Call for Panels
Group Practice, Systems and Quality:
Examining the Linkages
Chair: L. Gregory Pawlson, M.D., M.P.H.
Sunday, June 3 • 11:00 a.m.-12:30 p.m.
Panel Overview: Overview: In addition, number
of key individuals doing work in related fields will
be invited to lead the question/discussion
portion of the presentation. Building on findings
from current and past research by groups at
Harvard, UC Berkeley, NCQA and others, the
panel members will examine the growing
evidence related to critical relationships between
the presence and use of systems in practice, the
organization of practice, for example, group
structure, and quality as measured by clinical
and patient experience of care measures. The
work that will be discussed includes data from a
national study of large physician organizations,
studies of solo and single site small group. The
exploration of correlations between
characteristics of group practice, the presence
and use of practice systems, and higher quality
of care can help guide practice improvement
work and provide a basis for future research to
better understand systems-quality-practice
organization linkages.
ƒ Do integrated medical groups provide
higher-quality medical care than individual
practice associations?
Ateev Mehrotra M.D., M.P.H., Arnold M.
Epstein, M.D., M.A., Meredith B. Rosenthal
Ph.D.
Presented By: Ateev Mehrotra, M.D., M.P.H.,
Assistant Professor, RAND, 4570 Fifth Street,
Suite 600, Pittsburgh, PA 15213, Phone: (412)
683-2300 x4894, Email: mehrotra@rand.org
Research Objective: To examine the
associations between the organizational
structure of physician groups, use of quality
improvement strategies, and health care quality.
Study Design: Survey of 119 California physician
groups who care for approximately 1.7 million
enrollees of PacifiCare. Quality of care assessed
using six HEDIS quality measures:
mammography, Papanicolaou smear screening,
Chlamydia screening, diabetic eye examination,
use of an asthma controller medication, and
beta-blocker after acute myocardial infarction.
Supplemented by another survey of 79
Massachusetts physician groups and an
examination of factors that are associated with
use of quality improvement strategies.
Principle Findings: In our California survey,
controlling for other factors, Integrated Medical
Groups, as compared to IPAs, had a 11% higher
mammogram rate, 36% higher Pap smear rate,
10% higher Chlamydia screening rate, and 15%
higher diabetic eye screening rate (all p<0.05).
Leaders of IMGs were more likely to report using
EMRs (37% vs. 2%; P < 0.001) and quality
improvement strategies, but these
characteristics explained little of the quality
differences between IMGs and IPAs. The only
other predictor of higher quality was volume of
patients seen by the physician group for the
quality measure. . In our Massachusetts survey
of physician groups we found that pay-forperformance incentives did have an association
with the use of quality improvement strategies
(OR 1.6 [CI 1.0-2.4]). But several organizational
factors had a stronger association including use
of employed physicians (OR 3.2 [CI 1.5-7.2]),
being part of a larger network (OR 2.3 [CI 1.1.4.8]), and larger size in terms of number of
physicians (OR 2.2 [CI 1.0-4.9]).
Conclusions: Integrated medical groups in
California appear to provide a higher level of
clinical quality than independent practice
associations. Patients cared for in IMGs
generally received higher quality primary care
than those cared for in IPAs. Having an EMR
and implementation of quality improvement
strategies did not explain the differences in
quality. Among Massachusetts physician groups
larger integrated groups were more likely to use
quality improvement strategies. These findings
suggest that physician group type influences
health care quality.
Implications for Policy, Delivery, or Practice:
Together the findings of both studies would
imply that encouraging more integration in
physician groups might result in greater use of
quality improvement strategies and higher
quality. The second study potentially indicates
that this even true in the setting of pay-forperformance incentives
Funding: California HealthCare Foundation and
the Commonwealth Fund.
ƒ Development, Testing & Correlation with
Selected Quality Measures, of a Survey
Instrument to Assess Systemness in Office
Practice
Sarah Scholle, Greg Pawlson M.D., M.P.H., Leif
Solberg, M.D., Sarah Shih, M.P.H., Steve Asche,
M.S.
Presented By: Sarah Scholle, Dr.P.H., M.P.H.,
Assistant Vice President, National Committee for
Quality Assurance, 2000 L Street NW, Suite 500,
Washington, DC 20036, Phone: (412) 624-1825,
Fax: (412) 624-2360, Email: scholle@ncqa.org
Research Objective: The creation and validation
of a tool that assesses the presence and use of
systems in office practice. The elements included
in the tool are related to the Wagner Chronic
Care model.
Study design: The development of an
assessment tool used an expert panel, literature
review and pilot testing to construct a survey
related to four domains of Wagner chronic
illness model: decision support, delivery system
design, clinical information systems and patient
self management support. To validate the
survey, we obtained 120 respondents of varying
job titles, medical director, practicing physicians,
nurse coordinators, from 40 sites in 11 medical
group practices in Minnesota followed by an
onsite audit by trained observer. Correlation
studies are based on responses from 41 medical
directors representing distinct medical groups
completing a revised survey. Data on clinical
performance measures using chart review and
administrative data was obtained from the
Minnesota Measurement Collaborative.
Principle Findings: Some of the early findings
using the instrument which have been published
in the AJMC 2006 11: 789-96, indicate that at
least among groups in Minnesota, the presence
of practice systems in general was not strongly
influenced by whether or not the group had
implemented an EMR. Dr. Scholle will also
discuss findings from the study that address the
validity of the data collection using a survey
versus the gold standard of an on site audit,
correlations between the use of practice systems
in general, and specific areas of the chronic
illness model with clinical performance
measures and surveys of patient experience of
care. Preliminary findings suggest significant
correlations between overall systems use and
clinical quality of care measures in diabetes.
Data from similar concurrent studies being done
in California and Massachusetts, comparing
systems use to patient perceptions of care, will
also be presented. Finally, she will discuss the
use of a web enabled version of the survey that is
coupled with requirements for documentation of
the systems as a basis for payment for physician
practice systems in the Bridges to Excellence
program and elsewhere.
Conclusions: A survey of systems that are
related to the elements of the chronic disease
model has been developed and validated against
the use of an onsite audit.
Implications for Policy, Delivery, or Practice:
The survey may be useful not only as a research
tool, but as a basis of payment for the use of
systems in ambulatory care physician practice.
Fundering Source: Robert Wood Johnson
Foundation
ƒ The 2nd National Study of Physician
Organizations and Management of Chronic
Illness
Diane R. Rittenhouse, M.D., M.P.H., Stephen
Shortell, Ph.D., Lawrence Casalino, M.D., Ph.D.,
James Robinson, Ph.D., Robin Gillies, Ph.D.
Presented By: Diane R. Rittenhouse, M.D.,
M.P.H., Assistant Professor, Department of
Family and Community Medicine, University of
California, San Francisco, 500 Parnassus
Avenue, BOX 0900, Room MU 308-E, San
Francisco, CA 94143-0900, Phone: (415) 5024048, Fax: (415) 476-0705,
Email: rittenhouse@fcm.ucsf.edu
Research Objective: The National Study of
Physician Organizations and Management of
Chronic Illness (NSPO) was conducted in
2000/2001 and established an association
between the use of organized care management
processes (CMPs) and higher levels of external
incentives and information technology (IT)
capacity. We are currently in the field with the
2nd National Study of Physician Organizations
and the Management of Chronic Illness (NSPO
II). The goals of this study are (1) to describe the
extent to which physician organizations are
currently employing CMPs in the routine care of
patients; (2) to describe the current practice
environment, both internal (eg, adoption of
information technology, payer mix, size,
structure) and external (eg. external incentives to
improve quality, relationships with health plans);
and (3) to determine the extent to which these
factors, and their interrelationships, have
changed over time. In a related study we will be
linking these data to outcome measures to better
understand the relationship between the
structure and processes of physician
organizations and improved patient outcomes.
Study Design: 25-minute telephone survey. The
survey has been in the field since Spring 2006.
Targeted response rate: 70% by March 2007.
Population Studied: The Medical Director or
CEO of all physician organizations (medical
groups and independent practice associations)
with 20 or more physicians that care for patients
with asthma, CHF, depression and/or diabetes.
(n=1,216)
Principal Findings: Results will be finalized in
preparation for the 2007 Annual Meeting. Crosssectional data will be presented, including
descriptive statistics for all physician groups
across 8 major content domains: organizational
background and history (size, specialty, number
of sites, etc); information systems (presence of
electronic health record, measures of
functionality and interoperability); care
management and clinical practice (disease
registries, reminders, clinical practice guidelines,
feedback to physicians, self-management skills,
practice re-design, etc); preventive care and
health promotion; relationships with health
insurance plans; performance incentives
(financial and non-financial); revenue sources
and compensations methods; and organizational
culture. Results of regression models predicting
CMP use will also be presented. Comparisons
will be made to baseline data from 2000/2001.
Conclusions: These new cross-sectional data
provide valuable information on the practice
environment and quality improvement efforts of
large physician organizations nationwide.
Implications for Policy, Delivery, or Practice:
These data inform the effort by the Institute of
Medicine and others to improve the delivery of
chronic illness care through delivery system
improvements.
Funding Source(s): Robert Wood Johnson
Foundation; Commonwealth Fund; California
HealthCare Foundation.
Call for Panels
Medicare Advantage Private Plans: Costs &
Benefits in 2007
Chair: Stuart Guterman, M.A.
Monday, June 4 • 11:00 a.m.-12:30 p.m.
Panel Overview: The Medicare Modernization
Act (MMA) of 2003 included provisions intended
to greatly increase the role of private health plans
in Medicare. Consistent with this objective,
enrollment in Medicare Advantage (MA) plans
increased 24 percent between December 2005
and December 2006. Private plans now manage
benefits for over 17 percent of Medicare
beneficiaries, the largest share ever. Due to this
rapid increase in enrollment as well as recent
research findings indicating that MA plans do
not contain overall Medicare costs as often
claimed but in fact increase costs by 12 percent
per plan enrollee, MA private health plans have
become a major Medicare issue. Press
commentaries report that the future role of MA
plans will be a key item on the agenda of the new
Congress. Reports from health services
researchers on the costs and benefits of MA
private plans have played an important role in
this debate. The panel would present findings
from current research on the status and effects
of the MA program.
1. The first presentation summarizes how
Medicare Advantage (MA) plan availability,
benefits and premiums, and enrollment have
changed between 2005 and 2007 by the
Medicare Drug, Improvement, and
Modernization Act (MMA) of 2003.
2. The second presentation addresses payments
to MA plans with a focus on the specific MMA
policies that result in MA payments in excess of
average fee-for-service costs. Projected amounts
and patterns of extra payments in 2006 and
beyond would be discussed.
3. The third presentation compares private
health plans’ costs of delivering Medicare
benefits under the MA program with the costs of
delivering those benefits in the fee-for-service
program and investigate how private plans’ costs
vary across geographic areas and assess the
implications for converting Medicare to a
premium support system.
4. The fourth presentation addresses the
relationship between the legislatively determined
payment rates to Medicare managed care plans
and MA program enrollment, the availability of
particular benefits and plan choices for
beneficiaries.
Call for Panels
Assessing Quality of Care of Medicare
Managed Care Health Plans and Frailty
Adjustment Using the Medicare Health
Outcomes Survey and the Medicare Health
Outcomes Survey-Modified
Chair: Beth Ellis, Ph.D.
Sunday, June 3 • 11:00a.m.-12:30p.m.
Panel Overview: With the advent of Medicare
Part D Programs, enrollment in Medicare
managed care health plans is increasing rapidly.
Evaluation of health quality and appropriate
frailty adjustment for payment purposes are
critical to the success of beneficiary health and
continued managed care health plan
participation. This panel will discuss uses of the
Medicare Health Outcomes Survey (HOS) and
the HOS-Modified (HOS-M), which are
sponsored by the Centers for Medicare &
Medicaid Services. The Medicare HOS assesses
a health plan’s ability to maintain or improve the
physical and mental health functioning of its
beneficiaries over a two-year period of time. The
HOS-M instrument, which is a modified version
of the Medicare HOS, assesses health
functioning of vulnerable program members who
are at greatest risk for poor health outcomes to
generate information for payment adjustment.
We propose five panelists for the 90-minute
session. The first panelist will present the
background of the Medicare HOS and HOS-M
surveys; their history, goals, and purpose. The
second speaker will present information on the
new quality measures recently added to the
HOS, which include questions on falls,
height/weight, physical activity, and
osteoporosis. The third panelist will present
HOS case-mix adjustment, based on the
Veteran’s 12-Item Health Survey for the
evaluation of managed care health plan
performance. The fourth panelist will present
information on frailty adjustment, which is used
for CMS payment purposes. The final speaker
will provide the results of analyses that examine
differences in response rates and other
outcomes associated with various HOS-M
protocols.
ƒ The Medicare Health Outcomes Survey &
the Medicare Health Outcomes SurveyModified: An Introduction
Vijit Chinburapa, Ph.D., C.P.H.Q., Beth Hartman
Ellis, Ph.D., MaryAnne D. Hope, M.S., Laura
Giordano, R.N., M.B.A.
Presented by: Vijit Chinburapa, Ph.D., C.P.H.Q.,
Senior Healthcare Researcher, Surveys, Research
& Analysis, Health Services Advisory Group,
1600 Northern Avenue, Phoenix, AZ 85020,
Phone: (602) 665-6185, Fax: (601) 241-0757,
Email: vchinburapa@azqio.sdps.org
Research Objective: The objective of this
presentation is to provide an introduction to the
national Medicare Health Outcomes Survey
(HOS) and the national Medicare Health
Outcomes Survey-Modified (HOS-M).
Study Design: The Medicare HOS assesses a
managed care health plan’s ability to maintain or
improve the physical and mental health of its
members over time. The Medicare HOS 2.0
instrument incorporates the Veterans 12 Item
Health Survey (VR-12), demographics, chronic
medical conditions, activities of daily living, the
number of unhealthy days, urinary incontinence,
depression and other characteristics.
The Medicare HOS-M survey is a modified
version of the Medicare HOS. Originally entitled
the Program of All-Inclusive Care for the Elderly
Health Survey (PACE), the HOS-M is
administered to vulnerable Medicare
beneficiaries at greatest risk for poor health
outcomes. The instrument assesses the physical
and mental health functioning of Program
members to generate information for payment
adjustment.
Population Studied: The HOS measure is
administered at Baseline to a randomly selected
sample of beneficiaries from each managed care
plan that holds risk and cost contracts. The twoyear follow up sample includes beneficiaries that
are identified from the Baseline sample who are
alive, enrolled in their original managed care
plan, and have sufficient VR-12 data to derive a
physical and mental component summary
(PCS/MCS) score. Multivariate statistical
methods are used for case-mix adjustment so all
plans will be as equal as possible in terms of
demographic and socioeconomic characteristics,
chronic medical conditions, and other
covariates. The HOS-M was fielded for the first
time in the spring of 2005. Program members
are enrolled in specialized Medicare plans-----PACE programs, Minnesota Senior Health
Options, Minnesota Disability Health Options,
Wisconsin Partnership Program, and
Massachusetts MassHealth Senior Care
Options.
Principal Findings: The PCS and MCS scores
derived from the HOS surveys reflect normbased standardized scores with a mean of 50
and a standard deviation of 10, relative to the
U.S. general population. The national PCS and
MCS scores for the Medicare beneficiaries
across the eight cohorts of study from 1998-2005
were relatively stable. The PCS scores ranged
from 40.3 to 42.0 with a weighted mean of 41.3.
The MCS scores ranged from 51.6 to 52.6 with a
weighted mean of 52.0.
Conclusions: For all HOS Cohorts, a Baseline
Report and a Performance Measurement Report
are prepared following data analysis activities.
Seniors (age 65 or older) who have a calculatable
PCS and MCS score at Baseline are eligible for
Performance Measurement. Both Baseline and
Performance Measurement Reports are provided
to all participating plans as well as all Quality
Improvement Organizations. HOS-M reports
will be distributed in 2007 and plan data will be
available in 2008.
Implications for Policy, Delivery or Practice:
The Medicare HOS program gathers valid and
reliable health status data in the Medicare
managed care population for use in quality
improvement activities, plan accountability,
public reporting, and improving health.
Funding Source(s): CMS
ƒ Impacts of Applying Frailty to Medicare
Advantage Plans
Thomas Kornfield, M.P.P., Sean Creighton, M.Sc.
Presented by: Sean Creighton, M.Sc., Research
Analyst, Division of Program Analysis and
Performance Measurement, 7500 Security
Boulevard, C4-13-01, Baltimore, MD 21244,
Phone: (410) 786-9302, Fax: (410) 786-1048,
Email: screighton@cms.hhs.gov
Research Objective: Evaluate the impact of
applying frailty adjustment factors to the entire
Medicare Advantage program, based on the
Medicare Health Outcomes Survey (HOS) and
HOS-Modified (HOS-M) data.
Study Design: Cross-sectional analysis of survey
data, combined with prospective risk adjustment
model. The risk adjustment model was
developed by linking data from the 2003
Consumer Assessment of Healthcare Providers
and Systems (CAHPS) with Medicare claims
data in order to estimate new frailty factors. We
included only non-ESRD, non-managed care
Medicare enrollees who responded to all six
Activities of Daily Living (ADL) questions on the
2003 Fee-for-Service (FFS) CAHPS in the
modeling process. For each beneficiary in the
sample, we obtained demographic and
enrollment information from the Medicare EDB,
and used the Data Extract System (DESY) to
acquire Medicare claims. We also linked data on
long-term institutionalization from the Minimum
Data Set, and information on ADL difficulty from
the CAHPS. Based on the HOS and HOS-M,
data, we calculated contract level frailty factors.
We restandardized the ratebook so that the net
impact of frailty would be zero. We then
adjusted contract bid data from health plans for
the 2007 contract year in order to account for
the new frailty factors. Finally, we estimated the
payment impacts of applying frailty to these
contracts.
Population Studied: The frailty factor developed
from the Medicare FFS CAHPS data was applied
to respondents to the HOS and HOS-M data,
and a selected group of July 2006 Medicare
managed care enrollees.
Principal Findings: We used diagnostic
information from base year claims for hospital
inpatient, outpatient, and physician sources to
calculate annualized predicted expenditures
based on the CMS-HCC model. We calculated
residual expenditures as the annualized actual
expenditures minus the annualized predicted
expenditures. We then used the residual
expenditures in an equation in order to
determine the additional costs associated with
one of four ADL difficulty groups: no ADLs, 1-2
ADLs, 3-4 ADLs, and 5-6 ADLs. We developed
two frailty models: one in which the frailty factors
depended on Medicaid status during the
payment year, and the other in which the frailty
factors were independent of Medicaid status.
For both models, we calculated frailty factors for
non-ESRD, non-institutionalized individuals aged
55 and over. We evaluated the impacts of
applying frailty to the entire Medicare Advantage
program to analyze whether or not applying the
frailty factor would more accurately pay these
contracts for their frail population.
Based on bidding data from health plans for the
2007 contract year, we estimated the payment
impacts of applying the frailty factor to these
plans. We estimated impacts at the contract and
contract/plan level.
Conclusions: The Medicare HOS and the HOSM measure health status and functional
limitations for the entire Medicare managed care
population. The availability of the ADL data
under HOS and HOS-M provides avenues for
exploring new payment options under frailty
adjustment.
Implications for Policy, Delivery or Practice:
The findings have implications for payment
policy decisions.
Funding Source(s): CMS
ƒ Comparing Two Approaches to Surveying
Frail Enrollees in the 2006 Medicare Health
Outcomes Survey - Modified (HOS-M)
Edith Walsh, Ph.D., G. Khatutsky, M.S., Ph.D.
Candidate
Presented by: Edith Walsh, Ph.D., Senior Health
Policy Researcher, Aging, Disability and LongTerm Care, RTI International, 1440 Main Street
Suite 310, Waltham, MA 02451, Phone: (781) 4341754, Fax: (781) 434-1701, Email: ewalsh@rti.org
Research Objective: Due to high frailty and
cognitive impairment rates of HOS-M
respondents, HOS-M sampling and
administration protocols include special efforts
to collect detailed contact information for
potential proxies in order to increase response
rates. However, in 2006 this process was waived
for some plans as their enrollments were too
large for this process to be feasible. Enrollees in
these plans were sampled without making any
special provisions to obtain proxy information.
This study evaluates these two survey
approaches by comparing survey results and
discussing future implications for HOS-M
administration.
Study Design: Examines the differences in
response rates, demographic, health, and
functional status characteristics between 2006
HOS-M respondents in samples with proxy
contact information and those without.
Population Studied: Frail dually eligible
Medicare beneficiaries aged 55+ residing in the
community and enrolled in PACE or Dual Eligible
Demonstration Plans in Minnesota, Wisconsin
and Massachusetts, and who responded to the
2006 HOS-M.
Principal Findings: The overall response rate for
plans without proxy contact information was
68.7%, comparing favorably with 74.3% for
enrollees in plans surveyed using a protocol
including collecting detailed contact information.
However, proxy response rates and mean
characteristics between the groups were very
different. Respondents in the first group reported
less functional and cognitive impairment with
33% reporting no limitations in activities of daily
living (ADLs), 2 ADL impairments on average
and 29% reported cognitive impairment
compared to only 19% with no ADL
impairments, 3 ADL impairments on average
and 45% cognitively impaired in the small plans
using the full, extended survey protocol.
Conclusions: The overall response rates and
lower level of impairment observed in the plans
without proxy contact information suggest that
implementing burdensome extra efforts to
increase response rates may not be warranted
for these plans. Nonetheless, it is possible that
the HOS-M protocol misses the frailest
beneficiaries enrolled in these plans due to nonresponse. In previous HOS-M analyses we found
the mean ADLs for self respondents to be
substantially lower than for those with proxy
respondents (2.5 vs 3.8). Given the greater
degree of impairment in the small plans and the
high proxy response rates, the extended protocol
may be necessary to ensure accurate
representation of their enrollment. In addition,
any future declines in the response rate could
reflect changing enrollment patterns and warrant
consideration of using the extended protocol for
all plans participating in the HOS-M.
Implications for Policy, Delivery or Practice:
Creating detailed contact data for the HOS-M
sample is burdensome for plans and adds to the
overall costs of conducting the HOS-M, yet it
does provide enhanced response rates and
potentially ensures representation of more
impaired beneficiaries. The results of this
analysis provide information that policy makers
can use to evaluate approaches to data collection
for frailty adjustment.
Funding Source(s): CMS
ƒ New Measures in the Medicare Health
Outcomes Survey
Judy Ng, Ph.D., Lok Wong, M.H.S., Sarah
Scholle, Dr.P.H.
Presented by: Judy Ng, Ph.D., National
Committee for Quality Assurance, 2000 L Street,
NW, Washington, DC 20036, Phone: (202) 9553599, Email: ng@ncqa.org
Research Objective: The Medicare Health
Outcomes Survey assesses health and functional
status outcomes of Medicare beneficiaries
enrolled in Medicare Advantage. We discuss
the background and rationale for the
development of a new set of Health Outcomes
Survey measures addressing prevalent problems
in older populations, including falls,
osteoporosis, urinary incontinence, and physical
activity. We use descriptive data from the 2006
Medicare Health Outcomes Survey to evaluate
the extent of these problems, and to examine
whether differences exist by vulnerable
populations.
Study Design: The Health Outcomes Survey is a
mail survey with telephone follow up,
administered to a nationally representative
sample of Medicare beneficiaries in Medicare
Advantage. The survey questionnaire includes
the VR12 and other items addressing chronic
conditions, physical and functional limitations,
and demographics. Importantly, this survey
makes it possible to learn whether patients are
receiving advice and counseling on important
topics related to prevention and functioning for
older adults.
Population Studied: Community dwelling
Medicare beneficiaries, aged 65 and older,
enrolled in over two hundred US Medicare
Advantage plans.
Principal Findings: 2006 Medicare Health
Outcomes Survey data will be used to evaluate
the overall prevalence of problems with falls,
osteoporosis, urinary incontinence and physical
activity among elderly beneficiaries, as well as
physician counseling and advice for these
conditions. The data will also be used to
perform subgroup analyses of select populations
who are particularly susceptible to health
problems, such as non-white racial minorities,
and those with lower education or income.
Conclusions: The Health Outcomes Survey is
the largest scale effort to evaluate health
outcomes among the elderly Medicare
population, and serves as a platform for
evaluating the quality of care and addressing
comprehensive health needs of the elderly,
particularly vulnerable groups.
Implications for Policy, Delivery or Practice:
The Health Outcomes Survey is a valuable tool
in targeting quality of care in Medicare managed
care. The interpretation of Health Outcomes
Survey data provides key information in
maintaining and improving these efforts.
Funding Source(s): CMS
ƒ Case-Mix Measures for Monitoring System
Performance
Kazis Lewis, Sc.D, Alfredo Selim, M.D., M.P.H.,
William Rogers, Ph.D., Shirley Qian, M.S.,
Xinhua S. Ren, Ph.D.
Presented by: Kazis Lewis, Sc.D, Director,
Boston University, Center for the Assessment of
Pharmaceutical Practices, 715 Albany Street,
Boston, MA 02118, Phone: (617) 414-1418,
Email: lek@bu.edu
Research Objective: This presentation will give
methodologies for case mix measures using the
Veterans RAND 12 Item Health Survey (VR-12)
as the outcome. The VR-12 is modified from the
Veterans RAND 36 Item Health Survey and the
RAND SF-36. This measure of functional status
and wellbeing spans the domains of physical to
psychological functioning. The VR-12 has
recently been included as one of the
performance measures in the HEDIS 2006
Information Data Set. The objectives of this
presentation will be to examine measures of
health at baseline and at follow-up for purposes
of case mix adjustment using the VR-12 PCS
(physical function) and MCS (mental function)
summary scores for measuring the performance
of health plans over two years in the Medicare
Advantage Program (MAP).
Study Design: The Medicare Health Outcome
Survey (HOS) is a cohort design with a baseline
sample and a follow-up 2 years later. Cohorts
using the VR-12 began in 2006 round 7 at followup. The SF-36 version 1.0 was used in prior
cohorts beginning in 1998. The recent addition
of the VR-12 for the transition year cohort 7
affords researchers the opportunity to investigate
risk adjustments for monitoring changes in
health using the 12 item versions including the
RAND SF-12 and the VR-12. The case mix
methodology for performance measurement
(monitoring outcomes of health over 2 years)
uses separate case mix models for death, change
in PCS and change in MCS. The models for
change in PCS or (MCS) are the same or better
versus worse. The case mix measures for the
change models include demographics (age,
gender, race, Medicaid status) and socioeconomic (education, income) status. The
mortality models add chronic medical
conditions, functional status and design
variables (mode of administration). The two
models are then combined to calculate the
probability of being alive and having PCS (or
MCS) the same or better. Scores are reported as
the percentage of respondents within a plan who
are better, same or worse for a two year period.
Population Studied: The Medicare Advantage
Program (MAP) cohort 7 is from the HOS that
randomly selected 1000 beneficiaries who were
continuously enrolled for at least 6 months in
each of the MAP plans nationally.
Principal Findings: Based upon prior cohorts,
the SF-36 PCS and MCS outcome measures have
been validated by evaluating the performance of
health plans nationally. Recently, parsimonious
models using the VR-12 with adjusted PCS and
MCS scores are developed capturing fewer case
mix adjusters for monitoring performance
among the plans. Separate models include the
use of adjusted VR-12 PCS and MCS scores
cross-sectionally to capture health. Results
suggest that cross-sectionally adjusted scores
are a function of disease burden with less
variability explaining performance.
Conclusions: Results of this work provide an
important basis for case mix approaches to
evaluate plan performance.
Implications for Policy, Delivery or Practice:
This study has important implications for the
monitoring of system performance among the
plans.
Funding Source(s): CMS
Call for Panels
Issues in Public Health Finance
Chair: Peggy A. Honore', D.H.A., M.H.A.
Sunday, June 3 • 3:00 p.m.-4:30 p.m.
Panel Overview: Public Health Finance has been
described as a field of study that examines the
acquisition, utilization, and management of
resources for the delivery of public health
functions and the impact of those resources on
population health and the public health system.
This field is both an economic concept and a
profession. It is an economic concept, grounded
in public finance theories where the focus is on
policies to guide resource utilization and
financing strategies and also a profession, where
financial management concepts should be
applied to the utilization of scarce resources.
This panel will examine a continuum of issues
that lie within the definition of public health
finance. Finance topics presented by the panel
will include the challenges in public health
agencies for financing legislatively mandated
state newborn screening programs; assessing
two different financing strategies for increasing
dental sealant prevalence among Medicaid
children; and the development of a
comprehensive, evidence-based list of
competencies for the public health financial
management workforce.
ƒ Financing Newborn Screenings: Sources,
Issues, and Future Considerations
Michele Lloyd-Puryear, M.D., Ph.D., Donna
Williams, B.S., Kay Johnson, M.P.H., Marie Y.
Mann, M.D., M.P.H., Lauren Raskin Ramos,
M.P.H.
Presented by: Michele Lloyd-Puryear, M.D.,
Ph.D., Chief, Genetic Services Branch, Division
of Services for Children with Special Health
Need, MCH Bureaus, HRSA, 5600 FIshers Lane,
Rm 18-A-19, Rockville, MD 20857, Phone: (301)
443-8604, Fax: (301) 443-1080,
Email: MPuryear@hrsa.gov
Research Objective: State Public Health
agecnies currently face the challenges of
financing state-of-the-art newborn screening
systems, which feature tests for more conditions,
use of new laboratory technology, staff with skills
to use new technology, and more effecetive
follow-up with families. The pressures on State
public health agencies reflect the coming
together of parental concern, private sector
marketing, and public opinion. The objective of
this resaerch was to document the sources and
uses of funding for newborn screening
programs.
Study Design: Two types of data are reported.
