differences, however, in the types of practices
being used and what factors are associated with the adoption
of a specific practice including some potentially modifiable
factors such as practitioner certification and culture.
Implications for Policy, Delivery, or Practice: Developing
strategies to encourage the use of proven infection prevention
practices is an important issue in creating a safer patient
environment. Through this study we have identified several
factors that appear to influence the adoption of certain key
practices. This analysis provides a macro level perspective
about the use of infection prevention practices. Future work is
necessary to more fully understand the complex set of
decisions and activities that affect whether and how these
practices are implemented to effectively decrease hospitalacquired infection.
Primary Funding Source: VA
●Small Area Variation in the Diffusion of Percutaneous
Coronary Intervention, 1991-2003
Gregory B. Kruse, MSc, M.P.H., J. Sanford Schwartz, M.D.,
John B. Kimberly, Ph.D., Mark V. Pauly, Ph.D.
Presented By: Gregory B. Kruse, MSc, M.P.H., Health Care
Systems, The Wharton School, University of Pennsylvania,
3641 Locust Walk, Philadelphia, PA 19104; Tel: 215-868-2553;
Fax: 215-898-0229; Email: gkruse@wharton.upenn.edu
Research Objective: It is well known that there are
considerable differences among small geographic areas in the
utilization of healthcare technologies, but it is not clear how
these differences change as a new technology diffuses into
clinical practice. Wider variation has more substantial quality
implications.
Study Design: Using hospital discharge data from 1991-2003,
we calculated the rate of percutaneous coronary interventions
(PCI, i.e., angioplasty and/or coronary stenting) performed
among patients admitted for acute coronary syndromes in
each hospital referral region (HRR) in Pennsylvania, New
Jersey, and Maryland. We divided these states' 23 HRRs into
tertiles based on the procedure rates in 1991. We next fitted
logistic regression models with receipt of a percutaneous
coronary intervention within 90 days of index admission as
the dependent variable. Independent variables included
patient-level predictors such as acute myocardial infarction,
the presence of chronic ischemic heart disease, clinical
comorbidities, age, sex, race, and whether the patient was
hospitalized in an early-uptake versus late-uptake HRR based
on the utilization of percutaneous coronary interventions in
1991. We also included year-HRR tertile interactions to test for
differences in utilization time trends among the three HRR
tertiles.
Population Studied: Patients hospitalized for acute coronary
syndromes in the 23 hospital referral regions with populations
that were at least 90% composed of residents of
Pennsylvania, Maryland, or New Jersey.
Principal Findings: We found that HRRs in the lower tertiles
of PCI utilization rates in 1991 continued to have lower
utilization rates in 2003, although the relative difference in
utilization between the lowest tertiles and the highest tertile
had substantially diminished. The utilization rate in the lowest
HRR tertile was initially 57% lower than the rate of utilization
in the highest HRR tertile, but this difference had narrowed to
22% lower by 2003 (test for trend, p<0.01). The uptake rate in
the lowest tertile, corresponding to 2.4% additional acute
coronary syndrome patients receiving PCI per year, was
significantly greater than the uptake rate in the highest tertile,
where 2.1% additional acute coronary syndrome patients
received PCI per year (p=0.046). These differences in uptake
rates between tertiles persisted despite adjustment for clinical
and demographic differences among patients in the
multivariate model. This model also suggested that hospitals
in rural areas were less likely than urban areas (odds ratio
0.91, 95% confidence interval 0.84-1.00, p=0.06) to utilize PCI
in acute coronary syndrome patients.
Conclusions: There was substantial variation among small
geographic areas in three mid-Atlantic states in the utilization
and uptake rates of percutaneous coronary interventions for
acute coronary syndrome patients hospitalized between 19912003. The earliest time period was associated with the
greatest geographic differences in utilization rates.
Implications for Policy, Delivery, or Practice: Technologies
in their earliest stage of diffusion are at greatest risk of being
used at varying rates among small geographic areas. This
potentially introduces an important difference in the quality
and outcomes of care for potential recipients of new
technologies.
Primary Funding Source: VA, University of Pennsylvania
Department of Medicine
●Medical Record Documentation Regarding Drug-Drug
Interactions
Jennifer Elston Lafata, Ph.D., Janine Simpkins, MA, Scott
Kaatz, DO, John R. Horn, PharmD, Marsha A. Raebel,
PharmD, Lonni Schultz, Ph.D.
Presented By: Jennifer Elston Lafata, Ph.D., research scientist,
Center for Health Services Research, Henry Ford Health
System, One Ford Place, 3A, Detroit, MI 48202; Tel: 313-8745480; Fax: 313-874-1883; Email: jlafata1@hfhs.org
Research Objective: Previous efforts document the
prevalence of drug-drug interactions in ambulatory care. Yet,
the prescribing of drugs that interact may be indicated if their
benefits are judged to outweigh their risks. We describe the
extent to which prescribers document potential risks and the
delivery of patient education regarding these risks when drugs
that interact are prescribed as well as the extent to which
clinical management is changed to minimize such risks.
Study Design: We abstracted information from medical
records on documentation of provider knowledge of drug
benefit/risk-benefit ratio, provision of related patient
education, and relevant clinical management changes for the
6 week period preceding and 2 week period following the first
date of evidence of co-prescribing found in the medical record.
Taking into account the clustering effect of site, we estimate
the percent of (and 95% confidence intervals [CI] for) patients
with documented provider knowledge, patient education, and
changes in clinical management.
Population Studied: The study population was drawn from
the ten organizations comprising the HMO Research
Network’s Center for Education and Research on Therapeutics
(CERTs). Each organization randomly selected medical
practice-based samples until a total of approximately 200,000
individuals was identified (N=2,020,037). From among these
patients, we identified a random sample (stratified by
organization and drug-drug interaction) of individuals aged 18
and older with prescription drug coverage in 2000 initiating a
‘new’ co-dispensing of known prevalent interacting drugs:
warfarin with a non-steroidal anti-inflammatory drug (n=97),
digoxin with verapamil or diltiazem (n=100), or
lovastatin/simvastatin with diltiazem or verapamil (n=89).
Principal Findings: Most patients (range = 84% for those
prescribed digoxin to 63% for those prescribed
lovastatin/simvastation) had medical record documentation
indicating both drugs were prescribed during the same visit at
least once in the observation period. Among those prescribed
an interacting drug pair, prescriber knowledge of risk and the
provision of patient education were documented less than 14
percent of the time for each of the three interacting drug pairs
(all corresponding 95% CI ranging < 23%). Documentation
regarding clinical management changes was more common,
ranging from 64% (95% CI: 47-81%) among patients
prescribed lovastatin/simvastatin to 79% (95% CI: 60-99%)
among patients prescribed warfarin.
Conclusions: Among ambulatory patients co-prescribed
interacting drugs, a substantial proportion lack any medical
record documentation of the risks associated with the
concomitant use of the interacting pair. These patients are
also unlikely to have medical record documentation of a
discussion of the risks associated with concomitant use or
related patient education. Documentation regarding changes
in clinical management was more common.
Implications for Policy, Delivery, or Practice: Our findings
call into question the extent to which physicians and patients
are making informed decisions regarding the concomitant use
of interacting drugs in the outpatient setting. At a minimum
opportunities exist to improve medical record documentation
regarding the use of interacting drugs in the outpatient
setting.
Primary Funding Source: AHRQ
●Improving Osteoporosis Screening among at Risk
Women
Jennifer Elston Lafata, Ph.D., Deneil Kolk, M.P.H., MSW,
Edward Peterson, Ph.D., Bruce McCarthy, M.D., M.P.H., Bruce
Muma, M.D.
Presented By: Jennifer Elston Lafata, Ph.D., research scientist,
Center for Health Services Research, Henry Ford Health
System, One Ford Place, 3A, Detroit, MI 48202; Tel: 313-8745480; Fax: 313-874-1883; Email: jlafata1@hfhs.org
Research Objective: Routine osteoporosis screening for
women 65 years and older is recommended in evidence-based
guidelines. Yet, less than a third of elderly women receive
bone mineral density (BMD) testing. We present results from
a clustered, randomized trial evaluating the use of patient
mailed reminders, alone and in combination with physician
prompts, to improve osteoporosis screening and treatment.
Study Design: Primary care clinics (n=15) were randomized to
either: usual care, patient mailed reminders alone, or patient
mailed reminders with physician prompts. Using automated
clinical and pharmacy claims data, information was collected
on BMD testing, pharmacy dispensings, and patient sociodemographic and clinical characteristics. Unadjusted and
adjusted differences in testing and treatment were assessed
using generalized estimating equation approaches to logistic
regression.
Population Studied: Study patients were females aged 65-89
years receiving primary care from the salaried medical group
staffing the 15 clinics.
Principal Findings: In the 12-months following the
intervention, unadjusted screening rates were 10.8% in usual
care, 24.1% in mailed reminder only, and 32.3% in mailed
reminder with physician prompt. Adjusted results indicated
that mailed reminders alone improved testing rates compared
to usual care, and that mailed reminders with prompts
improved testing compared to both other arms. However, the
magnitude varied by age. For example, among women 75
years of age, those who received mailed reminders alone were
2.08 (95% CI 1.51-2.77) times more likely to be tested than
women who received usual care, and those who received
mailed reminders with physician prompts were 3.46 (95% CI
2.66-4.48) times more likely to be tested than women who
received usual care. On the other hand, among women 85
years of age, those who received mailed reminders alone were
2.89 (95% CI 1.93-4.31) times more likely to be tested than
women who received usual care, and those who received
mailed reminders with physician prompts were 5.70 (95% CI
3.89-8.33) times more likely to be tested than those who
received usual care. Unadjusted osteoporosis treatment rates
among women tested were 2.3% in usual care, 4.9% in mailed
reminders, and 5.2% in mailed reminders with prompt. Both
adjusted and unadjusted results found no difference in
treatment rates between those receiving mailed reminders
alone or in combination with physician prompts, but women
in either of these arms were significantly more likely to be
treated compared to those in usual care.
Conclusions: The use of mailed reminders, either alone or
with physician prompts, can significantly improve the
osteoporosis screening and treatment rate among insured
primary care patients. This is particularly true among those of
advanced age, or those most at risk of osteoporosis.
Implications for Policy, Delivery, or Practice: With the
continual enhancements to and proliferation of electronic
medical record capabilities, primary care practices have an
increased number of options and tools available to improve
the early detection and treatment of diseases such as
osteoporosis which can have profound sequelae if not
detected and treated early.
Primary Funding Source: Other Funding
●Community Correlates of Pneumonia Hospitalization
and Mortality, Among Persons 65+, along the TexasMexico Border (Border)
Frank Lemus, BA, MA, Daniel H. Freeman, Jr., Ph.D., Jean L.
Freeman, Ph.D.
Presented By: Frank Lemus, BA, MA, Graduate Assistant,
Preventive Medicine and Community Health, University of
Texas Medical Branch, 700 Harborside Drive, Ret. 1153,
Galveston, TX 77555-1153; Tel: (409) 762-5140; Fax: (409) 7625140; Email: fclemus@utmb.edu
Research Objective: An increasing body of literature is
developing to indicate neighborhood or contextual factors
influence the health of an individual in a community. The
mechanisms linking neighborhood advantage/disadvantage to
better/poorer health outcomes are complex. They may be
related, in part, to community variations in quality of care.
This project applied the Evans and Stoddart determinants of
health (field) model to examine the effect of health care quality
on bacterial pneumonia hospitalization rates among persons
65+ in Texas counties along the Texas-Mexico Border region.
Quality of health care was measured using one AHRQ quality
of health care indicator: bacterial pneumonia hospitalization
rates (prevention quality indicator-PQI) among persons 65+.
Hospitalization rates were obtained from public use files of
the Texas Health Care Information Council’s hospital
discharge abstracts for years 1999-2001. Population counts
and community (county) characteristics were obtained from
the 2000 U.S. Census summary files. This study will provide
baseline estimates towards a better understanding of
pathways through which community social and physical
environments, together with health care quality, influence
population health. Such knowledge will inform policy debates
and guide public health interventions for those adversely
impacted by the variability and disparities in health along the
Texas-Mexico Border.
Study Design: In this descriptive study we used the exact ICD9-CM codes AHRQ identified for the PQI to extract the data,
and we followed precisely the methodology used in Healthy
People 2010 to calculate hospitalization rates. We developed
a subset of data from the hospital discharge records for the 32
Texas Border Counties for years 1999-2001 (10,749 records).
The numerator for our rate calculation was the number of
hospital discharges for bacterial pneumonia in the target
county multiplied by 10,000. We derived our denominator
from the 2000 U.S. Census (number of persons age 65+
multiplied by 3). We used this multiplier because the 2000
U.S. Census served as a mid-point in our rate determination,
and represented one year, while the numerator consisted of 3
years of data.
Population Studied: We used the ten ICD-9-CM codes for
bacterial pneumonia specified by AHRQ to identify the
number of hospitalizations among persons 65+ in the 32
Texas Border counties where the code was the first listed
(principal) diagnosis.
Principal Findings: We found disparities between nonHispanic Whites and Latinos in all 32 Border Counties in rates
of bacterial pneumonia. Also, our results show income
gradients correlated with rates of bacterial pneumonia.
Conclusions: Our estimate of baseline rates for the quality
health indicator, bacterial pneumonia hospitalization rates
among persons 65+ in the 32 Texas Border Counties, sets the
basis for future rate estimates. It also presents a basis for
future comparability health studies with other Texas counties,
4 U.S. and 6 Mexico Border States.
Implications for Policy, Delivery, or Practice: Use of state
hospital discharge data with AHRQ’s quality of care indicators
demonstrates a methodology for other communities to better
assess, measure and make decisions about the quality of
health care provided in their community and how this impacts
disparities in their population’s health outcomes. This study
has immunization implications for Texas that address Healthy
People 2010 and Healthy Border 2010 objectives. This study
also offers an opportunity to directly address one AHRQ high
priority area: strengthen health care quality measurement and
improvement. This methodology supports improvements in
health outcomes by translating research findings into practice
with interventions that result in decreased health care
expenditures and health disparities through preventable
hospitalizations. This study demonstrates use of hospital
discharge data for community based health services research.
Primary Funding Source: UTMB, PMCH Educational
Enhancement Fund
●The Relationship between Patient Satisfaction and
Inpatient Nursing Care
Yu-Fang Li, Ph.D. RN, Sharp, Nancy D, Ph.D., Greiner,
Gwendolyn T, M.P.H. MSW, Lowy, Elliott, Ph.D., Liu, ChuanFen, Ph.D. M.P.H., Sales, Anne E, Ph.D. RN
Presented By: Yu-Fang Li, Ph.D. RN, Research Health Science
Specialist, HSR&D, VA Puget Sound Health Care System, 1100
Olive Way, Suite 1400, Seattle, WA 98101; Tel: (206) 768-5383;
Fax: (206) 764-2935; Email: yufang.li@va.gov
Research Objective: Patient satisfaction can be defined as the
extent to which patients receive care as they expect it. Prior
studies suggest that changes in hospital and health care
settings have affected nurse staffing, job satisfaction, and
nurse perception of the practice environment, which could
impact the quality of nursing care provided to patients. Our
objective was to report on associations between patient
satisfaction and the context of inpatient nursing care; the
structure of care; and the process of care among veterans
admitted to VA medical centers (VAMCs), measured at the
nursing unit-level.
Study Design: Data for analyses were drawn from: 1) an
inpatient patient satisfaction survey administered by the
Veterans Health Administration; 2) nursing staff survey data
on job satisfaction, burnout (Maslach Burnout Inventory), and
practice environment (Practice Environment Scale (PES))
collected for a larger cross-sectional study examining the
relationships between nurse staffing, nursing staff job
satisfaction, and patient mortality in VA; 3) the VA Decision
Support System labor input file; and 4) VA administrative
databases for patient characteristics, health outcomes, and
facility characteristics. We used a 2-step multilevel regression
model to analyze the associations between patient satisfaction
and patient, nursing unit, and hospital level data corrected for
clustering at the nursing unit and hospital levels.