The primary data comes from a survey of State
public health agencies conducted from
Novemebr 2004 - February 2006. These survey
data are supplemented with findings from seven
case studies: California, Maryland, Minnesota,
Mississippi, New York, Oklahoma, and Oregon.
The authors particiapted in the design of both
the survey and the case studies, as well as
analysis of the data reported by the States.
Population Studied: The survey population in
this research included state public helath agency
newborn screenign programs.
Principal Findings: Concurrent to the expansion
in newborn screening panels has been the
increased dependence of newborn screening
programs on fees. Findings suggest that the
general trend in the amount of funding for
newborn screening programs is increasing.
Additional laboratory costs for technology,
expansion of testing panels, increased follow-up
resulting from additional screening tests, and
costs for associated public health services
accounted for the need for additional funds.
Funding increases emanate primarily from fees
(22), and to a lesser extent from Medicaid (10),
Title V MCH Block Grant (7), and State general
revenue funding (3). Twenty States reported no
change in the level of Medicaid or Title V fiscal
support, and 23 had no increase from State
revenues.
Conclusions: Regardless of the source of funds,
the available evidence indicates that states are
committed to maintaining their programs and
securing the necessary financing for the initial
screening through diagnosis. Use of federal
funding is currently limited; however, pressure to
provide dedicated federal funding would likely
increase if a national recommendation for a
uniform newborn screening panel were to be
issues.
Implications for Policy, Delivery or Practice:
With recent increases in costs, fees, and
government support, newborn screening has
attained greater health care financing visibility.
One major misperception is that tax dollars are
the primary source of newborn screening funds.
When State public health officials request
additional funding support from legislatures and
governors, they are often confronted with budget
constraints. Yet in most States the majority of
program funding comes from fees and not from
State or Federal sources. A valid funding concern
in many States relates to matching funds that are
required to support screening for newborns
receiving Medicaid assistance. These concerns
are of greater concern in the States with larger
percentages of Medicaid births. Additional
funding concerns for program infrastructure and
follow-up arise in States where laboratory
services may be contracted to private sources
without proper attention to the continuing public
health follow-up and service activities that may
arise from the testing.
ƒ Impact of Inceasing Medicaid Dental
Reimbursement and Implementing School
Sealant Programs on Sealant Prevalence
Susan O. Griffin, Ph.D., Kari A. Jones, Ph.D.,
Stuart Lockwood, D.D.M., M.P.H., Nicholas G.
Mosca, D.D.S., Peggy A. Honore, D.H.A.
Presented by: Susan O. Griffin, Ph.D., Health
Economist, Division of Oral Health, Centers for
Disease Control and Prevention, 4770 Buford
Highway, Mailstop F-10, Atlanta, GA 30341,
Phone: (770) 488.6064, Fax: (770) 488.6080,
Email: sig1@cdc.gov
Research Objective: Delivering dental sealants
to lower-income children has been shown to be
cost-saving. Sealant prevalence among these
children (25%), however, is well below the
HP2010 goal of 50%. This research examined
the impact of two financing strategies on dental
sealant prevalence among Medicaid children.
The two strategies were 1) increasing Medicaid
reimbursement for sealant delivery to increase
dentists' willingness to supply sealants in clinical
settings levels and 2) changing the delivery
setting from private dental offices to schools to
increase demand for sealants by reducing
parental travel and time costs.
Study Design: Multivariate time series with
comparison group
Population Studied: 7- to 9-year-old children
enrolled in Medicaid in Alabama (AL) or
Mississippi (MS) in 1999 or 2003.
Principal Findings: Increasing the sealant
reimbursement rate was associated with a 39%
and a 102% increase in sealant prevalence in AL
and MS, respectively. Implementing school
sealant programs more than doubled sealant
prevalence - 153% in AL and 200% in MS.
Conclusions: Depending upon supply and
demand conditions in dental markets, both
strategies can be effective in increasing the
prevalence of sealants. In areas with dental
shortages and where time and travel costs are
likely to be especially taxing on parents, and for
low income populations at greatest risk for
disease, delivery in a school setting becomes
increasingly cost-effective.
Implications for Policy, Delivery or Practice:
Policies for providing sealants in school settings
to lower-income children is likely to be the most
cost-effective mode of sealant delivery.
ƒ Developing Competencies for the Public
Health Financial Management Workforce
Julia F. Costich, J.D., Ph.D.
Presented by: Julia F. Costich, J.D., Ph.D., Chair,
Dept. of Health Services Management, College
of Public Health, University of Kentucky,
Lexington, KY 40536-0003, Phone: (859) 2576712, Fax: (859) 257-3909,
Email: jfcost0@email.uky.edu
Research Objective: The development of a
comprehensive, evidence-based roster of
competencies for public health financial
management.
Study Design: Two strategies were used to
garner both formal statements of practice
standards and the experience of experts in the
field. First, a review of competency assessments
from cognate disciplines included the work of
federal and state finance officers’ professional
organizations, health care financial management
groups, public health educational and
professional associations, and position
descriptions in which competency statements
were embedded. Second, the principal
investigator called on a group of nationally
recognized experts in public health finance and
related areas to participate in a series of
conference calls in which proposed
competencies were developed and reviewed.
Population Studied: Public health financial
managers
Principal Findings: Competencies were
identified in three broad areas: (a) financial
management, analysis and assessment; (b)
Policy development and program planning; and
(c) general administration. Typical activities were
listed in associated with each identified
competency. The level of competency required
for public health finance staff, management, and
senior management was ranked as proficient,
knowledgeable, or aware.
Conclusions: The work of financial management
in public health lies at the intersection of two
broad, well-developed disciplines: public sector
finance and health care finance. The demand for
financial transparency and accountability in
taxpayer-funded entities raises the need for more
uniformity and accessibility in public health
financial data. Meeting these expectations will
require high levels of performance and greater
investment in financial expertise.
Implications for Policy, Delivery or Practice:
The next step towards improvement in public
health financial management is an assessment
of current managers’ training needs from their
own perspective as well as that of their
supervisors. Once these needs have been
identified, existing educational resources should
be assessed and where necessary, new materials
developed to raise the level of public health
financial management performance.
Funding Source(s): RWJF
Call for Panels
Improving Safety Culture in Healthcare
Organizations
Chair: Sara Singer, M.B.A.
Sunday, June 3 • 3:00p.m.-4:30p.m.
Panel Overview: Since the Institute of Medicine
publicized in 1999 significant lapses in safety in
US hospitals, efforts to improve patient safety
have received considerable attention from
policymakers and practitioners. Based on
evidence from organizations that face hazardous
conditions akin to healthcare delivery yet achieve
strong safety performance, many believe that
initiatives that seek to strengthen an
organization’s culture of safety are among the
most promising for reforming healthcare
delivery. In recent years, research has made
significant progress in defining and measuring
safety culture, understanding how it varies
among and within organizations, and using data
to direct and evaluate safety performance
enhancing improvements. Our panel will review
some of the most important research-based
initiatives in this field. Four presentations will
address this topic. The first presentation will
address considerations in choosing the right
safety culture survey instrument for one’s own
organization or system by presenting a
comprehensive review and comparison of safety
culture measurement instruments. The second
presentation will discuss the importance of
comparing safety culture in one’s own hospital
to others and will present comparative data from
the AHRQ Hospital Survey on Patient Safety
benchmarking database. The third presentation
will explore internal variations in safety culture
for targeting improvement efforts based on
findings from an investigation of safety culture
survey results from a random sample of 92 US
hospitals. The fourth presentation will describe
one of the most promising interventions
evaluated to date for improving safety culture
and performance reliability: the comprehensive
unit-based safety program.
ƒ Safety Climate: Variation by Management
Status and Profession
Sara Singer, M.B.A., Laurence Baker, Ph.D.,
David Gaba, M.D., Alyson Falwell, M.P.H.,
Tobias Rathgeb, B.A., Shoutzu Lin, M.S.
Presented by: Sara Singer, M.B.A., Senior
Research Scholar, Center for Health
Policy/Primary Care and Outcomes Research,
Stanford University, 117 Encina Commons,
Stanford, CA 94305, Phone: (617) 495-5047, Fax:
(617) 325-7451, Email: ssinger@hbs.edu
Research Objective: Understanding how safety
culture varies within organizations is critical for
designing patient safety improvement efforts.
Using more aggregated measures of safety
culture perceptions and often focusing on
narrow groups of hospitals, previous research
has identified variations across broad classes of
hospital personnel, showing for example that
senior managers’ perceptions about safety
culture are more optimistic than those of front
line employees and that physicians are more
optimistic than nurses. To use this information
to fashion culture improving strategies, further
detail is needed. This paper examines in a
nationwide hospital sample whether variations
between (1) senior managers and front line
personnel and (2) nurses and physicians persist
when more detailed measures of safety culture
are examined. We also study whether patterns
of variation are observed among groups of
varying ages, gender, tenure, and work area.
Study Design: Individual-level cross sectional
comparisons relate survey-based measures of
safety culture to individual characteristics. Safety
culture measures are derived from the Patient
Safety Climate in Healthcare Organizations
(PSCHO) survey, administered to personnel at a
nationwide sample of hospitals in 2004. Survey
results provide an overall measure of each
respondent’s perception of the strength of safety
culture at their institution, and eight sub-scale
measures. Psychometric analysis of response
patterns supports data validity. Differences in
perceptions are examined using regressions that
recognize sample dependence due to the nesting
of individuals within work areas within hospitals
(e.g. hierarchical and fixed effects modeling).
We focus on variations by management status
(senior manager, front line worker) and
profession (physician, nurse, other staff).
Regressions control for hospital, work area, and
respondent’s age, gender, and job tenure.
Population Studied: All senior managers, all
physicians, and a random sample of 10% of
other staff at 92 hospitals. The sample includes
hospitals of a wide range of sizes and hospitals
in all census regions.
Principal Findings: While nurses reported more
problems than physicians in some areas,
physicians reported more problems in other
areas. For example, physicians were more likely
to experience shame when they needed to ask for
help. Also, differences among nurses and
doctors depended on additional characteristics
of personnel. For example, nurses who were
senior managers or near retirement were more
optimistic than their younger and more junior
counterparts. We also found that while hospital
senior managers were more optimistic than nonsenior managers consistently across all
dimensions of safety climate, this was not true in
all hospitals. In addition, senior managers who
were young or who were clinicians were more
pessimistic than older, non-clinician senior
managers.
Conclusions: Physicians do not uniformly
perceive safety culture more optimistically than
nurses. There are many nuances and
interactions that influence safety culture
measurement.
Implications for Policy, Delivery or Practice:
Differences in perceptions of problems between
nurses and physicians suggest the need for
different interventions for each group.
Consistent differences in perceptions between
senior managers and front line employees
suggest that interventions that engage senior
managers and expose them to experiences of
front line care could improve safety culture.
Funding Source(s): AHRQ
ƒ Results from the AHRQ Survey on Patient
Safety Culture
Joann Sorra, Ph.D., Veronica Nieva, Ph.D.,
Theresa Famolaro, M.P.S., Naomi Dyer, Ph.D.
Presented by: Joann Sorra, Ph.D., Westat, 1650
Research Boulevard, Rockville, MD 20850,
Phone: (301) 294-3933, Fax: (301) 315-5912,
Email: joannsorra@westat.com
Research Objective: The AHRQ Survey on
Patient Safety Culture Hospital Database
represents the largest compilation of data from
the AHRQ Hospital Survey on Patient Safety
Culture currently available and serves as a useful
reference for benchmarking and comparison.
This paper presents results from the 2007
database, identifying areas of strength and
patient safety culture improvement across the
database hospitals. Survey results by hospital
characteristics, and respondent staff position
and unit are also presented.
Study Design: The database consists of survey
data from 382 hospitals with 108,621
respondents. Hospitals voluntarily submitted
their data to the database. An average of 284
surveys were submitted per hospital, with an
average hospital response rate of 56%. Paper
surveys had a higher average response rate
(62%) than web (43%) or mixed mode (53%).
Most hospitals (79%) administered the survey to
a census of staff or a sample from all hospital
departments. The database hospitals are
generally consistent with the national
distribution of AHA-registered hospitals (by bed
size, teaching status, region).
Population Studied: The 2007 database
consists of data from 108,621 respondents
across 382 U.S. hospitals.
Principal Findings: The average percent of
positive response on each of the survey’s 42
items and 12 patient safety culture composites
was calculated across database hospitals.
Teamwork Within Units was the composite with
the highest average positive response (78%);
Nonpunitive Response to Error had the lowest
positive response. On average, the majority of
respondents within hospitals (70%) gave their
unit a patient safety grade of ‘‘Excellent’’ or ‘‘Very
Good.’’ However, the majority of respondents
within hospitals (53%) had reported no events
over the past year. Smaller hospitals (49 beds or
fewer) had the highest average positive response
on all 12 patient safety culture composites.
Respondents in Administration/Management
had the highest average positive response on 11
of the 12 composites. Rehabilitation had the
highest percent of respondents who gave their
unit a patient safety grade of ‘‘Excellent’’ or ‘‘Very
good’’ (81%); Medicine had the lowest percent
(58%). ICU (any type) had the highest percent of
respondents who had reported one or more
events in the past year (69%); the lowest percent
reporting was from Anesthesiology (41%).
Percentile scores are also presented on the items
and composites to show the variability of scores
and enable more precise comparisons for
hospitals.
Conclusions: There is a fair amount of variability
on the survey across database hospitals.
Averages and percentiles provide useful
information to enable hospitals to compare their
results. The database also helps identify key
areas of strength and patient safety culture
improvement for most hospitals.
Implications for Policy, Delivery or Practice: It
is important for hospitals administering a patient
safety culture survey to be able to compare their
results against other hospitals. The 2007
database presents statistics across a large
number of hospitals and respondents to enable
hospitals to compare their results, and provides
general conclusions about patient safety culture
results across hospitals, and by staff positions
and work areas.
ƒ Considerations for Choosing a Safety
Climate Survey Instrument
Barry Kitch, M.D., M.P.H., Eric Campbell, Ph.D.
Presented by: Barry Kitch, M.D., M.P.H.,
Institute for Health Policy, Massachusetts
General Hospital, 50 Staniford Street, Boston,
MA 02114, Phone: (617) 726-4201,
Email: bkitch@partners.org
Research Objective: Safety culture is important
to patient safety and measurable. Yet numerous
surveys have been developed in the absence of a
clear consensus on required or desirable survey
content or a gold standard against which to
judge the performance of the instruments. To
assist potential users we undertook a review of
existing patient safety culture surveys and
compared content (e.g. areas or dimensions of
safety addressed), survey development, and
other features of relevance to their use for quality
improvement and research purposes.
Study Design: Surveys were identified through a
systematic review, which included a MEDLINE
and Internet search, expert input, and review of
references from relevant articles. We compared
the content of surveys by assigning each item in
each survey to the area or dimension of safety
addressed by that item. For each instrument we
also examined target populations, number of
questions, reliability and validity testing, and
availability of benchmarking.
Population Studied: 13 instruments were
selected for detailed review.
Principal Findings: The instruments varied
substantially in content, emphasis, and length.
Individual surveys addressed as many as 19
dimensions of safety or as few as 4. While most
of the surveys had substantial face validity,
limited validation of the instruments by
comparison with qualitative measures of patient
safety culture, such as in depth interviews or
observation, has been done, and little data exists
on their ability to predict other existing patient
safety outcomes or indicators.
Conclusions: Available safety culture
instruments vary substantially in content,
emphasis, and length. Questions about which
dimensions a safety culture survey should
include as well as how best to analyze and
interpret results remain.
Implications for Policy, Delivery or Practice: In
choosing an instrument potential users need to
be aware of substantial differences between
surveys; selection should depend on the
intended use, target population, desire for
benchmarking, and specific areas of safety that
one would like to assess as well as an
appreciation for differences across instruments
in the rigor and transparency of the development
process. An awareness of the differences
between the instruments, as well as their
collective limitations, should facilitate the
process of assessing safety culture and
underscores the need for further research.
Funding Source(s): AHRQ
ƒ Safety Culture at Work: Assessing &
Improving
J. Bryan Sexton, Ph.D., Peter Pronovost, M.D.,
Ph.D., Eric Thomas, M.D., Martin Makary, M.D.
Presented by: J. Bryan Sexton, Ph.D., Assistant
Professor, Johns Hopkins Quality and Safety
Research Group, Department of Anesthesiology
and Critical Care Medicine, The Johns Hopkins
University School of Medicine, 1909 Thames,
Suite 200, Baltimore, MD 21231, Phone: (443)
287-2274, Fax: (410) 502-3235,
Email: sexton@jhmi.edu
Research Objective: Here we review 10 years of
cross sectional and longitudinal administrations
of the Safety Attitudes Questionnaire (SAQ), to
elicit frontline caregiver assessments of safety
culture. Only recently has research on safety
culture in healthcare blossomed, after the
Institute of Medicine*s recent suggestion that
healthcare organizations improve their safety
cultures. Despite the large and growing body of
research in other high risk work settings, safety
culture research in healthcare is still in the early
stages. Early evidence demonstrates that safety
culture can be measured, is linked to clinical and
operational outcomes, and is improvable with
structured interventions. Executives, managers
and frontline clinician leaders in quality and
safety have widely varying levels of familiarity and
comfort with safety culture data, but there is
increasing pressure from boards, quality
improvement organizations, and the media to
do something about the safety culture.
Study Design: Using standard psychometric
tests such as multi-level confirmatory factor
analyses, ICCs, and Rwgs, we demonstrate that
consensus assessments of frontline caregivers
within a patient care area can be represented by
SAQ scale score. Using multi-level structural
equation modeling, we demonstrate the
relationship between safety culture and clinical
outcomes. Using paired samples t-tests, we
demonstrate how safety culture at the care area
level is responsive to interventions.
Population Studied: This review includes 200
USA hospitals between 2004 and 2006,
including teaching and non-teaching, urban and
rural, and bed sizes ranging from fewer than 50
to over 800. Specific interventions at the care
area/nursing unit level that were targeted to test
for improvement in safety culture were the
Comprehensive Unit-based Safety Program
(CUSP), Executive WalkRounds, and Operating
Room Briefings.
Principal Findings: These three interventions
are relatively popular despite a lack of robust
evidence in how to implement and monitor
them. There is evidence for teamwork and safety
climate improving after implementing CUSP.
Also, improvements in safety climate were
associated with rigorously conducted Executive
WalkRounds, while teamwork climate
improvements were associated with rigorously
conducted OR Briefings. Care areas with fewer
than 40 caregivers generally improved more than
their medium (40-80) and larger (over 80)
counterparts. Many caregivers report that they
are not aware of the intervention being
implemented in their unit, and there was
considerable ambiguity around use of terms like
Executive WalkRounds vs., e.g., Safety Rounds.
Conclusions: Quality and safety interventions
are often implemented without committing to
the methodological rigor needed to demonstrate
impact. However, when methodological rigor is
utilized, the association between interventions
and safety culture can be evaluated.
Implications for Policy, Delivery or Practice:
Given the limited quality and safety resources
available to most hospitals and the limited
attention afforded to quality by hospital leaders,
changes to care delivery processes and
structures need to be contingent upon evidence
of positive impact, rather than the justification
that it needs to be done because it is for quality.
There is great promise of safety culture research
to inform this evidence and provide actionable
information relevant to implementing
innovations in healthcare, and tracking the
impact of those interventions over time.
Call for Panels
Lessons from the VA on Patient Safety: From
Research to Policy & Practice Implications
Chair: Amy Rosen, Ph.D.
Sunday, June 3 • 5:00 p.m.-6:30 p.m.
Panel Overview: The purpose of this panel is to
highlight recent findings on patient safety from
the Veterans Administration (VA) that impact
both policy and practice. The VA is considered a
national leader in patient safety, with numerous
initiatives including medication bar coding,
computerized physician order entry, and
nationwide incident reporting systems. Yet, little
is known as to whether these system-wide
activities have resulted in improvements in care
delivery, patient outcomes, and safety culture.
We present results from several current research
projects and initiatives in the VA in order to fill
this evidence gap. Presentations are focused on:
1) current research priorities and activities of the
VA National Center for Patient Safety: assessing
their impact on VA policy and practice; 2) the
change in resident work hours and its impact on
quality and safety in the VA; 3) the variation in
safety culture across VA facilities and differences
in perceptions between job type and work areas:
comparison to the non-VA setting; 4) racial
disparities in rates of Patient Safety Indicators
(PSIs) across the VA: comparison to the non-VA
setting; and 5) processes and structures of care
from site visits to low- and high-performing
facilities on PSI rates: organizational themes and
implications for best safety practices.
ƒ Efforts to Achieve a Culture of Safety in the
Veterans Health Administration
Edward Dunn, M.D., M.P.H.
Presented by: Edward Dunn, M.D., M.P.H.,
Director of Policy and Clinical Affairs, VA
National Center for Patient Safety, Department
of Veterans Affairs, 24 Frank Lloyd Wright Drive,
PO Box 486, Ann Arbor, MI 48106-0486, Phone:
(734) 930-5872, Fax: (734) 945-0199,
Email: Edward.Dunn@va.gov
Research Objective: To evaluate the
effectiveness of policy changes and national
programs designed to enhance patient safety
culture across the Veterans Health
Administration.
Study Design: Before/after analyses of patient
safety culture from surveys performed in 2000
and 2005, and descriptive analyses of rationale
for pursuing identified interventions in an effort
to achieve a culture of safety.
Population Studied: Employees of the Veterans
Health Administration.
Principal Findings: The VA National Center for
Patient Safety (NCPS) is guided by a mission of
preventing harm to patients. Toward that end,
NCPS has established a structure of Patient
Safety Managers in 154 VA medical centers and
Patient Safety Officers in the 21 VA networks who
facilitate program implementation throughout
the Veterans Health Administration. NCPS has
pursed a multi-pronged strategy to improve the
culture of safety within the VA system that
includes implementation of patient safety
training, patient safety toolkits, medical team
training, alerts and advisories, and VA directives.
Over the time period examined, measures of
patient safety culture have improved in some
areas, but have not changed in other areas.
Conclusions: The NCPS has made progress in
improving a culture of safety within the Veterans
Health Administration. Culture is likely to
change slowly, so repeat surveys are warranted.
Implications for Policy, Delivery or Practice:
Future efforts to enhance patient safety culture
should target areas where the greatest need for
improvement has been identified. Ongoing,
regular collection and analysis of patient safety
culture can be an effective strategy to identify
areas for improvement and monitor progress
toward achieving a culture of safety.
Funding Source(s): VA
ƒ Safety Culture in VA Hospitals
Priti Shokeen, M.S., Amy Rosen, Ph.D., Shibei
Zhao, M.P.H., Mark Meterko, Ph.D., Alyson
Falwell, M.P.H., David Gaba, M.D.
Presented by: Priti Shokeen, M.S., Health
Science Specialist, Center for Health Quality
Outcomes and Economic Research, Department
of Veterans Affaris - Edith Nourse Rogers
Memorial Veterans Hospital, 200 Springs Road
(152), Bedford, MA 01730, Phone: (781) 687-3258,
Fax: (781) 687-3106, Email: ptrivedi@bu.edu
Research Objective: A culture of safety is
recognized as a key to reducing adverse events
in hospitals. The perceived safety culture may
vary within institutions, with potential
implications for efforts to improve safety culture.
Although the VA (Veterans Administration) is a
national leader in patient safety, little is known
about the current safety culture in VA hospitals.
This study explores the strength and uniformity
of safety culture within VA hospitals and
compares safety culture between VA and non-VA
hospitals.
Study Design: We measured safety climate
using the Patient Safety Climate in Healthcare
Organizations (PSCHO) survey, administered to
employees at a nationwide sample of 29 VA
hospitals between December 2005 and June
2006. The PSCHO consists of 42 Likert-scale
questions and six demographic questions. For
each question, a "problematic response" (PR)
suggests a weak safety climate. We calculated
average percent PR rates and 95% confidence
intervals (CIs), accounting for sample proportion
and patterns of non-response for each question,
job-type, and work area across VA hospitals. A
lower percent PR represents a stronger safety
climate. T-tests were performed to check for
significance between job-types and work areas.
We compared results to 105 non-VA hospitals
that administered the PSCHO survey during
2004.
Population Studied: At each hospital, 100% of
hospital-based physicians, 100% senior
managers, defined as department heads and
above, and 10% random sample of all other
personnel were surveyed. At 10 hospitals, we
surveyed 100% staff whose primary location of
work was a high-hazard unit.
Principal Findings: The overall response rate
was 49% (4,504 individual responses) and the
average overall PR rate across all hospitals was
18.1%; this rate was similar to the private sector
(17.3%). Rates of PR varied widely across
institutions and by questions. The items eliciting
the highest and lowest PR rate were ‘‘My unit
recognizes individual safety achievement
through rewards and incentives’’ (49.6%) and ‘‘If
I make a mistake that has significant
consequences and nobody notices, I do not tell
anyone about it,’’ respectively (4.4%). Senior
managers had a significantly lower PR rate
compared to supervisors, frontline staff, and
physicians. Nurses had the highest PR rates of
any job-type (CI=17.0%-24.7%). Employees in
high-hazard units had a higher PR rate (22.6%)
than other employees (17.6%). In general, our
results are comparable to non-VA hospitals.
Conclusions: Both strength and uniformity of
safety culture are important in assessing
hospitals’ commitment to patient safety.
Consistent with the private sector, we found that
a) perceptions of safety culture differed among
hospitals, job-type, and work area, b) senior
managers have a more positive perception of
safety culture than frontline staff, and c)
employees working in high-hazard areas have a
more negative perception of safety culture.
Implications for Policy, Delivery or Practice:
Despite the positive overall perception of safety
culture among VA employees, variation in
problematic response warrants further
investigation. Efforts to create a uniform safety
culture within an institution are necessary to
improve patient safety. Activities that reduce
differences in perceptions between senior
managers and frontline workers, as well as those
that focus on high-hazard units, are critical to
enhancing efforts to improve safety throughout
an organization.
Funding Source(s): VA
ƒ Racial Disparities in Patient Safety Indicator
(PSI) Rates in the Veterans Health
Administration
Stephanie Shimada, Ph.D., Maria Montez-Rath,
M.S., Susan Loveland, M.A.T., Shibei Zhao,
M.P.H., Nancy Kressin, Ph.D., Amy Rosen, Ph.D.
Presented by: Stephanie Shimada, Ph.D.,
Postdoctoral Health Services Research Fellow,
Center for Health Quality Outcomes and
Economic Research, Department of Veterans
Affaris, 200 Springs Road (152), Bedford, MA
01730, Phone: (781) 687-2000 x6698, Fax: (781)
687-3106, Email: shimada@bu.edu
Research Objective: Minorities often experience
reduced access to care, poorer quality of care,
and worse outcomes. Recent findings from
AHRQ using data from the Healthcare Cost and
Utilization Project (HCUP), a national sample of
inpatient discharges, suggest that there are
some disparities in Patient Safety Indicator (PSI)
rates between whites and ethnic minorities.
However, there is evidence within VA that quality
of care does not differ by race and may even be
better for minorities because there are fewer
financial barriers to access compared with other
healthcare systems. We examined whether there
are racial/ethnic disparities in the incidence of
AHRQ’s Patient Safety Indicators (PSIs) within
VA.
Study Design: We utilized FY2000 ---- FY2005
inpatient discharge data from 128 acute-care VA
hospitals (N=2,281,286 hospitalizations) to
calculate PSI events and obtain data on patient
characteristics. We then supplemented inpatient
race with data from the VA outpatient files to
lower the overall rate of missing race to 3.6%.
To determine whether race was significantly
associated with the likelihood of a PSI event, we
ran logistic regression models for fifteen PSIs,
controlling for patient age, gender, and
comorbidities, adjusting for clustering at the
hospital level. Patients were excluded from the
analyses if they had missing inpatient race data
(3.4 - 4.9% of hospitalizations, depending on the
PSI).
Population Studied: Veterans who received
inpatient care at one of 128 acute-care hospitals
in the Veterans Health Administration between
FY2000 and FY2005 (N=93,488 ---- 2,281,252
hospitalizations, depending on the PSI).
Principal Findings: The odds of decubitus ulcer
(OR=1.35, 95%CI [1.20,1.53]) and postoperative
pulmonary embolism and deep vein thrombosis
(PE/DVT) (OR=1.23, 95%CI [1.13,1.34]) were
higher for African-Americans than for whites. In
contrast, African-Americans had significantly
lower odds than whites for 3 PSIs: complications
of anesthesia (OR=0.63, 95%CI [0.44,0.91]),
postoperative wound dehiscence (OR=0.71,
95%CI [0.53,0.97]), and accidental
puncture/laceration (OR=0.69, 95%CI
[0.56,0.84]). Latinos also had significantly higher
odds of decubitus ulcer (OR=1.60, 95%CI
[1.02,2.50]) but had significantly lower odds of
postoperative hemorrhage/hematoma (OR=0.73,
95%CI [0.55,0.97]) compared with whites. No
significant racial/ethnic differences were found
for the remaining 9 PSIs.
Conclusions: The risk of PSI events varied
across racial/ethnic subgroups. Many PSIs
showed no significant differences across
race/ethnicity; the differences found showed that
minorities were at higher risk for some patient
safety events and at lower risk for others.
However, two high-frequency PSI events
(decubitus ulcer and postoperative PE/DVT)
were more likely to occur in minorities.
Implications for Policy, Delivery or Practice:
We found fewer differences between ethnic
minorities and whites in PSI rates than have
been documented in non-VA settings. However,
because decubitus ulcer and postoperative
PE/DVT are high-frequency adverse events,
quality improvement initiatives targeting these
conditions, in addition to research exploring the
pathways leading to these disparities, should be
initiated within the VA. Future research should
examine patient- and hospital-level factors that
might explain variation in PSI events within the
VA and between the VA and non-VA settings.