Population Studied: The analyses included 11,451 patients
discharged between 2/03 and 6/03 from 241 acute
medical/surgical and 188 intensive/critical care units (ICUs) in
126 VAMCs who responded to the patient satisfaction survey.
2,706 registered nurses (RNs) from the same units responded
to the nursing staff survey, a response rate of 29%.
Principal Findings: 61% of patients reported they were very
satisfied with the nursing care received. We found a positive
association between patient satisfaction with nursing care and
being surgical (vs. medical) patients (OR=.82), having lower
predicted risk of developing complications (OR=.96), staying
on units with higher RN staffing (OR=1.18), and units where
RNs responded positively to PES measures on perceptions of
quality nursing care (OR=1.32) and nurse-physician relations
(OR=1.14), and staying in less complex hospitals (OR: 1.291.41). However, we did not find an association between patient
satisfaction with nursing care and non-RN staffing, RN job
satisfaction, burnout, and perception of participation in
hospital affairs, nurse manager leadership, and
staffing/resource adequacy, facility case mix, and the
proportion of baccalaureate prepared RNs at the hospital
level.
Conclusions: In addition to patient characteristics and health
status, adequate RN staffing level and a practice environment
supporting quality nursing care and collegial nurse-physician
relations are important factors in promoting health care
experience for patients.
Implications for Policy, Delivery, or Practice: Patient
satisfaction is an indicator of organizational effectiveness.
Nursing is the largest workforce in healthcare settings and the
front line healthcare provider for inpatients. Implementing
policies and practice rules to support an effective and efficient
practice environment for nurses is important to improve
patient satisfaction with care received.
Primary Funding Source: VA,
●How do health care organizations adapt to legislation
that protects adverse event reviews from disclosure?
Anu MacIntosh-Murray, Ph.D., , , , , , , , , , , ,
Presented By: Anu MacIntosh-Murray, Ph.D., Adjunct Faculty,
Health Policy, Management, & Evaluation, University of
Toronto, 122 Langley Avenue, Toronto, M4K 1B5; Tel: (416)
732-1381; Fax: ; Email: anu.macintosh@utoronto.ca
Research Objective: To understand how hospitals have
reacted to a law designed to protect adverse event information
from use in litigation. How do they adapt their structures and
activities in response to such legislation and how are these
decisions made?
Study Design: Longitudinal, multiple-case study approach
based on in-depth study of a focal case organization and
comparison with three peer organizations. Data are based on:
observation over two years of monthly meetings of the Quality
of Care Committee in the focal organization; semi-structured
interviews with committee members and the committee Chair;
two sets of semi-structured interviews with risk managers in
three peer organizations; document review of terms of
reference, minutes, organizational charts, reporting templates,
and communication material. Qualitative methods,
interpretive and discourse analysis are used for the interviews
and observations, the focus of this presentation.
Population Studied: Four large, urban academic health
sciences centres in the province of Ontario, where The Quality
of Care Information Protection Act, 2004 (QCIPA) came into
effect in November, 2004. (The law allows information, such
as critical incident reviews and root causes analyses, reported
to Quality of Care Committees designated by health care
organizations to be protected from use in legal proceedings.)
Principal Findings: 1. Ostensibly similar organizations are
interpreting and implementing the QCIPA provisions very
differently. Some have chosen a “narrow” approach: drafting
terms of reference for ad hoc Quality of Care Committees that
will be called into action on a case-by-case basis only for
limited types of critical incident reviews, as an addition to their
incident, safety and quality improvement structures and
processes. In contrast, the focal case organization is taking an
“umbrella” approach: renaming an existing committee as the
parent Quality of Care Committee and bringing a wide
network of local committees and review processes under the
protection of the Act by designating them all as quality of care
committees. 2. Interview participants differed markedly in
their reactions to QCIPA. In the focal organization (taking the
umbrella-approach), members embraced the Act describing it
as a positive force for change to develop “robust, QCIPAdriven processes and line accountability.” They expect to see
more open participation by staff in incident reviews. In the
narrow-approach organizations, participants expressed
disappointment and frustration with the legislation, describing
it as “a step backward.” They cited the overly restrictive nature
of the disclosure prohibitions, saying that implementing the
protections would hamper their ability to freely share
information about incidents and lessons learned with patients,
families, and staff. 3. A number of participants stated that
factors other than the legislation were having a greater impact
on their efforts to expand learning from incidents, for example,
implementation of improved electronic incident reporting
systems.
Conclusions: Two conflicting themes emerge that highlight
the difficulty of balancing potentially competing values of
disclosure and protection. Disclosure aims for transparency of
critical incident analysis for all involved, including patients.
However, protection aims for transparency of review for
internal stakeholders and to encourage participation in
reviews. This is complicated by the ambiguity experienced in
interpreting the legislation and uncertainty because it has yet
to be tested in court. Organizations’ responses can vary
significantly and appear to be influenced by multiple factors,
including a) who leads the information seeking and
interpretation of the legislation in-house (clinicians or
lawyers), and b) the extent to which organizations are
predisposed to sharing critical incident review information
with patients.
Implications for Policy, Delivery, or Practice: Legislation
intended to facilitate critical incident review and learning from
adverse events may provoke conflicting interpretations,
resource-consuming changes to structures and processes,
and uncertainty about prospects for successful outcomes.
Primary Funding Source: Canadian Health Services Research
Foundation Postdoctoral Fellowship
●Interruptions in Nursing Work Environments and Patient
Safety Outcomes
Linda McGillis Hall, Ph.D., Cheryl Pedersen, MSc
Presented By: Linda McGillis Hall, Ph.D., Associate Professor
& CIHR New Investigator, Faculty of Nursing, University of
Toronto, 155 College Street, Suite 215, Toronto, M5T 1P8; Tel:
(416) 978-2869; Fax: (416) 978-8222;
Email: l.mcgillishall@utoronto.ca
Research Objective: Little or no research has been aimed at
understanding and acting on the complex processes and
interruptions in the nursing work environment that may
contribute to patient safety incidents. The purpose of this
research was to investigate interruptions in nursing work and
develop innovations aimed at redesigning the nursing work
environment to reduce these interruptions in the future. The
objectives of this study were to: a) identify the types of
interruptions that are occurring in nursing work environments,
b) determine the systems-related environmental factors that
contribute to work interruptions for nurses and c) identify the
patient safety outcomes of interruptions to nursing work.
Study Design: A multi-site comparative research design was
used in this study comprised of three phases. Phase one
utilized work sampling techniques to observe nursing
interruptions as they would be experienced on a typical
nursing shift to identify the types of interruptions in the
nursing work environment. Phase two involved focus groups
to determine nurses’ perceptions of the types of interruptions
that were found on the study units, the job stressors that
nurses’ experience, and the causes of interruptions in the work
environment. Phase three examined the patient safety
outcomes that resulted from interruptions in the nursing work
environment and the development of selected strategies
aimed at reducing interruptions on the units.
Population Studied: The study took place in 3 adult acute
care hospitals on 6 similar adult medical surgical units in
Ontario, Canada. A total of 27 nurses were observed and data
on 1,687 interruptions were collected. The data were
categorized using Jett & George’s (2003) framework for
interruptions.
Principal Findings: The majority of interruptions that
occurred were intrusions (n=785; 47%), followed by
discrepancies (n=494; 29%), distractions (n=217; 13%), and
breaks (n=191; 11%). The study also captured information
related to 1) the sources of interruptions to nurses work; 2)
the types of interruptions that occur (i.e. a phone call; respond
to patient call bell; meet with physician); 3) the timing of the
interruption (i.e., time of day); 4) the length of time that the
interruption takes the nurse away from their original work; and
5) the outcome of the interruption (i.e. delays, omissions)
from the perspective of patient safety.
Conclusions: Close to half of nursing interruptions were
related to intrusions. A number of key environmental factors
were highlighted as having a substantial impact on
interruptions including technological factors and
environmental noise. Several of these formed the basis for the
development of work redesign initiatives that involved all
levels of staff.
Implications for Policy, Delivery, or Practice: The
importance of developing an understanding of systems factors
in the work environment such as interruptions in nurses’ work
that may contribute to patient safety was underscored.
Organizations are able to redesign nursing work environments
to decrease nursing work interruptions and patient safety
occurrences. Strategies for decreasing these interruptions are
outlined and future work at a National level is discussed. The
importance of a multidisciplinary approach, combined with
working with health care decision makers and leaders is
highlighted.
Primary Funding Source: Connaught Fund, Canada
●Use of the Quality Promotion Model to Evaluate
Interventions by Quality Improvement Organizations
Linda McKibben, M.D., M.P.H., DrPH, Jane C. Banaszak-Holl,
Ph.D., Richard A. Hirth, Ph.D., Abigail Shefer, M.D.
Presented By: Linda McKibben, M.D., M.P.H., DrPH, Senior
Advisor, Health Services Research, Division of Healthcare
Quality Promotion, CDC (Atlanta), 1600 Clifton Road,
Mailstop A-07, Atlanta, GA 30333; Tel: 404-639-4039; Fax: 404639-4043; Email: lmckibben@cdc.gov
Research Objective: We developed the Quality Promotion
Model, or QPM, by merging Pfeffer and Salancik’s resource
dependency and Rogers’ organizational innovation
perspectives and used QPM concepts to investigate the
effectiveness of interventions by Quality Improvement
Organizations, or QIOs. Research objectives were to validate
the QPM and investigate factors affecting vaccination
program innovation within nursing facilities.
Study Design: From 1999 to 2003, a controlled before-andafter program evaluation was conducted by the Centers for
Medicare & Medicaid Services, or CMS, and CDC to
determine the effectiveness of QIOs, state-based private
contractors of CMS. Government intervention consisted of
varying intensity of QIO promotion of standing orders
programs, or SOPs, to facilities and relaxation of federal rules
prohibiting SOPs. Staff of facilities were surveyed before and
after government intervention on use of SOPs and
perceptions of implementation barriers. These data were
combined with federal administrative data on the structure
and resources of facilities. We developed measures of
resource dependency factors, including how intense the QIO
intervention was and government-resource dependency for a
facility; and organizational innovativeness, including resource
abundance within the facility and perceived barriers to SOP
implementation. Control variables included prevalence in
states of facilities that had adopted SOPs at baseline,
prevalence of such facilities in counties, total number of
facilities in each state, and whether a facility was governmentowned, hospital-administered or a member of a multi-facility
organization. Analyses included evaluation of state-level
variation in adoption rates and logistic models of adoption at
the facility level, including interaction terms to assess the
effect of government intervention in the presence of QPM
factors. Econometric methods were incorporated to interpret
interaction terms.
Population Studied: Selected using administrative criteria,
QIOs mailed surveys to all Medicaid- or Medicare-licensed
facilities in twelve states, and promoted SOP implementation
to facilities in seven states. Pre- and post-intervention survey
response rates were eighty percent of 4,200 eligibles.
Analyses of the likelihood of SOP adoption were done using
the 2,133 matched NFs not using SOPs at baseline and for
which complete data were available.
Principal Findings: Facilities in states with higher prevalence
of SOP adoption at baseline were more likely to adopt SOPs
during the study period. In addition, six of twenty-eight
interaction terms attained statistical significance for some
proportion of observations: government intervention was
effective in the group of 310 facilities in which ninety percent
received the highly-intense SOP promotion and privately
owned facilities were more prevalent than the study
population; but was ineffective in the group of 169 facilities
with low nursing resources and government-resource
dependency, despite receipt of high-intensity promotion.
QPM factors also influenced SOP adoption in groups not
receiving government intervention.
Conclusions: Evidence indicates both resource dependency
and organizational innovation factors are influential as posited
by the QPM. SOP adoption is facilitated by greater intensity of
QIO interventions except in facilities with low dependence on
government resources and nurse staffing levels.
Implications for Policy, Delivery, or Practice: Policy
implications include the need for government to alleviate
barriers to innovations, increase intensity of QIO programs,
focus on facilities that accept innovations and consider
alternative policies in resistant populations.
Primary Funding Source: CDC, Centers for Medicare &
Medicaid Services
●Physician Characteristics Associated with Potentially
Avoidable Hospitalizations for Medicare Beneficiaries
Ann S. O'Malley, M.D., M.P.H., Hoangmai H. Pham, M.D.,
M.P.H., Deborah Schrag, M.D., M.P.H., Peter B. Bach, M.D.,
MAPP
Presented By: Ann S. O'Malley, M.D., M.P.H., Senior
Researcher, Suite 550, Center for Studying Health System
Change, 600 Maryland Ave. S.W., Washington, DC 200242512; Tel: (202) 554-7569; Fax: (202) 484-9258;
Email: aomalley@hschange.org
Research Objective: To identify physician and practice
characteristics associated with ambulatory care sensitive
condition (ACSC) hospitalizations among Medicare
beneficiaries with a usual physician.
Study Design: We conducted a cohort study using time-toevent analysis of three years (2000-2002) of Medicare claims
for beneficiaries treated by respondents to the nationally
representative 2000-2001 Community Tracking Study (CTS)
Physician Survey. The beneficiary’s usual physician was
identified in 2000 and claims were examined to determine
whether an ACSC hospitalization occurred in 2001 or 2002.
Physician characteristics (from CTS) of primary interest
included: years in practice, board certification, US vs.
international medical school attendance, and specialty type.
Practice characteristics of primary interest were practice size
and practice revenue derived from Medicaid and managed
care. Outcomes were potentially avoidable hospitalizations for
six prevalent ACSCs (congestive heart failure, chronic
obstructive pulmonary disease (COPD), long term
complications of diabetes, bacterial pneumonia, dehydration
and urinary tract infection.) Potentially avoidable
hospitalizations were assessed for each condition
independently as well as for the composite outcome of any
ASCS.
Population Studied: 522,061 Medicare beneficiaries, age 65
and older, whose usual physicians (n=6,862) were primary
care generalists or specialist respondents to the CTS physician
survey. Institutionalized beneficiaries and those with end stage
renal disease were excluded. Those who entered nursing
homes, hospice, HMOs or who died were censored.
Principal Findings: In multivariate models, beneficiary
socioeconomic status, prior ACSC hospitalizations, and
comorbidities had the strongest associations with ACSC
hospitalizations (p<.0001 for each); and, controlling for these
attenuated the associations between physician/practice
characteristics and hospitalization. There was a graded
association between physician years in practice and ACSC
hospitalizations. Having a usual physician who had been in
practice for less than 10 years (vs. 20 years or more) (Hazard
Ratio 1.11: 95% Confidence Limits 1.05-1.20), or who had
received medical education outside the U.S. (HR 1.09: 95%CL
1.01-1.18), was associated with higher risk of an ACSC
hospitalization. Beneficiaries whose usual physician was a
primary care generalist had lower rates of hospitalization for
COPD and bacterial pneumonia. Beneficiaries with a specialist
as usual physician had lower rates of hospitalization for
diabetic complications. Beneficiaries cared for in a practice
that derived a higher percentage of its revenue from managed
care (>30%) and a lower percentage from Medicaid (<5%) had
lower ACSC hospitalization rates overall (p<.0001 for both).
There were no significant associations between board
certification or practice size and ACSC hospitalization.
Conclusions: Although socioeconomic and baseline health
status are most influential, physician training and experience,
and sources of practice revenue account for a meaningful
degree of variation in older adults’ risk of hospitalizations for
ambulatory care sensitive conditions.
Implications for Policy, Delivery, or Practice: The ability to
link ACSC hospitalizations to physician and practice
characteristics enables quality improvement initiatives to
target those providers and patients with the greatest potential
for gain.
Primary Funding Source: NIA
●Randomized Trial to Improve Appropriate Aspirin Use in
Patients with Diabetes
Stephen Persell, M.D., M.P.H., Therese A. Denecke-Dattalo,
MS, APRN, FNP, Daniel P. Dunham, M.D., M.P.H., David W.