Funding Source(s): VA
ƒ Moving Patient Safety Improvement
Practices to the Next Level: Closing the
Organizational Learning Loop
Peter Rivard, Ph.D., Victoria Parker, D.B.A., Amy
Rosen, Ph.D.
Presented by: Peter Rivard, Ph.D., Postdoctoral
Fellow, Center for Organization, Leadership and
Management Research, Department of Veterans
Affairs - VA Boston Healthcare System, 150
South Huntington Avenue (152M), Boston, MA
02130, Phone: (857) 364-5691, Fax: (857) 3544438, Email: rivardp@bu.edu
Research Objective: The Veterans Health
Administration (VA) is recognized for its
initiatives to improve patient safety. Many such
initiatives follow the quality-improvement
organizational learning model: planning and
implementation of safety improvements is
driven by collection and analysis of data from
within the organization. Examples include
adverse event reporting, Root Cause Analyses
(RCA) and Healthcare Failure Mode Effects
Analyses (HFMEA). The purpose of this study
was to identify variation in implementation of
patient safety improvement practices between
facilities with higher and lower patient safety
performance.
Study Design: Our research entailed exploratory
case studies of four VA acute care hospitals. The
sites were selected for their (a) high and low
performance on a proxy for patient safety
outcomes, and (b) geographic diversity. Safety
performance was approximated by the AHRQ
Patient Safety Indicators (PSIs), which use
administrative data to identify potentially
preventable adverse events in acute inpatient
care. A subset of PSIs was applied to Fiscal Year
2001-2004 VA discharge data to generate a
composite ranking of facilities. At the selected
sites, data on safety improvement structures and
practices were obtained through face-to-face
semi-structured interviews conducted in 2005
with executives, managers, and others involved
in patient safety improvement. Interviewers and
interviewees were blinded as to facility safety
performance on the PSIs. Data analysis explored
facility-level variation in implementation of
patient safety improvement. Interview
transcripts were coded for both a priori
constructs and emergent themes.
Population Studied: VA facilities that provide
acute inpatient care.
Principal Findings: The two facilities with
patient safety performance that was either better
overall or more improved 2001-2004 showed
evidence of organizational structures for patient
safety learning and improvement that were more
comprehensive on two dimensions: (1) Projectlevel evaluation: more follow-up assessment of
specific patient safety interventions through
monitoring of implementation success and
measurement of changes in process and
outcomes; (2) Program-level evaluation: more
(a) executive-level assessment of the overall
effectiveness of the patient safety program and
(b) attention to integration across various
patient safety initiatives. Our study describes
these evaluation practices and presents
exemplars.
Conclusions: While organizational safety
learning practices such as RCAs and HFMEAs
help to elevate organizational response to safety
problems from first-order ‘‘quick fixes’’ and
workarounds to more effective second-order
system improvement, this study suggests that
higher-performing facilities then progress to a
higher order of organizational learning, where
they systematically assess the effectiveness and
efficiency of their interventions. This is
consistent with findings from implementation
research: organizations with systems to monitor
implementation of innovations and evaluate
their impact are more likely to assimilate and
sustain the innovation.
Implications for Policy, Delivery or Practice:
Further research should test the propositions
generated by this study, i.e., that a next level of
patient safety improvement, after
implementation of focused system improvement
practices such as event reporting and RCA,
consists of systematic assessment of the
effectiveness of these interventions, along with
steps to prioritize, coordinate and integrate
safety improvement initiatives across the facility.
Furthermore, our findings suggest that an
organization’s progress to this level may not be
automatic and may require additional
intervention.
Funding Source(s): VA
ƒ The Impact of the ACGME Duty Hour Rules
on PSI Rates in VA Hospitals
Kevin Volpp, M.D., Ph.D., Amy K Rosen, Ph.D.,
Paul R. Rosenbaum, Ph.D., Patrick Romano,
M.D., Orit Even-Shoshan, M.S., Anne
Canamucio, M.S., Susan Loveland, Jeffrey H.
Silber, M.D., Ph.D.
Presented by: Kevin Volpp, M.D., Ph.D., Staff
Physician and Assistant Professor, CHERP Center for Health Equity Research and
Promotion, Philadelphia VA Medical Center
CHERP, (University of Pennsylvania School of
Medicine and the Wharton School), 1232
Blockley Hall, Philadelphia, PA 19104-6021,
Phone: (215) 573-0270, Fax: (215) 573-8778, Email:
volpp70@mail.med.upenn.edu
Research Objective:
In response to concern about deaths in
American hospitals from medical errors, the
Accreditation Council for Graduate Medical
Education (ACGME) released rules effective July
1, 2003 that restricted duty hours for all ACGMEaccredited residency programs. The objective of
this study was to determine the impact of the
duty hour regulations, one of the largest
interventions ever undertaken to improve patient
safety in teaching hospitals nationwide, on rates
of Patient Safety Indicators (PSIs) in VA
hospitals. We use one high-frequency wellvalidated PSI, the failure to rescue (FTR) rate
(death after surgical complications), to examine
this question. As lower (better) FTR rates are
generally believed to be associated with early
detection of complications, more rested
residents may recognize and respond to
complications more rapidly, thus reducing FTR.
However, more handoffs may also create
situations in which residents do not know the
patients they cover as well as in the past, thereby
lengthening the time before intervention after
complications and possibly worsening FTR rates.
Study Design:
Observational study using interrupted time
series analysis with data from July 1, 2000 - June
30, 2005. Logistic regression was used to
examine the change in the FTR rate for patients
in more versus less teaching-intensive hospitals
before and after duty hour reform, adjusting for
patient comorbidities, common time trends, and
hospital site. Data were obtained from the VA
Patient Treatment hospital discharge file, the VA
Beneficiary Identification and Record Locator
Subsystem file, and the VA Office of Academic
Affiliations.
Population Studied:
All unique patients (n=66,456) admitted to
acute-care VA hospitals from July 1, 2000 to June
30, 2005 with complications from general,
orthopedic or vascular surgery.
Principal Findings:
There were no significant relative changes in the
FTR rate for patients who suffered complications
from general surgery, orthopedic surgery or
vascular surgery in either post reform year 1 or 2.
Combining the 3 groups of patients to form a
‘combined surgery group’ did not result in
significant relative change in the FTR rate in
either post-reform year 1 (OR 0.87, 95% CI [0.66,
1.15]) or post-reform year 2 (OR 0.88, 95% CI
[0.66, 1.17]). Finally, combining both years of
post data also indicated no significant relative
change in the FTR rate (OR 0.87, 95% CI [0.70,
1.09]).
Conclusions:
The ACGME duty hour regulations were not
associated with significant changes in the FTR
rate for patients in VA hospitals.
Implications for Policy, Delivery, or Practice:
We found no statistically significant indication
that the new duty hour standard harmed or
helped surgical patients after developing
complications. In contrast, our earlier work
indicated that for some medical conditions,
there were significant reductions in mortality
post-reform. A single duty hour standard may
affect trainees in different specialties differently,
and further research is needed to better
understand why there are differences in the
impact of duty hour reform on medical vs.
surgical patients.
Primary Funding Source: VA
Call for Panels
Emerging Findings in Public Health Systems
Research
Chair: Kate Papa, M.P.H.
Sunday, June 3 • 5:00 p.m.-6:30 p.m.
Panel Overview: The tragedy of 9-11 exposed the
long-neglected public health infrastructure in the
United States. Bioterrorism threats that
followed, along with a series of natural disasters
such as the hurricanes of 2004 and 2005 that
devastated parts of Florida and the Gulf Coast,
have further underscored the inadequacies of
public health systems at all levels of government.
The emerging field of Public Health Systems
Research (PHSR) is responding by growing the
evidence base and investigating the public health
system at many levels. This field of inquiry
examines the organization, financing,
performance, and impact of health systems ---defined as the constellation of governmental and
non-governmental actors that influence
population health, including health care
providers, insurers, purchasers, public health
agencies, community-based organizations, and
entities that operate outside the traditional
sphere of health care. This panel will present
findings from four projects funded under the
Robert Wood Johnson Foundation’s Special
Topic Solicitation in PHSR funded through the
Changes in Health Care Financing and
Organization (HCFO) initiative. The first study
explores the definition of measurement of the
value of services provided by governmental
public health systems and explores alternative
methodologies for measuring value. The second
study examines local public health service
delivery in small jurisdictions. The third study
examines regionalization in public health
systems, examining differences in development
and performance. Finally, the last study
examines the relationships between spending
and outcomes. Together, these papers will
provide a comprehensive snapshot of public
health systems, the expectations of these
systems, and their effect on preparedness and
health outcomes.
ƒ Causes and Consequences of Change in
Local Public Health Spending
Glen P. Mays, Ph.D., M.P.H.
Presented by: Glen P. Mays, Ph.D., M.P.H.,
Associate Professor, Chair Pro Tem, Department
of Health Policy and Management, Fay W.
Boozman College of Public Health, University of
Arkansas for Medical Sciences, 4301 W.
Markham, #820, Little Rock, AZ 72205, Phone:
(501) 526-6647, Email: gpmays@uams.edu
Research Objective: A growing body of evidence
indicates that the availability and quality of public
health services varies widely across
communities, but relatively little is known about
the factors that give rise to this variation. Public
health activities in the U.S. are supported
through a patchwork of funding streams that
vary considerably across communities and
change over time in response to economic and
policy shifts. This variability in spending may
have important consequences for community
health. The objectives of this analysis are to: (1)
examine how local public health spending levels
and funding sources changed over the past
decade; (2) identify the types of communities
most likely to experience disparities in public
health spending; and (3) determine whether
changes in public health spending are associated
with changes in community health status and
disease burden.
Study Design: A longitudinal cohort design is
used to analyze changes in spending patterns
and population health within service areas of the
nation’s 3000 local public health agencies
between 1993 and 2005. The National
Association of County and City Health Officials
(NACCHO) collected data on the organizational
and financial characteristics of these agencies
through census surveys fielded in 1993, 1997,
and 2005. We linked these data with
contemporaneous information on community
characteristics, federal and state spending, and
public health disease burden from other data
sources. Multivariate regression models for
panel data are estimated to test for changes in
spending patterns over time and differences
across communities defined by demographic,
economic, organizational and geographic
characteristics. Instrumental-variables methods
are used to identify associations between
spending levels and community health while
controlling for unmeasured factors that jointly
influence spending and health.
Population Studied: The study includes all U.S.
agencies meeting NACCHO’s definition of a
local health department: an administrative or
service unit of a local or state government that
has responsibility for the health of a jurisdiction
smaller than a state. There were 2875 such
agencies in 1993 (response rate 77%) and 2864
in 2005 (response rate 80%).
Principal Findings: Local public health agency
spending increased from a median of $20 per
capita in 1993 to $29 in 2005 in nominal dollars,
indicating an annual growth rate of less than 4
percent. More than 20% of agencies
experienced reductions in per capita spending.
Larger increases in spending were observed
among agencies serving metropolitan
communities, agencies operating as units of
local government, and agencies governed by
local boards of health (p<0.01), even after
adjusting for other agency and community
differences. Increases in spending were
associated with significant improvements
(p<0.05) in several key measures of community
health, including infant mortality and case rates
for clamydia, gonorrhea, hepatitis, and
tuberculosis. Instrumental-variables estimates
indicate that even stronger relationships exist
after accounting for unmeasured characteristics
that jointly influence spending and health.
Conclusions: Public health spending varies
widely across U.S. communities, and this
variability has persisted over time. Differences in
spending may contribute to differences in
community health outcomes.
Implications for Policy, Delivery or Practice:
Efforts to improve local public health resources
and infrastructure may help to address
important differences in health status across
communities.
Funding Source(s): RWJF
ƒ Regionalization in Local Public Health
Systems: Variation in Rationale,
Implementation, and Impact on Public Health
Preparedness
Michael A. Stoto, Ph.D.
Presented by: Michael A. Stoto, Ph.D., Professor
of Health Services Administration and
Population Health, Georgetown University
School of Nursing & Health Studies, 3700
Reservoir Road, NW, Box 571107, Washington,
DC 20057-1107, Phone: (202) 687-3292,
Email: stotom@georgetown.edu
Research Objective: The objective of this study
is to provide a better understanding of the
regionalization of public health systems in order
to inform the many state and local health
departments currently developing regional
structures.
Study Design: In order to learn from areas that
have adopted a regional approach we have used
comparative case studies to (1) document the
variation in the rationale for creating regional
public health structures; (2) understand how
these structures have been organized,
implemented, and governed; and (3) assess the
current and likely impact on of regional
structures on public health preparedness and
public health systems more generally.
Population Studied: Massachusetts, Northern
Illinois, Nebraska, and the Washington D.C.
National Capital Region.
Principal Findings: The impetus for forming
regions was some combination of the following:
a crisis or perceived need for a coordinated
response, a need to build local public health
capacity, or an effort to use federal preparedness
funds more efficiently. The relationship between
public health regions and geopolitical
jurisdictions in which they sit was complex.
Some involve combinations of local health
departments, and one set up new health
departments for groups of counties. The
National Capital Region has multiple definitions,
including some that crossed state lines. The
regions vary in terms of their congruence with
regional structures for partner agencies such as
emergency management agencies as well as
hospital and health services markets and
organizational structure. Some of the regions
focus on building formal organizational
relationships to coordinate and sometimes
standardize preparedness and response
activities, and sometimes build regional capacity,
while others focused on building informal
professional networks. It is not clear which
approach will have the largest effect on building
social capital. Whatever the approach is chosen,
however, it is clear that strong leadership and
trust are required for effective planning,
emergency response, and sustainability.
Conclusions: Does regionalization improve
emergency preparedness? Logically,
regionalization allows for more efficient use of
resources and since disease outbreaks do not
respect geopolitical boundaries coordination is
needed. The case studies demonstrated
progress in terms of planning and coordination;
memoranda of understanding; regional capacitybuilding, training, and exercises; and
development of professional networks. The case
studies also showed effects in terms of the
public health response to the flu vaccine
shortage in 2004 and anthrax and tularemia
alarms in the National Capital Area in 2005.
Does regionalization improve public health
generally? Logically, the same arguments about
sharing resources and coordinated response
apply. Also, the case studies identify a number
of areas where regional capacities address other
needs. It seems likely that preparedness
concerns are forcing communities to think about
public health structures in a way that has not
been done in decades, and that network
development may be creating social capital that
helps with other concerns. But concerns that
preparedness efforts were drawing resources and
attention from other areas of public health were
also heard.
Implications for Policy, Delivery or Practice:
Most states have responded to the increased
interest in and funding for public health
preparedness by setting up regional structures,
but the rationale for these structures, the way
they are implemented, and presumably the
impact of this organizational change vary
considerably. As states receive this increase in
funding, they will be under increased pressure to
demonstrate accountability. These findings will
inform states and communities as they develop
rationale for such regionalized strictures.
Funding Source(s): RWJF
ƒ Measuring the Value of Public Health
Services
Peter D. Jacobson, J.D., M.P.H.
Presented by: Peter D. Jacobson, J.D., M.P.H.,
Professor of Health Law and Policy, Director,
Center for Law, Ethics, and Health, University of
Michigan School of Public Health, 109
Observatory, Ann Arbor, MI 48109-2029,
Email: pdj@umich.edu
Research Objective: 1) To understand how the
value of the services provided by governmental
public health systems (GPHSs) can be defined
and measured.
2) To determine what methodologies (e.g.,
contingent valuation, cost-benefit or costeffectiveness analyses) can be used to measure
value.
3) To develop an analytical framework for
measuring the value of GPHSs.
Study Design: First, we conducted a critical
synthesis of the literature to: 1) review the
methodological literature to examine how
various methodologies have been applied to
similar efforts (such as disease prevention); 2)
examine how other systems (public and private)
define and measure the value of their services;
and 3) assess general applicability of the
literature to the public health system. Second,
we interviewed key national, state, and local
public health practitioners, policymakers, and
elected officials to ascertain participants’ views
of how the public health system can define and
measure value, as well as the feasibility of doing
so. Altogether we conducted more than 40
interviews, including practitioners in three states.
Third, we conducted an extensive search for and
analysis of economic evaluations of GPHS
activities to: identify key gaps in the data;
critically evaluate each of the methodologies
used to value public health services; and
investigate alternative ways for quantitatively
measuring the value of GPHS services.
Population Studied: The populations studied
are national, state, and local public health
practitioners.
Principal Findings: 1) Despite a large number of
articles across different fields, very few focus on
the relevant concept of value. Most articles did
not explicitly define or outline how value was
measured.
2) Limited research on measuring value in
related fields.
3) Interview results:Very few current attempts to
formally define and measure value; Promising
cost accounting-deliberative democracy strategy;
Disagreement on the merits of focusing on
preventing disease as the measure of value;
Limited data available; Criteria emerged to
develop a framework for defining and measuring
value; Interviews suggest the importance of
communication in thinking about value.
4) Methodological results: Several approaches
have been used to value public health and
prevention strategies, including willingness to
pay (contingent valuation), human capital,
quality-adjusted life years (QALYs), life-years
gained, rate of return on investment, and the use
of ‘‘natural’’ health units; None emerged as an
optimal method for measuring value; Despite
the large number of studies, there are notable
variations in the quality and completeness of the
literature.
Conclusions: We conclude that a framework for
analysis can offer ways for GPHSs to define and
communicate the value of public health services
to various constituencies. Public health
practitioners will need to use a combination of
methods to measure value.
Implications for Policy, Delivery or Practice: It
is essential that GPHSs allocate resources to
those activities likely to achieve maximum value
for improving the public's health. The
framework we have developed is the first step
toward enabling public health practitioners,
policymakers, and elected officials to explain the
overall benefits of GPHS to the public.
Funding Source(s): RWJF
ƒ Structural Capacities, Processes, and
Performance of Essential Public Health
Services by Small Local Public Health
Systems
Susan Zahner, Dr.P.H., M.P.H., R.N.
Presented by: Susan Zahner, Dr.P.H., M.P.H.,
R.N., Associate Professor, School of Nursing,
University of Wisconsin-Madison, H6/246
Clinical Sciences Center, 600 Highland Avenue,
Madison, WI 53792-2455, Phone: (608) 263-5282,
Email: sjzahner@wisc.edu
Research Objective: Contemporary local public
health service delivery is conceptualized as being
provided by a local public health system that
includes, but is not limited to, the local health
department. Studies of the performance of local
public health systems are limited although
research has demonstrated a relationship
between local health department size and
performance. This study is being conducted to
examine the capacities, processes, and
performance of local public health systems
serving small jurisdictions (under 25,000
population).
Study Design: The study is being conducted in
two phases. Phase 1 included qualitative, on-site
interviews with key informants (n=37) from
organizations included in four small, local public
health systems. Phase 2 includes survey
instrument development and data collection
from 34 small local public health systems.
Population Studied: Small, local public health
agencies in Wisconsin.
Principal Findings: Phase 1 results will be
summarized and presented. Progress to date in
survey instrument development, sample
determination, data collection, and analysis
plans will be shared.
Conclusions: Study methods used may
contribute to the development of public health
system performance research methodologies
particularly in evaluating the organization,
financing, and delivery of public health services
and in measuring evidence about the
effectiveness of public health services and
interventions.
Implications for Policy, Delivery or Practice:
Study findings may be useful to policymakers
and administrators in targeting resources and
technical assistance in building structural
capacities in key areas associated with public
health system performance.
Funding Source(s): RWJF
composite measures of patient experience, in
different types of hospitalization, on overall
rating of hospital. To ensure that comparisons
across hospitals are fair and valid, CMS
implemented a randomized experiment to
measure and adjust for the effects of survey
mode. The third presentation offers new
empirical findings from this mode experiment.
A full-scale dry run of HCAHPS conducted in
2006 in over 2800 hospitals resulted in key
lessons learned in survey implementation and
data submission, the topic of the fourth
presentation. The fifth presentation examines
public reporting of HCAHPS results. Fresh
results from consumer testing of the Hospital
Compare website are offered, including preferred
options for coherently integrating the patient
perspective, clinical and mortality elements of
the website.
ƒ Developing and Implementing HCAHPS,
the First National, Standardized, Publicly
Reported Survey of Patients' Perspectives of
Hospital Care
William Lehrman, Ph.D., Elizabeth Goldstein,
Ph.D.
Call for Panels
Designing, Collecting, Analyzing and Publicly
Reporting Patient Assessments of the Quality
of Hospital Care: The HCAHPS Survey
Chair: William Lehrman, Ph.D.
Presented by: William Lehrman, Ph.D., Health
Insurance Specialist, Center for Beneficiary
Choices, Centers for Medicare & Medicad
Services, S1-13-05, 7500 Security Boulevad,
Baltimore, MD 21244-1850, Phone: (410) 7861037, Fax: (410) 786-8004,
Email: william.lehrman@cms.hhs.gov
Monday, June 4 • 9:00 a.m.-10:30 a.m.
Panel Overview: The first national, standardized
survey of hospital patients' perspectives of care
was launched by Centers for Medicare &
Medicaid Services (CMS) in October 2006. The
CAHPS® Hospital Survey (HCAHPS) produces
hospital-level information on quality of care for
public reporting. Beginning in July 2007,
HCAHPS will become an ingredient in ‘‘pay-forreporting’’ with the potential to affect billions of
dollars of CMS payments to hospitals. HCAHPS
will have a direct impact on patient choice,
consumer information, public perception, and
public policy. This panel describes the
development and scope of HCAHPS and
presents new empirical findings. The first
presentation provides an overview of the
purpose, composition, development, and
evolving scope of HCAHPS. Drawing on data
collected from 50,000 patients in the HCAHPS
pilot study, the second presentation examines
the differential impact of the survey’s seven
Research Objective: This presentation
introduces and explores the newly launched
CAHPS® Hospital Survey, also known as
HCAHPS: the first national, standardized,
publicly reported survey of patients' perspectives
of their hospital care. HCAHPS (Hospital
Consumer Assessment of Healthcare Providers
and Systems) unites patient-centered care with
public reporting of hospital quality information.
It is intended to augment consumer choice,
increase hospital accountability, provide
incentives for quality improvement, and add
greater transparency, all through the public
reporting of meaningful, valid and standardized
quality measures.
Study Design: After years of development and
testing, HCAHPS was launched by CMS in
October 2006. HCAHPS was developed to
produce comparable hospital-level information
for public reporting purposes. Currently, over
2800 acute care hospitals, and hundreds of
thousands of patients, are voluntarily
participating. However, beginning in July 2007,
HCAHPS will become an ingredient in ‘‘pay-forreporting’’ with the potential to affect billions of
dollars of CMS payments to hospitals. By mid2007, nearly all eligible hospitals (~4000) are
expected to participate. The first public reporting
of hospital results should attract wide attention
from the healthcare industry and the general
public. This presentation will provide the most
up to date information on the content,
development and scope of HCAHPS.
Content and Implementation
We present information on HCAHPS’ content
(seven areas of vital importance to patients are
tapped through 27 items: 18 on experience of
care; five demographic; four screeners),
participating hospitals (‘‘general,’’ acute-care),
eligible patients (adult; alive; overnight stay;
discharged; medical, surgical or maternity care),
the process and options for implementation (by
hospitals or approved survey vendors; four
survey modes; standardized protocols; 300
completed surveys per year), how results will be
publicly reported (on the Hospital Compare
website; integrated with clinical and mortality
measures), and how oversight will be conducted.
The second part of the presentation discusses
the process of developing, testing and refining
the survey, from its inception in 2002 to its
current form. The purpose of the survey has
evolved from the original goal of public
reporting, to encompass pay for reporting, and
eventually pay for performance. HCAHPS has
been developed in a open and participatory
manner, which included several opportunities for
public scrutiny and comment, independent
reviews and endorsements. We will discuss how
interests of key stakeholders, particularly
hospitals, survey vendors and consumer groups,
have been accommodated to the degree possible
while maintaining fidelity to original goals.
Finally, the impact of HCAHPS on patients,
consumers, hospitals, and the healthcare
industry will be broached. Because its full
impact will not be apparent for some time, we
will invite the audience to speculate on possible
outcomes, consequences, and avenues for
research.
Conclusions: By offering new information to
consumers and encouraging hospitals to
improve the care they provide, HCAHPS will
have a significant and direct impact on patient
choice, consumer information, public
perception, and public policy. This presentation
will also provide context and background for
other presentations on HCAHPS.
Implications for Policy, Delivery or Practice:
HCAHPS is the result of a fresh approach to
quality improvement, will produce valuable, new
publicly reported information, and promises to
be a catalyst for change.
Funding Source(s): CMS
ƒ Components of Care Vary in Importance for
Overall Patient-Reported Experience by Type
of Hospitalization in the HCAHPS Survey
Marc Elliot, Ph.D., David Kanouse, Ph.D., Carol
Edwards, Lee Hilborne, M.P.H., M.D.
Presented by: Marc Elliot, Ph.D., Senior
Statistician, RAND Corp., PO Box 2138, 1776
Main Street, Santa Monica, CA 90407-2138,
Phone: (310) 393-0411, Email: elliot@rand.org
Research Objective: To determine whether the
contributions of composite measures of patient
experience to overall hospital ratings on the
HCAHPS Survey (also known as the CAHPS®
Hospital Survey) vary by type of hospitalization.
Study Design: We defined 24 hospitalization
types using a combination of major diagnostic
category and service line (medical, surgical, or
obstetrical), pooling across categories when
necessary to achieve adequate sample size (n >
180). We examined seven composite measures
of patient experience: Doctor Communication
with Patients, Nurse Communication with
Patients, Staff Responsiveness to Patients’
Needs, Room Clean and Quiet, New Medicines
Explained, Pain Control, and Post-Discharge
Information. To assess the importance of each
composite for each type of hospitalization, we
calculated simultaneous partial correlations
between composite scores and an overall
hospital rating, controlling for patient
demographics.
Population Studied: 49,812 English- or Spanishspeaking adult patients with non-psychiatric
primary diagnoses discharged to home between
December 2002 and January 2003 after an
inpatient admission lasting at least one day in
132 general acute care hospitals in three states.
Principal Findings: Nurse communication was
the most important composite overall, with an
average partial correlation of 0.34 (range 0.170.49); it was statistically significant (p<0.005)
and among the three largest partial correlations
for all 24 hospitalization types. Discharge
information was the least important overall, with
an average partial correlation of 0.05 that was
statistically significant for only 10 of 24 types
(p<0.05) and among the three most important
for only one (r=0.16). Partial F-tests of
interactions found significant (p<0.05) variation
in partial correlations by hospitalization type for
5 of 7 composites, with the largest variations
observed for Nurse Communication, Doctor
Communication, and Pain Control (F >2,
p<0.005). For example, for Kidney/Urinary
Surgery, New Medicines, Pain Control, and
Nurse Communication were most important (in
that order), whereas for infectious disease
hospitalizations, such as tuberculosis, Nurse
Communication was much more important than
the next two (Pain Control and Discharge).
Doctor Communication was most important for
non-surgical stays for metabolic disorders; Pain
Control was most important for ENT
hospitalizations.
Conclusions: After controlling for patient
demographics and other aspects of care, Nurse
Communication is important to overall patient
experience for all types of hospitalizations, but
especially for some types; Responsiveness and
Cleanliness/Quiet are uniformly of moderate
importance; Doctor Communication and Pain
matter for most types and are especially
important for at least six of 24 hospitalization
types; Medications and Discharge matter for half
of 24 hospitalization types but are important
only for a few.
Implications for Policy, Delivery or Practice:
Results have implications for measurement,
reporting, and quality improvement efforts.
Because some aspects of patient experience are
relevant only for certain types of hospitalization,
measuring those aspects for all types adds noise;
quality measures might be improved by targeted
administration or by restricting administration or
scoring to relevant hospitalizations. Web-based
reporting could highlight the most important
composites for consumers who anticipate a
particular hospitalization type. Finally, quality
improvement efforts can be targeted to those
aspects of patient experience that matter most
for each type of hospitalization.
Funding Source(s): AHRQ
ƒ Findings from the HCAHPS Mode
Experiment
Marc Elliott, Ph.D., Elizabeth Goldstein, Ph.D.,
William G. Lehrman, Ph.D., Katrin
Hambarsoomians, M.S., Laura A. Giordano,
R.N., M.B.A., Mary Anne Hope, M.S.
Presented by: Marc Elliott, Ph.D., Senior
Statistician, RAND Corp., PO Box 2138, 1776
Main Street, Santa Monica, CA 90407-2138,
Phone: (310) 393-0411, Email: elliott@rand.org
Research Objective: To determine the effects of
mode of survey administration on patient
responses to the CAHPS® Hospital (HCAHPS)
Survey.
Study Design: Within each of 45 randomly
sampled hospitals, a total of 27,229 patients
were randomized in equal proportions to four
modes of survey administration: Mail Only,
Telephone Only, Mixed Mode (mail with
telephone follow-up), or Active IVR (interactive
voice response, in which patients respond via
telephone keypads). These patients completed
the HCAHPS (Hospital Consumer Assessment
of Healthcare Plans and Systems) Survey, in
which recently discharged patients evaluate
aspects of their hospital care. All surveys were
administered by a single vendor.
Population Studied: English- or Spanishspeaking adult patients with non-psychiatric
primary diagnoses discharged alive after at least
one overnight stay in a general acute care US
hospital with at least 1200 annual discharges.