Baker, M.D., M.P.H.
Presented By: Stephen Persell, M.D., M.P.H., Assistant
Professor, Medicine, Northwestern University, 676 N. St. Clair
St. Suite 200, Chicago, IL 60611; Tel: 312-695-1524; Fax: 312695-0951; Email: spersell@nmff.org
Research Objective: Information technology interventions
targeting patients and physicians could increase appropriate
prescribing beyond that achieved through physician-directed
approaches alone. We sought to compare a patient-directed
quality improvement approach with clinician reminders alone
for patients with diabetes who were not taking aspirin as
recommended by guidelines.
Study Design: We performed a cluster-randomized trial by
randomizing physicians to two approaches: 1) computerized
reminders presented to clinicians at office visits, and 2)
reminders plus an active intervention. The active intervention
consisted of 1) emailing physicians to ask whether aspirin was
indicated for each patient, 2) mailing an information sheet to
patients followed by a nurse telephone call to patients whose
physician approved aspirin use and 3) prescribing aspirin by
the nurse if she did not identify a contraindication. We
randomized 19 physicians caring for 334 eligible patients. We
assessed patient-reported regular aspirin use (daily or every
other day) by phone at 4 to 6 months. Because the
interventions were not expected to change aspirin use in
current users or patients with contraindications, this outcome
was also assessed for the subgroup of patients who did not
report long-term aspirin use or a contraindication to aspirin.
These intention-to-treat analyses used bivariable logistic
regression with generalized estimating equations to account
for clustering by physician.
Population Studied: Using data from the electronic medical
record (EMR), we selected patients of participating physicians
from a large urban internal medicine primary care practice
who had diabetes mellitus; were over age 40 years; did not
have aspirin, another antiplatelet drug, or warfarin on their
medication list; and did not have an allergy to aspirin or a
non-steroidal anti-inflammatory drug.
Principal Findings: Intervention group physicians answered
97% of emails and approved the intervention 64% of the time.
We completed outcome assessment interviews for 242
patients (73% of intervention and 71% of reminder only
patients). At follow up, regular aspirin use was reported by 60
of 130 (46%) intervention patients and 44 of 112 (39%)
reminder only patients, a non-significant 7.2% difference (95%
CI, -3.9 to 18 percentage points, P=0.20). In the subgroup of
patients who reported no aspirin use at baseline and no
contraindications, 33 of 76 (43%) intervention and 22 of 74
(30%) reminder only patients used aspirin, a 10% difference
accounting for clustering (95% CI, 2.2 to 18 percentage points,
P=0.013). Of intervention patients interviewed by the nurse
who were advised to start aspirin, 33/46 (72%) were taking it
at the follow up assessment. Obstacles included: difficulty
reaching patients by phone, real or perceived
contraindications (29% of nonusers reported a medical
reason for not taking aspirin), and failure to follow the nurse´s
advice.
Conclusions: A patient-directed intervention modestly
increased aspirin use among diabetes patients beyond that
achieved using computerized clinician reminders for ideal
candidates. This effect is in addition to that achieved using
clinician reminders alone. Logistic barriers, failure to follow
the nurse´s recommendation and contraindications to aspirin
(real or perceived) limited the intervention´s effectiveness.
Implications for Policy, Delivery, or Practice: Using EMRs
to identify patients for targeted interventions to improve
quality or safety has great potential. However, future studies
are needed to explore ways to overcome the barriers we
identified so that these interventions can be routinely done in
a cost-effective way.
Primary Funding Source: Department of Medicine,
Northwestern University
●How a new United Kingdom Hospital met its Length of
Stay Target
Margaret Peterson, Ph.D., Jeanne Marie Cioppa-Mosca, MBA,
Eileen Finerty, Suzanne Graziano, RN, Sue King, Nicola Judge,
MSC, Jane Harrison, MSCP, Thomas J. Sculco, M.D.
Presented By: Margaret Peterson, Ph.D., Director Applied
Statistics Core, Research, Hospital for Special Surgery, 535
East 70th St, New York, NY 10021; Tel: 212 606 1916; Fax: 212
606 1160; Email: petersonm@hss.edu
Research Objective: In 2000, The Royal College of Surgeons
of England and the British Orthopaedic Association published
a report on the long waiting lists for elective surgery.
Concentrating on total hip replacement, they reported the
length of stay for 57% of the patients as 8 to 10 days and
waiting times of more than 6 months for an operation. They
recommended reducing the length of stay so that available
beds would reduce the waiting list. They also recommended
consulting with an institution whose length of stay was shorter
and whose reputation was high. The decision was made to
build a new hospital to replace the five NHS trusts whose data
was used in the report, and Hospital for Special Surgery, New
York City (HSS), was selected as the resource partner. The
object of this analysis is to demonstrate that the length of stay
for total hip surgery was reduced successfully to approximately
half the former value for the new hospital.
Study Design: This is a descriptive study with a historical
control and comparison to the current norms at the resource
partner. Data were collected during the first twelve months of
the hospital operation.
Population Studied: The data was collected on patients
undergoing total hip replacement at the South West London
Elective Orthopaedic Center, SWLEOC, from January to
December 2004. SWLEOC is a new hospital replacing the 5
NHS trusts whose data was reported in 2000 and serves the
same population. SWLEOC opened in the middle of January,
so to allow for start up, data was analysed from February
through December 2004, n=615 primary unilateral total hip
patients. The results are compared to the values reported in
the 2000 report and to equivalent data from HSS, n=1506
primary unilateral total hip replacements in the same period
2004. All data was collected and handled in accordance with
the HIPPA and European privacy regulations. The key quality
concepts transferred from the HSS to the U.K. team were
surgical and hospital throughput, the hospital facility design, a
standardized Interdisciplinary Preoperative Patient Education
Program, the infection control standard and a standardized
rehabilitation model.
Principal Findings: In 2004, the average length of stay in the
new hospital was 6.1+ 3.0 days. This was down from the
values reported in 2000 for the years 1996-97, p<0.001. In the
earlier period 15% stayed 8 days or less, compared with 86%
in 2004. The overall length of stay for all procedures at the
new institution was on target at less than 5 days. In addition
the infection rate by MRSA in the new hospital was zero
compared to a reported rate of 9% in the 2000 report.
Conclusions: The international cooperation was successful in
its aim to reduce length of stay. Despite cultural differences it
was possible to institute a “best practices” model, which
maximized hospital operational flow and throughput.
Implications for Policy, Delivery, or Practice: Cross-cultural
exchanges can be successful at an international level, with
careful planning and a rigorous structure of the
implementation of new practices.
Primary Funding Source: No Funding
●Why Sexual Assault Nurse Examiners Matter in the
Emergency Department
Stacey Plichta, Sc.D., Paul Clements, Ph.D., APRN, BC, DFIAFN, Clare Houseman, Ph.D. APRN BC
Presented By: Stacey Plichta, Sc.D., Graduate Program
Director, College of Health Sciences, Old Dominion
University, 204 Spong Hall, Norfolk, VA 23546;
Tel: (757) 683-4989; Email: splichta@odu.edu
Research Objective: This study seeks to describe the different
models of care for sexual violence victims in the emergency
department (ED) and examines how these models relate to
the quality of the care provided.
Study Design: A cross-sectional survey of the Nursemanagers for all 82 EDs in the Commonwealth of Virginia was
conducted. Potential participants were mailed a survey about
services for sexual violence victims; non-respondents were
invited to participate via telephone, e-mail or fax. Overall, 62
responded (Response Rate = 76%).
Population Studied: Respondent EDs were similar to EDs
across Virginia in terms of size and ownership. Sligthly over
one-third (36%) were small (25-99 beds), 32% were medium
(100-199 beds) and 32% were large (200+ beds). Over half
(58%) were part of a network of hospitals. The majority (65%)
were not-for-profit, 27% were for-profit and 8% were
educational and/or gov't owned hospitals.
Principal Findings: Four models of care were identified based
on the presence of sexual assault nurse examiners (SANEs): 1)
no specific SANE services (27.5%), 2) sub-contracting of
SANE services (16.1%), 3) having SANEs employed by the
hospital, only available sometimes (14.5%), and 4) having
SANEs employed by the hospital and always available (41.9%).
The larger hospitals and the for-profit hospitals were much
more likely to employ the fourth model. Those with the fourth
model provided a higher quality of care as measured by both
structural and process characteristics. Subcontrating out
SANE services or providing partial coverage did not perform
as well as expected.
Conclusions: Those hospitals with full-time SANE coverage
were more likely to have a formal training plan for all ED staff,
to provide annual training to their staff, to have linkages to
rape crisis centers and to participate on community-wide
Sexual Assault Response Teams.
Implications for Policy, Delivery, or Practice: Ideally, all EDs
would have SANEs on staff and available at all times. This is
not likely to be feasible, especially for the smaller hospitals.
Those EDs employing other models of care should seek to
ensure that consistent care is provided for victims, that annual
training is provided to staff and that linkages to rape crisis
centers are build and maintained.
Primary Funding Source: Other Government
●Propoxyphene: Is it Really Inappropriate?
Mary Jo Pugh, Ph.D., RN, Monica Horton, M.D., Francesca
Cunningham, PharmD, Jacqueline Pugh, M.D., Dan R.
Berlowitz, M.D., M.P.H.
Presented By: Mary Jo Pugh, Ph.D., RN, Research Health
Scientist, VERDICT, South Texas Veterans Healthcare System,
7400 Merton Minter Blvd (11C6), San Antonio, TX 782294404; Tel: 2106175300 x7193; Fax: 210-567-4423;
Email: pughm@uthscsa.edu
Research Objective: Propoxyphene, a weak opioid, is
commonly used for pain in the elderly. Some experts argue
that the risk of adverse effects outweighs the benefit of pain
relief that is comparable to non-narcotic agents. While this
drug is included in the new HEDIS list of drugs to avoid in the
elderly, and is the most commonly used potentially
inappropriate drug in the elderly, its inclusion is hotly debated.
Older patients report that propoxyphene provides better pain
relief than non-narcotic agents with fewer adverse effects than
slightly stronger narcotics such as codeine. Because studies
assessing the analgesic efficacy of propoxyphene were
conducted on younger patients, it is possible that older
patients are more sensitive to propoxyphene, as sensitivity to
opioids tends to increase with age. This study begins to
assess the risk-benefit ratio for propoxyphene by comparing
adverse patient outcomes associated with use of
propoxyphene, codeine (a similar mild opioid), and an
unexposed control group.
Study Design: This retrospective cohort study used
administrative and pharmacy data from the Veterans Health
Administration (VA) to identify older patients who had no
prior use of propoxyphene or codeine in fiscal year (FY) 1999,
and had new use of propoxyphene or codeine respectively
during the first half of FY 2000. A control group with no
exposure to either drug was also identified. Outcomes for
codeine and propoxyphene groups were assessed after initial
exposure through September 30, 2000; outcomes for the
control group were assessed from the median start date for
propoxyphene and codeine (12/07/99) through September 30,
2000. Injury outcomes (hip fracture, other fracture, head
injury, falls) were identified using validated ICD-9 algorithms.
Hospital and emergency care was identified in VA
administrative data. Multivariable logistic regression analyses
determined if propoxyphene patients had higher rates of injury
or healthcare utilization than codeine and control groups after
controlling for age, sex, race and disease burden defined as a
count of 30 physical and 6 mental health diagnoses, presence
of diagnoses that increase risk of falls (e.g. Parkinson’s
disease, epilepsy, dementia), the number of unique drugs
prescribed in FY00, and FY99 healthcare utilization (hospital,
emergency, geriatric care).
Population Studied: US Veterans >65 years on October 1,
1999, with at least two VA outpatient visits during FY99.
Principal Findings: Patients in the control group were far less
likely than propoxyphene or codeine exposed patients to
experience any measured adverse outcomes. Patients
prescribed codeine were 50% more likely to have an
emergency visit (OR 1.5; 95% CI 1.3-1.8) and 30% more likely
to have a fracture than those prescribed propoxyphene (OR
1.3; 95% CI 1.2-1.5); hospitalization rates were similar (OR 1.1;
95% CI 1.1- 1.2) Propoxyphene users were more likely to have
a fall than codeine users (OR 1.13; 95% CI 1.03-1.30).
Conclusions: Propoxyphene appears to have fewer adverse
outcomes than codeine.
Implications for Policy, Delivery, or Practice: Despite being
classified as inappropriate for the elderly by the Beers criteria
in 1991, controversy has surrounded this classification, and
rates of propoxyphene use have remained stable at about 6%.
Research comparing the analgesic properties of these drugs in
an elderly population would help clarify the risk-benefit ratio of
propoxyphene.
Primary Funding Source: VA
●Cross-National Comparison of Three Versions of the
ICD-10 Charlson Index
Hude Quan, M.D. Ph.D., Vijaya Sundararajan, M.D., M.P.H.,
FACP, Patricia Halfon, M.D., K. Fushimi, M.D., Ph.D., William
A. Ghali, M.D., M.P.H.
Presented By: Hude Quan, M.D. Ph.D., Assistant Professor,
Community Health Sciences, University of Calgary, 3330
Hospital Dr. NW, Calgary, T2N 4N1; Tel: (403) 283 5307;
Email: hquan@ucalgary.ca
Research Objective: This study was conducted to compare
the properties of the Australian, Canadian and Switzerland
versions of ICD-10 coding algorithms using data from four
countries.
Study Design: This is an analysis of hospital discharge data
from four countries.
Population Studied: Hospital discharge data from Alberta,
Canada (2002/2003, up to 16 diagnosis codes); Canton de
Vaud, Switzerland (1999/2001, unlimited number of
diagnoses), Victoria, Australia (2000-2001, 25 diagnoses) and
Japan (2003, 11 diagnoses) were used for the analysis. For
patients with more than one admission, only the first
admission with a length of stay of 2 days or longer for each
patient was included. Three ICD-10 coding algorithms were
applied in these four datasets to define Charlson
comorbidities. Logistic regression was fitted using in-hospital
mortality as the dependent variable and individual
comorbidities as independent variables and then was fitted
again using weighted Charlson index score as independent
variables. C-statistic and its 95% confident interval (CI) were
employed to evaluate model performance.
Principal Findings: Inpatients from Alberta, the Canton de
Vaud and Victoria were similar in age, in-hospital mortality
and length of stay, whereas Japan´s inpatients were older, with
higher mortality and longer length of stay. Within each
locality´s data, the distribution of comorbidity levels was
similar across the three translations of the coding algorithms.
The models with either individual comorbidity or Charlson
score as independent variables produced slightly higher Cstatistic for Canadian version than for Australia and Swaziland
version in each dataset. For example while fitting the logistic
model using individual comorbidities in Japan data, C
statistics was 0.709 for Australian coding algorithm, 0.712 for
Canadian algorithm and 0.694 for Switzerland algorithm.
However, the difference was not statistically significant.
Conclusions: Our analyses show that although all three
versions of the ICD-10 Charlson Index coding algorithms have
good to excellent discrimination in their ability to predict inhospital mortality, the Canadian algorithm consistently
demonstrates slightly higher discrimination, not only in
Canadian data but also in data from Australia, Switzerland and
Japan. Use of the individual comorbidities in model building is
preferable to the use of the weighted index.
Primary Funding Source: No Funding
●Patient Perceptions of Patient Safety: A New Lens for
Viewing Quality
Cheryl Rathert, Ph.D., Douglas R. May, Ph.D.
Presented By: Cheryl Rathert, Ph.D., Assistant Professor,
Health Management and Informatics, University of MissouriColumbia, 324 Clark Hall, Columbia, MO 65211; Tel: (573) 8843719; Fax: (573) 882-6158;
Email: RathertC@health.missouri.edu
Research Objective: To explore relationships between
patient-centered care and patient safety by developing a
reliable, valid patient perception measure of patient safety.