Principal Findings: The 9 HCAHPS outcomes
consist of 2 overall ratings and 7 composites
constructed from 16 report items. Linear
regression was used to model each outcome
from fixed effects for mode, hospital identifiers,
and patient characteristics. Substantial and
statistically significant (p<0.05) mode effects
were found for both overall ratings and 4 of 7
composites (Cleanliness & Quiet,
Responsiveness of Hospital Staff , Pain
Management, and Discharge Information);
mode effects were smaller for the 3 composites
involving communication (with doctors, nurses,
and about medications). Patients provided more
positive evaluations in the Telephone Only and
Active IVR modes than in the Mail Only and
Mixed modes. Differences between Telephone
Only and Active IVR responses were small, and
there were very few differences in Mail Only and
Mixed Mode responses. When measured in
terms of hospital-level standard deviations, the
Telephone Only and Active IVR scores for the 6
most affected outcomes noted above were at
least 0.4-0.5 standard deviations higher and
sometimes as much as 1 standard deviation
higher than scores from Mail Only and Mixed
modes. These mode effects varied little by
hospital. Response rates varied strongly by
randomized mode (p<0.0001), ranging from
41.2% for Mixed Mode to 20.7% for Active IVR;
these patterns were consistent across hospitals.
Conclusions: Because a hospital’s choice of
vendor or survey mode may be confounded with
factors related to underlying quality, an external
mode experiment is necessary to estimate mode
effects for subsequent fieldings of the survey. We
conclude that adjustments for mode effects are
necessary to make the reported HCAHPS scores
comparable; in the absence of such adjustments
a hospital that would have ranked at the 50th
percentile in the Mail Only mode would be
ranked at the 66th to 84th percentile in the
Telephone Only mode for a majority of
outcomes. More surveys must be fielded in the
Active IVR mode than in the Mixed mode to
achieve the target of 300 completed surveys per
hospital.
Implications for Policy, Delivery or Practice:
CMS will employ adjustments derived from this
experiment, in conjunction with adjustment for
patient-level characteristics, that allow hospitals
to administer the HCAHPS Survey in any of four
modes without affecting the comparability of
their scores with other hospitals, national or
regional averages, or their own past
performance.
Funding Source(s): CMS
ƒ Lessons Learned from the ‘Dry Run’ of the
HCAHPS Survey
Laura Giordano, R.N., M.B.A., Mary Anne Hope,
M.S., Elizabeth Goldstein, Ph.D., William G.
Lehrman, Ph.D.
Presented by: Laura Giordano, R.N., M.B.A.,
Vice President, Surveys, Research & Analysis,
Health Services Advisory Group, 1600 East
Northern Avenue, Phoenix, AZ 85020-3983,
Phone: (602) 665-6162,
Email: lgiordano@azqio.sdps.org
Research Objective: In preparation for national
implementation of the CAHPS® Hospital
Survey, also known as HCAHPS, a ‘‘Dry Run’’ for
data collection and submission was conducted
from April through June 2006. Careful analysis
of the processes and procedures was then
conducted. The primary purposes of the Dry
Run were to provide an opportunity for all
participating hospitals ("general," acute care)
and approved survey vendors to establish and
fully experience the data collection and
submission processes. In addition, the Dry Run
allowed the Centers for Medicare & Medicaid
Services (CMS) to establish the final
administrative protocols, in advance of full
national implementation, which began on a
voluntary basis in October 2006. The data
collected for the Dry Run will, however, not be
used for public reporting purposes.
Study Design: The Dry Run provided an
opportunity for hospitals and survey vendors to:
select a mode of Survey administration; create a
sample frame of eligible patient discharges;
administer the Survey to the eligible patients;
collect the data; create XML data files or use the
online tool for data submission; and utilize the
QualityNet system. In addition, the Dry Run
provided an opportunity for CMS to: implement
and test systems for Survey data storage and
retrieval through the QualityNet Exchange
system; collect and process Survey data; and
develop and implement protocols for providing
technical assistance.
Population Studied: HCAHPS data were
collected for all HCAHPS-eligible patients:
English- or Spanish-speaking adults with nonpsychiatric primary diagnoses who were
discharged alive after at least one overnight stay
in a participating general acute care hospital.
Over 2,800 hospitals, and tens of thousands of
patients, participated in the Dry Run.
Principal Findings: The primary issues
encountered in the Dry Run included: the
creation of XML format files for data submission;
deployment of an HCAHPS on-line tool as an
alternative method of data submission for
smaller hospitals; unsuccessful data
submissions; establishing correct accounts in
the QualityNet Exchange data warehouse; and
data coding questions. The Dry Run allowed
hospitals, survey vendors and CMS to identify
and resolve problematic issues before the
national implementation, resulting in more
efficient data collection and submission
processes and improved data integrity.
Conclusions: It is essential that a standardized
process for data collection and submission be
consistently applied by all hospitals and survey
vendors participating in HCAHPS. Successful
completion of a Dry Run by hospitals and survey
vendors greatly improves the standardization of
data collection and submission across hospitals,
enhances the quality of the data, and provides a
critical opportunity for all affected parties to gain
valuable experience before HCAHPS results will
be publicly reported or play a role in pay for
reporting. The lessons learned by participating
hospitals, survey vendors, the data warehouse
and CMS have resulted in a smoother national
implementation of the survey.
Implications for Policy, Delivery or Practice:
Successful completion of a Dry Run by hospitals
and survey vendors will continue to be required
prior to participation in HCAHPS.
Funding Source(s): CMS
ƒ Public Reporting HCAHPS in Context
David Miranda, Ph.D., Jeanne McGee, Ph.D.,
Kristin Carman, Ph.D.
Presented by: David Miranda, Ph.D., Social
Science Research Analyst, Center for Beneficiary
Choices, Centers for Medicare & Medicaid
Services, S1-13-05, 7500 Security Boulevard,
Baltimore, MD 21244-1850, Phone: (410) 7867819, Fax: (410) 786-8004,
Email: david.miranda@cms.hhs.gov
Research Objective: This presentation examines
public reporting of the CAHPS® Hospital
Survey, also known as HCAHPS (Hospital
Consumer Assessment of Healthcare Providers
and Systems), results and integration into a
public website with performance information
that includes clinical process and outcomes (30day mortality) measures as well. Prior research
had indicated the importance of meeting the
needs and concerns of both patients and
physicians in hospital quality reports. Prior
research also had indicated consumer interest in
patient survey data. However, preliminary
research on integrating HCAHPS into Hospital
Compare found considerable difficulties that
consumers have understanding the distinction
between outcomes, process measures, and
patient survey data, let alone combining and
weighting the different types of data to fit their
needs. The current study involved iterative
rounds of qualitative research to design a
website with accurate and useful information
about hospitals that meets the needs of
consumers, physicians, and other audiences
including hospitals.
Study Design: The research team conducted
qualitative research in the form of three rounds
of one-on-one interviews and focus groups with
patients, their family members, physicians and
hospital quality improvement personnel.
Interactive web mockups of draft versions of
HCAHPS information allows for both structured
interviewing and unstructured browsing
observed by the research team.
Population Studied: People with Medicare,
people with a hospitalization experience,
consumers who help family members and
friends make hospitalization decisions,
physicians who refer to or practice in hospitals,
hospital quality improvement personnel.
Principal Findings: Creative design elements
and careful use of drill-down and other
navigation features allow for a successful
integration of HCAHPS into a website tool with
clinical process and outcomes data in a fashion
that meets the needs of both physicians and
consumers. Careful explanations allow
consumers to understand the data well enough
to play an active role in making informed
hospital care decisions in consultation with their
physician. To foster physician buy-in, detailed
explanations of composite construction, case
mix and mode adjustments, survey fielding and
data collection and analysis processes help
insure trust and confidence in the data.
Conclusions: Creative use of language and
design, in conjunction with iterative rounds of
qualitative testing, contribute to the successful
implementation of HCAHPS public reporting by
clarifying and responding to needs of multiple
audiences.
Implications for Policy, Delivery or Practice:
Application of textbook principles of web design
cannot substitute for on-going audience testing
research to insure that the public reporting
meets its goals and objectives. Policy maker and
researcher attention to multiple audiences and
how they interact around publicly reported
measures and the constructs they represent is
needed.
Funding Source(s): CMS
Call for Panels
Enhancing Research Through Collaborative
Community-Higher Education Partnerships:
Models of Community-Based Participatory
Research
Chair: Michelle Proser, M.P.P.
Monday, June 4 • 9:00 a.m.-10:30 a.m.
Panel Overview: National organizations,
funding agencies, and researchers have called for
a renewed focus on an approach to public health
research that recognizes the importance of
social, political, and economic systems to health
behaviors and outcomes. This renewed focus is
due to many converging factors, including
increased understanding of the complex issues
that affect health, the importance of both
qualitative and quantitative research methods,
and the need to translate the findings of basic,
interventional, and applied research into changes
in practice and policy. As a result, participatory
models of research, in which communities are
actively engaged in the research process through
partnerships with academic institutions, have
become central to the national prevention
research agenda as articulated by the Institute of
Medicine, the National Institutes of Health, the
Centers for Disease Control and Prevention
(CDC), Partnership for Prevention, the Public
Health Foundation, and the WK Kellogg
Foundation. Community-based participatory
research (CBPR) can lead to healthier
communities through collaborative solutions
that bring communities and institutions together
as equal partners and build upon the assets,
strengths, and capacities of each. Communitycampus partnerships involve communities and
higher educational institutions as partners, and
may address such areas as health care delivery,
research, health professions education,
community-wide health improvement,
community service, and community/economic
development. This panel will describe CBPR,
provide models and ‘‘how-tos’’ for establishing
effective and sustainable community-academic
partnerships, provide examples of CBPR
projects, identify helpful resources, and discuss
issues related to working with safety-net
community-based organizations.
ƒ A Review of Published Community-Based
Participatory Research (CBPR) and
Recommendations for Future Scholarly Work
Darius Tandon, Ph.D., Eric B. Bass, M.D.,
M.P.H.
Presented by: Darius Tandon, Ph.D., Assistant
Professor, The Johns Hopkins University School
of Medicine, 1620 McElderry Street, Rm. 203,
Baltimore, MD 21205, Phone: (410) 614-5281,
Fax: (410) 614-5431, Email: standon@jhmi.edu
Research Objective: First, we describe the
importance of, and examples of published
research in, eight areas of scholarly activity that
can promote health partnership research,
education, and action: 1. original research, 2.
works in progress and lessons learned, 3. policy
and practice, 4. theory and methods, 5.
education and training, 6. practical tools, 7.
systematic reviews, and 8. community
perspectives. Second, we provide
recommendations on issues, problems, and
topics within each area on which CBPR
manuscripts submitted to peer-reviewed journals
should focus.
Study Design: We used a group judgment
technique, Delphi Process, to elicit
recommendations of expert leaders. During the
first stage of the Delphi Process, the idea
generation, we sent an open-ended
questionnaire to the Associate Editors of a new
journal dedicated to the work of community
health partnerships, n equals 17, and asked them
to provide specific recommendations on the
most important issues, problems, or topics on
which articles published in Progress in
Community Health Partnerships, PCHP, should
focus for each of the eight areas above.
Recommendations were combined into specific
themes in each of the eight areas. During the
second stage of the Delphi Process, idea
prioritization, we gave the themes and specific
recommendations to PCHP’s diverse group of 17
external Editorial Board members, representing
academic, governmental, and community
perspectives, and asked them to check the topics
that are the highest priority for articles in each of
the eight areas of scholarly activity.
Population Studied: We obtained responses
from 17 Associate Editors and 12 Editorial Board
members.
Principal Findings: Stage 1 generated 318
recommendations across the 8 areas of scholarly
activity; these recommendations were collapsed
into 62 themes. Based on Stage 2, the most
commonly rated priority for Original Research
was translation of research into policy and
practice. Building community partnerships and
challenges in conducting CBPR were most often
selected for Work-in-Progress/Lessons Learned.
For Policy and Practice, engaging community
members in policy/practice was most commonly
selected. For Theory and Methods, research
methods was most commonly selected. For
Education and Training, CBPR curriculum and
graduate medical education reform and training
new investigators were most commonly selected.
The highest priorities for Practical Tools were
resources to develop community partners’ skills
and to evaluate projects. The highest priorities
for Systematic Reviews were reviews on CBPR
methods and CBPR effectiveness. For
Community Perspectives, the most commonly
given priority was community members’
perspectives on research usefulness.
Conclusions: The Delphi Process generated
many specific recommendations, as well as
priority areas, for future CBPR work. Academic
and community partners should consider
publishing work that can address these issues.
Implications for Policy, Delivery or Practice:
CBPR is an increasingly used approach for
conducting health research. Using Rogers’
diffusion of innovations theory as a framework,
future CBPR adoption will occur if CBPR is
perceived to have greater relative advantage,
compatibility, trialability, and observability, and
less complexity than other research approaches.
Future peer-reviewed publications, based on our
study recommendations, can influence
perceptions of CBPR that promote its adoption.
Funding Source(s): W. K. Kellogg Foundation
ƒ Conducting Research Using a CommunityBased Participatory Research Approach
Kristine Wong
Presented by: Kristine Wong, Program Director,
Community-Campus Partnerships for Health,
University of Washington, Box 354809, Seattle,
WA 98195-4809, Phone: (206) 543-7954,
Email: kristine@u.washington.edu
Research Objective: This presentation will (1)
review the principles of and logic behind
community-based participatory research (CBPR),
an emerging approach to research that increases
rigor in a way that a traditional research
approach may not; (2) provide models of CBPR
activities in community-based health care
settings as well as the continuum of such
models; (3) provide tips on how to establish
community-academic partnerships and how to
develop research designs that will not only serve
the community, but also increase the validity of
the research through authentic community
involvement; (4) discuss funding opportunities
and resources available to assist with creating
effective partnerships; and (5) demonstrate how
studying community-based primary care and
prevention at community health centers using a
CBPR approach will improve the delivery of care,
quality of care, and health outcomes for entire
communities.
Study Design: Community-Based Participatory
Research (CBPR) is a ‘‘collaborative approach to
research that equitably involves all partners in
the research process and recognizes the unique
strengths that each brings. CBPR begins with a
research topic of importance to the community
and has the aim of combining knowledge with
action and achieving social change to improve
health outcomes and eliminate health
disparities.’’ (W. K. Kellogg Foundation)
Population Studied: Community health centers
(community-based, safety net ambulatory care
providers that are run by patient-majority
governing boards) and academics/researchers
seeking to develop CBPR partnerships.
Community health center governing boards, as
well as community institutional review boards
established by some health centers, play
important roles in vetting researcher proposals.
Principal Findings: CBPR benefits both
partners. Not only does CBPR increase the rigor
of research through authentic community
involvement, but it also improves the health of
patients and the community. Patient and
community health improvement are achieved
through improved delivery of care and health
outcomes, reductions in health disparities, and
enhanced community capacity for future
research.
Conclusions: Researchers who wish to
successfully develop a CBPR study design may
need to develop new approaches and
perspectives towards working with communities
and the providers who serve them. Such
approaches are often detailed in community
institutional review board procedures. Many
foundations and government agencies are
increasingly dedicating funding streams to this
research approach, which will boost demand for
collaborative partnerships with communities. A
growing body of resources is available for those
interested in learning more about CBPR.
Implications for Policy, Delivery or Practice:
CBPR provides communities with opportunities
to decide what research questions are asked,
contribute to research designs that have the best
interests of their patients in mind, participate in
data collection and analysis, and establish longterm relationships with researchers for future
research. Researchers benefit from the inclusion
of community expertise and gain opportunities
to apply their expertise and skills for the greater
good, fulfill personal values and beliefs, and
cultivate external funding, as well as build a
foundation for new areas of scholarship.
Funding Source(s): WKK, NIH, CDC, Robert
Wood Johnson, EPA
ƒ Community-Based Participatory Research:
Practical Tips for the Research Community
Marshall Chin, M.D., M.P.H., Elbert S. Huang,
M.D., M.P.H., MidWest Clinicians’ Network
Team
Presented by: Marshall Chin, M.D., M.P.H.,
Associate Professor of Medicine, Section of
General Internal Medicine, University of Chicago,
5841 South Maryland Avenue, MC2007, Chicago,
IL 60637, Phone: (773) 702-4769, Fax: (773) 8342238, Email: mchin@medicine.bsd.uchicago.edu
Research Objective: Given from the ‘‘academic
institution perspective,’’ this presentation is
intended to provide examples of research
strengthened through community-based
participatory research (CBPR) and tips for
creating effective partnerships with communitybased organizations. The presentation will
highlight a major collaborative effort between the
University of Chicago and a multi-state
community health center network. The
collaboration sought to evaluate the impact of a
federally-supported chronic care initiative, known
as the Health Disparities Collaboratives, on
quality of care and outcomes for patients over
the seven-year span of the initiative, and to
determine ways to enhance the effectiveness,
sustainability, and spread of the initiative. With
the input of community members and providers,
the project will help inform future efforts to
improve chronic care management in health
centers.
Study Design: Between 2004 and 2005,
investigators studied approximately 200
community health centers in the Midwest and
West Central regions of the country. The study
and analysis were done collaboratively with the
community health centers involved. Investigators
conducted chart review and patient surveys for
performance indicators at a subset of 32 health
centers. On the universe of 200 health centers,
leaders and staff were surveyed to determine
incentives and assistance that would be helpful
in participating in the Health Disparities
Collaboratives, and semi-structured interviews
were performed on a sub-sample.
Population Studied: This CBPR example
studied low income community health center
patients with chronic illness who were
predominately uninsured or publicly insured.
Principal Findings: Community-based providers
have better understandings of the needs of the
populations they serve, and the barriers to care
they face. This presentation will provide
recommendations for creating successful CBPR
models, such as how to build relationships with
the community, when and how to be flexible,
working with senior management in community
settings, how to create equal partnerships, what
the research process should look like, and
meeting the interests of various funders.
Conclusions: The number of existing CBPR
models is small but growing. CBPR opens doors
to more research opportunities to study
traditionally hard-to-reach populations, but
requires the collaboration with the community
from start to finish in order to ensure the
research is rigorous, comprehensive, meets
complex societal problems, and improves
patient and community health.
Implications for Policy, Delivery or Practice:
Seamless partnerships between both academic
and community researchers ensures that
research is translated into practice or policy.
Funding Source(s): AHRQ, Robert Wood
Johnson Foundation, NIDDK
ƒ Community-Based Participatory Research:
Working with Community Health Centers
Michelle Proser, M.P.P.
Presented by: Michelle Proser, M.P.P., Director
of Policy Research, National Association of
Community Health Centers, 1400 Eye Street,
NW, Suite 330, Washington, DC 20005, Phone:
(202) 296-1960, Fax: (202) 296-3526, Email:
mproser@nachc.com
Research Objective: Using community health
centers as the example, this presentation will
highlight the benefits of working with
community-based safety net providers and
issues to consider when conducting communitybased participatory research (CBPR). It will also
shed light on the research capacity of health
centers, including the adoption of health
information technology and level of interest in
research. Lastly, this presentation will describe a
newly established research agenda for
community health centers and the medically
underserved.
Study Design: The community health center
research agenda was established through a
consensus-driven process bringing together over
100 experts in the field, including those from
health centers, health center networks, state
primary care associations,
researchers/academics, foundations, publishers,
government agencies, and national partners.
The agenda is being published in the February
2007 issue of Progress in Community Health
Partnerships: Research, Education, and Action
(Johns Hopkins University Press).
Population Studied: Over 5000 community
health center delivery sites serve nearly 16 million
patients in every state and territory. Health
center patients are disproportionately poor,
uninsured or publicly insured, and minority:
92% of patients are low income, 40% are
uninsured, 36% have Medicaid, and 63% are
members of racial and ethnic minorities. The
near-elderly make up the fastest growing patient
age group. Roughly 25% of encounters are for
chronic conditions. Most health centers have
been trained in and are employing the Chronic
Care Model through the federally-initiated Health
Disparities Collaboratives, including the use of
electronic patient registries. More and more are
adopting electronic health records.
Principal Findings: As community-directed
providers of care in medically underserved
communities, health centers have a history or
providing high-quality, patient-centered care to
traditionally hard-to-reach populations. Given
their expansive knowledge of the communities
they serve and the resources they bring to the
table, they are ideal partners in health services
research. A growing number of health centers
are utilizing health information technology,
including electronic patient records, making
research using patient-level data less
complicated and more robust. Health centers
are also organizing themselves into networks
across large regions, creating forums where
patient data are consolidated and case mix even
more diverse. In addition, they are actively
participating in health professional training
opportunities.
Conclusions: The number of health centers
participating in and interested in CBPR is
growing. They are ideal partners for CBPR given
their operations and service delivery approaches,
in-depth knowledge of their communities, and
expertise in delivering high quality care and
removing barriers to care. A research agenda for
health centers includes research questions on
determining their community value, impact on
health outcomes, and use of health information
technology to improve care.
Implications for Policy, Delivery or Practice:
Health centers are ideal settings to study
community-based primary care and prevention,
especially for underserved populations. Such
research can improve the delivery and quality of
care and health outcomes. Health centers have
worked collaboratively with researchers and
others to develop a research agenda, and stand
ready to engage in more research.
Funding Source(s): RWJF, AHRQ, HRSA, The
Commonwealth Fund, The Community Clinics
Initiative of the Tides Foundation and California
Endowment
Call for Panels
Implementation Issues in Medicare’s
Physician Quality Reporting Initiative (PQRI)
Chair: Myles Maxfield, Ph.D.
related to design and implementation of the
voluntary reporting program in the context of
recent legislation, and potential adaptation for
any future physician pay-for-performance
programs in Medicare. Teresa N. Clark, Office of
Clinical Standards and Quality, CMS, will
moderate. The first study: ‘‘Creating a PQRI
Data Validation Strategy that Includes
Opportunities for Ongoing Data Collection,
Program, and Measure Improvement,’’ provides
options for validating data underlying PQRI
measures, considering cost, effectiveness,
feasibility, scientific soundness, and flexibility,
while supporting continuous improvement in
program and measurement design. The second
study: ‘‘PQRI: Changes in Physician Office
Workflows for Quality Reporting: A Perspective
from the Trenches,’’ reports qualitative findings
on the impact of Medicare quality reporting on
physicians’ practice workflows, and the relatively
low associated time required for reporting based
on site visits to participating practices. The third
study: ‘‘PQRI: Performance Reporting and Payfor-Performance Considerations for Providers
Treating Vulnerable and Underserved Patients’’
reports interview, survey, and claims analysis
findings suggesting specific program design
options to limit potential unintended
consequences for disadvantaged patients,
including reduced access to physician services
and hindrance of quality improvement efforts.
The fourth study: ‘‘PQRI: Operational Insights
and Issues from Early Physician Practice
Adopters,’’ reports findings based on
discussions with participating physicians,
analysis of data from reporting of 2006 PQRI
starter set measures, and technical reviews of
PQRI performance measure specifications, to
examine operational issues related to physician
participation in the program. Researchers from
Mathematica Policy Research will be joined by
discussant Thomas Valuck, Office of Clinical
Standards and Quality, CMS.
ƒ PQRI: Creating a PQRI Data Validation
Strategy that Includes Opportunities for
Ongoing Data Collection, Program, and
Measure Improvement
Mary Laschober, Ph.D., Philip Renner, M.B.A.,
Zakiya Pierre, B.S., Sam Simon, Ph.D., Sarah
Davis, M.P.A., Teresa Clark, M.B.A., M.P.H.
Monday, June 4 • 11:00a.m.-12:30p.m.
Panel Overview: In 2007, CMS will begin to
offer physicians payment incentives for reporting
standardized quality measures on their Medicare
patients. This session explores several issues
Presented by: Mary Laschober, Ph.D., Senior
Researcher, Health, Mathematica Policy
Research, 600 Maryland Avenue, SW, Suite 550,
Washington, DC 20024, Phone: (202) 554-7572,
Fax: (202) 863-1763,
Email: mlaschober@mathematica-mpr.com
Research Objective: To provide options to CMS
for validating physician quality measurement
data for the 2006 Physician Voluntary Reporting
Program (PVRP) that take into account cost,
feasibility, scientific soundness, flexibility, and
applicability to Medicare policy. Options
provided should also ensure opportunity for
feedback to providers, CMS, and measure
developers to support ongoing improvements in
program design and measure development.
Study Design: Literature review and
environmental scan that included discussions
with leaders of measurement initiatives using
state-of-the-art strategies to validate data for
health care quality indicators, plus discussion
with a technical expert panel of thought leaders
in the field of measurement validation methods.
Population Studied: National and regional
physician-level and other health care provider
quality measurement and reporting programs in
the U.S.
Principal Findings: We identified 30 providerlevel initiatives with validation for the
environmental scan discussions, and the
literature review identified 12 reports highly
relevant to the study. There is a much deeper
literature on scientifically ensuring the validity of
quality measures, but few studies address
validation of quality measurement and reporting
systems. Most physician measurement
initiatives use measures based in administrative
or medical record data, and are limited to certain
payers or regions. Most of the validation
methods calculate either an agreement rate
between data submitted and re-abstracted chart
data, or an estimate of the effect of errors on
measure rates. All of the initiatives studied
sought to balance accuracy of validation with the
burden of validation.
Conclusions: An optimal strategy will likely
validate PVRP data at the primary nexus of
responsibility for ensuring accurate data ---- the
individual physician ---- but will employ a twostage sampling strategy to minimize provider
and CMS burden. A first-stage broad approach
would review a small sample of patients from an
approximately 1% random sample of providers
and a second-stage narrow approach would
review a larger sample of patients from a smaller
sample of outliers. Both stages would act to
detect systematic unintentional errors and to
deter system gaming. Because medical record
review (MRR) is the ‘‘gold standard’’ even for
electronic medical records, options suggested to
CMS will include retrospective MRR,
supplemented with Medicare claims analysis,
logical built-in edits, and benchmarking and
trending of measure results to detect errors at
different steps of the reporting system. It may
also be desirable to link PVRP validation with
Medicare’s existing Part B claims validation and
appeals activities. Additionally, CMS is
considering patient registries as a supplement to
administrative claims for collecting PVRP clinical
information, which may require a different
auditing operation.
Implications for Policy, Delivery or Practice:
There is no one perfect solution for validating
data underlying physician quality measures. As a
result, CMS will likely have to choose among
several options, each with their own advantages
and disadvantages for ensuring data accuracy,
consistency, completeness, reliability, and
feasibility. Because the PVRP is a new program,
all validation steps should promote learning and
improvement among all PVRP stakeholders,
including providers, CMS, other health care
payers/purchasers, and measure developers.
Funding Source(s): CMS
ƒ PQRI: Changes in Physician Office
Workflows for Quality Reporting, A
Perspective from the Trenches
Angela Merrill, Ph.D., Lorenzo Moreno, Ph.D.,
Whitney Scott, M.A., Martha Kovak, M.P.H., Jung
Kim, M.P.H., M.B.A., Stephanie Peterson, M.P.P.
Presented by: Angela Merrill, Ph.D., Senior
Researcher, Health, Mathematica Policy
Research, Inc., 955 Massachusetts Avenue, 8th
Floor, Cambridge, MA 02139, Phone: (617) 3018977, Fax: (617) 491-8044,
Email: amerrill@mathematica-mpr.com
Research Objective: The 2006 Physician
Voluntary Reporting Program (PVRP) was a first
step in collecting quality measures from
physicians for Medicare fee-for-service patients.
Practices participated by reporting specified ‘‘Gcode’’ procedure codes for quality measures on
their claims. The objective of this study was to
summarize the initial investments, changes to
workflows, and time associated with reporting
PVRP quality measures.
Study Design: Site visits with 15 physician
practices reporting PVRP measures in early
2006, conducted October 2006-January 2007.
Discussions about start-up investments,
workflow changes, time, and challenges to report
G-codes on Medicare claims.
Population Studied: Physician practices
reporting PVRP measures in early 2006. Practice
specialties included primary care, cardiology, and
surgery. Practice sizes were typically 3 or fewer
physicians, with two larger practices. Practices
were typically reporting 3 to 6 primary care
measures, while surgeons reported one
measure.
Principal Findings: The common initial steps to
participate in PVRP were: (1) learning about
PVRP through CMS website or conferences, (2)
developing worksheets with G-codes, and (3)
adding codes to billing software and resolving
issues with billing. These steps typically took less
than three hours, although some practices
reported many hours of negotiation with billing
companies to get them to accept G-codes in
claims. The workflow changes typically consisted
of three main steps: (1) identifying Medicare
patients and attaching a worksheet, (2) physician
checking off appropriate G-codes during the
visit, and (3) entering the codes to the billing
system. For surgeons reporting use of
antibiotics, measures were coded after the
surgery, through review of the medical record or
surgeon dictation. The median time required to
report a single G-code on a claim was less than 2
minutes, but ranged from 15 seconds to over 12
minutes for more complicated cases/measures.
The most time consuming step was typically
time for physician to choose the code. No
practices implemented electronic medical
records (EMRs) or other clinical management
systems to participate; and the two practices
with EMR did not use them to generate PVRP
reporting.
Conclusions: We found that the initial workflow
changes and time to report quality measures on
claims were relatively modest but did vary across
practices, depending on factors such as patient
complexity and the number of measures
reported. Workflow changes were basic, and did
not involve major changes. Addressing billing
issues was one of the biggest initial investments.
EMRs were not needed to report; however, one
practice that stopped reporting thought it would
be less time consuming once they had EMR.
Implications for Policy, Delivery or Practice:
This study provides information on the types of
workflow changes physician practices may need
to implement to participate in quality reporting
programs. Based on the experience of PVRP,
practices should be able to report on quality
measures with modest workflow changes, but
the cost of reporting will be variable, depending
on the number and complexity of reported
measures, and the mix of staff performing the
reporting.
Funding Source(s): CMS
ƒ PQRI: Performance Reporting
Considerations for Providers Treating
Vulnerable and Underserved Patients
Hoangmai Pham, M.D., M.P.H., Erin Fries
Taylor, Ph.D., Timothy Lake, Ph.D., Melanie Au,
M.P.P., Stephanie Peterson, M.P.P., Myles
Maxfield, Ph.D.