Although much patient safety research has been conducted
previously, reliable measures are still lacking. Even data that
seem “hard” are often subjectively defined and reported, and
are not consistent across organizations. Further, empirical
research on the patient perception of patient safety is lacking.
The present study developed theory for why the patient
perception of safety is important. In addition, we developed a
measure and tested how patient safety serves as a mediator
between patient-centered care and patient satisfaction.
Study Design: A grounded theory approach was used for
development of theory and measures. Qualitative research
was conducted using focus groups. We then developed a
measure of patient safety, and used a survey methodology to
test the measure and its relations to patient-centered care and
patient satisfaction. A smaller version of the scale was
subsequently included on a large national survey.
Population Studied: Focus groups included adult patients,
family members, nurses, and physicians. The first survey
included adult inpatients (previous ninety days). These
patients were randomly selected from three metropolitan
hospitals in the eastern U.S. (n= 1040). The larger study
included approximately 150,000 respondents from across the
U.S.
Principal Findings: The qualitative phase of the study
indicated that there are attributes of hospital care processes,
that patients can perceive, that are related to patient safety.
These processes were revealed through patient stories of their
experiences. Nurses and physicians independently validated
these perceptions. The initial quantitative survey phase
resulted in several important findings. First, the measure of
patient safety was empirically distinct from measures of
patient-centered care and patient satisfaction. Second,
patients who perceived they were recipients of patientcentered care reported significantly higher levels of patient
safety. These patients also were significantly more satisfied
with their hospital experiences. Patient safety partially
mediated the relationship between patient-centered care and
patient satisfaction. It appears that patients who received
patient-centered care were more satisfied because they felt
more safe. When the measure was used for the large scale
study, it discriminated at the hospital-level of analysis, and
was also correlated with other important hospital-level
outcomes.
Conclusions: Patient-centered care processes appear to be
antecedents of patient perceptions of safety, and perceptions
of safety appear to be necessary for overall satisfaction. In
fact, our data suggest that patient perceptions of safety are
better predictors of satisfaction than are actual experiences of
adverse events. This measure was demonstrated to be useful
for research at the individual- and organization-levels of
analysis.
Implications for Policy, Delivery, or Practice: Experts have
called for health care organizations to “partner” with patients
in order to improve patient safety. Our data suggest that
patient measures of safety would be valuable for revealing
troublesome processes of care before they manifest in adverse
events within the hospital. Furthermore, patient perception
measures would be useful for benchmarking at the hospital
level.
Primary Funding Source: NRC+Picker
●Development of Standardized Measures of Quality of
Cancer Care within Integrated Delivery Systems
Phil Renner, M.B.A., Robert A. Hiatt, M.D. Ph.D., Flory Nkoy,
M.D., MS, M.P.H., David Mosen, Ph.D., Min Gayles Kim,
M.P.H., Mark Hornbrook, Ph.D.
Presented By: Phil Renner, M.B.A., AVP, Quality
Measurement, Quality Measurement, National Committee for
Quality Assurance, 2000 L Street NW, Washington, DC
20036; Tel: 202.955.5192; Fax: 202.955.3599;
Email: renner@ncqa.org
Research Objective: Cancer is the second-most-costly and
lethal disease in the U.S. causing one out of four deaths.
According to American Cancer Society (ACS) estimates, there
will be 1,334,100 new cases of cancer and 556,500 cancer
deaths in 2004. Despite the large cost and mortality
associated with cancer, there are few standardized quality
measures to evaluate the delivery of cancer care at the health
system or provider level in the United States. In addition, due
to the current fragmented nature of cancer care,
operationalization of measures of cancer quality will likely be
facilitated by access to linked data sources found in an
integrated system, such as inpatient and outpatient claims,
mammography data, pharmacy data, pathology, and tumor
registries. Our objective is to extend the learnings from recent
research in quality of cancer care measurement to identify
methods to develop and operationalize measures of cancer
care quality in integrated health care delivery systems.
Study Design: We identified key clinical concepts and areas
for potential evaluation and measurement based on a
literature review of cancer care recommendations and
guidelines, and through an expert panel consensus process.
Population Studied: Patients from two large integrated
delivery systems seeking cancer screening, diagnosis and
treatment.
Principal Findings: A two-dimensional framework for
measuring quality of care in cancer was developed. The first
dimension spans the type of care being measured, which
includes cancer site-specific technical clinical quality and
clinical measures that can be applied across cancer sites
including the experience of care by patients. The second
dimension consists of phases in the cancer continuum from
screening through diagnosis, treatment, and palliative care to
the end of life. Measure identification and development were
prioritized to the domains of breast cancer, colorectal cancer,
prostate cancer, and measures of symptom management for
people with cancer, due to their contribution to the overall
cancer burden and to remain consistent with other national
efforts. Between five and ten measures were selected for
testing in each of the four domains based on evidence and
expert panel consensus. Thereafter, needed data systems (e.g.
tumor registry, claims data, pathology data) as well as
necessary data elements for these measures were identified
and assessed for overall feasibility. For measures that require
data elements from multiple data sources, the unique patient
identifier index number available in each source provided a key
element to link these data sources. Validation tests for data
retrieval algorithms are being developed as well as an
assessment of whether the use of these measures is
associated with better outcomes
Conclusions: We have identified a meaningful set of cancer
quality measures that can be used for accountability, ongoing
quality assessment and improvement activities. Our approach
exemplifies principles needed to develop and operationalize
standard measures of cancer care quality
Implications for Policy, Delivery, or Practice: There is a
growing concern among purchasers, payers, regulators, and
consumers regarding the lack of standardized quality data in
cancer care. Creative solutions will be required to develop and
report meaningful, reliable quality measures for the
assessment of cancer care delivered to people in the United
States. Methods and measures can be identified and
developed that will be of value in quality improvement and
monitoring for both integrated health care systems and
eventually for use in other settings nationally.
Primary Funding Source: No Funding
●Profiling Providers of Outpatient Physical Therapy
Linda Resnik, Ph.D., PT, OCS, Dawei Liu, Ph.D., Vince Mor,
Ph.D., Sharon-Lise Normand, Ph.D., Dennis Hart, Ph.D.
Presented By: Linda Resnik, Ph.D., PT, OCS, Assistant
Professor, Community Health, Brown University/ Providence
VA Medical Center, 2 Stimson Avenue, Providence, RI 02906;
Tel: 401-863-9214; Email: Linda_Resnik@Brown.edu
Research Objective: Research to improve service delivery in
outpatient rehabilitation is a priority given the economic
impact and high volume of patients with low back related
syndromes. At this time we lack understanding of organization
level and service delivery factors related to quality of care in
physical therapy. The purposes of this research are to 1) profile
physical therapy clinics by quality of care 2) quantify the
relationship between clinic and caseload characteristics and
the quality profile that we develop and 3) examine the
relationship between clinic quality profile and service
utilization.
Study Design: Patient functional health status (FHS)
measured at discharge was modeled using a 3-level
hierarchical linear model (HLM) in which patients were
nested within therapists, therapists nested within clinic, and
confounding factors associated with outcomes (age, FHS at
intake, acuity, surgical history, gender, payer type, exercise
history and employment) added into the model. Inverse
probability weighting was used to control for bias due to
missing follow-up data. After modeling, residual scores were
aggregated by clinic. Clinics were classified into best (the
upper 25%), middle (26-75%) and worst (lower 25%) groups,
using aggregated risk adjusted residual scores. Relationships
between clinic group and staffing and caseload characteristics
were examined by logistic regression. Models included
variables measuring number of physical therapists (PTs) on
staff, ratio of PTs to physical therapist assistants (PTAs), years
of PT experience, proportion of patients with low back
syndromes, and number of new patients per month per PT.
The relationship between clinic group and number of visits per
treatment episode was examined in an HLM of patients
nested within therapists, and therapists nested within clinics,
controlling for patient characteristics.
Population Studied: The study sample consisted of 114
outpatient rehabilitation clinics participating with the Focus
On Therapeutic Outcomes, Inc. (FOTO) database in 20002001. Clinics treated 17,710 patients (40.6% male, age 49±6)
with lumbar syndromes. Each clinic had a minimum of 40
patients with low back pain (mean 155, sd 142)
Principal Findings: Mean FHS discharge scores for the best
clinic group were 73.5 ±18.0, the middle 62.5±21.0, and the
worst 69.1 ±19.2. The ratio of PTs to PTAs was predictive of
being in the best clinic category as compared to the worst
(P<.05). The best clinics had mean ratios of PT/PTA of 4.5
±3.4, middle 3.9 ±3.9 and the worst group 3.0 ±1.5. No other
staffing or caseload characteristics were associated with clinic
category. Clinic category was not associated with number of
visits per treatment episode.
Conclusions: We observed differences in patient outcomes as
well as staffing patterns in the best, average and worst clinics.
Clinics staffed with higher proportions of PTs as compared to
PTAs delivered the best quality care as gauged by risk adjusted
patient outcomes.
Implications for Policy, Delivery, or Practice: These findings
can be used by physicians who refer patients to physical
therapy, therapists seeking employment, and consumers when
evaluating characteristics of clinics likely to be associated with
higher quality care. These results can also inform managers of
rehabilitation services in decision making regarding optimal
staffing patterns.
Primary Funding Source: NICHD
●Parent-Centered Quality Improvement (QI): How a
Parent-Based Survey can be used to Design QI Efforts
Focused on Developmental Services for Young Children.
Colleen Reuland, MS, Christina Bethell, Ph.D., M.P.H., M.B.A.,
Rasjad Lints, M.D., Scott Shipman, M.D.
Presented By: Colleen Reuland, MS, Senior Research
Associate, Pediatrics, CAHMI, Oregon Health and Science
University, Mailcode CDRCP, 707 SW Gaines Street, Portland,
OR 97239-3098; Tel: 503-494-0456; Fax: 503-494-2475;
Email: reulandc@ohsu.edu
Research Objective: To assess how a consumer-centered
quality measurement tool--the Provider-Level Promoting
Healthy Development Survey (ProPH.D.S-- can be feasibly and
validly implemented in a health system in order to gather
information to design QI efforts at the system-, office- and
provider-level. To examine variations in ProPH.D.S findings by
office and provider characteristics in order to identify leverage
arms for QI. To understand how the ProPH.D.S findings
should be reported to providers and system leaders in order to
motivate and shape parent-centered QI efforts.
Study Design: The ProPH.D.S was administered to generate
quality information on eight communication-dependent
aspects of preventive and developmental services for young
children and to generate quality data for all pediatric offices
and individual providers in one large health plan within an
entire metropolitan area. Quantitative variation and
association of ProPH.D.S scores by office- and provider-level
descriptive information (including an office system inventory,
examination of electronic medical chart prompting systems
and parent education materials, and provider characteristics
(gender, FTE, age) was assessed. Informant interviews (N=11
providers, N=6 health system leaders) and presentations at
pediatric business meetings were conducted focusing on
reporting qualty findings to stimulate and shape QI and
assess the added value of a consumer-centered quality
measurement tool and improvement model.
Population Studied: The ProPH.D.S was administered to a
stratified, random sample of parents of children under age
four enrolled in a managed care organization who had at least
one well-visit. The sampling strategy was designed to allow for
office- and provider-level analyses (children sampled = 5003;
offices represented=10; providers represented = 56).
Principal Findings: Findings indicate significant
opportunities for improvement on each aspect of care
measured in the ProPH.D.S (scores: 25-81). ProPH.D.S scores
vary significantly (p < .05) across pediatric offices on 4 of 6
measures and vary significantly (p < .05) for all measures
across individual providers. No office scored the highest or
lowest on all measures. Offices with systems addressing
developmental services and providers with electronic medical
chart prompts and parent education materials focused on
developmental services scored higher on the PH.D.S. Key
barriers identified by providers include lack of office staff to
assist in providing developmental services and lack of
awareness about where parents are not having their
informational needs met. In order to stimulate/design QI
efforts, providers want reports that provide general
(composite-level) and specific (item-level) findings with
graphic and narrative presentation of the data. Benchmark
data adds value/incentive to improve. Health system leaders
want reports providing background evidence, how the PH.D.S
compares and contrasts to current quality data, and
office/provider characteristics associated with higher quality.
Conclusions: The ProPH.D.S provides quantitatively unique
and actionable information stimulating learning necessary for
improvement and ensures QI efforts are aligned with the
needs of parents and children.
Implications for Policy, Delivery, or Practice:
Communication-dependent aspects of clinical care, such as
recommended preventive and developmental services for
children, are essential to include in quality improvement
efforts and require the use of consumer survey data.
Integration of this data with utilization and system
information provides a powerful platform for QI.
Primary Funding Source: No Funding
●Improving Breast Cancer Care in a Managed Care
Population through Multi-faceted Collaboration
Nancy Rodriguez, RN, BSN, CPHQ, PAHM, Sandra L. White,
M.D.
Presented By: Nancy Rodriguez, RN, BSN, CPHQ, PAHM,
Senior Clinical Quality Compliance Administrator, Quality
Management, Blue Cross Blue Shield of Georgia, 3350
Peachtree Rd, NE; MSC G00608, Atlanta, GA 30326; Tel:
(404) 848-2334; Fax: (404) 842-8480;
Email: nancy.rodriguez@bcbsga.com
Research Objective: Radiation therapy following breast
conserving surgery has been shown to reduce the local
recurrence of early stage breast cancer; and is an element of
quality breast cancer care. Review of claims data revealed that
our members were not substantially better than the national
average in seeking this treatment. Since we know that it
should not be due to lack of insurance, this initiative seeks to
find out why these patients are not getting this effective
treatment; and will provide awareness and targeted
interventions to promote informed decision making; improve
appropriate treatment selections; evaluate disparities; and
measure impact on treatment outcomes.
Study Design: Two Advisory Panels, one of physicians and
one of layperson survivors and support organization leaders,
were formed to advise on program development. Key
messaging was identified for both the physician audience and
breast cancer patients. These were then considered during
the process of selecting and developing materials for this
effort, as well as in structuring the rollout plan. Interventions
are designed to provide complete information to patients to
support informed decision making, delivered in a 6-8 grade
reading level, and suitable to a diverse audience throughout
Georgia. Educational materials were distributed to patients
and physicians, and have been made available on our
company website. A feedback survey is included in each
packet mailed out.
Population Studied: This project has been developed for all
Blue Cross Blue Shield of Georgia lines of business,
encompassing HMO, POS, PPO and Indemnity products.
Patients diagnosed with breast cancer in 2004 and thereafter
were identified for inclusion in the project. Patients were
included regardless of state of residence, that is, on a national
basis as appropriate. Physicians in Georgia were also
included in the intervention. This included: Adult Primary
Care Physicians, OB/GYN physicians, General Surgeons,
Oncologists, Hematologists, Radiation Oncologists, and
Hematology Oncologists. At his point, we have distributed
educational materials to over 6,000 patients nationally; and
over 5,200 physicians at over 7,200 office locations
throughout Georgia. Our ongoing activity is expected to reach
over 3,000 patients annually.
Principal Findings: Initial evaluation of program materials
has been extremely positive. Physicians attending a Breast
Cancer Continuing Medical Education program were asked to
evaluate materials developed for the project, including: the
Resource List, the Patient Education Booklet, the Patient
Checklist, and the General Treatment Guidelines for
Physicians. Each piece of material was evaluated individually
as well as within an overall evaluation of the materials.
Physicians were asked to rate for each item, the level of
helpfulness or usefulness to the patient, and/or to themselves.
Questions were also asked to evaluate the desired outcome or
intended use of the materials. The average of all responses to
a given question fell in the highest rating category, strongly
agree; and all scored above the midpoint of the category,
above 4.50. The materials in general are deemed suitable for
patients, and physicians will use them in their practice. The
Resource List will be helpful in locating additional sources of
support or information. Physicians felt the Patient Education
Booklet covered topics important to both patient and
physician, would be helpful to both, and would help the
patients better understand breast cancer and how to cope
with it. The characteristics of the Patient Checklist made it
valuable to both the patient, to better understand her
treatment options and discuss with her doctor; and the
physician, to give guidance for discussing treatment options
with the patient. The General Treatment Guidelines gave
physicians enough information to discuss treatment options
and would be helpful to the physician in discussions with the
patient.