Presented by: Hoangmai Pham, M.D., M.P.H.,
Senior Researcher, Center for Studying Health
System Change, 600 Maryland Avenue SW, Suite
550, Washington, DC 20024, Phone: (202) 5547571, Fax: (202) 484-9258,
Email: mpham@hschange.org
Research Objective: We examined barriers to
performance reporting among providers treating
a disproportionate share of vulnerable and
underserved patients and considered strategies
to avoid unintended consequences of
performance reporting and pay-for-performance
for these patients.
Study Design: We reviewed literature on
prevalence of and barriers to performance
reporting among providers treating vulnerable
populations, and conducted over 20 semistructured discussions with experts in quality
reporting and performance-based incentive
programs, providers who treat vulnerable
populations, and other stakeholder groups. We
convened a meeting of 15 advisors, including
quality measurement and case-mix adjustment
experts, and representatives from provider
organizations, health plans, purchaser
organizations, and CMS. We also examined
claims data linked to the 2000-2001 Community
Tracking Study Physician Survey to compare
performance on quality measures (delivery of
preventive services) between providers who treat
more versus fewer patients of racial/ethnic
minorities, and of lower socioeconomic status.
Population Studied: This study focused on
vulnerable patients and the providers who treat
them. The question of what characteristics best
define such providers was answered through our
discussions. Although some stakeholders
preferred a more encompassing definition, most
concurred that providers’ status as those who
treat vulnerable patients should take into
account patients of low socioeconomic status,
those who belong to a racial/ethnic minority, and
those residing in underserved and/or rural areas.
Principal Findings: Stakeholders cited many
barriers to participation by providers treating
vulnerable populations in performance reporting,
although many of these barriers relate to small
practice size rather than the types of patients
they treat, per se. While financially-strapped
providers who treat vulnerable populations may
have fewer staff and some types of resources,
little evidence exists that providers necessarily
face higher reporting costs relative other
providers. In pay-for-performance programs,
however, providers treating vulnerable
populations likely face some distinct
disadvantages. The lower cumulative
reimbursement rates that some providers receive
for treating many Medicaid or uninsured
patients mean that such providers likely have
fewer practice resources available, which may
impair their ability to engage in systematic
quality improvement. In addition, some groups
of disadvantaged patients treated may be less
likely to adhere to recommended treatment,
although physicians share responsibility for
improving adherence. Finally, fewer support
services and resources among these patients,
along with lower education and health literacy
levels, may require providers who treat
vulnerable populations to devote more resources
to achieve quality outcomes that are comparable
to those of other providers. However, there
limited and mixed evidence on the relative
quality performance between these provider
groups, and more generally on the relationships
between patient characteristics, adherence, and
quality performance.
Conclusions: Providers who treat vulnerable
populations likely face some important
disadvantages under pay-for-performance
programs. If such disadvantages are not taken
into account, some providers may be less likely
to treat vulnerable and underserved patients-----given their potential effect on performance
scores and rewards.
Implications for Policy, Delivery or Practice:
Differential performance rewards under pay-forperformance may be warranted for providers
treating high volumes of patients with low
socioeconomic status. More extensive data
collection and analysis are needed before similar
program design decisions can be made
regarding providers treating other vulnerable
populations.
Funding Source(s): CMS
ƒ PQRI: Operational Issues and Insights from
Early Physician Practice Adopters
Timothy Lake, Ph.D., Melissa Neuman, M.S.,
Mary Laschober, Ph.D., Zakiya Pierre, B.S., Philip
Renner, M.B.A., Latousha Leslie, R.N., B.S.N.,
M.S.
Presented by: Timothy Lake, Ph.D., Senior
Researcher, Health, Mathematica Policy
Research, Inc., 600 Maryland Avenue SW, Suite
550, Washington, DC 20024,
Email: tlake@mathematica-mpr.com
Research Objective: To examine operational
and implementation issues related to physician
data collection and reporting of the starter set of
16 PVRP measures in 2006.
Study Design: Discussions with clinical and
administrative staff at 13 physician practices,
technical review of measure specifications, and a
descriptive analysis of PVRP reporting data
submitted from January 2006 through August
2006. Discussions were conducted in-person or
by telephone. Discussions centered on
physician decisions to participate in the program
and on which measures to report, assistance
with understanding the program’s data reporting
instructions and process, and ease of program
participation.
Population Studied: Primary care physicians
and specialists operating in solo, single specialty,
and multi-specialty group practices that had
submitted at least one PVRP measure during the
study period.
Principal Findings: Early findings suggest that
clinical and administrative staff at participating
practices had a limited understanding of selected
aspects of the 2006 PVRP measurement
specifications and were often unsure of how to
access additional program information and
support. Participants also generally did not
review specification updates over time. When
uncertainty existed regarding measure
instructions, practices generally chose coding
and reporting approaches that either appeared
most consistent with the specifications or that
were most feasible given particular practice
circumstances. As a result, reporting processes
varied across practice settings. For example,
practice staff members were often unsure when
and how often to report intermediate outcome
measures, such as HbA1C screening test results
for diabetes. They also expressed uncertainty
about whether a diabetes diagnosis was always
required to be on a claim containing diabetes
measures. One surgical practice submitted
measure information (antibiotic prophylaxis)
related to a procedure that was not covered in
the existing specifications. Other areas of
confusion included when to use exclusion codes,
submission of claims with zero-dollar amounts,
and interpretation of numerator definitions
(such as what constitutes a ‘‘falls assessment’’).
Variation and issues arose at several points
during the reporting process, including 1)
reporting system set-up, 2) patient identification
and selection, 3) coding of measures based on
clinical service delivery and diagnostic
information, and 4) reporting of measures on
Medicare claims. Analysis of Medicare claims
data indicate that reported measures did not
always contain the relevant diagnostic or service
delivery information.
Conclusions: Limited understanding of measure
specifications, and variation in reporting
approaches, may affect program participation,
the consistency of reported data, and conformity
with clinical measure specifications in future
quality measurement programs.
Implications for Policy, Delivery or Practice:
Increased communication and technical
assistance resources made available to practices
can help ensure the accuracy and consistency of
the information collected in quality
measurement programs. This could include 1)
broader outreach to practices considering
participation and 2) increased availability of
technical consultation services to participating
practices. Existing measurement specifications
may also be refined or modified in certain areas
to address the varying information needs of
practice staff involved in different aspects (for
example, clinical coding or billing) of the
reporting process.
Funding Source(s): CMS
Call for Panels
Translating Research into Practice: Case
Study of Improving Heart Attack Care
Chair: Carolyn Clancy, M.D.
Monday, June 4 • 2:30 p.m.-4:00 p.m.
Panel Overview: The panel, moderated by
Carolyn Clancy, MD, Director of the Agency for
Healthcare Quality and Research, examines a
model of translating research evidence into
practices to improve quality and save patients’
lives. As a case study of the proposed model of
translation, speakers will describe recent
experiences to gather, analyze, publish, and
disseminate research on how to reduce door-toballoon time in patients with ST-segment
elevation myocardial infarction. Prompt
treatment is critical for patient survival after this
kind of myocardial infarction, and less than 40%
of hospitals meet current guidelines, which
recommend a median of 90 minutes or less for
door-to-balloon times, i.e., the time between
patient presentation at the hospital and balloon
inflation to unblock coronary arteries. Panelists
represent a range of perspectives central to
current efforts to improve hospital performance
in door-to-balloon times nationally. Panelists will
include leaders from the National Institute of
Heart, Lung, and Blood and the Agency for
Healthcare Research and Quality; key
investigators who produced the research
evidence; the central architect from the American
College of Cardiology, who oversaw the national
campaign to create changes at the level of
practicing cardiologists and their hospitals; and
the Chief Medical Officer of Blue Cross/Blue
Shield who will provide the payer perspective on
the importance of research and its diffusion into
practice. Taken together, the panelists will
illustrate an effective approach for performing
and disseminating research to promote
evidence-based changes in quality of care.
Implications for quality improvement efforts
nationally will be discussed.
ƒ Strategies to Reduce Door-to-Balloon Time
in Myocardial Infarction
Elizabeth Bradley, Ph.D., Jeph Herrin, Ph.D.,
Jeptha Curtis, M.D., Brahmajee Nallamothu,
M.D., David Magid, M.D., M.P.H., Harlan
Krumholz, M.D., S.M.
Presented by: Elizabeth Bradley, Ph.D.,
Professor of Public Health, Epidemiology and
Public Health, Yale School of Medicine, 60
College Street, New Haven, CT 06520, Phone:
(203) 785-2937, Fax: (203) 785-6287, Email:
Elizabeth.Bradley@yale.edu
Research Objective: Prompt reperfusion
treatment is essential for patients with STsegment elevation myocardial infarction
(STEMI); however, the majority of U.S. hospitals
fail to meet national guidelines, which
recommend median door-to-balloon time of 90
minutes or less for primary percutaneous
coronary intervention (PCI). Door-to-balloon
time, which is important for patient survival, is
the time between the patient’s presentation to
the hospital and subsequent balloon inflation for
the procedure to unblock the coronary arteries.
We sought to identify the hospital strategies that
were significantly associated with faster door-toballoon times.
Study Design: We surveyed 365 hospitals to
identify hospital strategies that were correlated
with faster door-to-balloon times. We used
hierarchical generalized linear models and
patient-level data from the Centers for Medicare
& Medicaid Services to determine the
association between hospital strategies and
door-to-balloon time.
Population Studied: Patients with STEMI
treated during 2006 in 365 randomly selected
U.S. hospitals performing primary PCI.
Principal Findings: In multivariable analysis, six
strategies were significantly associated with
faster door-to-balloon times, including having
emergency medicine physicians activate the
catheterization laboratory (mean reduction in
door-to-balloon time 8.2 minutes), activating
with a single call to a central page operator (13.8
minutes), activating based on pre-hospital
electrocardiograms while the patient is still en
route to the hospital compared with waiting until
the patient had arrived (15.4 minutes), expecting
interventional cardiologists to arrive within 20
versus within more than 30 minutes of being
paged (19.3 minutes), having an attending
cardiologist always on site (14.6 minutes), and
using real-time data feedback with emergency
department and catheterization laboratory staff
(8.6 minutes). More strategies resulted in faster
times. Despite their effectiveness, these
strategies were used by a minority of hospitals.
Conclusions: Our findings highlight a discrete
set of effective, feasible, and currently underused
strategies that hold promise for substantially
improving patient care.
Implications for Policy, Delivery or Practice:
The findings provided a foundation for a national
campaign to improve hospital care for patients
with STEMI, using evidence-based practices. The
work identifies priorities for quality improvement
efforts to improve health care delivery and
enhance national progress toward meeting the
policy recommendation that all hospitals have
median door-to-balloon times of 90 minutes or
less. Organizational changes needed to
accomplish this level of performance may
generate benefits for other areas of hospital
practice as well.
Funding Source(s): National Heart, Lung, and
Blood Institute
ƒ Moving from Research to Practice - D2B: An
Alliance for Quality
John Brush, M.D., Wayne Bachelor, M.D., Eva
Kline-Rodgers, R.N., M.S., Brahmajee
Nallamothu, M.D., Elizabeth Bradley, Ph.D.,
Harlan Krumholz, M.D., S.M.
Presented by: John Brush, M.D., Chair, Quality
Strategic Directions Committee, President,
Virginia Chapter, American College of
Cardiology, 844 Kempsville Road, Suite #204,
Norfolk, VA 23502, Phone: (757) 466-6100, Fax:
(757) 466-1597, Email: jebrush@earthlink.net
Research Objective: The American College of
Cardiology (ACC), through Guidelines Applied in
Practice (GAP) programs, has sought to identify
gaps in clinical practice and to develop
mechanisms to motivate positive change in
cardiology to improve the quality of care. D2B:
An Alliance for Quality is a GAP program
involving a national campaign launched by the
ACC to save time and save lives by reducing the
door-to-balloon times in U.S. hospitals
performing primary PCI. Currently, only 40% of
hospitals achieve median door-to-balloon times
of 90 minutes or less. The broad goal of the D2B
Alliance is to translate knowledge about how to
improve time to treatment into practice. The
specific objective of the national campaign is to
have all participating hospitals achieve door-toballoon times of 90 minutes or less for at least
75% of their patients treated with primary
percutaneous coronary intervention (PCI) for STsegment elevation myocardial infarction
(STEMI).
Study Design: The ACC assembled the ACC
D2B Work Group and 5 subgroups comprised of
physician, nurse, and researcher volunteers to
summarize the existing evidence on strategies
for reducing door-to-balloon time, to develop
tools and educational materials to support
hospitals’ efforts to adopt effective strategies,
and to enroll hospitals into the national
campaign to improve times. State chapters of
the ACC provided a network of chapter
presidents whow were commissioned to recruit
physician champions and hospitals across the
country. In addition, the ACC effort also
partnered with more than 20 additional
collaborators and supporters including but not
limited to the American Heart Association, Blue
Cross/Blue Shield Association, Institute of
Healthcare Improvement, the Society for
Academic Emergency Medicine, and the
National Association for EMS Physicians.
Population Studied: Patients with STEMI
treated with PCI in hospitals that enroll in the
GAP-D2B national campaign.
Principal Findings: More than 150 hospitals
enrolled in the GAP-D2B campaign within the
first 2 months of its launch, and enrollment is
continuing. The participants reflect differing
geographical regions across the country, and
include hospitals of varying size, teaching and
non-teaching status, and ownership types. All
enrolled hospitals complete a baseline survey to
document their current processes to facilitate
timely care for patients with STEMI. Hospitals
also receive a Tool Kit and Change Package
designed by the ACC D2B Work Group based on
existing research identifying about effective
practices.
Hospitals’ door-to-balloon times will be tracked
using existing data sources, including publicly
report data from Center for Medicare &
Medicaid Studies.
Conclusions: Practical efforts, based on
research-based findings about what works to
reduce door-to-balloon time, are being used to
translate research into practice and improve
quality of care. The role of professional
organizations, in collaboration with health
services and outcomes researchers, is essential
to providing the needed impetus to disseminate
best practices and influence the quality of clinical
care.
Implications for Policy, Delivery or Practice:
The D2B national campaign demonstrates a
model of moving research findings quickly into
improved clinical practice. In this model, the
research focused not on ‘‘what’’ to do but "how
to do it. Close interaction between research and
dissemination teams ensured that the research
addressed questions of practical importance
while the dissemination efforts remained
evidence-based. This approach of translating
research into practice with the use of effective
networks of leading clinicians, researchers,
professional associations, and policy-makers
may be helpful in promoting national quality
improvement efforts in other clinical areas as
well.
ƒ From Retail to Wholesale in Safety and
Quality Improvement
Irene Fraser, Ph.D., Irene Fraser, Ph.D., Allen
Korn, M.D., F.A.C.P.
Presented by: Irene Fraser, Ph.D., Director of
the Center for Delivery, Organization, and
Markets, Agency for Healthcare Research and
Quality, 540 Gaither Road, Rockville, MD 20850,
Phone: (301) 427-1364, Fax: (301) 427-1542
Research Objective: The door-to-balloon time
initiative shows that examination and
codification of processes used by top-of-class
providers can provide replicable models for
substantial leaps in quality and safety. This is
tremendous news. But the methods for
achieving this breakthrough are slow and
somewhat expensive, and require identifying
steps specific to each clinical procedure.
Moreover, it is not clear whether the
combination of procedure-specific safety
strategies will add up to success at the hospital
level. Is there a way to move from retail to
wholesale in identification of winning safety
strategies, or must we must continue to make
improvements on a ‘‘one-off’’ basis? Our
objective is to identify efficient strategies for
spreading safety improvements to more sites
and more processes.
Study Design: This presentation considers the
experiences and potential of five potential
approaches for moving from retail to wholesale:
1. Create a national campaign to promote takeup of one process redesign.
2. Replicate the approach and methodology in a
second study of a second procedure.
3. Spread the methodology itself.
4. Seek system or organizational approaches that
will support improvements in multiple clinical
outcomes at once
5. Align purchaser (government, employer, plan)
incentives to support evidence-based strategies
such as this.
Population Studied: Patients, providers, and
organizations involved with diverse approaches
for spreading evidence-based improvements.
Principal Findings: Each approach has shown
some potential to improve spread of evidencebased improvements,
Conclusions: Approaches have varying
advantages and disadvantages in design,
implementation, and evaluation.
Implications for Policy, Delivery or Practice:
Using one or more of these strategies will enable
funders, payers, and providers to move from
retail to wholesale in safety improvement.
ƒ The Door-to-Balloon (D2B) Project as Part
of Translation Research at the National Heart,
Lung, and Blood Institute (NHLBI)
Lawton Cooper, M.D., M.P.H., Barbara Wells,
Ph.D., Lawrence Fine, M..D, Dr.P.H., Denise
Simons-Morton, M.D., Ph.D.
Presented by: Lawton Cooper, M.D., M.P.H.,
Medical Officer, Division of Prevention and
Population Sciences, National Heart, Lung, and
Blood Institute, 6701 Rockledge Drive, Room
10108, Bethesda, MD 20892, Phone: (301) 4350419, Fax: (301) 480-1773,
Email: cooperls@nhlbi.nih.gov
Research Objective: The NHLBI has invested
enormous research resources to identify safe
and effective CVD preventive, diagnostic, and
therapeutic strategies. However, mechanisms
are lacking to ‘‘close the loop’’ from research
results to clinical care, i.e., to track actual clinical
use of these therapies, to identify factors related
to their adoption, and to assess associations of
adoption vs. non-adoption with patient
outcomes.
Study Design: This abstract describes how the
D2B project relates to NHLBI research activities
in the area of translation of clinical practice
knowledge into practice.
Principal Findings: The NHLBI supports a
number of grants through various funding
mechanisms that address the gap between
clinical knowledge and practice. The study on
door-to-balloon time led by Drs. Krumholz and
Bradley that is featured in this panel is an
outstanding example. Some features that make it
outstanding include the innovative design that
merges qualitative and quantitative methods, the
focus on concrete actions that could affect the
outcome measure, and the dissemination plan
that directly engages U.S. hospitals performing
PTCI. It thus serves as a model for others
studying the translation into practice of other
clinical guidelines. NHLBI currently funds
translation grants that were initiated by
investigators, and others that responded to
specific funding announcements, such as RFAs
or PAs. One RFA supports grants evaluating
interventions in clinical practice to improve
implementation of national, evidence-based
clinical practice guidelines for the treatment of
heart, lung and blood diseases and conditions.
Another RFA funds several grants studying
clinically feasible interventions related to medical
care delivery that increase the proportion of
treated hypertensive African American patients
whose blood pressure is controlled according to
JNC guidelines. Recently the NHLBI released an
RFA entitled ‘‘Cardiovascular Research Network
(CVRN) in Community-Based Care’’ (RFA-HL-07011), one of whose objectives is to promote
research on clinical practice and quality of care.
At the NIH level, the new consortium funded by
Clinical and Translational Science Awards
(CTSAs) is viewed by the NHLBI as an
opportunity to reduce the delay between clinical
knowledge generation and adoption, by
encouraging broader community involvement in
research to increase general awareness of
ongoing studies and increase interest in research
outcomes.
Conclusions: The D2B project is a prime
example of NHLBI’s history of supporting
research activities in the area of translating
clinical research into practice. The Institute’s
Strategic Plan, currently available for public
comment, emphasizes the important role of
these activities within the NHLBI’s overall
mission, along with traditional basic and clinical
research.
Implications for Policy, Delivery or Practice:
Findings emerging from NHLBI-funded research
on the translation of clinical guidelines into
practice have significant implications for public
health policy and health care delivery. Welldesigned dissemination efforts such as the
national D2B campaign promise to take the
findings of translational research and translate
them into practice.
Call for Panels
Improving Health and the Use of Medical
Services: Analyzing the Contributions of
Medicaid and SCHIP
Chair: Richard Kronick, Ph.D.
Monday, June 4 • 2:30 p.m.-4:00 p.m.
Panel Overview: In their role as a ‘‘laboratory for
democracy,’’ states have taken a variety of
approaches to providing public coverage to low
income children and families. States differ
significantly in choices made about eligibility,
service coverage, and provider payment policies
for the Medicaid and State Children’s Health
Insurance Program (SCHIP) programs. This
panel presents research that explores the impact
of some of these different decisions on access to
health care and health outcomes, using several
datasets from the National Center for Health
Statistics. The first paper examines the effect of
reductions in state Medicaid physician payment
levels on access to care. Understanding these
effects is important when considering trade-offs
among provider payments, eligibility levels,
benefit packages and other demands on state
budgets. The second paper questions whether
increasing generosity in Medicaid eligibility helps
newborns and mothers receive care in the same
hospitals as the privately insured in the face of
residential segregation, public provision of care,
and possibly inadequate public payment of
providers. The third paper analyzes the effects of
SCHIP implementation on the use of health
services among low-income children. The final
paper investigates the effect of Medicaid and
SCHIP eligibility and other state policies on
children’s use of health services and health
status, considering the effects of expansions in
children’s eligibility on outcomes for both
children and families. The implications of these
papers for the prospects of further reducing the
gap in health between low- and high-income
children and families will be discussed during
the panel.
ƒ Medicaid Physician Payment and
Ambulatory Care of Medicaid Patients
Sandra Decker, Ph.D.
Presented by: Sandra Decker, Ph.D., Senior
Service Fellow, Division of Health Care Statistics,
National Center for Health Statistics, 3311 Toledo
Road, HYATTSVILLE, MD 20782, Phone: (301)
458- 4748, Email: sdecker@cdc.gov
Research Objective: Although states commonly
cut or freeze physician fees in an attempt to
control Medicaid spending, little work has
assessed the impact of a reduction in physician
fees on the care of Medicaid patients. Previous
research suggests that reductions in fees may
reduce the number of physicians willing to
accept Medicaid patients, possibly reducing the
quantity of care patients receive and/or leading
them to seek care in hospital emergency
departments or other sites, but the evidence is
not strong. This paper furthers our
understanding of the relationship between the
level of state Medicaid fees and use of
ambulatory care among Medicaid patients.
Study Design: I estimate the effect of the
generosity of Medicaid physician payment levels
on the volume and site of ambulatory care
received by Medicaid patients compared to other
patients, controlling for age group, gender, race,
and state and year effects. The use of state fixed
effects and the comparison to treatment of
privately-insured patients serves to separate
Medicaid's effect on access to care from any
correlation between the Medicaid fee and other
attributes of the state in which a patient lives.
Data on state Medicaid physician payment levels
in 1993, 1998 and 2003 were collected by the
Urban Institute, and data on the use of health
services come from several National Center for
Health Statistics (NCHS) surveys in 1993/1994,
1998/99, and 2003/2004. Data on the selfreported total number of physician visits in the
past year come from the National Health
Interview Survey (NHIS), a nationally
representative sample of households in the U.S.
Information on site of visits comes from the
National Ambulatory Medicare Care Survey
(NAMCS) and the National Hospital Ambulatory
Care Survey (NHAMCS), nationally
representative surveys of visits to office-based
physicians and hospital outpatient and
emergency departments (OPDs and EDs).
Population Studied: Data on the 317,711
individuals in the NHIS and the 462,359 visits in
the NAMCS/NHAMCS among those under age
65 and not on Medicare.
Principal Findings: Reductions in Medicaid
physician fees lead to statistically significant
reductions in the number of visits for Medicaid
patients compared to privately insured patients.
Compared to privately insured patients,
Medicaid patients rely much more heavily on
OPDs and EDs and less on physician offices,
though differences in site of care between
Medicaid and privately insured patients are
smaller in states with more generous Medicaid
physician payment. Reductions in fees lead to a
statistically significant shift away from physician
offices and toward both EDs and OPDs, though
the shift is larger toward OPDs than EDs. Firstlisted diagnoses for which site of care shifts are
most pronounced include diabetes,
hypertension, asthma, migraine, noninfectious
enteritis and colitis, and chest pain.
Conclusions: This paper finds a significant link
between state policy on Medicaid physician
payment levels and both the volume and site of
care of Medicaid patients, a link that is
significantly stronger and larger than estimated
in previous studies using older data.
Implications for Policy, Delivery or Practice:
Since Medicaid enrollees in states with low
Medicaid physician fees receive significantly less
care at different sites than patients in other
states, future research on the implications of
these differences in care for the health status of
Medicaid patients is particularly important.
ƒ Medicaid Eligibility and Hospital Care for
Mothers and Newborns
Kosali Simon, Ph.D., Robert Kaestner, Ph.D.,
Sandra Decker, Ph.D., Kosali Simon, Ph.D.
Presented by: Kosali Simon, Ph.D., Assistant
Professor, Department of Policy Analysis and
Management, Cornell University, N227 MVR
Hall, Ithaca, NY 14853, Phone: (607) 255-7103,
Fax: (607) 255-4071, Email: kis6@cornell.edu
Research Objective: When enacted in 1965, the
Medicaid program sought to finance medical
care for the poor. It was not designed to create a
‘‘two-tiered’’ system of care where publiclyinsured and privately-insured patients receive
care from largely different providers. This paper
assesses the extent to which Medicaid provides
pregnant women with access to the same
hospitals and same medical care that is obtained
by privately insured mothers. If women with
different types of insurance largely receive care at
separate facilities, or receive different care within
facilities, this could partly explain differences in
maternal and infant health by insurance status.
Study Design: We compare the characteristics
of hospitals and the care received in those
hospitals by women with different types of health
insurance coverage. To address the nonrandom nature of insurance status, we use the
state and time variation in insurance status
caused by differences in the timing and
magnitude of Medicaid eligibility expansions
across states between 1988 and 2003.
Specifically, we use a representative sample of
women from the Current Population Survey and
Medicaid eligibility rules to calculate the fraction
eligible for Medicaid in each state and year that
is independent of economic and health
conditions in the state. We then use this
simulated eligibility measure to predict insurance
status and identify the effect of insurance status
on the type of facility where a pregnant woman
receives care.
Population Studied: Discharges from short-stay
hospitals from the National Hospital Discharge
Survey (NHDS), 1988-2003, linked with
additional hospital characteristics from the
American Hospital Association (AHA) Annual
Survey Database, and with simulated Medicaid
eligibility by state and year as explained above.
Principal Findings: We examine the effect of
insurance status on the share of privatelyinsured deliveries in the hospital where a woman
gives birth, the probability that the mother gives
birth in a hospital with a neo-natal intensive care
unit (NICU), the probability that a women gives
birth in a proprietary, non-profit or government
hospital, the distance between the woman’s
home and the hospital she gives birth in,
whether or not the child is born by Cesarean
section, and mother’s length of hospital stay.
Results imply that individuals in low income
areas and those on Medicaid largely receive
health care at different facilities than others.
State expansions in Medicaid eligibility may
affect where women give birth and the type of
hospital care received.
Conclusions: This research points to the need
for further investigation of whether or not
separate care is equal in quality and in turn
whether it is related to differences in health
outcome.
Implications for Policy, Delivery or Practice: If
future research shows that separate care is
different in quality, options for eliminating the
‘‘two-tiered’’ system of care should be evaluated.
ƒ Effects of the State Children’s Health
Insurance Program (SCHIP) on Access to
Medical Care and Use of Medical Services
Hua Wang, Ph.D., Edward Norton, Ph.D., Gary
Rozier, Ph.D., Sally Stearns, Ph.D.
Presented by: Hua Wang, Ph.D., Research
Associate, Department of Policy Analysis and
Management, Cornell University, 432A Martha
Van Rensselaer Hall, Ithaca, NY 14853, Phone:
(607) 255-2505, Fax: (607) 255-4071,
Email: hw227@cornell.edu
Research Objective: To provide national
estimates of implementation effects of the State
Children's Health Insurance Program (SCHIP)
on medical care access and use for low-income
children.
Study Design: The study design is based on
variation in the timing of SCHIP implementation
across states and among children observed prior
to and after implementation. Two analyses were
conducted. The first estimated the total effect of
SCHIP implementation on measures of access
to care (dichotomous variables denoting whether
a child had a usual place for care and whether a
child had experienced in the previous 12 months
any delayed medical care, unmet need for
medical care, prescription drug and mental care)
and medical services utilization (dichotomous
variables indicating whether a child had visited a
general doctor and specialty doctor in the past
year) for low-income children (family income
below state SCHIP eligibility thresholds) using
county and time fixed effects models. The
second analysis estimated differences in medical
care access and use among low-income children
with SCHIP or Medicaid coverage and their
uninsured counterparts, using instrumental
variables methods to control for selection bias.
Selected instruments for insurance coverage
include state SCHIP program features and
variables representing whether any family
member receives Food Stamps and whether
health insurance was offered to a child’s parents
through the workplace. Both analyses
controlled for child and family characteristics.
Population Studied: National representative
sample of children in U.S. households from the
1997-2002 National Health Interview Survey.
Principal Findings: When SCHIP had been
implemented for more than one year, the
probability of having visited a general doctor for
low-income children increased by 7.4 percentage
points. Compared with their uninsured
counterparts, those who had SCHIP or Medicaid
coverage were more likely to have a usual place
for care (23.4%), less likely to report delayed
(?12.3%) or unmet medical care need (?7.7%),
and more likely to have visited a general (25.7%)
and specialty doctor (8.8%) within 12 months.
SCHIP program type had no differential effects.
Policy simulations suggest further expansion of
public insurance may not close the gap between
low- and high-income children if the
effectiveness of public health programs in
improving the use of health services remains the
same.
Conclusions: Consistent results from two
analytical approaches provide evidence that
SCHIP implementation significantly increased
use of medical services for low-income children
in the U.S. Children enrolled in SCHIP or
Medicaid had substantially increased access to
and use of medical care than the uninsured.
Implications for Policy, Delivery or Practice:
SCHIP did make a difference in improving
medical care access and use for low-income
children, though the effectiveness of public
health programs in improving the use of health
services may need to improve in order to
eliminate disparities in medical care access and
use between the low- and high-income children.
Funding Source(s): National Center for Health
Statistics (NCHS)/AcademyHealth Health Policy
Fellowship
ƒ Effect of Recent State Policies
(SCHIP/Medicaid Expansions, Welfare
Reform and EITC) on Child Health and Health
Care Use
Maki Ueyama, M.P.A., Rachel Dunifon, Ph.D.,
Kosali Simon, Ph.D., Maki Ueyama, M.P.A.
Presented by: Maki Ueyama, M.P.A., Ph.D.
candidate, Department of Policy Analysis and
Management, Cornell University, Ithaca, NY
14853, Phone: (607) 255-7978, Fax: (607) 2554071, Email: mu24@cornell.edu
Research Objective: To examine the effect of
state-level social assistance policies on child
health and health care use. Specifically, we use
the 1990-2002 National Health Interview Survey
accessed through the NCHS Research Data
Center to study the impact of SCHIP/Medicaid,
welfare reform, and the Earned Income Tax
Credit (EITC) on children.
Study Design: To examine the impact of
Medicaid/SCHIP expansions on children, we use
an instrumental variables approach,
instrumenting for Medicaid participation and
eligibility with an index of state Medicaid/SCHIP
generosity created from the Current Population
Survey (CPS) at the state/age/year level. We do
this both at the child level as well as the family
level, since making a larger fraction of the family
eligible may have considerable spillovers. We
first estimate the effect of any public health
insurance eligibility on health insurance status to
compare the effect with that reported in other
data sets. We then estimate the effect of public
health insurance participation (instrumented in
the same manner described above) on children’s
use of health care and health status (e.g. a
parental report of child’s health, whether the
child has any limitation of activity, number of
doctor visits in past 12 months, number of
school days lost to illness in past 2 weeks and 12
months, and a dummy indicating whether the
child needed prescription drugs but could not
get them in past 12 months). To examine the
effects of welfare reform, we use a treatment and
control group approach to study the health and
health care outcomes for children of single lowskilled mothers before and after welfare reform
relative to children of married mothers. In the
case of the EITC we take advantage of withinstate changes over time in the generosity of state
EITC supplements, as well as the fact that
children in different families (married vs. single
for example) are differentially affected by these
supplements.
Population Studied: Low-income children from
the 1990-2002 National Health Interview Survey.
Principal Findings: Our preliminary analysis
shows that public insurance eligibility increased
for all children from 1997-2002; eligibility
increased especially between 1998 and 1999 with
the greatest increase for children ages 6-19. We
also find a greater increase in public insurance
among children living at 100-200% of the
Federal Poverty Line, particularly among older
children.
Conclusions: This study adds to our
understanding of the effects of important social
assistance programs on children’s health. In
particular, we examine both parents’ and
children’s public health insurance eligibility and
participation using more recent data, as well as
focus on children’s health status and other
health-related outcomes.
Implications for Policy, Delivery or Practice:
The U.S. has seen a slow improvement in child
health over the past several decades (Federal
Interagency Forum on Child and Family
Statistics (FIFCFS) 2004). Although the health
gap between low- and high-income children
decreased slightly over the years, low-income
children remain consistently and considerably
less healthy compared to high-income children
(FIFCFS 2004). In light of this, it is critical to
understand how changes in social welfare
policies during the 1990s may have influenced
child health.
Call for Panels
Improving the Quality of Depression
Treatment for Ethnic/Racial Minorities
Chair: Margarita Alegria, Ph.D.
Monday, June 4 • 4:30 p.m.-6:00 p.m.
Panel Overview: This panel presents findings
from four papers that analyze the provision of
quality depression treatment to ethnic/racial
minorities and the economic burden of
depression in minority populations. All of the
papers are based on new data from the
combined National Latino and Asian American
Study (NLAAS) and the National Comorbidity
Survey Replication (NCS-R), one of the most
comprehensive data sources available that can
be used to study these topics. The objectives of
the four papers are: 1) to compare the quality of
depression treatment that ethnic/racial
minorities receive relative to what non-Latino
whites receive; 2) to analyze differences in rates
of co-morbidity between chronic physical health
conditions and depression across racial/ethnic
groups; 3) to compare rates of retention in
depression treatment for minorities compared to
non-Latino whites; and 4) to estimate and
compare the effects of depression on labor
market outcomes in minority and majority
populations. Findings from the first three papers
will supply mental health service providers,
researchers, and policy makers with the most
recent and best available data related to
optimizing assessment, treatment, and retention
of ethnic/racial minority patients with
depression. Results from the fourth paper will be
useful in demonstrating to decision makers that
expanding access to high-quality patientcentered depression treatment is not only
justified on equity grounds but also on economic
grounds. As a whole, findings from this session
can be used to target resources, develop public
policies, and inform guidelines aimed at making
depression prevention and treatment services
responsive to the needs of ethnic/racial
minorities.
ƒ Appropriateness and Intensity of
Depression Treatment for Ethnic/Racial
Minorities
Margarita Alegria, Ph.D., Pinka Chatterji, Ph.D.,
Norah Mulvaney-Day, Ph.D., Lisa Fortuna, M.D.,
David Takeuchi, Ph.D.
Presented by: Margarita Alegria, Ph.D., Director
and Professor, Center for Multicultural Mental
Health Research, 120 Beacon Street, 4th Floor,
Somerville, MA 02143, Phone: (617) 503-8447,
Fax: (617) 503-8430,
Email: malegria@charesearch.org
Research Objective: This paper compares the
quality of the depression treatment that
ethnic/racial minorities receive to that of nonLatino whites, as well as identifies other
individual correlates of receiving quality
depression treatment.
Study Design: Our analytic sample is limited to
NLAAS/NCS-R respondents with last year DSMIV diagnosis of major depression or dysthymia
and respondents who are sub-threshold cases of
depression. We construct quality measures that
capture the appropriateness and the intensity of
treatment. Appropriateness of treatment is
measured by a dummy variable indicating if
either of the following two criteria is met during
a 12-month period: 1) four or more visits of at
least five minutes of counseling, or 2)
antidepressant medication with medical
supervision. Intensity of treatment is captured by
a continuous index which proxies the average
value of treatment services used in the past year.
This index is constructed by multiplying the
quantity of mental health services reported by
the respondent in the 12-month period by the
national average prices paid for services,
obtained from the 2002 Medical Expenditure
Panel Survey (MEPS).
We conduct regression analyses to evaluate the
effect of ethnicity/race on quality of treatment.
Demographic and socioeconomic status (SES)
covariates included in the regression model
include race/ethnicity (non-Latino white as the
baseline, Latino, Asian, African-American,
Other), poverty level, age category (18-34 as the
baseline, 35-49, 50-64, 65+), gender, marital
status, education level, region, nativity, language
of interview, insurance (uninsured as the
baseline, private insurance, Medicare,
Medicaid/other public programs) and a HMO
indicator based on whether the respondent
belongs to an HMO.
Population Studied: The combined National
Latino and Asian American Study (NLAAS) and
the National Comorbidity Survey Replication
(NCS-R) (pooled, N = 14,146) is one of the most
comprehensive national data sources available
that can be used to study depression treatment
for ethnic/racial minorities. Both the NLAAS and
NCS-R studies are part of the NIMH
Collaborative Psychiatric Epidemiology Surveys
(CPES), which focus on collecting
epidemiological information on the risk factors
for mental disorders as well as extensive
information on service usage among the general
population with special emphasis on minority
groups.
Principal Findings: Preliminary findings indicate
that after controlling for other factors, AfricanAmericans in need of depression treatment are
much less likely to receive appropriate care
compared to non-Latino whites in need of
depression treatment (OR: 0.33, p-value = 0.03).
Respondents classified as ‘‘Other’’ had similarly
low odds of receiving appropriate care compared
to non-Latino whites (OR: 0.20, p-value=0.03).
Medicaid insurance coverage (versus being
uninsured) increased the likelihood of receiving
appropriate care, as did being in an older age
group (compared to the 18 to 34 age group).
Models of treatment intensity are in progress.
Conclusions: These findings suggest that there
are important racial/ethnic disparities in receipt
of appropriate depression treatment among
depressed individuals.
Implications for Policy, Delivery or Practice:
This information will help policymakers make
multi-tiered recommendations to address gaps
in service provision for depression for minorities,
and to generate concrete solutions for improving
quality of care.
Funding Source(s): RWJF
ƒ Comorbid Chronic Physical Conditions and
Depression Across Racial and Ethnic Groups
Norah Mulvaney-Day, Ph.D., Margarita Alegria,
Ph.D., Pinka Chatterji, Ph.D., Lisa Fortuna, M.D.,
David Takeuchi, Ph.D.
Presented by: Norah Mulvaney-Day, Ph.D.,
Associate Director, Center for Multicultural
Mental Health Research, 120 Beacon Street 4th
Floor, Somerville, MA 02143, Phone: 617-5038448, Fax: 617-503-8430, Email: nmulvaneyday@charesearch.org
Research Objective: The objective of this paper
is to analyze racial/ethnic differences in
comorbid depression and other chronic physical
health conditions, and test for differences in
service patterns across groups.
Study Design: This descriptive analysis
compares the rates of comorbidity between
depressive disorder and seven chronic
health/pain conditions across race/ethnicity
(non-Latino whites, Latino, Asian, AfricanAmerican, and Other), controlling for age and
gender. Health conditions considered were
asthma, diabetes, hypertension/cardiovascular
disease, any chronic health condition, back/neck
pain, headaches and any chronic pain condition.
Bivariate tests were conducted to identify
significant differences across racial and ethnic
groups for all comorbid depression and health
conditions. Differences in service patterns for
those with comorbid conditions were analyzed,
including use of antidepressant medication and
whether the individual spoke with a primary care
physician about mental health concerns.
Population Studied: The combined National
Latino and Asian American Study (NLAAS) and
the National Comorbidity Survey Replication
(NCS-R) (pooled, N = 14,146) is one of the most
comprehensive national data sources available
that can be used to study depression treatment
for ethnic minorities. Both the NLAAS and NCSR studies are part of the NIMH Collaborative
Psychiatric Epidemiology Surveys (CPES), which
focus on collecting epidemiological information
on the risk factors for mental disorders as well as
information on service usage among the general
population with special emphasis on minority
groups.
Principal Findings: There were few differences
in rates of comorbid depression and any of the
three chronic health conditions (diabetes,
asthma, or cardiovascular disease) across ethnic
and racial groups. Asians and Latinos had lower
rates of comorbid depression and any chronic
illness while those who where classified as
‘‘Other’’ had higher rates of comorbidity,
compared to non-Latino whites. Across chronic
pain conditions, Latinos, Asians and African
Americans tended to have lower rates of
comorbidity with depression, while those
classified as ‘‘Other’’ had higher rates than nonLatino whites. Looking at the service patterns,
Asians, Latinos and African Americans with
comorbid depression and chronic pain were less
likely to be asked about their mental health by a
primary care physician than non-Latino whites
with a similar profile. Although all ethnic/racial
minority categories had lower rates of
antidepressant use than non-Latino whites for
both the chronic illness and chronic pain
categories, these differences were not significant
in the bivariate models. Multivariate models
looking more closely at antidepressant use are
currently in progress.
Conclusions: This descriptive analysis suggests
lower rates of comorbid chronic health/pain
problems and depression among racial and
ethnic minorities, with the exception of those
who are classified as ‘‘Other.’’ However,
differences exist in primary care physicians’
assessment of comorbidity in minority patients.
Implications for Policy, Delivery or Practice:
Quality improvement programs for depression
screening in primary care should target multiple
chronic health conditions across racial and
ethnic groups. Physicians who treat minority
patients with chronic pain should inquire about
their mental health.
Funding Source(s): RWJF
ƒ Staying in Treatment: Factors Associated
with Retention in Depression Care Among
Ethnic/Racial Minorities
Lisa Fortuna, M.D., Margarita Alegria, Ph.D.,
Pinka Chatterji, Ph.D., Norah Mulvaney-Day,
Ph.D., David Takeuchi, Ph.D.
Presented by: Lisa Fortuna, M.D., Psychiatrist,
Center for Multicultural Mental Health Research,
120 Beacon Street, 4th Floor, Somerville, MA
02143, Phone: 617-503-8485, Fax: 617-503-8430,
Email: lfortuna@charesearch.org
Research Objective: The objective of this paper
is to identify patient and health system factors
associated with retention in depression care
among ethnic/racial minorities in the U.S. as
compared to non-Latino whites.
Study Design: Our analytic sample is limited to
NLAAS/NCS-R respondents with sub-threshold
depressive disorders or fulfilling full criteria for
depressive disorders in the last year and at least
one mental health visit during the last year or no
visits but stating that they dropped out of care.
We conduct a series of regression analyses to
evaluate the main effect of ethnicity/race on
retention in treatment as well as sequentially
adding measures of potential mechanisms
related to retention in care. Our dependent
variable of interest is a binary outcome of
retention in care defined by at least 4 mental
health care visits in the last 12 months or fewer
visits but currently reporting being in care coded
as one; and zero otherwise. Demographic and
socioeconomic status (SES) covariates included
in the regression model include race/ethnicity
(non-Latino white as the baseline, Latino, Asian,
African-American, Other), poverty level, age
category (18-34 as the baseline, 35-49, 50-64,
65+), gender, marital status, education level,
region, nativity, language of interview, insurance
(uninsured as the baseline, private insurance,
Medicare, Medicaid/other public programs) and
a HMO indicator based on whether the
respondent belongs to an HMO. We use data
from all racial/ethnic categories to fit the best
empirical model, reporting differences in
retention in care for depression in African
Americans, Asians, Latinos, and Other group
relative to non-Latino whites.
Population Studied: The combined National
Latino and Asian American Study (NLAAS) and
the National Comorbidity Survey Replication
(NCS-R) (pooled, N = 14,146) is one of the most
comprehensive nationally representative data
sources available that can be used to study
depression treatment for diverse groups of
ethnic minorities. Both the NLAAS and NCS-R
studies are part of the NIMH Collaborative
Psychiatric Epidemiology Surveys (CPES).
Principal Findings: Preliminary findings indicate
that after controlling for other factors, Latinos
(OR: 2.22, p-value .06) with depression or
dysthymia are more likely to be retained in
depression care as compared to their non-Latino
white counterparts. Asians, on the other hand,
were considerably less likely than non-Latino
whites to be retained in care (OR: .40, p-value:
.087). These effects are statistically significant at
the .10 level. Having Medicaid (OR: 2.7, pvalue=.02) or Medicare (OR: 4.4, p-value=.04) is
associated with higher likelihood of retention in
treatment.
Conclusions: These findings suggest that there
are important racial/ethnic differences in
retention in depression care with both patient
and systems factors, including type of insurance
coverage, associated with this outcome.
Implications for Policy, Delivery or Practice:
Identifying important correlates of retention in
treatment is likely to be useful in clinical care and
service planning for ethnic/racial minority
patients with depression.
Funding Source(s): RWJF
ƒ Estimating the Effects of Depression on
Labor Market Outcomes in Diverse
Populations
Pinka Chatterji, Ph.D., Margarita Alegria, Ph.D.,
Norah Mulvaney-Day, Ph.D., Lisa Fortuna, M.D.,
David Takeuchi, Ph.D.
Presented by: Pinka Chatterji, Ph.D., Health
Economist, Center for Multicultural Mental
Health Research, 120 Beacon Street, 4th Floor,
Somervile, MA 02143, Phone: 617-503-8449, Fax:
617-503-8430, Email: pchatterji@charesearch.org
Research Objective: The objective of this paper
is to estimate and to test for racial/ethnic
differences in the association between
depression and labor market outcomes.
Study Design: We use probit models to
estimate the association between depression
and labor market outcomes. Because of the
differences in the labor market pathways of men
and women, we run all analyses separately by
gender. Our dependent variables are the
following dummy indicators: currently employed
full-or part-time, absent from work at least once
in the past month among those employed, and
received any form of public assistance in the last
year. Our main covariate of interest is a dummy
variable indicating whether or not the
respondent meets criteria for depressive
disorders in the past year. All models include
controls for age, marital status, education,
number of household members under age 18,
physical health, and psychiatric co-morbidities.
Since reverse causality and unobserved
heterogeneity are concerns, we test whether
main findings are sensitive to these issues by
creating gender/race sub-samples for sensitivity
analyses. Using these sub-samples, we estimate
bivariate probit models, in which the
employment/public assistance and depression
equations are estimated simultaneously.
Population Studied: The combined National
Latino and Asian American Study (NLAAS) and
National Comorbidity Survey Replication (NCSR) (Pooled, N= 14,146) is one of the most
comprehensive national data sources available
that can be used to study depression treatment
for ethnic minorities. Both the NLAAS and NCSR studies are part of the NIMH Collaborative
Psychiatric Epidemiology Surveys (CPES), which
focus on collecting epidemiological information
on the risk factors for mental disorders as well as
extensive information on service usage among
the general population with special emphasis on
minority groups.
Principal Findings: Preliminary findings based
on NLAAS indicated that having depression or
dysthymia in the past 12 months is associated
with statistically significant reductions in the
probability of being employed of about 10
percentage points for Latino males and about 23
percentage points for Latino females. There are
no statistically significant associations between
depressive disorders and employment among
Asian models of other outcomes and bivariate
probit models based on the combined
NLAAS/NCS-R sample are in progress.
Conclusions: As a whole, the findings indicate
that depression imposes labor market costs on
Latinos that are at least as large and as
important as those found in studies based on
mostly non-Latino white samples.
Implications for Policy, Delivery or Practice:
Our results suggest that there may be important
labor market benefits associated with public
health policies targeted at the prevention and
effective treatment of depressive disorders in
racial/ethnic minority groups.
Funding Source(s): RWJF
Call for Panels
Post Acute Care: Changing with the Times
Chair: Marilyn Moon, Ph.D.
Monday, June 4 • 4:30 p.m.-6:00 p.m.
Panel Overview: Medicare policies and post
acute provider responses changed subtly over
the past few years. New payment policies went
into effect for each of the providers since 1998,
attention to IRF certification policies was
renewed, and access to services shifted for
certain populations as the new PPS underwent
incremental adjustments. These changes have
taken place in an environment concerned with
developing pay for performance initiatives that
reward good outcomes and create efficiencies.
This panel examines the impact of these policy
changes on post-acute payments, utilization, and
provider specialization for skilled nursing
facilities, inpatient rehabilitation facilities, home
health agencies, and long-term care hospitals.
The four papers presented here explore the
impact of these changes both within and across
post-acute care settings. The first paper
examines whether differences in severity exist
among admissions to different settings. The
second paper focuses on the rehabilitation
population and their trajectory following
discharge. The third paper addresses whether
IRFs specialize in certain populations, and
discusses alternative measures of specialization
that focus on economies of scope and scale. The
fourth paper looks at the more medically
intensive populations treated in both acute
hospitals and Long Term Care Hospitals and
examines margins for similar cases treated in
these settings. These papers are part of a larger
CMS effort to re-examine the incentives in
Medicare’s PAC payment policies and how they
may affect Medicare program costs and
beneficiary outcomes. These issues are
important as insurers call for better information
on which to develop pay for performance
methods.
ƒ A New Era: Post Acute Use Under PPS
Barbara Gage, Ph.D., Melissa Morley, Ph.D.,
Jeremy Green, B.A.
Presented by: Barbara Gage, Ph.D., Deputy
Director, Aging, Disability and LTC, RTI
International, 1440 Main Street, Suite 310,
Waltham, MA 02451, Phone: (781)-434-1717,
Email: bgage@rti.org
Research Objective: The purpose of this study
was to examine how case mix severity affects
variations in post-acute utilization.
Study Design: Retrospective claims analysis of
episodes of post acute care. Episodes of care
were created for the first acute hospital discharge
of the year and included claims from subsequent
inpatient rehabilitation, long term care hospitals,
skilled nursing facilities, home health agency,
and outpatient therapy services. The All Patient
Refined DRG, APR-DRG, software was used to
assign a minor, moderate, major, or extreme
severity of illness subclass to each hospital
admission. Binomial and multinomial logistic
regression models were used to test factors
predicting the probability of using any PAC
following hospital discharge, and to predict the
first site of post-acute care.
Population Studied: The study population
included Medicare beneficiaries with an acute
hospital discharge in 2004. The data source was
the 2004 Medicare claims and denominator files
for a 5-percent sample of beneficiaries.
Principal Findings: About one-third of all
inpatient hospital discharges in 2004 used some
type of PAC. Of those who used PAC, the
majority were discharged to a SNF, 41.5 percent.
Another 36 percent were discharged to HHA;
13.5 percent to IRFs and 2.2 percent to LTCHs.
The average severity of illness level for PAC users
was slightly higher than for non-PAC users. Over
36 percent of PAC users were in the two highest
APR-DRG severity of illness subclass groups
while the non-PAC users had 20.3 percent in the
higher severity groups. Analysis of discharges by
APR-DRG severity index revealed that those with
the highest severity were more likely to be
discharged to an institutional setting, particularly
a SNF, while those with the lowest severity were
more likely to use ambulatory services. After
controlling for demographics, severity of illness,
medical conditions, and supply of services, the
regression results suggested that the most
important factors for predicting PAC use and
first site of PAC care included specific DRGs, the
APR-DRG severity of illness measure, and age.
Beneficiaries ages 85 and over were 4.8 times as
likely to use PAC services compared to younger
beneficiaries. The odds of any PAC use for
beneficiaries in the major severity subclass were
57 percent higher than beneficiaries in the minor
severity subclass. Beneficiaries classified as
extreme severity were 8.3 times more likely to
use LTCH than beneficiaries classified as minor
severity.
Conclusions: These analyses reveal significant
differences in the populations using each type of
PAC setting. While many of the services in PAC
settings are similar, the multinomial model
identifies certain factors that clearly distinguish
differences in the probability of using each type
of setting. The severity measures provide
important additional insight into clinical
distinctions that may explain differences in PAC
level of care needs.
Implications for Policy, Delivery or Practice:
These findings will be useful as the Medicare
program continues to refine its payment systems
and the definitions of which populations are
appropriate for admission to each of the various
PAC settings.
Funding Source(s): CMS
ƒ Changes in IRF Use: Pre and Post PPS
Melissa Morley, Ph.D.
Presented by: Melissa Morley, Ph.D., Research
Associate, RTI International, 1440 Main Street,
Suite 310, Waltham, MA 02451, Phone: (781)-4341773, Email: mmorley@rti.org
Research Objective: The goal of this study is to
investigate whether the populations treated at
inpatient rehabilitation facilities, IRFs, have
changed in intensity since the implementation of
the prospective payment policies in 2002.
Study Design: Episodes of care were created for
IRF discharges and included claims from
subsequent IRF, acute hospital, long term care
hospital, skilled nursing facility, home health
agency, and outpatient therapy services. The All
Patient Refined DRG, APR-DRG, software was
used to assign a minor, moderate, major, or
extereme severity of illness subclass. Bivariate
and multivariate analyses were used to examine
differences in diagnoses, severity, length of stay,
and payment between 2000 and 2003.
Population Studied: The study population
included Medicare beneficiaries with an IRF
discharge and a prior acute hospitalization in
2000 and 2003. The data source for the analyses
was Medicare claims and denominator files for
100% of beneficiaries admitted to IRFs in 2000
and 2003.
Principal Findings: The volume of IRF
admissions increased by 26 percent from 2000
to 2003, corresponding with a 7 percent increase
in the number of IRF providers nationwide.
Overall, the severity of beneficiaries admitted to
IRFs was slightly higher in 2003 compared to
2000. In 2003, 34 percent of beneficiaries were
in the two highest severity of illness groups
compared to 32 percent in 2000. Despite the
increase in severity, lengths of stay in IRFs were
shorter in 2003 compared to 2000, 12.2 days
versus 13.8 days. The types of admissions to
IRFs also shifted slightly between 2000 and
2003. The three most common types of
admissions were DRG 209:Major Joint & Limb
Reattachment Procedures of Lower Extremity;
DRG 014:Specific Cerebrovascular Disorders
except TIA; and DRG 210: Hip & Femur
Procedures except Major Joint. The volume and
relative ranking of admissions for heart-related
conditions including DRG 015: Transient
Ischemic Attack & Precerebral Occulsuion and
DRG 127: Heart Failure & Shock increased from
2000 to 2003. As expected, length of stay and
payments increased with severity of illness. In
2003, 27 percent of IRF admissions for DRG 210
were major or extreme severity compared to 22.8
percent in 2000. The IRF admission lengths of
stay for these admissions decreased in 2003.
Similar patterns were also observed in other
DRGs. Discharge from a short-stay acute
hospital was also found to be related to longer
IRF length of stay.
Conclusions: The number of Medicare IRF
admissions and providers continued growing
between 2000 and 2003, despite the move from
TEFRA-based payments to PPS. After the
implementation of prospective payment, the
relative mix of diagnoses shifted somewhat,
severity of admissions increased, and lengths of
stay decreased suggesting that the population
treated in IRFs increased in intensity from 2000
to 2003. Other administrative changes in
Medicare’s coverage policies may have
contributed to these changes.
Implications for Policy, Delivery or Practice:
The increase in severity of illness in the IRF
settings is important for policymakers to
understand when considering quality of care in
IRF settings and further post-acute care payment
reform.
Funding Source(s): CMS
ƒ Medicare Margins Analysis for Long-Term
Acute Care Hospitals
Kathleen Dalton, Ph.D., Barbara Gage, Ph.D.
Presented by: Kathleen Dalton, Ph.D., Senior
Research Economist, RTI International, 3040
Cornwallis Road, Research Triangle Park, NC
27709, Phone: (919) 541-5919,
Email: kdalton@rti.org
Research Objective: 1.To document the impact
of prospective payment on long-term acute,
LTCH financial performance by analyzing
variation in LTCH Medicare margins across
facilities and types of cases. 2. To evaluate the
new LTCH cost weights by identifying systematic
but unintended differences in profitability across
DRGs.
Study Design: Retrospective analyses of
payments, charges and costs from facility cost
reports and Medicare LTCH claims. Claims-level
costs are estimated using service and facility
specific, cost/charge ratios. Claims-level
regression models identify adjusted differences
in payment-to-cost ratios by DRG after
controlling for patient attributes as well as fixed
facility effects.
Population Studied: Facility-level samples
include all LTCHs with cost reports filed from FY
2001 through FY 2004. Claims-level samples
include all post-PPS Medicare discharges that
could be matched to LTCH cost reports,
restricted to facilities that chose to receive
payment under 100 percent federal rather than
partially phased-in PPS rates.
Principal Findings: Medicare and overall facility
margins rose dramatically after PPS
implementation. Public LTCHs had significantly
lower margins both before and after PPS.
Median PPS margin was higher among forprofits than others and higher for facilities that
were recently certified than for older facilities. At
the individual case level, margins varied
substantially across DRGs even after stratifying
to remove the effects of high-cost outliers, 6
percent of sample, or short-stay outliers, 40
percent of sample. By DRG, average margins
were lowest for Rehabilitation, -0.1percent, and
highest for Ventilator Support, +21.3 percent.
Aggregate margin across all non-outlier cases
was +17.4 percent, compared to +13.8 percent for
short-stay outliers and -14.3 percent for high-cost
outliers. Variation in profitability across DRGs
was even greater in the multivariate models than
in the descriptive statistics computed across
cases. Regression results also confirm high
underlying levels of profitability; for a base non-
outlier case in 2005 that was admitted as an
acute hospital transfer and discharged to home
and had a DRG weight equal to 1.00, predicted
PPS payment was 23.2 percent above cost.
Conclusions: Unusually high average
profitability in the first two years of LTCH PPS
indicates that the initial PPS base rate was more
generous than intended. Variation in profitability
across DRGs also suggests substantial bias in
the DRG weights, where weights have been
systematically overstated for cases using
relatively more ancillary services and understated
for cases relying more on nursing care.
Implications for Policy, Delivery or Practice:
Smaller updates in the base rate were
implemented for LTCH PPS rules in FY 2006
and 2007, and may have brought underlying
LTCH PPS profitability back in line with other
Medicare PPS. DRG-level variation suggests
additional work is needed for estimating costbased weights, such as those being phased in
between 2007 and 2010 for short-stay DRGs. As
improved cost-based weights are implemented,
the profitability of LTCHs specializing in
respiratory care, in particular, may be strongly
challenged.
Funding Source(s): CMS
ƒ Inpatient Rehabilitation Facilities:
Alternative Definitions of Hospital
Specialization
Roberta Constantine, Ph.D.
Presented by: Roberta Constantine, Ph.D.,
Research Associate, RTI International, 1440 Main
Street, Suite 310, Waltham, MA 02451, Phone:
(781) 434-1711, Email: rconstantine@rti.org
Research Objective: Inpatient Rehabilitation
Facilities, IRFs, must meet the 75 percent rule to
qualify for IRF PPS rates in the Medicare
program. This study analyzes alternative
definitions of IRF specialization that could be
used to determine appropriateness of IRF
admissions, including measures of volume or
intensity rather than types of conditions.
Study Design: Retrospective analyses of
Medicare claims, Inpatient Rehabilitation Facility
Patient Assessment Instrument, IRF-PAI, and
Medicare certification data to examine variations
in hospital specialization in treating certain types
of patients. Twenty-one Rehabilitation
Impairment Categories, RICs, were utilized to
categorize admissions. Specialization was
defined in terms of 3 definitions: proportion of
total admissions in RIC groups, volume of
admissions in RIC groups, and proportion of all
RIC cases nationally per provider. These
measures were used to examine the relationship
between specialization and functional outcomes
at discharge.
Population Studied: 100 percent of Medicare
IRF admissions with discharge dates between
June 2004 and June 2005 that could be matched
to an IRF-PAI report.