Conclusions: Feedback from the distribution of materials has
demonstrated that this program is valuable to patients and
physicians. Additional study is warranted to determine
physician practice characteristics and to identify possible
disparities in care and treatment.
Implications for Policy, Delivery, or Practice: We are
undertaking a retrospective, health plan administrative
database and linked chart abstraction study, investigating the
diagnosis, treatment patterns, and healthcare resource
utilization in Blue Cross Blue Shield of Georgia patients with
breast cancer. The analysis is projected to be a cross-sectional
5-year study in a large managed care database. The research
will complement further development of ethnically diverse
breast cancer prevention initiatives and effective strategies for
breast cancer diagnosis and treatment among ethnically
diverse populations.
Primary Funding Source: Blue Cross Blue Shield of Georgia
Quality Management Budget
●Identifying Risk Factors for Predicting Incidence of
Hospitalization for Home Health Care Patients Within the
First 60 Days of Admission
Robert Rosati, Ph.D., Liping Huang, MA
Presented By: Robert Rosati, Ph.D., Director of Outcomes
Analysis and Research, Center for Home Care Policy &
Research, Visiting Nurse Service of New York, 5 Penn Plaza,
12th Floor, New York, NY 10001; Tel: 212-609-5776; Fax: 212290-3756; Email: Robert.Rosati@vnsny.org
Research Objective: In home health care there is limited
research that focuses on identifying the risk factors related to
who is likely to be hospitalized after admission. The focus of
the current study was to identify risk factors that are
associated with an event of acute care hospitalization within
60 days of admission based on start of care information.
These factors would be validated and then used in a model
that would identify patients who are at risk for hospitalization
at the time of admission. In addition, we plan to explore
whether early clinical intervention with high-risk patients will
reduce the rate of hospitalization.
Study Design: A case-control study was designed based on
retrospective data. Logistic regression analysis was performed
to predict the likelihood of hospitalization within the first 60
days of admission. Validation was performed using a
separate sample and an assessment of whether patients
identified as at risk would be similarly identified by clinicians.
Data included administrative information (e.g. demographics,
payer, etc), OASIS start of care assessments, medication
utilization information at start of care and the physician’s plan
of care (e.g. certified nursing, home health aides, and social
work visits). When available clinical information from prior
episodes of care was also included (e.g. had an event of
hospitalization during the prior 6 months). In addition,
electronic medical record information was included in the
model (e.g. temperature, pulse, blood pressure, BMI, blood
glucose level, dialysis, severity of diabetes).
Population Studied: Patients who were admitted to an adult
acute care program at a large urban home health care agency
from September 2004 to April 2005. The first event of
hospitalization occurring within the first 60 days of care was
used as a response variable. The total number of cases in the
study was 46,159. In the study sample 19.9% had a
hospitalization within the first 60 days of care.
Principal Findings: The risk model developed had a C
statistic equal to .71 and included significant variables from all
of the domains of data. Some of the strongest predictors of
increased risk of hospitalization included higher number of
certified nursing visits by the physician in the plan of care,
having specific diagnoses (e.g. cancer, CHF, HIV), unhealed
wounds, degree of IADL assistance needed and abnormal vital
signs. Sensitivity and specificity were in the moderate range.
Analysis of the validation sample showed the results could be
replicated and the model did not vary. When the risk values
assigned by the model were compared to clinical judgments in
a blinded review, kappa statistics for each clinician were > .75.
Conclusions: The results of this study show that it is possible
to identify home health care patients who were at risk for
hospitalization using information available at start of care.
Further, the risk identification was shown to concur with
clinical judgments. In the next phase of the study, we will be
using an automated approach to calculating risk values at the
time of admission then assessing whether early intervention
with the highest risk patients can reduce hospitalization rates.
Implications for Policy, Delivery, or Practice: High-risk
screening is an effective approach to allocating appropriate
resources to patients. The present study shows promising
results that indicate it is possible to identify patients at risk for
hospitalizations within the first 60 days of home health care.
The next important step will be testing whether this type of
information can be used to improve patient outcomes.
Primary Funding Source: No Funding
●Linking IT Innovations and Quality of Care: Improving
Proscribing Patterns for Public Mental Health Clients
Aileen Rothbard, Sc.D., Eri Kuno, Ph.D.
Presented By: Aileen Rothbard, Sc.D., Research Professor,
Psychiatry & School of Social Policy, University of
Pennsylvania, 3535 Market Street, room 3014, Philadelphia, PA
19106; Tel: (215) 349-8707; Fax: (215) 349-8715;
Email: rothbard@mail.med.upenn.edu
Research Objective: To evaluate the impact of automated
prescription systems and feedback report mechanisms on the
quality of psychiatric prescribing practices for public sector
clients.
Study Design: As part of a four year research initiative on
health disparities, this study involves four participating mental
health agencies along with a consortium of research centers
and university participants. State of the art automated
prescription software and equipment has been introduced to
public sector community agencies in a large urban setting
environment to facilitate enhanced quality of care. Using a
clinician intervention technique which is similar to academic
detailing, medical directors at two of the four agencies
employing the new technology, provide feedback reports to
clinic psychiatrists regarding prescribing practices considered
to be questionable with respect to evidence based practices.
The reports are generated by research staff monthly using data
obtained from the automated prescribing system. Each report
identifies providers and clients by disorder whose prescribing
patterns are outside the standards considered to be
appropriate. Pre-post analyses are done for each site and the
intervention sites are compared to the control sites on a series
of quality indicators derived from the American Psychiatric
Association, PORT Studies and various other algorithms
found in the literature. A cost analysis was done on
implementing the technology.
Population Studied: The population consists of clinic
patients receiving public assistance (Medicaid) or subsidized
care through county/state mental health funds. Each of the
four mental health clinics has ~1000 persons per year in
treatment for serious mental illness. The majority are minority
patients between the ages of 18-64 that have a diagnosis of
schizophrenia or major depression or both.
Principal Findings: A pre-post comparison during the first
year of the electronic intervention showed decreases in
polypharmacy for schizophrenia patients (8% to 5%) and
increases in use of combination typical/atypical drugs (9%11%). For patients with affective disorders, increases occurred
in combination new/old antidepressants (9% to 25%).
Information has also been gathered, via survey, on the
reasons for prescribing medications (type, dose,
combinations) that are not considered to be evidence-based
practice such as: out-of-range doses, polypharmacy within and
between medication class, discontinuity in prescribing of
antipsychotics or antidepressants. The initial costs of
implementing the technology ranged from $20,572 to $27,549
per agency. Annual ongoing costs were ~ $10,000 per agency.
On a per psychiatrist based expenditure, initial costs per
prescribing physician averaged $2385 and the ongoing cost
per physician averaged $1100.
Conclusions: Improving the quality of prescribing in public
sector clinics is possible using electronic technology. The
intervention must involve full participation by the medical
director and consensus amongst the psychiatrists as to what
constitutes appropriate care. Polypharmacy is reduced
considerably using the feedback reports.
Implications for Policy, Delivery, or Practice: Electronic
prescribing is an efficient and effective means of monitoring
patient medication patterns and physician performance and
reducing inappropriate drugs. The on-going cost is
reasonable and the technology is well accepted by staff
following the transition/learning period.
Primary Funding Source: Other Government
●State Estimates of Risk-Adjusted In-Hospital Mortality
after Stroke.
David Samson, MS, Nancy Hollinger-Samson, Ph.D., Edward
L. Hannan, Ph.D.
Presented By: David Samson, MS, Epidemiology, University
at Albany, 19 Scenic Drive, Poughkeepsie, NY 12603; Tel: (845)
462-4828; Email: david.samson@bcbsa.com
Research Objective: Disparities in stroke outcomes across
geographic areas could be explained by both differences in
case mix and in quality of care. Estimation of risk-adjusted
outcomes can remove some of the contribution of case mix
differences. Remaining disparities can suggest areas in need
of quality improvement. Using a national sample of
discharges from community hospitals, this investigation
sought to identify predictors, available on admission, of stroke
in-hospital death. We examined disparities in in-hospital
mortality across states, before and after adjustment using a
multivariable model.
Study Design: This study used an administrative database to
develop a multivariable model serving the purpose of risk
adjustment. Data came from the Agency for Healthcare
Research and Quality (AHRQ) Healthcare Cost and Utilization
Project (H-CUP). The 2003 Nationwide Inpatient Sample
(NIS) contains data for nearly 8 million discharges from 994
hospitals in 37 states. The analysis was restricted to NIS
records for which the principal diagnosis was acute
cerebrovascular disease and the admission was made on an
emergency basis. Candidate predictor variables included age,
gender, AHRQ comorbidities, Medstat disease stages, payer
type and median household income for a patient’s ZIP code
(by quartile). A variable was entered into the stepwise
multivariable logistic regression model if its presence as a
single variable was associated with increased risk of death and
the p value was < 0.2.
Population Studied: The 2003 NIS is designed to
approximate a 20% sample of U.S. community hospitals. It
samples from all nonfederal, short-term, general and other
specialty hospitals. Sampling probabilities are proportional to
the number of U.S. community hospitals in each of several
strata defined by hospital characteristics. Discharge abstract
data are provided for 100% of that hospital’s 2003 discharges.
Principal Findings: NIS records with a principal diagnosis of
acute cerebrovascular disease and emergency admission
numbered 91,229. Exclusions totaled 6,091, based on missing
data for in-hospital death and disease stage (all cases from
PA), thus the final data set included 85,138 records from 36
states. The total number of deaths was 9,053, giving a total
mortality of 10.633%. Significant predictor variables in the
final model included age 81-90, age > 90, congestive heart
failure, coagulopathy, drug abuse, liver disease, renal failure,
disease stages 3-5, Medicare as payer, self-pay, and the lower
two income quartiles. The c statistic value was 0.769. Seven
states (AZ, OH, MD, IA, FL, MI, IL) were found to have riskadjusted in-hospital mortality that was significantly lower than
total mortality and 9 states had significantly higher riskadjusted mortality (TX, NC, TN, NY, MA, SC, KY, RI, WV).
Observed (unadjusted) mortality varied across states by a
factor of 6.5; risk-adjusted mortality varied by a factor of 6.1.
Conclusions: There is considerable geographic variation in
observed and risk-adjusted in-hospital mortality after stroke.
Predictors include age, selected comorbidities, disease stage,
payer type and ZIP code income quartile. Limitations in
administrative data, particularly unmeasured variables,
suggest caution in interpreting findings.
Implications for Policy, Delivery, or Practice: Identified
disparities between states in risk-adjusted in-hospital mortality
after stroke should provoke further investigation of the
structures and processes of acute stroke care in need of
quality improvement.
Primary Funding Source: No Funding
●Informatics Techniques for Harmonizing Laboratory Test
Values Across the Mayo Health System
Patricia Schryver, BS, Saralyn Pruett, BA, Bruce Bjorgum, BA,
Peter Amadio, M.D., Kari Toft, MA, George Klee, M.D., Ph.D.
Presented By: Patricia Schryver, BS, Senior Analyst
Programmer, Health Care Policy & Research, Mayo Clinic, 200
1st Street SW, Rochester, MN 55934; Tel: 507-266-4608;
Email: schryver.patricia@mayo.edu
Research Objective: To reconcile the laboratory data at the
Mayo Health System (MHS) sites and at the Mayo Clinic
Rochester (MCR) site, which reside on varied computer
systems, into one repository, the Integrated Clinical Data
Repository (ICDR). To assure harmonization of test values for
patients receiving care across this network. Two approaches
were used; first to standardize the equipment, and secondly to
standardize the reference ranges. In addition, we wanted to
develop a method to confirm and monitor the laboratory data
to ensure that it continues to perform similarly.
Study Design: MHS consists of 13 sites, having 3 different
laboratory information systems in place. Minimal common
test name terminology, or test identification numbering
system existed between the sites. It needed to be determined
which tests compared to one another. In some cases the
reporting units differed, and conversion factors had to be
established to manage a difference in units of measure. To
evaluate whether the test system was giving equivalent data,
the distributions of patient test values were used. The data
between the sites were compared using reference ranges,
measures at the percentages, the percentage of patient values
outside of the references ranges, and the percentage of patient
values that were outside of the key decision ranges. When a
given test result range varied according to the patients gender
or age, a separate analysis on the specified group was
performed.
Population Studied: The analysis population was developed
from MCR electronic laboratory system on patients having
visits between October 2004 and December 2004. At the
MHS sites the last 1000 laboratory test results or all available
results on the requested tests were collected electronically.
Only patients 18 years of age or older were included in the
analysis.
Principal Findings: Tests that were run on the same
instrumentation compared closely and could be integrated as
long as the reference values matched. Tests that used different
methods, or instrumentation and were reported in different
units than MCR, needed further review on how the test should
be reported and whether they could be merged. Some of the
test values had unusual distributions across the reference
ranges, and the results could be seen to change or trend over
time. In cases such as these the methods needed further
review, or new reference range studies needed to be
conducted before proceeding with a merge of data.
Conclusions: In order to merge data into the ICDR a method
of analyzing and standardizing the data had to be established.
A test naming and numbering convention was established. By
calculating the 25th, 50th, and 75th percentiles, and doing a
comparison of percentages outside of the reference range
recommendations could be given for reference values, and
whether test values could be merged.
Implications for Policy, Delivery, or Practice: The
standardization of test values in the ICDR will allow providers
and clinicians within the Mayo organization seamless access
to patient data at any MHS or MCR facility, and enable
laboratory data to be interpreted consistently.
Primary Funding Source: Mayo Foundation
●The Effect of Medicare's Prospective Payment System on
Patient Satisfaction: An Illustration with Four Inpatient
Rehabilitation Facilities
Parag Shah, M.D., Allen W. Heinemann, Ph.D., Larry M.
Manheim, Ph.D.
Presented By: Parag Shah, M.D., Postdoctoral Fellow,
Institute for Healthcare Studies, 339 E. Chicago Avenue Rm
712, Chicago, IL 60601; Tel: 312-503-0448;
Email: p-shah7@northwestern.edu
Research Objective: Congress recently implemented a
prospective payment system to reimburse inpatient
rehabilitation facilities (IRFs). Payment systems provide
incentives that may affect patient outcomes including the
quality of patient care. Patient satisfaction is well established
as an important outcome in health care and has recently been
touted as a measure of health care quality. The objective of
this study was to compare patient satisfaction pre and post
implementation of Medicare’s Prospective Payment System
(PPS).
Study Design: This was a prospective study using a telephone
survey to examine the effects of Medicare’s IRF PPS
implementation. The analysis compared the percent of
patients reporting excellent overall satisfaction ratings pre and
post implementation of PPS. Logistic regression methods
adjusted satisfaction scores for age, gender, respondent,
functional status, and facility.
Population Studied: The patients were selected from four
affiliated Midwestern IRFs, including one rural unit, one
trauma center, one urban center, and one suburban center.
Patients were eligible to complete a telephone survey one
month post discharge. Inclusion criteria for this analysis were
age 18 years and older, IRF length of stay greater than 4 days,
and alive one month post-discharge. 1805 patients discharged
November 2000 – December 2001 (pre) were compared with
2515 patients discharged January 2003 – December 2003
(post).
Principal Findings: Patient characteristics varied across
facilities, and several characteristics were related to overall
satisfaction including admission functional status, gender,
and respondent (patient or proxy). Patient respondents had a
1.26 (95% CI: 1.11,1.43) times higher odds of reporting
excellent satisfaction compared to proxy respondents, and
men were 1.19 (95% CI: 1.06,1.33) times more likely to report
excellent ratings. Admission functional status had a small but
statistically significant positive effect on the likelihood of
reporting an excellent rating. We found an increase in
satisfaction from 56.8% to 63.1% (p<0.05) after PPS
implementation despite a decrease in FIM scores. This
increase is slightly greater (56.1% vs. 63.5%) after risk
adjustment.