Principal Findings: This study showed that the
average FIM score at admission was 65
suggesting patients were fairly impaired at the
time of admission. These admission scores
ranged but the highest mean and modal scores
were between the mid-70s and 80 for
populations having Non-Traumatic Spinal Cord
Injuries and Replacement of the Lower Extremity
Joint. The lower end of impairment scores at
admission were in the 50s for Stroke cases and
the trauma-related cases.
The average length of stay was 14 days but the
mean varied by RIC from 10 days for
Replacement of the Lower Extremity Joint to 21
days for Traumatic Spinal Cord Injury and
Guillain-Barre.
Virtually all IRFs had at least one Stroke
admission, 99.8 percent, but relatively few, 12.2
percent, had at least one Burn admission.
Examining volume of admissions by RIC,
admissions ranged from a high of 5 admissions
for Guillain-Barre to 1,612 for total admissions.
As a percent of total admissions to an IRF, the
percent of admissions for a RIC ranged from less
than 1 percent for Burns to 22.4 percent for
Replacement of Lower Extremity Joint. The third
definition of specialization, by the proportion of
all RIC cases nationally per provider, ranged
from less than 1 percent to as high as 18 percent
for Burns. Next, we created four variables for
each RIC based on whether the IRF had a
proportion of total admissions in certain RIC
groups and volume of admissions in RIC group.
Finally, we tested whether specialization had an
affect on FIM score.
Conclusions: This analysis attempts to define
specialization for an IRF; an important step in
defining what an IRF does and whether this has
an affect on patient outcomes. The results
showed that outcomes were better for certain
populations in facilities specializing in these
populations, particularly for those conditions
with smaller volume.
Implications for Policy, Delivery or Practice:
IRFs are the only Medicare hospitals certified on
the basis of admitting certain diagnoses. These
results will be useful for examining alternative
definitions of IRF certification.
Funding Source(s): CMS
Call for Panels
Impact of Florida Legislative and Regulatory
Innovations on Nursing Home Turnover,
Tenure, Financial Performance and Quality of
Care
Chair: Kathryn Hyer, Ph.D., M.P.P.
Tuesday, June 5 • 9:00 a.m.-10:30 a.m.
Panel Overview: Florida is a harbinger of
national long-term care issues because of its
proportion of residents in the 85+ age group.
Florida ranks sixth nationally in nursing home
beds, has the nation’s highest mandated nursing
home staffing levels and Florida’s bed growth
rate is ten times the national average of 2%.
Competition from assisted living facilities for
residents and staff is keen. From 1999-2007
Florida’s State Legislature implemented several
initiatives designed to improve quality in nursing
homes. Legislative innovations include financial
incentives to increase nurse staffing, mandates
to increase paraprofessional staffing over 5-years,
improved Medicaid reimbursement, tort reform,
enhanced monitoring and new reporting
requirements. Current Population survey data is
used to explore national nursing turnover and
tenure in long term care industry with special
attention to Florida. Reconciling data for from
three Current Population Survey reports this
paper compares turnover rates of LTC workers
within the LTC industry and with similarly skilled
workers in competing industries. In addition to
comparing turnover and tenure by nursing
levels, the relative importance of compensation,
training, and benefits on turnover are calculated.
Nursing home data (Florida State reports,
Medicare and Medicaid Cost Reports, Minimum
Data Set (MDS), the On-line Survey Certification
of Automated Records (OSCAR) file, and the
Area Resource File (ARF)) is matched at the
facility level to provide comprehensive
understanding of how quality innovations and
legislative mandates impact administrative
tenure and turnover, staffing and financial
performance, stability of workforce and quality
outcomes in Florida’s long-term care facilities.
ƒ Mandates Matter: Florida Nursing Home
Provider Response to Financial Incentives to
Increase Nurse Staffing
Kathryn Hyer, Christopher E. Johnson, Ph.D.,
Jeffery Harman, Ph.D., Mishu Popa, M.S.G.,
Robert Weech-Maldonado, Ph.D., M.B.A.
Presented by: Kathryn Hyer, 13301 Bruce B.
Downs Boulevard MHC 1323, Tampa, FL 33612,
Phone: 813-974-3232, Email: khyer@cas.usf.edu
Research Objective: Florida’s State Legislature
provided $40 million (annualized) in the
Medicaid Direct Care Staffing Adjustment
(DCSA) as an add-on to the patient care
component of the Medicaid per diem rate in
nursing homes. The financial incentives allowed
providers freedom in spending money as long as
total resources for direct care staffing increased
over baseline. This pre- post- study assesses how
the nurse staffing incentives beginning in April
2000 (compared to 1999) were used by nursing
homes to increase hours per resident day,
increase wages of current staff or to change use
of agency staff. We examine impact for the RN,
LPN and CNA staffing levels, and analyze if the
incentive resulted in sustained higher staffing
levels. The impact of nurse staffing change on
quality is measured by deficiencies and quality
measures.
Study Design: The data comes from Florida
State reports, Medicare and Medicaid Cost
Reports, Minimum Data Set (MDS), the On-line
Survey Certification of Automated Records
(OSCAR) file, and the Area Resource File (ARF)
for the period of 1999-2001. Facilities were
categorized into quartiles (N=145) receiving
most incentive money and least incentive
money. The dependent variable, quality of care is
measured by deficiencies, chronic (ADL
improvement, fractures, pain, pressure sores),
and post-acute outcomes (pain, walk
improvement, pressure sores). Independent
variables consist of structure (staffing ratios),
process (restraints and catheters), and control
variables (ownership/chain affiliation, Herfindahl
index, payer mix, case mix, and size). Quality
indicators were calculated for the six month
period immediately preceding implementation
and for six months following the incentive. A loglinear multivariate regression model assessed
the impact of incentive payments on quality
indicators.
Population Studied: Medicaid and medicare
nursing homes in Florida from 1999-2002. The
analysis of 595 Florida homes certified as
Medicaid homes that participated in the financial
incentive programs
Principal Findings: When offered incentives,
providers in top quartile of incentives responded
differentially from providers in lowest quartile.
Quartile receiving the most money ($68,117)
increased total nurse staffing from 1.77 hours per
resident day to 2.08 hours per resident day,
wages stayed stable and use of agency increased.
Facilities receiving least money ($34,163)
decreased from 2.54 to 2.33 hours per resident
day, increased wages for paraprofessionals and
decreased use of agency. The lowest staffed
highest Medicaid facilities increased RN, LPN
and CNA staffing but higher staffed homes
decreased highest skill nurses and substituted
lower cost nurses. Difference in difference
results indicate pressure sore quality measures
were significant.
Conclusions: Providing incentives results in
absolute dollar increases allocated to direct care
staff but skill mix decreases and total hours per
resident day may not increase. Difference in
differences shows mixed quality outcomes and
suggests threshold needs to be reached before
quality differences can be seen.
Implications for Policy, Delivery or Practice:
Legislatures wanting increased direct care
staffing per resident need mandates. Providers,
when given choice respond differentially but
substitute lower skilled workers for higher skilled
workers and increase wages for current workers.
The implications of skill mix on quality outcomes
are uncertain, especially since staffing per
resident day remain below CMS preferred levels
Funding Source(s): CWF, Administration on
Aging
ƒ Employee Turnover in the Long-Term Care
Industry
David Macpherson, Ph.D., William Even, Ph.D.
Presented by: David Macpherson, Ph.D., Rod
and Hope Brim Eminent Scholar in Economics,
Director of the Pepper Institute on Aging and
Public Policy, Florida State University, Mail Code:
2180, Florida State University, Tallahassee, FL
32306, Phone: (850) 644-3586, Email:
dmacpher@mailer.fsu.edu
Research Objective: The presentation will
discuss several potential causes of understaffing
in the long-term care (LTC) industry. First, we
present measures of the extent of turnover in
several key occupations within the LTC industry
and discuss the importance of compensation,
training, worker characteristics and working
conditions on the extent of turnover. We
compare the turnover rates of LTC workers with
similarly skilled workers in competing industries
and explain the source of differentials in the
turnover rate. Second, the presentation will
provide evidence on the source of workers
moving into the LTC industry and the destination
of workers moving out of the industry.
Study Design: Current Population survey data is
used to explore national nursing turnover and
tenure in long term care industry with special
attention to Florida. Registered nurses, licensed
nurse, health, personal and home care aides and
nursing aides, orderlies and attendants------are
tracked from 1989-2002. Employee turnover can
be reduced with a variety of incentives that are
not equally effective. Direct comparison of cost
effectiveness of approaches to reducing turnover
to policies increasing wages or fringe benefits is
not possible. Reconciling data from three
Current Population Survey reports, this paper
compares turnover rates of LTC workers within
the LTC industry with similarly skilled workers in
competing industries. In addition to comparing
turnover and tenure by nursing levels, the
relative importance of compensation, training,
and benefits on turnover are calculated.
Population Studied: Detailed examination of
worker turnover in nursing and personal care
industry for four key occupations: RN, LPH,
nurse aides and health aides. Three different
data sets were employed and several measures
of turnover are examined.
Principal Findings: Different measures of
turnover can lead to different conclusions
regarding whether workers have higher turnover
than other groups. A single measure does not
address all concerns. Regression analysis used
to determine the influence of worker and job
characteristics on turnover reveal improved
pension and health insurance reduces turnover
in all nursing levels. Unionism and job training
reduce turnover but environmental disamenities
or physical demand increase turnover. In a
comparison of increased wages, pension
coverage or health insurance, pension coverage
is the lowest cost approach to reducing turnover.
Conclusions: Employee turnover can be reduced
with a variety of incentives that are not equally
effective. Direct comparison of cost effectiveness
of approaches to reducing turnover to policies
increasing wages or fringe benefits is not
possible
Implications for Policy, Delivery or Practice:
This information is important to understanding
how future changes in the U.S. economy are
likely to influence the supply and demand for
long-term care workers. While this study
provides evidence on how to reduce employee
turnover questions remain about how turnover is
affected by working conditions and staffing levels
and how low-cost facilities may impact turnover.
Funding Source(s): Administration on Aging
ƒ Impact of Top Management Tenure on
Nursing Home Turnover and Quality
Christopher Johnson, Ph.D., Kathryn Hyer, Ph.D.,
Jeffery Harman, Ph.D., Robert WeechMaldonado, Ph.D., M.B.A., Mishu Popa, M.S.G.
Presented by: Christopher Johnson, Ph.D.,
Associate Professor, Department of Health
Policy and Management, School of Rural Public
Health, Room 311 SRPH Building, College
Station, TX 77843-1266, Phone: (979) 458-4165,
Email: cejohnson@srph.tamhsc.edu
Research Objective: The objective of the is
research is to examine the impact of nursing
home top management (administrator and
director of nursing) tenure on nursing staff
turnover and quality of care in Florida nursing
homes. We hypothesize that organizations with
higher administration turnover rates have higher
nursing staff turnover and poorer quality of care
outcomes
Study Design: We analyzed the impact of
nursing home administrator and director of
nursing (DoN) tenure from 2002-2004 using
Florida Medicaid Nursing Home Staffing
Reports for 580 Medicaid nursing homes, CMS’
Online Survey Certification and Reporting
(OSCAR), Medicare Cost reports, State of Florida
Medicaid Cost reports, CMS Minimum Data Set
(MDS) and Area Resource File (ARF). Tenure
was defined as the administrator or DoN being
the same for the 12-month period of the Nursing
Home Staffing Reports for that year. The
dependent variables include 1) turnover (certified
nursing assistant and licensed nurses) and 2)
change in quality of care from previous year
(MDS activities of daily living, fractures, and
pressure sore chronic care indicators). The
independent variables include 1) administrator
and DON salary, 2) administration tenure, 3)
facility resources, 4) facility characteristics, 5)
county demographic variables, and 6) county
market factors. These variables were analyzed
using a two-staged least squares regression to
adjust for potential endogeniety between staffing
and administration tenure, and to analyze the
impact of the adjusted variables on nurse staff
tenure and quality of care.
Population Studied: All Medicaid certified
nursing homes in Florida.
Principal Findings: Tenure trends in nursing
home administrator and DoNs in Florida
followed similar patterns of sharp decline in
2003 (from roughly 22% being present for one
year to 18%) and modest increases during 2004
(to roughly 20%), but not to the higher 2002
levels. Higher certified nursing assistant and
licensed nurse turnover rates and poorer quality
of care were associated with higher tenure rates
among top management
Conclusions: These results raise questions
about the impact that top management stability
has on nursing home performance in Florida’s
facilities. While some top management turnover
may be required (in order to remove ‘‘bad
apples’’), there may be a threshold above which
facilities will experience decreasing negative
results across nursing home performance
measures
Implications for Policy, Delivery or Practice:
Much attention has been spent on turnover
within nursing staffs within policymaker efforts
to improve quality of care. Approximately 80%
of the nursing homes in Florida did not have the
same administrator and/or DoN for the entire
year during the time period of this study. States
and organizations may need to think of ways of
increasing tenure in nursing home
administrations as part of an overall nursing
staff and facility quality
Funding Source(s): CWF, Administration on
Aging
ƒ The Relationship between Staffing, Quality,
and Financial Performance: For-Profit versus
Not-for-Profit Nursing Homes
Robert Weech-Maldonado, Ph.D., M.B.A.,
Christopher Johnson, Ph.D., Zhou Yang, Ph.D.,
Alex Laberge, M.B.A., Kathryn Hyer, Ph.D.
Presented by: Robert Weech-Maldonado, Ph.D.,
M.B.A., Associate Professor, Department of
Health Services Research, University of Florida,
College of Public Health and Health Professions,
101 S. Newell Drive, Gainesville, FL 32610-0195,
Phone: (352) 273-6080, Fax: (352) 273-6075,
Email: rweech@phhp.ufl.edu
Research Objective: Nursing homes are
experiencing increased revenue constraints as a
result of changing reimbursement policies and
declining demand for nursing home care. At the
same time, nursing homes are facing increasing
costs pressures as a result of higher nurse
staffing and liability insurance costs. As the
financial performance of nursing homes
deteriorates, there are concerns that quality of
care will also decline. Using Donabedian’s
structure, process, and outcomes framework of
quality assessment, this study examines the
relationship between staffing, quality, and
financial performance, and whether this
relationship varies by ownership (for-profit
versus not-for-profit status).
Study Design: The data for this study comes
from the Medicare and Medicaid Cost Reports,
Minimum Data Set (MDS), the On-line Survey
Certification of Automated Records (OSCAR) file,
and the Area Resource File (ARF) for the period
of 2000-2004. The dependent variable consists
of financial performance tier: low (bottom
quartile), medium (2nd and 3rd quartiles), and
high (upper quartile), based on the operating
margin. The independent variables consist of
structure (RN, LPN, and CNA staffing ratios),
process (use of restraints and catheters), chronic
care outcomes (ADL improvement, fractures,
pain, pressure sores), and post-acute outcomes
(pain, walk improvement, pressure sores).
Control variables consist of ownership/chain
affiliation, payer mix, case mix, and size. Ordered
probit regression is used to model the
relationship between financial performance tier
and the quality measures, with year as a fixed
effect. Standards errors are adjusted for
clustering due to repeated observations for
facilities over time using Huber/White
correction. The models are run for the overall
sample, and separately for the for-profit and notfor-profit samples.
Population Studied: All Medicare and Medicaid
certified nursing homes in Florida for the period
of 2000-2004, approximately 575 facilities per
year
Principal Findings: For the overall sample,
better financial performance was characterized
by lower RN/LPN ratios but higher CNA ratios,
but there were no significant differences in
process and outcomes of care by financial
performance tier. Among not-for-profit nursing
homes, better financial performance was
characterized by lower RN staffing, higher CNA
staffing, lower fractures, and lower pressure
sores (chronic care). While among for-profit
homes, those with better financial performance
were characterized by lower RN staffing, lower
restraints, lower catheters, and higher ADL
improvement.
Conclusions: The relationship between quality
and financial performance varies by type of
quality measure. Both for-profit and not-profit
nursing homes that have better financial
performance have better resident outcomes in
some areas, yet have lower RN staffing ratios
than those with worse financial performance.
However, there are some differences by
ownership type. Not-profit homes with better
financial performance have higher CNA staffing
ratios than not-for-profit nursing homes with
worse financial performance, and for-profit
homes with better financial performance have
better processes of care than for-profit nursing
homes with worse financial performance.
Implications for Policy, Delivery or Practice:
Further research is needed to examine the
management strategies of the higher performing
nursing homes by which they are able to achieve
better processes and outcomes of care with
lower RN staffing ratios
Funding Source(s): Administration on Aging
Call for Panels
Medication Use by Aged and Disabled
Medicare Beneficiaries across the Spectrum
of Morbidity
Chair: Stuart Guterman, M.A.
Tuesday, June 5 • 10:45 a.m.-12:15 p.m.
Panel Overview: The multiplicity of Part D plans
and benefit designs available to Medicare
beneficiaries raises serious issues regarding the
quality of medication use under the new
prescription drug benefit. One important and
under-studied aspect of medication quality is
how overall disease burden affects physician
prescribing and patient utilization behavior. This
issue is the subject of a Commonwealth Fund
sponsored monograph to be published in April
2007 entitled, Chartbook of Medication Use by
Aged and Disabled Medicare Beneficiaries across
the Spectrum of Morbidity. The Chartbook
provides descriptive benchmarks showing how
drug utilization varies with overall disease
burden for 8 common chronic diseases:
diabetes, depression, dementia, COPD, arthritis,
hypertension, ischemic heart disease, and heart
failure. A major aim of the project collaboration
is to generate hypotheses for future research on
the relationship between polymorbidity and
polypharmacy. This proposed ARM panel
includes 5 related papers that follow up on
hypotheses generated during the Chartbook
collaboration. The first paper describes the
project and presents major descriptive findings.
The second paper addresses geriatricians’
concerns about overly aggressive treatment at
end of life by examining medication profiles for
survivors and decedents with diabetes, COPD
and CHF. The next two papers examine
disparities in medication use for therapies
recommended for dementia and depression.
They find both racial and age-related disparities
that grow wider with increasing disease burden.
The final paper addresses a common problem at
the nexus of polypharmacy and polymorbidity;
namely, growing side effects of use of analgesics
and anti-inflammatory agents to treat arthritis.
ƒ Charting Medication Patterns of Medicare
Beneficiaries with Rising Disease Burden:
Summary of Findings
Research Objective: This paper summarizes
findings from the Commonwealth Fund
Chartbook of Medication Use by Aged and
Disabled Medicare Beneficiaries across the
Spectrum of Morbidity scheduled for publication
in April 2007. The Chartbook provides
nationally-representative benchmarks of drug
utilization by Medicare beneficiaries with 8
common chronic diseases: diabetes, depression,
dementia, COPD, arthritis, hypertension,
ischemic heart disease, and heart failure. It is a
reference work designed to help policy makers
and health services researchers better
understand the complex relationship between
medication use and disease burden and to
develop a research agenda for addressing
suboptimal medication utilization patterns in the
Medicare population. A key feature of the
Chartbook is that all major classes of
pharmaceuticals are profiled for each disease
cohort, thereby demonstrating how medication
regimens evolve with accumulating morbidity.
Study Design: Data come from the 2002
Medicare Current Beneficiary Survey. The study
sample comprises fee-for-service communitydwelling beneficiaries divided into 8 diseasespecific cohorts using ICD-9 codes from
Medicare claims and self/proxy reports (for
selected conditions). Each cohort is divided into
deciles or quintiles (depending on sample size)
based on beneficiaries’ cumulative annual
spending on all medical care services. Medical
spending is a useful proxy for overall disease
burden and has the further advantage that the
study methods that can be readily applied in any
payor setting. Variables include: drugs classified
by UPS pharmacological and therapeutic
category, counts of medication-sensitive
conditions from the CMS RxHCC risk
adjustment model, drug spending, total medical
spending, demographic and socioeconomic
characteristics, health status measures, and
health system contact variables. Charts for each
disease state show how medication intensity
changes with quintile/decile of medical
spending. Medication intensity measures
include: drug spending as a fraction of total
medical spending, prescription fills per RxHCC,
prevalence and level of drug use by USP
category, and prevalence and utilization rates for
drugs specifically recommended to treat each of
the 8 conditions. The Charts are augmented by
extensive appendix tables. All results are
weighted to be representative of the Medicare
population.
Population Studied: Aged and disabled
Medicare beneficiaries in 2002 (N = 8,455), with
disease-specific cohorts ranging in size from 521
(dementia) to 4,617 (hypertension).
Principal Findings: Selected charts from
among the 82 contained in the Chartbook will be
presented. A consistent finding across the 8
diseases is that medication intensity increases
with spending to a plateau and then declines.
This inverted U shape pattern is observed in
each of the medication intensity measures, albeit
the steepness of the rise and fall as well as the
peak threshold points varies by disease and drug
measure. Suboptimal use of recommended
agents is found at both ends of the spectrum.
Conclusions: The Chartbook presents only
unadjusted cross-tabulations and further
research is necessary to identify the reasons
behind the unusual inverted U shape patterns in
medication use observed among Medicare
beneficiaries with the 8 diseases.
Implications for Policy, Delivery, or Practice: The
Chartbook provides an overview of the
‘‘medication landscape’’ pre-MMA that will help
policy makers design and target quality initiatives
under Part D.
ƒ Medication Use by Community Dwelling
Medicare Beneficiaries at the End of Life: Is it
Too Much Too Late?
Research Objective: Geriatricians argue that
aggressive medication regimens appropriate for
healthy seniors should be tempered late in life,
particularly when time-to-benefit exceeds life
expectancy. Geriatric guidelines for diabetes and
certain other chronic diseases recognize this
issue, but it is unclear how these
recommendations are implemented in practice.
The objective of this study is to empirically
determine the extent to which Medicare
beneficiaries in their last year of life receive less
intense medication regimens than survivors, all
else being equal.
Study Design: The study uses data from the
2002 Medicare Current Beneficiary Survey
(MCBS) compiled as part of the Commonwealth
Fund Chartbook project. Annualized differences
in medication use between survivors and those
who died in 2002 are analyzed for the
community-dwelling Medicare population and
for 3 nonexclusive disease cohorts with diabetes,
COPD, and congestive heart failure. Aggregate
medication intensity is measured as prescription
fills per medication-sensitive condition using the
CMS RxHCC risk adjustment model. Medication
intensity is also compared for drugs specifically
recommended for each of the three conditions.
All estimates are weighted to be nationallyrepresentative and adjusted for the complex
MCBS survey design.
Population Studied: A nationally representative
sample of fee-for-service community-dwelling
Medicare beneficiaries (n=8,455) and three
subgroups of beneficiaries with diabetes
(N=1,956), COPD (N=1,016), and CHF
(N=1,140) in 2002.
Principal Findings: 3.6% of the overall sample
died during 2002. Mortality rates in the
diabetes, COPD, and CHF cohorts were 4.6%,
11.4%, and 13.2%, respectively. Annualized rates
of prescription fills per RxHCC were consistently
higher for decedents compared to survivors: 5.5
versus 4.4 for the whole population; 5.6 versus
5.5 for the diabetes cohort; 5.2 versus 4.2 for
those with COPD; and 5.0 versus 4.2 for the CHF
cohort. Unadjusted rates for recommended
medications were lower in patients at end of life
but the differences disappear in multivariate
analyses adjusting for mean survival time,
sociodemographic characteristics, health status,
and health system contact variables. For
example, diabetes patients at the end of life had
insignificantly higher odds of receiving any
antidiabetic medication [OR 1.61 95% CI (0.69 3.77)], and among users, the number of fills was
insignificantly higher among decedents [0.30 fills
95% CI (-2.86, 3.47)]. Multivariate analyses for
the COPD and CHF groups are ongoing.
Conclusions: We find no evidence that
medication utilization rates are lower among
Medicare beneficiaries in their last year of life
and some evidence that they may be higher.
Implications for Policy, Delivery, or Practice:
The study findings have important clinical and
policy relevance given the recent implementation
of the Medicare drug benefit. Part D improves
access to medications in general and may
potentially increase polypharmacy in patients late
in life. There is an urgent need to develop
explicit guidelines to facilitate joint decision
making among physicians, patients, and families
on when it is appropriate to consider
withholding or discontinuing medications in
patients late in life.
ƒ Racial and Ethnic Disparities in Treatment
of Dementia among Medicare Beneficiaries
Research Objective: To determine whether
there are systematic racial and ethnic differences
in the treatment of dementia among communitydwelling Medicare beneficiaries. Specifically, we
examine the prevalence and intensity of use of
antidementia drugs across racial/ethnic groups
and by level of disease burden.
Study Design: Cross-sectional study building on
data files constructed for the forthcoming
Commonwealth Fund Chartbook. Antidementia
drugs include cholinesterase inhibitors available
for use in 2002 (tacrine, donepezil, galantamine,
rivastigmine). Antidementia drug prevalence is
defined as the presence of one or more
prescriptions for an antidementia drug during
2002. Antidementia drug intensity among users
is based on number of prescription fills for an
antidementia drug during the year. Beneficiaries’
overall disease burden is categorized by quintile
of cumulative annual medical spending from all
payor sources (including out-of-pocket) with
quintile #1 representing the lowest disease
burden, and quintile #5 the highest. As of this
writing, descriptive analyses are complete and
multivariate analyses are underway. Projected
time for completion of the paper is the end of
March, 2007.
Population Studied: Data were obtained from
the 2002 Medicare Current Beneficiary Survey
(MCBS), a nationally-representative database on
Medicare beneficiaries’ use and cost of health
services. Community-dwelling persons were
selected based on either a self/proxy report of
dementia or the presence of a claim with an ICD9 code indicating dementia. The final MCBS
sample with dementia comprised 521
individuals, representing approximately 1.5
million beneficiaries nationally in 2002.
Principal Findings: Over one-fourth of
community-dwelling beneficiaries with dementia
had a prescription fill for an antidementia drug in
2002. Antidementia drug prevalence was highest
among those in quintile #2 (38%) and lowest
(18%) in the group with the least disease burden
(quintile #1), suggesting that use of these agents
rises with low to moderate disease burden, then
drops off with increasing disease burden to just
23% in quintile #5. White, non-Hispanics had
higher prevalence of antidementia drug use
overall (28.8%) and in every spending quintile
compared to nonwhite Medicare beneficiaries
(18.8%). The difference was nearly two-fold in
quintile #2, with prevalence among whites at
43% compared to other races at 21%. However,
among beneficiaries who used antidementia
drugs during the year, medication intensity did
not vary among the race/ethnicity categories,
with the mean number of antidementia
prescriptions similar overall (whites=5.7,
nonwhites=5.5) and across spending quintiles.
Conclusions: These descriptive results suggest
important racial differences in who receives
treatment with antidementia drugs, but no
evidence that the intensity of use varies by
race/ethnicity. Logistic models are being
estimated to determine if the racial/ethnic
disparities remain in adjusted analyses
controlling for a rich array of demographic,
socioeconomic, and health status variables
available in the Chartbook analytic data files.
Implications for Policy, Delivery or Practice:
Understanding the etiology and implications of
these observed racial differences will enable
policy makers to assess the human and
economic impact of racial differences in
dementia treatment.
ƒ Differentials in Medication Treatment of
Aged and Disabled Medicare Beneficiaries
with Depression
Research Objective: Depression is common
among both aged and SSDI disabled Medicare
beneficiaries, but as demonstrated in the
Commonwealth Fund Chartbook of Medication
Use by Aged and Disabled Medicare
Beneficiaries across the Spectrum of Morbidity,
medication treatment patterns are far different
for the two groups. Specifically, utilization rates
for antidepressants, anxiolytics, opioid
analgesics, and sedative hypnotics are all much
greater among disabled beneficiaries with
depression, and the difference increase with
rising comorbidity. The purpose of this study is
to further investigate the source and clinical
significance of these differences.
Study Design: Data are from the 2002 Medicare
Current Beneficiary Survey (MCBS). Descriptive
univariate and bivariate analyses of drug
utilization patterns for study subjects focus on
the following medication classes:
antidepressants (SSRI and other), anxiolytics,
sedative-hypnotics, and opioid analgesics.
Prevalence of use and counts of prescription fills
for each type of medication are computed for the
population with depression as a whole and
stratified by aged (65+) and SSDI disabled (<65)
eligibility status. Overall disease burden is
proxied by quintile of cumulative total medical
spending. Negative inflated binomial models
are used to test for significant aged/disabled
differences in utilization of these agents with
interaction terms for overall disease burden
(quintile of spending), controlling for
sociodemographics, health status measures, and
other comorbidities.
Population Studied: The sample comprises of
910 community-dwelling beneficiaries
categorized through ICD-9 codes and/or selfreport as having major depression. More than
one-quarter of the sample (27.5%) are under age
65 SSDI beneficiaries.
Principal Findings: Utilization rates for all drug
classes studied were much higher for SSDI
disabled beneficiaries compared to aged
beneficiaries with depression, and the
differences mount with growing disease burden.
For example, antidepressant use by the SSDI
subset varies from 7 fills in quintile 1 to 13.3 in
quintile 5. By contrast, antidepressant use for
the aged subset is 6.2 fills in quintile 1 and 6.4 in
quintile 5. Prescription fills for anxiolytics vary
from 6.5 to 7.5 for both subsets of beneficiaries
between quintiles 1 and 3, but then diverge
sharply, with disabled beneficiaries filling 3.8
more anxiolytic prescriptions than aged
beneficiaries by quintile 5. Similar patterns are
noted for the other drug groups. Multivariate
analysis is in progress.
Conclusions: Preliminary findings identify stark
differences in the use of prescription
medications known to ease the emotional and
physical suffering associated with depression
among aged and disabled Medicare
beneficiaries. Additional analysis is necessary to
determine whether these differences remain after
controlling for sociodemographics and health
status measures.