Conclusions: Patient characteristics including admission
motor and admission cognitive FIM scores, gender, and
respondent type were associated with patient satisfaction in
inpatient rehabilitation. Using this method of risk adjustment,
percent excellent satisfaction improved at these four facilities
after PPS implementation despite declines in admission
motor and cognitive FIM scores. Aggressive efforts to
improve service quality and assure discharge planning may
have countered deleterious effects due to shorter stays.
Implications for Policy, Delivery, or Practice: A worry about
the effect of a prospective payment system is that it creates
incentives for reducing length of stay and lowering costs.
These incentives may result in lowering patient’s perception of
the quality of care they received and therefore their overall
satisfaction. This was not the case with the advent of this PPS
at four IRFs. Patient satisfaction rose with the payment
system, possibly due to the implementation of various
programs by facilities to improve service quality in anticipation
of PPS.
Primary Funding Source: National Institute on Disability and
Rehabilitation Research
●Designing and Testing a Dental Care Survey for TRICARE
Dental Plan Enrollees
Karen K. Shore, Ph.D., Susan D. Keller, Ph.D., Roger E. Levine,
Ph.D., Christian Evensen, MS, Gary C. Martin, DDS, M.P.H.,
Robert H. Mitton, DDS, M.P.H.
Presented By: Karen K. Shore, Ph.D., Sr. Research Scientist,
Health Program, American Institutes for Research, 1070
Arastradero Rd., Suite 200, Palo Alto, CA 94304; Tel: 650-8438121; Fax: 650-858-0958; Email: kshore@air.org
Research Objective: To design a survey to measure dental
care experiences for individuals enrolled in TRICARE dental
plans, thereby enabling benchmarking of the performance of
these dental plans and identification of opportunities for
quality improvement interventions, using CAHPS instrument
design principles and procedures.
Study Design: To develop a dental care survey based on
CAHPS design principles (e.g., use best scientific evidence
available, measure only things for which the respondent is the
best or only source of information, base the assessment on
respondents’ experiences with specific provider behaviors), we
conducted a literature review, interviewed key informants
(dental care leaders), conducted focus groups with dental plan
enrollees, designed draft surveys, and cognitively tested the
draft survey instruments. Beginning in February 2006, the
surveys will be field-tested and mailed to 6,500 enrollees, with
results available in June 2006.
Population Studied: Twelve key informant interviews were
conducted with dental care experts and stakeholders. A total
of 72 TRICARE dental plan enrollees participated in twelve
focus groups, which were conducted in two geographic
locations (CA and NC). Two rounds of cognitive interviews on
the draft surveys were conducted with a total of 16 dental plan
enrollees. Surveys are being sent to a sample of all eligible
enrollees (i.e., those enrolled in the dental plan for at least 12
months, with a dental visit in the last 6 months).
Principal Findings: Based on the literature review, leadership
meetings, and focus groups, we identified domains of care
and performance that should be assessed. Items were drafted
to allow creation of composite measures for the following
domains: 1) dentist communication, 2) technical quality of
care, 3) office waiting time, 4) office cleanliness, 5) access to
necessary care, 6) access to care quickly, and 7) adequacy of
plan information. Single items were developed for the
following domains: 1) treatment outcome, 2) global rating of
dental care, 3) global rating of dental plan, 4) recommending
the dental plan, and 5) describing the plan as a good value.
Additional demographic items were also developed. A 50-item
survey was developed for active duty family members,
reservists, and family members of reservists; a 47-item survey
was developed for retirees and their family members. During
the cognitive testing, we identified several important issues
associated with developing dental survey items for these
populations, and these issues will be discussed. Preliminary
analyses of the pilot test will be presented, including
descriptive and multivariate analyses of the measures.
Conclusions: The relevant domains for a dental care survey
were identified through a literature review, meetings with
dental care leaders, and focus groups with consumers. This
information was then used to develop survey instruments that
were cognitively tested with consumers. A field test of the
surveys is underway with results expected in June 2006.
Implications for Policy, Delivery, or Practice: These dental
surveys are based on CAHPS design principles and can be
used to provide a standard measure of dental care (in civilian
as well as military settings) that can be used to make national
comparisons across different dental care delivery systems.
Primary Funding Source: TRICARE Management Activity
●Relationship of Organizational Characteristics to Safety
Climate in U.S. Hospitals
Sara Singer, M.B.A., Laurence Baker, Ph.D., Alyson Falwell,
M.P.H.
Presented By: Sara Singer, M.B.A., Doctoral Candidate, Ph.D.
Program in Health Policy, Harvard University, Sherman Hall
303-1, Boston, MA 02163; Tel: (617) 495-5047; Fax: (617) 4964397; Email: ssinger@hbs.edu
Research Objective: The presence of a hospital culture that
emphasizes patient safety is increasingly recognized as a key
to improving patient safety. However, relatively little is known
about hospital safety cultures and their correlates, even about
basic things such as general relationships between safety
culture and hospital characteristics. This AHRQ-funded study
explores relationships between safety culture measures and
hospital characteristics of size, tax status, teaching status, and
being in an urban as opposed to a non-urban area.
Study Design: We performed a cross sectional comparison of
the relationship between hospital characteristics and safety
culture measures derived from the Patient Safety Climate in
Healthcare Organizations survey. The survey was
administered to senior administrators, staff, and physicians at
a stratified random sample of 105 hospitals nationwide in
2004, and produces empirical estimates of the strength of
safety climate, i.e., the perception among personnel of safety
culture at a point in time, overall and along a number of
specific dimensions. Response rates and psychometric
analysis of response patterns support the validity of the data.
Culture measures were combined with data on hospital size,
urban status, tax status, and teaching status from the 2001
AHA survey. Regression analysis related culture measures to
organizational characteristics of interest, controlling for
hospital region and demographic and other characteristics of
the respondents in each institution. We also explored
interactions between hospital size and urban, tax, and
teaching status.
Population Studied: A stratified random sample of 105
hospitals, representing all 4 regions of the U.S. and 3 size
categories. In each hospital, 100 percent of senior managers,
100 percent of active physicians, and 10 percent of other staff
were surveyed for a total of 18,316 individual responses.
Principal Findings: Hospital bed size is statistically
significantly, but non-linearly, associated with changes in
safety climate at p<0.05. Medium-sized hospitals, those with
300-600 beds, have better climate scores than smaller and
larger hospitals. There is no statistically significant
relationship between climate scores and tax, teaching, or
urban status main effects. Tests for interactions between size
and other characteristics of interest indicated that, among
hospitals with more than 300 beds, those with non-profit
status had significantly better climate scores than for-profit
hospitals. No other interactions were significant.
Conclusions: Despite hypothesized relationships between
safety culture and multiple hospital characteristics, we found
evidence of a relationship only between hospital size and
climate. The better climate scores observed in medium size
hospitals may be related to the fact that these hospitals often
enjoy less hierarchy and better communication than larger
hospitals and more safety systems and technologies than
smaller hospitals. There is also evidence for an effect of tax
status in large hospitals. Large for-profit hospitals appear to
face the greatest challenge to achieving strong safety culture.
Prior research found that these hospitals offer more high
technology/high profit services, which may also be more
hazardous.
Implications for Policy, Delivery, or Practice: While room
exists for improvement in safety culture among all types of
hospitals, results suggest that small and large hospitals could
benefit the most from investments in safety culture. Within
large hospitals, particular attention should be focused on forprofit hospitals.
Primary Funding Source: AHRQ
●A Survey of Children with Special Health Care Needs in
the WVa Mountain Health Trust Medicaid Managed Care
Program
Nancy Smith, MS, Joseph Lynch, RN, CPHQ, Linda Oliver,
RN, Shelley Baston, RNC, M.B.A.
Presented By: Nancy Smith, MS, Scientist, Delmarva
Foundation for Medical Care, 9240 Centreville Road, Easton,
MD 21601; Tel: 410-763-6236; Fax: 410-822-7971;
Email: smithn@dfmc.org
Research Objective: The West Virginia Bureau for Medical
Services, Office of Medicaid Managed Care, began a Medicaid
managed care program in 1996 called Mountain Health Trust.
This program is based upon the concept that each Medicaid
consumer will have a medical home. As WV’s Medicaid
Program, MHT funds a variety of medical services for
approximately 130,000 of the state’s most vulnerable children
and adults. BMS has a responsibility to ensure that MHT
beneficiaries enrolled in contracted managed care plans
receive comprehensive services.
The best available prevalence estimate for the West Virginia
CSHCN population is 16.7 percent. This compares to a
nationwide prevalence estimate of 12.8 percent. Delmarva
Foundation for Medical Care undertook a survey on behalf of
West Virginia BMS to address whether children with special
health care needs who were enrolled in the MHT program had
medical homes and access to needed services according to
the Healthy People 2010 goal.
Study Design: A targeted sample size of 170 children
identified as CSHCN out of the 630 surveys conducted per
managed care plan, was calculated based on 80% power, an
alpha of 0.05 and an effect size of 0.15. Children with special
health care needs were identified according to the Child and
Adolescent Health Measurement Initiative CSHCN screener.
The screener consisted of five questions that were designed to
identify children with special health care needs according to
three definitional domains, i.e. dependency, service use and
functional limitations. A select set of questions were taken
from the National Survey of Children with Special Health Care
Needs to evaluate whether eligible children have medical
homes and access to needed services as well as a parent or
guardian’s satisfaction with the coordination of a child’s
medical and non-medical care.
Population Studied: The eligible survey population included
the pediatric Medicaid population in West Virginia who were
both less than 18 years and met continuous enrollment criteria
and were pre-screened according to the Child and Adolescent
Health Measurement Initiative list of ICD-9 diagnostic codes.
Principal Findings: The targeted sample size of 170 children
per plan who were identified as a CSHCN via the CAHMI
screener was in fact surpassed. Overall 42% of the surveyed
population were identified as CSHCN using the CSHCN
screener. A total of 19% of survey respondents were identified
as CSHCN based on service use, 35% based on prescription
medication use, 13% on functional limitation use, 6% on
special therapies and 15% on mental health services. Overall,
93% of respondents said that their child had a personal doctor
or nurse and 98% said that their child had a usual place that
they went for medical care. Interestingly, 72% of respondents
across the three plans, reported that a parent or guardian was
responsible for coordination of the child’s medical care while
19% reported that a physician was responsible for
coordination of the child’s medical care. A total of 44% said
that they were very satisfied and 41% were satisfied with
coordination of their child’s medical care. 30% of
respondents reported that a social worker coordinated their
child’s non-medical care, 19% reported that a physician
coordinated their child’s non-medical care and 12% reported
that a parent or guardian coordinated their child’s nonmedical care.
Conclusions: According to the preliminary results of this
survey, the West Virginia Mountain Health Trust managed
care plans appear to be making progress in assuring that its
beneficiaries have a medical home and are receiving
comprehensive care.
Implications for Policy, Delivery, or Practice: More states
are exploring managed care alternatives to meet the
specialized needs of this population. Because of the diverse
and sometimes intense care requirements of children with
special needs, the availability of managed care programs that
offer high-quality as well as comprehensive health services for
these children is often limited. West Virginia Medicaid
managed care plans may serve as a model for other states.
Primary Funding Source: WVA Bureau of Medical services,
Office of Medicaid Managed Care
●Rural Hospital Information Technology Implementation
for Safety and Quality Improvement: Lessons Learned
Mari Tietze, Ph.D., MSN, BSN, Marisa Galimbertti, M.D., MS,
Susan McBride, Ph.D., MSN, BSN, Kathy Mechler, MS,
CPHQ, Josie Williams, M.D., MMM
Presented By: Mari Tietze, Ph.D., MSN, BSN, Research
Project Manager, Data Initiative, Dallas-Fort Worth Hospital
Council, 250 Decker Drive, Irving, TX 75062; Tel: 469-6485034; Fax: 972-791-0284; Email: mtietze@dfwhc.org
Research Objective: The grant, entitled “Rural Hospital
Collaborative for Excellence Using Information Technology,”
is a three-year research grant that began in October 2004.
The research activities focus on the improvement of patient
care safety and quality in Texas rural and small community
hospitals through the use of technology and education.
Study Design: The technology component of the design
involves implementation of a Web-based business analytic
tool that allows hospitals to view, create reports and analyze
their hospital discharge data and their CMS quality indicator
data. Analysis of the hospital’s safety and quality data will be
compared in a pre- and post-implementation design. Partners
in the grant are: Texas A&M University System Health
Science Center, Rural and Community Health Institute;
Institute for Health Care Research and Improvement, Baylor
Health Care System; Texas Medical Foundation, Dallas-Fort
Worth Hospital Council Data Initiative, and Palo Pinto General
Hospital of Mineral Wells, Texas. The IT implementation
team consists of a core team of members from each partner
organization. The focus of this team is to recruit hospitals
participants and to implement the technology at each hospital
site. In addition to the technology implementation, the
implementation team is responsible for hospital staff training
in use of the analytic tool to create reports for safety, quality
and market analysis. RCHI is then responsible for assisting
the hospital to initiate quality improvement programs
resulting from the data analysis.
Population Studied: Rural hospitals were defined as acute
care hospitals (General Medical Surgical service type as
indicated on the 2003 American Hospital Association Hospital
Survey) and located in a county with a population of less than
100,000. In addition, all Texas Critical Access Hospitals are
included. The initial total study population target was 188
hospitals.
Principal Findings: Sixty-three of the 66 targeted Texas rural
hospitals are in active implementation of patient safety and
quality technology in their hospitals. Lessons learned during
the IT implementation of these hospitals are reflective of the
unique culture, financial characteristics, organizational
structure and technology architecture of rural hospitals.
Specific steps such as recruitment, IT assessment, conference
calls for project planning, data file extraction and transfer,
technology training, use of email, use of phones, personnel
management and engaging IT vendors were found to vary in
difficulty among hospitals: Maximum use of technologies
such as Web-based interactive seminars (Webinars) and
video-based packaged presentations positively impacted
communication efforts. It was also noted that ability to
communicate with hospitals on-site had the most positive
impact but was also the most challenging to accomplish.
Conclusions: Conclusions to-date suggest that technology
implementation in rural hospitals continues to be a challenge.
An implementation team’s ability to manage these challenges
is key to success in rural hospital research.
Implications for Policy, Delivery, or Practice: These factors
are discussed as they lead to ongoing adjustments made in
the steps for implementation in hopes that other IT rural
hospital initiatives might benefit. Regardless of the patient
safety and quality final results, implications for policy delivery
and practice are evident from lessons learned.
Primary Funding Source: AHRQ
●Longer Visit Duration Increases Physician Report of
Providing Routine Symptom Evaluation to Breast Cancer
Patients in Los Angeles County
Diana Tisnado, Ph.D., Anjali Misra, MS, Jennifer Malin, M.D.,
Ph.D., May Tao, M.D., Patti Ganz, M.D., Katherine Kahn, M.D.
Presented By: Diana Tisnado, Ph.D., Adjunct Assistant
Professor, Division of General Internal Medicine and Health
Services Research, UCLA, 911 Broxton Plaza, Box 951736, Los
Angeles, CA 90095; Tel: (310)794-0711; Fax: (310)794-0732;
Email: dtisnado@mednet.ucla.edu
Research Objective: Symptom management is an important
component of breast cancer care. Structural aspects of care
may influence clinical processes and patient outcomes. We
used provider self-report to evaluate whether visit duration
impacts cancer specialists’ propensity to routinely evaluate
symptoms that patients may experience during breast cancer
treatment.
Study Design: Cross-sectional, observational study utilizing a
mailed survey of breast cancer providers. Performance of
symptom evaluation (weighted for non-response and adjusted
for clustering) was calculated as a summary score. One point
each was assigned to providers for routine symptom
assessment 1) at initial consult (or first post-op visit for
surgeons), and 2) during treatment, across seven breast
cancer and treatment symptoms (e.g. pain, fatigue,
menopausal symptoms). The Symptom Evaluation Summary
Score (SESS) ranges from 0-14 points. Bivariate and
multivariate ordinary least squares (OLS) regressions were
conducted using the SESS as the dependant variable. The
hypothesized main effect was visit duration: self-reported
average time scheduled (minutes) for a routine on-treatment
visit. Covariates included provider and practice characteristics
(physician gender, age, specialty, volume, and belief regarding
which specialist in the cancer care team bears responsibility
for monitoring symptoms; practice type, and size). We used
recycled predictions to calculate the effect on SESS of changes
in visit length adjusting for all covariates.
Population Studied: Medical oncologists, radiation
oncologists and surgeons practicing in Los Angeles County
identified by a population-based cohort of women with breast
cancer retrieved from the cancer registry (76% response rate,
n=346).
Principal Findings: The mean score for surgeons (4.7 points)
was found to be significantly lower than the score for medical
oncologists (9.9) or radiation oncologists (8.6)(p< 0.00).
Bivariate regressions resulted in a positive and significant
relationship between visit duration and SESS (p<0.00). In the
OLS multiple regression model, a significant positive linear
relationship was also found between visit length and SESS
(p<0.01). SESS increased from 6.54 to 7.65 when visit
duration increased from 10 to 20 minutes. Female gender (p<
0.00) and staff/group model HMO practice type (p<0.02)
were also positively associated with visit length, controlling for
all other covariates. Radiation oncologists (p<0.00) and
surgeons reported significantly shorter visits compared to
medical oncologists (p<0.01). As an alternative approach, a
two-stage propensity score analysis was utilized to control for
bias due to observable characteristics, achieving consistent
results.
Conclusions: Provider and practice characteristics have
significant impacts on physicians’ likelihood of providing
routine symptom evaluation to breast cancer patients. Longer
visits appear to have a positive effect on symptom evaluation.
Furthermore, certain physician and practice characteristics,
such as female gender and staff/group model HMO practice
type, appear to be associated with longer scheduled visits,
enabling more symptom evaluation service delivery. The
systems used to accomplish these goals and possible
tradeoffs require further study.
Implications for Policy, Delivery, or Practice: These are the
first results, to our knowledge, to show that investments into
the health care system in the form of increasing visit length
with patients are important for symptom evaluation.
Primary Funding Source: CA Breast Cancer Research
Program
●Measuring Quality of Diabetes Care in Acute Hospitals
of the National Healthcare Group, Singapore
Matthias Paul Han Sim Toh, MBBS, MMed, Bee Hoon Heng,
MBBS, MSc, Pui San Chan, BA, Chee Fang Sum, MBBS,
MRCP, FAMS, FRCP, FACE, Jason TS Cheah, MBBS, MMed,
MSc
Presented By: Matthias Paul Han Sim Toh, MBBS,
MMed(Public Health), Registrar, Health Services and
Outcomes Research, National Healthcare Group, 6
Commonwealth Lane #06-01 GMTI Building, Singapore,
149547; Tel: (65) 6471-8971; Fax: (65) 6471-1767;
Email: matthias_toh@nhg.com.sg
Research Objective: The aim of the survey was to establish
the standards of clinical practice, degree of practice variability
and identify the areas for quality improvement for the
management of diabetes mellitus in Specialist Outpatient
Clinics (SOCs) of a healthcare delivery system in Singapore.
Study Design: A retrospective review of patient medical
records at ten SOCs was conducted from December 2004 to
October 2005. Subjects were randomly chosen from a list of
patients who attended the selected SOCs and were prescribed
anti-diabetic medication during the study period. Study
parameters included process and outcome indicators based
on the 1999 Singapore Ministry of Health (MOH) Diabetes
clinical practice guidelines (CPG). The most recent glycated
haemoglobin (HbA1c), blood pressure and LDL-cholesterol
(LDL-c) levels were assessed. Trained staff who were not
directly involved in the management of these patients
collected data and checked by the survey coordinator for
completeness. Data was analysed by an independent research
unit using SPSSv13.
Population Studied: The study population included patients
with diabetes mellitus managed by one of these six medical
subspecialties: Cardiology(CVM), General Medicine(GM),
Geriatrics(GRM), Neurology(NEU), Rheumatology(RAI) and
Respiratory Medicine(RM). Patients must be on follow-up for
at least 15 months during Oct 2003 to Apr 2005 and
prescribed anti-diabetic medication from these clinics. The last
recorded visit was denoted as the 15th month visit.
Principal Findings: A total of 570 cases were sampled from
ten SOCs across 3 hospitals. The aggregate score for all ten
process parameters ranged from 46.3% to 59.7% across the
hospitals with large inter- and intra-hospital variance for eye,
foot examination and weight measurement.
Indicators scoring above 75% were: (a) BP measurement
within 4 months (98.4%); (b) Serum creatinine testing within
15 months (87.2%); (c) HbA1c testing within 6 months
(82.8%); and (d) Serum lipids testing within 15 months
(77.2%). Six indicators scored below 50%: (a) foot
examination (18.2%); (b) dietary education (25.4%); (c)
diabetes education (27.0%); (d) weight measurement (27.5%);
(e) eye examination (39.6%); and (f) urine protein test
(44.2%). The mean HbA1c of patients was 7.3%, and 51.2%
had good control. The majority of patients had blood pressure
below 140/90 mmHg (51.9%) and optimal control of LDL-c
(50.9%).
Conclusions: Most of the medical SOCs have not achieved
full adherence with the process indicators as stipulated in the
national CPG. There remained huge inter- and intra-hospital
variance with adherence of process indicators, especially in
eye, foot examination and weight measurement. There are
still gaps in the screening and early detection for major early
macro- & micro-vascular complications, including eye and
foot complications, and the provision of diabetes and dietary
education.
Implications for Policy, Delivery, or Practice: These findings
provide a baseline for further improvement in the processes
and outcomes of diabetes care. A taskforce will review and set
the targets for quality improvement of diabetes care and
reduce the variance in care provision in NHG through regular
audit. Specific efforts will be invested to increase the rates of
retinopathy and foot screening through capacity building and
reinforcement among clinicians.
Primary Funding Source: National Healthcare Group
●Seeking Improvements in Inpatient Stroke Care at the
National Healthcare Group Singapore
Matthias Paul Han Sim Toh, MBBS, MMed, Bee Hoon Heng,
MBBS, MSc, Ann Yin, BHSc, Venketasubramanian, MBBS,
MMed, FAMS, Jason TS Cheah, MBBS, MMed, MSc
Presented By: Matthias Paul Han Sim Toh, MBBS,
MMed(Public Health), Registrar, Health Services and
Outcomes Research, National Healthcare Group, 6
Commonwealth Lane #06-01 GMTI Building, Singapore,
149547; Tel: (65) 6471-8971; Fax: (65) 6471-1767;
Email: Matthias_TOH@nhg.com.sg
Research Objective: The National Healthcare Group (NHG)
initiated strategies to improve inpatient care for stroke
patients and performed 2 audits to evaluate quality of care in
the acute care hospitals.
Study Design: A retrospective review of patient medical
records in 2002 and 2003 for patients admitted with diagnosis
of stroke to the 4 acute care hospitals in 2001 and 2002
respectively. The audit framework was based on standards
listed in the Singapore Ministry of Health (MOH) Clinical
Practice Guidelines (CPG) for Stroke. The four key processes
assessed were brain scan within 24 hours, blood glucose
determination on admission, electrocardiogram (ECG)
monitoring and aspirin therapy for specific patients. Other
indicators include functional and social assessment, multidisciplinary involvement, discharge planning and
communication with patients and carers on prognosis and
risk reduction. Trained auditors reviewed medical records to
assess the various clinical domains. Data was verified by a
second independent auditor and analysed using SPSS.
Population Studied: Patients discharged from 2001 and 2002
respectively with a primary diagnosis of stroke (ICD 430-436)
were drawn from the NHG database. Random samples of 65
and 105 cases were drawn from each of the 4 hospitals for the
2002 and 2003 audits respectively.
Principal Findings: A total of 246/260 (94.6%) and 372/420
cases (88.6%) were audited in 2002 and 2003 audits
respectively. The rest were excluded because the final primary
diagnosis at discharge was non-stroke or patients were
admitted for diagnostic testing or surgical intervention.
Demographic and casemix characteristics of the 2003 sample
were similar and comparable with that of 2002. In 2003, brain
scanning within 24 hours of admission was achieved in 85.8%
of the cases and was significantly (p=0.001) higher compared
with 75.6% in 2002. Blood glucose testing on admission was
achieved in 97.6% of cases (95.9% in 2002) while ECG within
24 hours of admission was achieved in 92.7% of cases (90.7%
in 2002). Anti-platelet therapy was instituted in 83.9% of
patients with ischaemic stroke, significantly higher (p<0.001)
compared with 74.1% in 2002. There was an increase in the
proportion of stroke patients who had swallowing assessment
from 73.2% 2002 to 81.9% in 2003. Assessment by
physiotherapist and occupational therapist increased from
71.5% (2002) to 81.8% (2003) and 65.9% (2002) to 81.2%
(2003) respectively. Communication of patients’ prognosis to
relatives was documented in 84.1% of cases in 2003 (82.5% in
2002). Significant improvement was noted in the
documentation of communication of risk education to
patients and relatives, 48.8% (22.4% in 2002) (p<0.001).
Conclusions: Comparing to the MOH Stroke Audit in 2001,
subsequent audits demonstrated improvements in inpatient
stroke care with greater involvement of the multidisciplinary
team such as speech, occupational and physiotherapists.
More patients were started on anti-platelet therapy where
indicated. There was also better communication between
health provider and the patients and their caregivers on
diagnosis and prognosis.
Implications for Policy, Delivery, or Practice: Regular
clinical audit is necessary to identify practice variation and
gaps. It allows clinical teams to benchmark against one
another, set realistic targets and seek continuous quality
improvement.
Primary Funding Source: National Healthcare Group
●Nursing Staff Reductions, Workload Increases, and
Patient Outcomes in Florida Hospitals 1992-2004: New
Variable and Longitudinal Approaches
Lynn Unruh, Ph.D., Keon Lee, Ph.D., Ning Jackie Zhang, Ph.D.
Presented By: Lynn Unruh, Ph.D., Associate Professor, Health
Professions, University of Central Florida, 4000 Central
Florida Blvd, Orlando, FL 32816-2205; Tel: (407) 823-4237; Fax:
(407) 823-6138; Email: lunruh@mail.ucf.edu
Research Objective: This study examines the relationship
between nursing staff reductions, workload increases, and
patient adverse events in Florida hospitals, 1992-2004.
Hypotheses are: 1) An increase in nursing workload during a
one year period is associated with an increase in adverse
events in that same year; 2) A 5 percent or more reduction of
the licensed nursing staff during year ti-1 is associated with
higher rates of adverse events and worse patient outcomes in
year ti-1 and year ti; 3) A 5 percent or more reduction of the
licensed nursing staff during year ti-1 is associated with a
higher nursing workload in year ti-1 and year ti; 4) A hospital
characteristic such as for-profit status is associated with
higher nursing workloads, and vice versa for a characteristic
such as teaching status; and 5) Higher nursing workloads in
year ti are associated with higher rates of adverse events and
worse patient outcomes in year ti.
Study Design: "Nursing staff reductions" in year ti is a
dichotomous variable defined as a drop in licensed nursing
staff of 5 percent or more measured from the beginning of
time ti-1 to the beginning of time ti. "Nursing workload" is the
ratio of adjusted patient days of care to the numbers of RNs
and LPNs taken separately and together (licensed nurses).
Patient days of care are adjusted for out-patient care and for
patient turnover, which affects the intensity of nursing care.
Adverse events are hospital-level rates of nursing sensitive
events such as urinary tract infections, atelectasis, pneumonia,
decubitus ulcers, and failure to rescue. The patient outcomes
variable is a latent measure derived from the adverse events
rates in the measurement model. Other measures are patient
case mix, and hospital characteristics such as ownership, size,
location and teaching status. Multi-wave, multivariate, latent
growth curve modeling is used to define the relationships
between the trajectories of endogenous and exogenous
variables over 13 waves of data, time invariant and timevarying covariates included. The time invariant variables are
the hospital characteristics, while the time varying variables
are the rest of the variables. Rates of adverse events are
transformed into approximately normally distributed variables
prior to introduction into the model. Maximum likelihood
estimation methods are used.
Population Studied: Staffing measures and hospital
characteristics are from the American Hospital Association
Annual Survey. Adverse events and case mix are extracted
from patient discharge records obtained from the Agency for
Health Care Administration in Florida.
Principal Findings: Prior longitudinal studies have had mixed
findings regarding the relationship between nurse staffing and
patient outcomes (Mark, et al., 2004, Unruh, 2003). It is
expected that these findings will help clarify the relationship
and deepen our understanding of causal factors.
Conclusions: Pending.
Implications for Policy, Delivery, or Practice: Nurse
staffing/outcomes studies have primarily used common
staffing measures and cross-sectional or repeated measures
approaches. However, given these new variables, longitudinal
data, and the use of growth curve modeling, we can learn
more about the relationship between nurse staffing and
patient outcomes, and derive more causal conclusions. This
will provide a basis for policy recommendations.
Primary Funding Source: Internal university funding
●Impact of Following Antibiotic Guidelines for
Hospitalized Pneumonia Patients: Is a "Bundled" Measure
Better?
Monica VanSuch, M.B.A., James M. Naessens, M.P.H., Erin K.
McMurtry, M.S.
Presented By: Monica VanSuch, M.B.A., Quality Analyst,
Health Care Policy and Research, Mayo Clinic, 200 First Street
SW, Rochester, MN 55905; Tel: (507) 284-1166; Fax: (507) 2841731; Email: vansuch.monica@mayo.edu
Research Objective: To identify if compliance with
pneumonia antibiotic consensus guidelines on the
appropriate selection of antibiotic and administration time of
antibiotics as stipulated in the Joint Commission on
Accreditation of Health Care
Organizations(JCAHO)pneumonia core measure impacts
hospital length of stay, readmission or mortality.
Study Design: Using data from patients discharged from July
2004 through June 2005, a retrospective study was conducted
at single hospital system of 639 randomly chosen patients
with a principal diagnosis of community acquired pneumonia
or a principal diagnosis of septicemia or respiratory failure
with a secondary diagnosis of community acquired
pneumonia. Principal diagnosis was determined by ICD-9-CM
codes. Sampling was done using JCAHO methodology that
was current for the respective time periods. Data was analyzed
to determine compliance with three measures: length of time
to first dose of any antibiotic, receipt of recommended type of
antibiotic, and blood culture before antibiotic administration.
395 patients had at least one measure of interest. The
measures were also bundled to determine if receiving all
‘services’ resulted in improved outcomes.
Associations between guideline compliance and outcome:
length of stay, time to readmission and survival time, were
assessed with standard regression models, Kaplan-Meier
survival analysis and Cox regression models. Severity of illness
was assessed using APR-DRG methodology. Analyses were
performed for patients receiving recommended antibiotics as
well as patients receiving other antibiotics. Potential covariates
in multivariable models included patient demographics,
severity of illness, comorbidities and admission source.
Population Studied: The population is limited to patients 18
years of age and older with a working diagnosis of pneumonia
on admission. Patients transferred from another acute care or
critical access hospital including another emergency room are
excluded. Patients receiving comfort care measures only are
also excluded.
Principal Findings: When independently assessed, there are
no associations between blood cultures prior antibiotics and
timing of antibiotics for length of stay, mortality and any
readmission. Receiving recommended antibiotics is
significantly associated with decreased length of stay
(p=0.002) and increased survival (p= <0.001). Some
association is seen between receiving the recommended
antibiotics and increased time to pneumonia-related
readmission (p=0.079). When covariates are included in a
regression model, receiving recommend antibiotics is still
significantly associated with increased survival (p= <0.001)
and increased time to pneumonia-related readmission (p=
0.020). There is also some association with decreased length
of stay (p=0.080). When assessed as a “bundle” of measures,
there is a significant association with increased survival
(p=<0.001) for patients that receive all measures for which
they were eligible. There appears to be a “dose-response”
relationship with survival (p=0.009) for patients who received
more of the three measures. When covariates are included,
the associations remain significant with p= 0.008 and
p=0.023, respectively
Conclusions: Receiving recommended antibiotics is
associated with increased time to pneumonia readmission,
increased survival and decreased length of stay whether
assessed independently or as part of a bundle of services.
Significant association is seen between receiving
recommended antibiotics and survival, even after adjusting for
relevant covariates.
Implications for Policy, Delivery, or Practice: Following
guidelines for recommended antibiotics for pneumonia
appears to contribute to improved outcomes
Primary Funding Source: No Funding
●Performance Improvement in Certified Primary Stroke
Centers
Ann Watt, M.B.A., Elvira Ryan, M.B.A., Linda Hanold,
M.H.S.A., Jerod M. Loeb, Ph.D.
Presented By: Ann Watt, M.B.A., Project Director,
Performance Measurement, Division of Research, Joint
Commission on Accreditation of Healthcare Organizations, 1
Renaissance Blvd., Oakbrook Terrace, IL 60181; Tel: (630) 7925944; Fax: (630) 792-4944; Email: awatt@jcaho.org
Research Objective: In a growing trend, several state
legislatures are in the process of using certification as a
means to designate preferred providers of primary stroke care.
To date, Florida, Maryland, Texas, Massachusetts and New
York have active or pending legislation to this effect. This
paper describes preliminary work done to determine if primary
stroke center certification promotes improvement in quality of
care for primary stroke patients.
Study Design: Using self-reported standardized performance
measure data from 30 organizations that had obtained
primary stroke center certification from the Joint Commission
on Accreditation of Healthcare Organizations (JCAHO), we
compared changes in performance rates for the 11
standardized measures incorporated into the certification
process over a 12 month period (10/04 - 09/05). Data
elements used to compute the standardized measures were
collected on a monthly basis for each study hospital and
performance rates were computed and aggregated on a
monthly basis. Analysis was then performed to determine if
study hospitals’ performance relative to their baseline
performance had improved over time based on changes to
individual measure rates.
Population Studied: Study participants were health care
organizations certified in primary stroke care according to the
standards established by JCAHO. A significant component of
JCAHO's disease specific care certification program for
primary stroke centers is the collection and reporting of data
on standardized performance measures. Participating
organizations were oriented to the performance measure data
definitions and collection guidelines by JCAHO staff.
Participating organizations identified members of the study
population and collected and reported actual performance
data on a monthly basis. Performance rates were computed
and aggregated by JCAHO staff.
Principal Findings: Absolute performance measure rates
improved for 10 of 11 standardized measures over the course
of the study period. Regression analysis indicated that of the
10 standardized measures showing improvement,
improvement rates were statistically significant (p<.05) for
eight. Significant measure absolute improvement rates
ranged from 1.64% (plan for rehabilitation considered) to
18.26% (t-PA given). Significant improvement also occurred
for the following measures: smoking cessation counseling
(14.42%); DVT prophylaxis (12.52%); stroke education
(10.82%); screening for dysphagia (10.15%); lipid profile
(8.64%); and patients discharged on antithrombotics (5.45%).
Measures that did not show a significant change in rates
included atrial fibrillation patients receiving anticoagulation
therapy; and t-PA considered. The rate of change for the only
measure which showed a decline in rate, antithrombotic
medication within 48 hours of admission, also was not
significant.
Conclusions: Primary stroke centers that collected data on
standardized performance measures embedded within
JCAHO's certification initiative demonstrated significantly
improved performance from baseline over a 12 month period.
Implications for Policy, Delivery, or Practice: Participation
in a certification program for primary stroke centers appears
to lead to improved performance on standardized measures.
Since equivalent data are not currently available from noncertified primary stroke centers, comparisons with other
groups are not possible. If the preliminary findings of this
study are confirmed, disease specific care certification may be
an important stimulant for clinical quality improvement and
accountability.
Primary Funding Source: No Funding
●Developing a Health Plan Quality of Care Performance
Measure for Falls Risk Prevention in Medicare
Beneficiaries
Lok Wong, M.H.S., Min Gayles, M.P.H., Claudia Squire,
M.P.H.
Presented By: Lok Wong, M.H.S., Senior Health Care Analyst,
Quality Measurement, NCQA, 2000 L Street, NW, Suite 500,
Washington, DC 20036; Tel: 2029551784; Fax: 2029553599;
Email: wong@ncqa.org
Research Objective: To develop a quality of care measure for
health plans to improve fall risk prevention in seniors.
Study Design: A clinical expert panel was convened to
develop a quality of care HEDIS measure based on survey
questions. Based on evidence in the literature and geriatric
guidelines, two aspects of falls prevention were identified:
discussing fall risk and managing fall risk in seniors at risk for
falling. Cognitive testing was conducted to develop and test
the survey questions to identify the target population and
these two aspects of fall prevention and management. A total
of four rounds of cognitive testing were conducted with
respondents to develop and test the survey questions to
identify seniors at risk for falling, and to assess if their health
care provider discussed falls risk and recommended ways to
prevent falls.
Population Studied: Interviews were conducted with a total
of 24 Medicare beneficiaries 65 years and older from North
Carolina and Massachusetts who had a visit to a doctor and a
fall or problem with balance or walking during the past year,
representing a diverse selection of education, race, age, and
health conditions.
Principal Findings: Key phrases were identified to assist
seniors understanding of the concepts of falls and falls
prevention at the appropriate reading level for seniors.
Cognitive testing determined separate questions were needed
to identify the target population by two risk factors: previous
falls or problems with balance or walking. However
respondents also saw these two as related concepts as most
talked about both with their health provider. Based on
respondents’ feedback, examples of a variety of interventions
that may be recommended by providers to prevent falls were
modified. A shorter list of examples and shorter question was
found to be clearer to seniors. Four final questions were
developed and confirmed to be effective with seniors based on
the cognitive testing.
Conclusions: The questions were well understood and
resonated in the senior population. The study validated the
new survey question for a HEDIS performance measure that
will assess if Medicare beneficiaries enrolled in managed care
who are at risk for falls receive appropriate fall risk prevention
and management. The HEDIS measure Fall Risk Management
will be implemented in 2006 and the four survey questions
will be included in the Medicare Health Outcomes Survey
implemented by health plans. This survey measure assesses
the quality of care delivered by health care providers and
therefore also has relevance for geriatric care in primary care
practices. These questions have also been implemented in
practice settings to assess provider quality.
Implications for Policy, Delivery, or Practice: The
implementation of this measure will help to improve fall risk
prevention and management in Medicare beneficiaries
enrolled in managed care plans. Health plans can develop
quality improvement interventions to educate patients and
providers on the importance of fall risk prevention to reduce
the morbidity and mortality associated with falls in seniors. In
addition to deployment in health plans, the development of
this survey measure will provide a model for implementation
of similar quality of care measures in other health care
settings relevant for Medicare beneficiaries.
Primary Funding Source: CMS
●Using Multi-Agent Modeling to Explore Clinical
Communication Related to Patient Safety
Donna Woods, EdM, Josh Unterman, Uri Wilensky, Ph.D.
Presented By: Donna Woods, EdM, Research Assistand
Professor, Institute for Healthcare Studies, Northwestern
University, 339 E Chicago Ave 7th Fl, Chicago, IL 60611; Tel:
(847) 571-2593; Fax: (312) 503-2593;
Email: woods@northwestern.edu
Research Objective: Medical errors and related injuries are
an important problem in medical care. Clinician to clinician
communication has emerged as a significant issue related to
medical error in both hospital-based and ambulatory medical
care and was a contributing factor in 65% of the Sentinel
Events reported to the Joint Commission for the Accreditation
of Healthcare Organizations. In addition, clinician hand-offs
and transitions have increased since the institution of the 80
hour work week.
Study Design: A dynamic model of information transfer was
developed using the agent-based modeling software, NetLogo.
The model has four stages: 1) A high but not perfect level of
information transfer accuracy, 2) Intervention is added
concistent with many evaluations of these aspects of clinical
communication from culture surveys– evaluation of the
accuracy of information in 50% of instances and willingness to
correct, in 50% of instances, 3) The ability to evaluate and
willingness to correct are increased to 70% of instances, 4)
Harm is assessed, given a range in the margin-of-error and a
range of likelihood of clinician taking action upon information.
Population Studied: Clinican communication through multiagent modeling simulation.
Principal Findings: The initial model of information transfer,
with no intervention, results in a cascade of information
inaccuracy. When evaluation and correction of inaccurate
information are added to the model, this results in an initial
cascade, but then recovery of information accuracy as the
system is corrected. However, this requires more interactions
than normally occur in patient care. Increasing the ability to
evaluate and willingness to correct leads to stemming the flow
of inaccurate information transfer and quick recovery of
information accuracy, resulting in a potential reduction of
related harm.
Conclusions: This agent-based model demonstrates two key
principles of robust systems of safe communication: the
ability to evaluate the accuracy of information and the
willingness to correct inaccurate information. Development of
systems that improve the ability to evaluate and encourage
correction of inaccurate information will likely improve patient
safety and reduce harm.
Implications for Policy, Delivery, or Practice: Agent-based
model of clinician-to-clinician communication demonstrates
two key principles for improving the accuracy and patient
safety of clinical communication: Improving the ability to
evaluate and encouraging correction of inaccurate
information.
Primary Funding Source: No Funding
●Child Specific Factors for Customizing Safety
Interventions Design for Pediatric Patient Safety
Donna Woods, EdM, Ph.D., Jane L Holl, M.D. M.P.H.,
Munisha Bhatia, M.D., Edward Ogata, M.D. MM, Kevin Weiss,
M.D. M.P.H.
Presented By: Donna Woods, EdM, Ph.D., Research Assistant
Professor, Institute for Healthcare Studies, Northwestern
University, 339 E Chicago Ave 7th FL, Chicago, IL 60611; Tel:
(847) 571-2593; Fax: (312)503-2936;
Email: woods@northwestern.edu
Research Objective: Children are different than adults in
many ways this study describes these meaningful differences
and methods for application in pediatric patient safety design.
and must be attended to in the design and implementation of
patient safety interventions.
Study Design: Extensive review of the formal and informal
pediatric literature, including pediatric texts, published
reports, peer reviewed literature, websites of pediatric
healthcare professional organizations, conference
presentations and proceedings, was performed to identify
factors specific to children and/or children’s healthcare that
could contribute to patient safety problems in children’s
medical care. This review focused on specific characteristics
of children, the adaptations made by pediatrics to compensate
for these differences, and the different types of contact that
children have with the health care system (e.g., differences in
epidemiology of illnesses and treatments), as well as pediatric
patient safety studies. These factors were then assessed to
understand their impact on factors that have been shown to
increase the risk of error.
Population Studied: Children, birth through 20 years of age.
Principal Findings: Analysis of three epidemiologic studies of
pediatric patient safety each employing different source data
and methodologies together suggest specific contexts of risk.
Additionally, child-specific-factors contribute to patient safety
risk. These child-specific-factors include: I. Physical
Characteristics A. Small size, weight and morphology. B.
Varied physical characteristics II. Development A.
Physiological development and growth: B. Cognitive social
emotional development III. Minor Legal Status A. Decisionmaking and consent B. Parental responsibility for medical
management: C. Confidentiality: D. Supervision requirements.
These child-specific-factors create the following conditions
that have been shown to increase error risk. 1) Increased
variability (medication dosages, equipment and devise sizes,
in signs and symptoms, examination of infants versus
adolescents), 2) visual difficulty distinguishing differences in
sizes and amounts (i.e., equipment, devises, medications) 3)
increased complexity in medical care processes (medication
ordering, dispensing and administration), 4) decreased
response time for communication and coordination (increase
physiological volatility), 5) decreased information (children’s
limited ability to communicate), 6) increased technical
difficulty (insertion of intravenous line).In addition, childspecific-resiliency-factors were identified. Children’s
physiological resilience and healing ability can protect children
in the context of increased error risk. Most hospitalized
children’s underlying health status is essentially healthy, which
leads to a physiological resilience that protects from cascading
failure of multiple physiologic systems. However, those
children with multiple co-morbidities lose this resilience and
are at a greater risk for error related injuries. This review
demonstrated differences in children’s epidemiology of illness
and intervention. Procedures shown to be high risk in adults
are infrequently if ever performed in children.
Conclusions: This review identifies specific characteristics in
children and children’s health care that contribute to patient
safety risks in children. These factors must be accommodated
in the design and implementation of patient safety
improvement interventions to accomplish improvement in
pediatric patient safety.
Implications for Policy, Delivery, or Practice: The more we
learn through pediatric patient safety research, the more we
find that in order to successfully improve the safety of
children’s medical care, specific attention must be paid to
designing safe practices with the special characteristics of
children in mind.
Primary Funding Source: AHRQ
●Trends and Geographic Variation of Opiate Medication
Use in State Medicaid Fee-For-Service Programs 19962003
Judy Zerzan, M.D., M.P.H., Nancy E. Morden, M.D., Stephen
Soumerai, Sc.D., Dennis Ross-Degnan, Sc.D., Elizabeth
Roughhead, Ph.D., Sean Sullivan, Ph.D.
Presented By: Judy Zerzan, M.D., M.P.H., Fellow, Robert
Wood Johnson Clinical Scholars Program, Medicine, Puget
Sound VA/ Univerisity of Washington, 1914 E Republican St,
Apt E, Seattle, WA 98112; Tel: 206-616-8770;
Email: zerzanj@u.washington.edu
Research Objective: While studies have documented hospital
and surgical service geographic variability, prescription use
geographic variability is largely unknown. Opiate pain
medications are widely used, particularly since the
promulgation of clinical guidelines promoting aggressive pain
treatment. This study describes temporal and interstate
variability in aggregate prescription opiate medication claims
within US Medicaid programs.
Study Design: The defined daily dose (DDD) per 1000 fee-forservice Medicaid adult enrollees per day was calculated for all
opiate medication categories. A market basket of non-pain
prescription medications was constructed for comparison.
Rates, trends and the coefficient of variation were determined
overall, by year and for each state.
Population Studied: A dataset of 49 states’ fee-for-service
Medicaid aggregate prescription drug dispensing records
from 1996 to 2002 was compiled and used to quantify
medication dispensing examining all opiates and two specific
medications in the opiate class: controlled release oxycodone
and methadone.
Principal Findings: From 1996 to 2002, overall use of opiate
pain medications increased 184 percent. The market basket
use increased 55 percent. Total opiate dispensing varied widely
from state to state with a range of 6.9 to 44.1 DDD/1000/day
in 1996, and 7.1 to 165.0 DDD/1000/day (a 23 fold difference)
in 2002. The coefficient of variation was 49.6 in 2002.
Controlled release oxycodone and methadone had a greater
rate of increase compared to all opiates.
Conclusions: Dispensing of opiate medications in Medicaid
programs increased at almost twice the rate of non-pain
related medications during the seven-year study period. Large,
unexplained geographic variation in aggregate use exists.
Implications for Policy, Delivery, or Practice: The observed
variation raises questions about the quality of care. It seems
unlikely that the quality of care is equal among states with
such disparate use, especially in light of increasing, rather
than decreasing, variability. Additional study of this variation is
warranted. The impact of Medicaid cost-containment
strategies on this variation in utilization and on individual
patient outcomes should be investigated.
Primary Funding Source: VA, Robert Wood Johnson Clinical
Scholars Program