Implications for Policy, Delivery or Practice:
Most of the policy attention given to
understanding patterns of medication use by
Medicare beneficiaries has focused on those
aged 65 and older. The results from this paper
suggest that more attention should be given to
the under age 65 population.
ƒ Arthritis, Disease Burden, and the Impact
of Medication Complications
Research Objective: This paper examines
medication use and complication rates for
Medicare beneficiaries with osteoarthritis (OA)
and rheumatoid arthritis (RA) as a function of
rising overall disease burden. Previous research
demonstrates patients and clinicians must
carefully balance the benefits from arthritis drugs
with potential complications. This balancing
becomes increasingly problematic as overall
disease burden increases. Little prior research
has looked specifically at these issues for a
national sample of Medicare beneficiaries with
arthritis.
Study Design: Data come from the 2002
Medicare Current Beneficiary Survey. The study
sample comprises fee-for-service communitydwelling beneficiaries with OA or RA divided into
deciles based on beneficiaries’ cumulative
annual medical spending as a proxy for overall
disease burden. The principal independent
variables are drugs used to treat arthritis pain
and inflammation (nonsteroidal antiinflammatory drugs, corticosteroids, opioid
analgesics). The dependent variables are
inpatient admissions, drugs, and other health
services that are markers for potential
medication complications. Control variables
include counts of medication-sensitive
conditions from the CMS RxHCC risk
adjustment model, demographic and
socioeconomic characteristics, health status
measures, and health system contact variables.
Regression models test both the independent
effect of drug use and the mediating effect of
drug use and decile assignment on complication
rates. All results are weighted to be
representative of the Medicare population.
Population Studied: Aged and disabled fee-forservice community-dwelling Medicare
beneficiaries in 2002 who had two or more
Medicare claims for OA or RA or had one claim
and a self-report of arthritis (N = 2,066).
Principal Findings: In 2002 more than 24
percent of community-dwelling Medicare
beneficiaries were diagnosed with arthritis. The
comorbidity profile of this population shows a
strong positive association with annual medical
spending. In the lowest spending decile 1, the
most prevalent comorbidities were hypertension
(47%) diabetes (15%), and ischemic heart
disease (11%). By decile 10, 83% had
hypertension, 61% had IHD, 41% CHF, and 39%
COPD. Approximately 55% of Medicare
beneficiaries with arthritis use NSAIDs. Opioid
use is very high (36% in the OA subset and 44%
in the RA subset). Corticosteroid use is also
surprisingly high with 13% of persons with OA
and almost 25% of persons with RA using these
drugs. Rising disease burden is associated with
increased markers for antiarthritic medication
complications, particularly gastrointestinal
drugs.
Conclusions: Initial analyses raise several
questions about appropriate protection from
medication complications associated with
arthritis treatment for Medicare beneficiaries
with OA or RA. Ongoing multivariate analysis is
designed to address the prevalence of arthritis
medication complications on Medicare
beneficiaries with rising morbidity.
Implications for Policy, Delivery, or Practice:
The paper suggests opportunities for Medicare
to improve prescribing and reduce costs
associated with arthritis treatment through
appropriate management of pharmacotherapy
for the pain and inflammation associated with
the disease.
Call for Panels
Advancing Diabetes Self-Management
Chair: Joseph Burton, M.S.
Tuesday, June 5 • 10:45a.m.-12:15p.m.
Panel Overview: Chronic diseases, including
diabetes, place a significant burden on the
patient and family, the community, and the
health care system. Over the past 30 years,
patients have been increasingly encouraged to
take a more active role in their health care,
including managing chronic illness. Providers
increasingly need to reorient their care strategies
and systems from a focus on acute care model
to chronic illness care. With the goal of
demonstrating the feasibility of self-management
interventions in real-world settings, the Diabetes
Initiative of the Robert Wood Johnson
Foundation funded six grantees to develop and
implement self-management programs in
primary care settings and another eight grantees
to demonstrate partnerships between clinical
and community organizations to improve
resources and supports for self-management.
The programs take one or more different
approaches to promoting self-management:
improving nutrition, increasing physical activity,
use of lay health workers/promotoras, group
medical visits and support groups, and
expanded use diabetes registers. The evaluation
of the initiative includes case studies, the
collection of clinical data, and a three-wave
participant survey. Grantees conducted internal
evaluations, participated in work groups, and
hosted site visits. We surveyed participants
about their level of physical activity and diet and
nutrition behaviors and the extent to which they
participated in program related activities. We
also collected clinical data over time from the
medical registers of the grantees that maintained
and used them as part their care processes. This
series of papers discuss lessons learned from
the diabetes initiative and implications for
improving chronic illness care.
ƒ Managing Depression to Improve Diabetes
Self-Management
Daren Anderson, M.D., Carla Bann, Ph.D., Daren
Anderson, M.D., Claire Horton, M.D.
Presented by: Daren Anderson, M.D., Medical
Director, Community Health Center, Inc., 635
Main Street, Middletown, CT 06457, Phone:
(860) 347-6971, Email: andersdr@chc1.com
Research Objective: Improving diabetes selfmanagement includes both lifestyle changes
and, often, adherence to regimes for taking
prescription medications. Both the lifestyle and
medical management of diabetes can be
challenging for persons with diabetes mellitus
and can be particularly challenging for
individuals with depression. We examine the
differences in outcomes between persons who
exhibit depressive symptoms compared to those
who do not and review the successes of
interventions in improving the self-management
behaviors among both groups.
Study Design: As part of a three-wave,
comprehensive telephone survey, we assessed
the key self-management behaviors of diet,
exercise, smoking, and self-monitoring. We also
assessed mental health status through eight of
the nine questions composing the patient health
questionnaire (PHQ-9). The PHQ-9 is typically
used in practice settings to screen for depressive
symptoms; in this study we used it to gauge
overall mental health status among the
participants in the various diabetes programs.
We collected glycated hemoglobin levels and
blood pressure values over time from medical
registers and linked these to the survey
responses. We analyzed mental health status,
behaviors, and clinical outcomes through a
series of multivariate regression and survival
analyses.
Population Studied: The Robert Wood Johnson
Foundation funded 14 programs designed to
advance diabetes self-management and improve
community supports for diabetes care. The
programs varied in terms of the clinical settings,
populations served, age ranges, and racial/ethnic
backgrounds. Participants from 11 of the
programs were surveyed up to three times.
Principal Findings: Controlling for several
personal and program characteristics, we found
persons who reported poorer mental health
(more depressive symptoms) were less
confident in being able to solve problems related
to diabetes, had a family history of diabetes and
were more likely to smoke; persons reporting
better mental health tended to be married,
younger, and reported greater access to
resources and support for managing their
illness. Persons in better mental health exercised
more and ate better compared to those with
poorer mental health, but both groups reported
similar levels of self-monitoring. Eating
behaviors------perhaps the most challenging
dimension of change------were better among those
in good mental health even when controlling for
marriage, reported supports, and confidence.
Clinical indicators and depressive symptoms
improved slightly over time among those
followed longitudinally
Conclusions: The relationship between mental
health and behavior change is complex and
challenging to measure and particularly so
among persons with diabetes because of the
bidirectional nature of depression and diabetes.
Addressing depression is an important part of
diabetes care and, when treated, can lead to
improved self-management behaviors and,
ultimately, improved clinical outcomes.
Implications for Policy, Delivery or Practice:
Persons with diabetes from underserved,
disadvantaged or minority populations have high
rates of depressive symptoms, which might
affect their ability to self-manage their disease.
Mental health status appears to moderate
behavior change, suggesting that strategies for
improving mental health be a core component of
self-management programs. Understanding the
role of mental health in diabetes care and more
generally, chronic illness care, can lead to
improved programs for improving selfmanagement, lifestyle behaviors, and well-being.
Funding Source(s): RWJF
ƒ Enhancing Chronic Care Delivery in Primary
Care Settings Through Planned and MiniGroup Visits
Devin Sawyer, M.D.
Presented by: Devin Sawyer, M.D., Family
Physician, Providence St. Peter Family Practice
Residency, 525 Lilly Rd. NE, Olympia, WA 98506,
Phone: (360) 493-7230,
Email: devin.sawyer@providence.org
Research Objective: Planned visits are
performed by medical assistants prior to routine
visits and include blood pressure, weight and
foot checks, routine laboratory tests,
immunizations, referrals and patient goal
setting. Mini-group visits occur when a physician
meets with three patients at a time with the
same chronic illness (i.e., diabetes mellitus) for
approximately one hour. This research describes
the preparatory steps needed to implement
planned and mini-group visits including use of a
disease registry and staff training. We review the
purpose and benefits of planned and mini-group
visits, as well as the time commitment, billing
codes, and staff support needed to make these
visits efficient. In addition, we discuss the
characteristics of good visits, facilitation
challenges, group dynamics, and techniques for
intervention. Lastly, teaching the skills required
for effective mini-group visits within a residency
program are identified.
Study Design: The study synthesizes results
from case study of the residency program,
internal observations documented from the
program’s quality assurance efforts, a three-wave
patient survey, and a review of clinical indicators
collected over time. We tracked blood pressure,
glysocated hemoglobin, weight, and cholesterol
levels of patients participating in the visits. We
also summarize provider reports related to
medications, histories, challenges, and
successes.
Population Studied: Programs funded under
the Robert Wood Johnson Foundations Diabetes
Initiative were designed to advance diabetes self
management and promote community supports
for diabetes. This study investigates the care
processes and related outcomes of patients who
received one or both of interventions offered by
the Providence St. Peter Family Practice
Residency Program. A portion of the patients
also participated in a multi-wave survey
questionnaire administered as part of the
national Diabetes Initiative.
Principal Findings: Planned visits provide
physicians with current health measures and
patient goals, hence allowing more time for
dialogue during individual and mini-group
appointments. This improves the patient
provider dynamics and patient satisfaction with
the care process. We observed improvements in
several clinical indicators among the
participating patients. In addition, while the
primary purpose of mini-group visits is medical
care, patients also experience peer support and
increased confidence to manage their disease.
Conclusions: Medical assistants are an integral
part of a diabetes care team in a practice setting.
With appropriate training and provider buy-in,
medical assistants and other non-physician
professionals can lead patients in goal setting,
skill building, and problem solving. They also can
improve a practice’s efforts to provide
monitoring and ongoing support for patients
with chronic illnesses. Both planned and minigroup visits provide a structure to better
integrate a chronic disease care team within
billing guidelines.
Implications for Policy, Delivery or Practice:
With the number of people living with chronic
disease on the rise, primary care physicians are
ill-equipped to manage patients under the
traditional acute care system of medicine. As a
result, primary care organizations continue to
experience poor health outcomes, missed
opportunities to maximize care, and greater
medical costs. To answer these challenges, the
medical community has needed new models and
approaches to chronic care. Planned and minigroup visits are two potential solutions to
chronic care that may improve patient outcomes,
as well as promoting a patients ability to selfmanage their chronic illnesses.
Funding Source(s): RWJF
ƒ Measuring Diabetes Self-management:
Findings from the RWJF Diabetes Initiative
Lauren McCormack, Ph.D., Carla Bann, Ph.D.,
Pam Williams-Piehota, Ph.D., Joseph Burton,
M.S., Claudia Squire, M.S., Doug Kamerow,
M.D., M.P.H.
Presented by: Lauren McCormack, Ph.D., SSSHSER-PHE-HC, RTI International, 3040
Cornwallis Road, Research Triangle Park, NC
27709-2194, Phone: (919) 541-6277, Fax: (919)
990-8454, Email: lmac@rti.org
Research Objective: Assessing lifestyle changes
and support? for chronic illness care requires the
use of standardized indicators, and the
complexity of self-management in diabetes
poses special challenges for such assessment.
Measures must have content validity, empirical
support, be feasible to implement in a variety of
settings, and be understood by those with
various levels of health literacy. Here we describe
a new measure: ‘‘Resources and Supports for
Self Management’’ (RSSM), developed as part of
the Robert Wood Johnson Foundation’s
Diabetes Initiative. We describe the measures
developed and their association with the
behavioral and clinical outcomes that the
fourteen interventions sought to improve.
Study Design: We administered three waves of
a cross-site, telephone survey of participants that
sought to measure whether and how the
interventions funded contributed to participants’
perceived access to resources and supports for
self-management (RSSM). We also measured
key lifestyle behaviors including diet, exercise,
smoking, and medical management regimens.
RSSM indicators included (1) individualized
assessment, (2) collaborative goal setting, (3)
communication, (4) skills, and (5) provider
follow-up. These reflect a combination of items
from existing survey instruments and newly
developed items that fill important gaps and
ensure a patient-centered focus. We conducted
both cross-sectional and longitudinal analyses.
Population Studied: Programs funded under
the Robert Wood Johnson Foundations diabetes
initiative were designed to advance diabetes self
management and promote community supports
for diabetes. They focused on one of more of the
following: improving nutrition, increasing
physical activity, use of lay health workers or
promotoras, group medical visits, and support
groups. Over 700 [check] participants with
diabetes who participated in one of 12
interventions different interventions completed
at least two waves of the survey.
Principal Findings: The scale and its 18 items
demonstrated good psychometric properties.
Confirmatory factor analysis supported creation
of five subscales, representing the five
components of RSSM.
The overall scale and 5 subscales were internally
consistent Participants reporting higher levels of
RSSM checked their blood sugar and feet more
frequently, had better diet and nutrition
behaviors, and were more likely to be physically
active.
Conclusions: The RSSM scales developed are
shown to predict self-management behaviors
related to diet, exercise, and medical regimens.
In addition, the findings suggest that programs
to promote self-management skills may lead to
improved health behaviors which, in turn, can
lead to improved clinical outcomes. Messages
for improving self-management can be targeted
at patients who are less likely to engage in a
healthier lifestyle.
Implications for Policy, Delivery or Practice:
The RSSM is a promising new instrument. Its
ecological orientation and link to the Chronic
Care framework confer potential for broader
application including the potential to serve as a
benchmark for national performance
measurement. In addition, the RSSM constructs
offer guidance and lessons learned for designing
and improving care processes for chronic illness.
Funding Source(s): RWJF
ƒ Synthesizing Lesson Learned from Fourteen
Diabetes Initiatives
Joe Burton, M.S.
Presented by: Joe Burton, M.S., Researcher, RTI
International, 1440 Main Street, Waltham, MA
02451, Phone: (781) 434-1710, Fax: (781) 4341701, Email: joeburton@rti.org
Research Objective: The Diabetes Initiative of
the Robert Wood Johnson Foundation includes
14 projects in primary care settings and in
community-clinical partnerships. The primary
purpose of the Initiative was to demonstrate
feasible and sustainable approaches to
promoting diabetes self management in primary
care and community settings. Projects serve
predominantly indigent populations from varied
cultural and linguistic backgrounds in urban,
rural, and frontier settings. This paper describes
the Initiative, its ecological perspective on self
management, and implications for program
development and sustainability in primary care
and community settings. It also discusses
methodological challenges in assessing multisite interventions.
Study Design: The evaluation of the initiative
includes case studies, the collection of clinical
data, and a three-wave participant survey.
Grantees conducted internal evaluations,
participated in work groups, and hosted site
visits from both the National Program Office and
an external evaluator.
Population Studied: Participants in 14 different
interventions designed to advance diabetes self
management and promote community supports
for diabetes.
Principal Findings: Several of the Initiative’s
grantees observed improvements in key selfmanagement behaviors and clinical indicators
among engaged patients over time. Overall,
reported rates of physical activity were higher
than national levels. Greater community
resources for physical activity were associated
with higher levels of physical activity. Individuals
who were older, more confident, less depressed,
ate most of their meals at home, or had greater
program participation tended to eat healthier.
Participants reported very high levels of
perceived support from their diabetes care
teams. Programs were able to adapt and evolve
based on sharing ideas through a learning
network and active workgroups.
Conclusions: Ecological perspectives stress
varied levels of influence ranging from
individuals to communities and policies. The
Initiative identified key components of Resources
and Supports for Self Management
(Individualized Assessment, Collaborative Goal
Setting, Enhancing Skills, Follow Up and
Support, Community Resources, and Continuity
of Quality Clinical Care). These constructs are
predictive of self-management. Lessons learned
include the central roles of Community Health
Workers, integration of healthy coping and
attention to negative emotion and depression in
self management, community partnerships,
approaches to ongoing follow up and support,
organizational factors in sustaining programs,
and the utility of a "Collaborative Learning
Network" for program development.
Sustainability stresses organizational and policy
supports for program.
Implications for Policy, Delivery or Practice:
The Diabetes Initiative demonstrates that
effective self management programs and
supports can be implemented in real-world
clinical and community settings, providing
models of worthwhile, sustainable programs.
Funding Source(s): RWJF
ƒ Ongoing Follow Up and Support in
Diabetes Management and the Critical Role of
the Community Health Worker
Edwin Fisher, Ph.D., Carol Brownson, Ph.D.,
Mary O'Tolle, Ph.D., Victoria Anwuri, M.A.,
Gowri Shetty, Ph.D., Kia Davis, Ph.D.
Presented by: Edwin Fisher, Ph.D., Professor
and Chair, Department of Health Behavior and
Health Education, University of North Carolina
at Chapel Hill, Campus Box 7440, Chapel Hill,
NC 27599-7440, Phone: (919) 966-6693,
Email: fishere@email.unc.edu
Research Objective: Ongoing follow up and
support is a critical determinant of sustained
behavior change across many areas of health
and health behavior, including chronic disease
and diabetes management.
Study Design: The study included collaboration
with grantees to identify key features of ongoing
follow up and support, key approaches to
providing these, and the contributions of
Community Health Workers to this.
Population Studied: Grantees, programs, and
program participants of the Diabetes Initiative of
the Robert Wood Johnson Foundation.
Principal Findings: From discussions with
grantees of key features of ongoing follow up
and support, key characteristics include: (a)
being centered on a personal relationship with a
member of the diabetes care team, (b)
availability and accessibility as needed or desired
by the participant, as opposed to according to
limited schedules or venues, (c) including
opportunities for engagement that do not
require high levels of investment to behavior
change by participants, (d) proactive
maintenance of contact with participants (at
least every 3 ---- 4 mos) to identify changes in
status or needs for further care and to provide
sense of continuity of contact and concern by the
care team, (e) unrestricted in time and topics
discussed and responsive to the needs and
interests of the participant as opposed to being
limited by protocol. Across these features
emerges the importance of choices for
participants in the ways in which they maintain
contact with the care team and are monitored by
them. Grantees have provided these key
features, including participant choice, through a
wide variety of approaches, ranging from open
events and activities such as ‘‘breakfast clubs,’’
culturally specific activities such as inclusion of
diabetes issues in ‘‘Talking Circles’’ in American
Indian settings, to contact of patients not seen
for over several months by teams of Community
Health Workers and nurses. Community Health
Workers make major contributions to ongoing
follow up and support by being available to
participants, having time to talk about diverse
topics, coaching participants on adapting
management plans to their individual living
circumstances, providing emotional support,
and facilitating ongoing linkage between the
individual and the rest of the diabetes care team.
Conclusions: Ongoing follow up and support
for diabetes management entails a range of key
features and the importance of participant
choice and, correspondingly, requires diverse
programmatic approaches. Community Health
Workers can make substantial contributions to
these types of services.
Implications for Policy, Delivery or Practice:
Practice settings can arrange a variety of
activities and types of services to provide
ongoing follow up and support for diabetes
management. The range and inclusion of
choices among diverse approaches may be as
important as any single program component.
Including Community Health Workers in the care
team can facilitate implementation of this range
of types of service and contact.
Funding Source(s): RWJF
Call for Panels
Monitoring the Healthcare Safety Net at the
State and Local Levels
Chair: John Billings, J.D.
Tuesday, June 5 • 10:45 a.m.-12:15 p.m.
Panel Overview: The healthcare safety net in the
United States is crucial to the health of
underserved populations and the financial
stability of the health sector. The nation’s large
and growing uninsured population combined
with the lack of consensus on how to reduce the
number of uninsured highlights the importance
of maintaining a strong and sustainable safety
net. The proposed research panel will present a
broad range of perspectives on surveillance and
improvement of safety net services at the state
and local levels. Presentations representing four
diverse regions of the United States will describe
recent efforts to collect and standardize data,
quantify the supply and demand for safety net
services, define the boundaries of safety net care,
and use data to improve service delivery. The
following overlapping themes will be addressed:
coordination of multiple health and social
services, definition of safety net concepts,
determination of service areas, data assembly
and dissemination, qualitative versus
quantitative information, and partnerships with
consumers of research. In addition to presenting
findings from their research, presenters will
address challenges that were encountered and
what actions are needed to better meet these
challenges in the future.
ƒ Use of Charity Care Records to Optimize
Care for the Uninsured
Derek DeLia, Ph.D.
Presented by: Derek DeLia, Ph.D., Assistant
Professor/Senior Policy Analyst, Rutgers Center
for State Health Policy, 55 Commercial Ave, New
Brunswick, NJ 08901-1340, Phone: (732)9324671, Fax: (732)932-0069,
Email: ddelia@ifh.rutgers.edu
Research Objective: Use hospital administrative
records in New Jersey (NJ) to document the
healthcare needs of the low-income uninsured
and highlight areas where cost management
strategies may be successfully employed.
Study Design: Charity care claim records are
used to determine the most prevalent and costly
services provided to charity care patients.
Concentration of costs and potentially avoidable
utilization are analyzed.
Population Studied: Users of hospital-based
charity care in NJ in 2003.
Principal Findings: Prenatal care and delivery
are the most common services provided to NJ’s
charity care population. Mental health and
substance abuse services are also heavily used
by this population. A wide variety of other charity
care services are provided including treatment
for circulatory disorders, hypertension, and
diabetes. Charity care costs are highly
concentrated among a small number of patients
with 50% of costs attributable to 5% of patient
encounters (i.e., outpatient visits plus inpatient
admissions). Since high-cost patients are
expected to have multiple encounters, the costs
are likely to be even more concentrated among
individuals. Finally, 10% of charity care costs are
attributable to emergency department visits and
inpatient admissions for conditions that may
have been avoided with more timely access to
primary care (e.g. episodes of asthma,
congestive heart failure).
Conclusions: The scope of services used by NJ’s
charity care population reveals a wide safety net
held by hospitals across the state. Nevertheless,
a number of charity care services provided by
hospitals might be redirected or better managed
to improve patient care or reduce costs. For
example, many pregnant women currently
relying on charity care may be eligible for
Medicaid or the state’s Family Care Program. In
addition, the prevalence of utilization for mental
health, substance abuse, and primary care
sensitive services suggests that many of these
medical episodes could be prevented with earlier
intervention. The concentration of costs within a
small number of users points to an area where
targeted management of high-cost cases may
generate substantial savings.
Implications for Policy, Delivery or Practice:
Study findings are especially timely as state
officials in NJ are currently considering ways to
restructure the hospital sector and reduce the
costs of charity care subsidies. The findings
highlight the need to thoroughly screen
individuals who may be eligible for coverage
before funding their services through charity
care. A great deal of potential exists for reducing
costs through focused management of high-cost
uninsured patients. Better coordination of
services for the uninsured outside of the hospital
such as mental health, substance abuse, and
primary care may improve outcomes and
efficiency as well. This coordination may be
challenging as charity care users include a
complicated mix of individuals who are
employed, unemployed, homeless, transient,
undocumented, and moving in and out of the
criminal justice system. An important barrier to
using the charity care claims database as a care
management tool is its lack of identifiers for
patient race/ethnicity, immigration status, area
of residence, and repeat utilization. However,
these identifiers, or reasonable proxies, may be
added to charity care records with little
incremental cost.
Funding Source(s): Johnson & Johnson
ƒ Challenges of Urban Safety Net Evaluation
Glenn Landers, M.B.A., M.H.A.
Presented by: Glenn Landers, M.B.A., M.H.A.,
Senior Research Associate, Georgia Health Policy
Center, 14 Marietta Street, Suite 221, Atlanta, GA
30303, Phone: (404) 463-9562,
Email: glanders@gsu.edu
Research Objective: Evaluate the supply of and
demand for safety net health care services in
seven metropolitan Atlanta counties and identify
challenges and opportunities in data collection.
Study Design: Secondary analysis of individual
characteristics, community level variables, and
health care access and outcomes for seven
counties. Primary data collection and analysis
using survey, administrative, and interview data
from 35 safety net organizations representing 101
service delivery sites.
Population Studied: Safety net health care
delivery sites in seven metropolitan Atlanta
counties, defined as:
o Public hospitals that participate in the
Disproportionate Share Hospital program (DSH)
o Hospital-based clinics
o Federally Qualified Health Centers (FQHC)
o Community Health Centers (CHC)
o Independent clinics or Freestanding Primary
Care Clinic, including faith-based clinics
o County health departments
o Community Service Boards and affiliated
mental health providers
o Non-traditional providers, including those that
focus on specific diseases, conditions, or
populations
o Individuals not covered by public or private
insurance.
Principal Findings: The determination of the
supply of and demand for health care safety net
services at the community or regional level is
challenged by:
1. Lack of common data definitions and
reporting across service providers.
2. The potential shortcomings of self-reported
data.
3. The lack of reliable estimates of the uninsured
at the county or community level.
4. Administrative idiosyncrasies across service
sites.
5. The time and effort involved in primary data
collection.
A qualitative study of community safety net
providers can surface potential partnerships,
collaborations, and joint ventures that might not
otherwise be evident.
Analysis across communities can also point to
common needs (root causes) such as stable
housing and employment opportunities not
apparent in publicly available data sets.
Conclusions: In spite of the challenges
encountered in measuring the supply of and
demand for health care safety net services, the
information gained from such an effort can
contribute to a greater understanding of
community-specific needs and opportunities.
Implications for Policy, Delivery or Practice:
Safety net providers are focused on their core
business of health care delivery, not the accurate
and reliable collection of administrative data that
might be used to create system-wide efficiencies
and promote collaborative partnerships. There
are, however, important opportunities that might
be exploited for safety net populations through
the coordination and alignment of safety net
services across providers within communities or
regions. Such opportunities might include
economies of scale or more efficient referral
patterns. Toward the better coordination of
safety net services, states have an interest in
encouraging partnerships and promoting
efficiencies. Such partnerships and efficiencies
can be supported through the establishment of
standardized data collection and reporting
requirements that enable reliable measurement
of safety net service supply and demand. States
might consider supporting sustainable safety net
monitoring systems that enhance the capacity
and efficiency of individual providers and that
promote partnership and collaboration across
communities.
Funding Source(s): Community Foundation of
Greater Atlanta
ƒ Community Data Needs Assessment for
Monitoring Portland’s Health Care Safety Net
Jennifer Lin, M.D.
Presented by: Jennifer Lin, M.D., General
Internal Medicine Fellow, P3-Health Services
Research Division, Portland VA Medical Center,
P.O. Box 1034, Portland, OR 97207, Phone: 503220-8262, Fax: 503-273-5367, Email:
linje@ohsu.edu
Research Objective: This study was designed to
assess the information and data needs from
organizations comprising the infrastructure of
Portland’s health care safety net (HCSN). Our
objectives were to understand the current use of
available information and data to support
decision-making; to understand perceived
information and data needs as identified by
decision makers within the HCSN; and to
identify recent initiatives to address these needs.
Study Design: We purposively sampled
organizations from a Portland-based HCSN
umbrella organization. We conducted semistructured interviews to elicit information in
three domains which paralleled our objectives.
Interview findings were hand scribed and
analyzed for thematic content.
Population Studied: We interviewed 22
executive-level decision makers in clinic and
hospital systems, associations of primary care
and hospitals, funding agencies, advocacy and
support organizations, health plan
organizations, state planning agencies, and a
state advisory council; three directors of IT within
these organizations; and we reviewed 11
supporting documents.
Principal Findings: Everyone acknowledged our
deficit in understanding the health needs of
patients receiving care through the safety net
and the viability of the HCSN at the community
level. Those in decision-making roles who
routinely used external data sources did not find
them very useful (e.g. not available at the local
level, not current, not in a useful format, and no
support staff to conduct secondary data
analysis). Those who routinely used internal
data had well-staffed IT departments, but
infrequently shared their data to non-clients,
primarily due to lack of requests for their data.
Most people felt that there is wealth of potential
information and data, and an emerging regional
health information infrastructure. Many
expressed skepticism about using this data in
decision making because of current limitations,
including lack of: (1) transparency of existing
measures; (2) coordination and integration of
data (i.e. across health systems, state/county
borders, and sectors of health care); (3) policies
to address political, legal, and technical
challenges of sharing data; and (4) skilled staff
and monies to support data management and
analysis. Other limitations include: (1) data gaps
in specific health care sectors (i.e. dental, mental
health, social services, pharmacy and
prescription drug use), (2) data gaps in specific
populations (i.e. uninsured, migrant workers,
homeless), and (3) gaps and concerns in
measurement development (e.g. valid costbased measures, quality of care measures,
meaningful outcome-based measures, and the
importance of tracking qualitative measures like
local politics and leadership capabilities and
community dialogue within the HCSN).
We identified several routine and ad hoc survey
mechanisms for future data collection and
community and state-based initiatives to support
data sharing and integration, and measurement
development.
Conclusions: There appears to be a wealth of
information and data that could better inform
decision making, and potential leadership and
mechanisms in our community to foster data
collection, data coordination, and measurement
development.
Implications for Policy, Delivery or Practice:
This study highlights many priority concerns for
policy making and targeting solutions within our
HCSN and identifies underutilized existing
resources that should shape ongoing or future
initiatives.
ƒ Bridging State and Local Safety Net Data
Gaps: Strategies and Initial Results
Jeff Bontrager, M.S.P.H., Reid Reynolds, Ph.D.,
Pamela Hanes, Ph.D.
Presented by: Jeff Bontrager, M.S.P.H., Research
Analyst, Colorado Health Institute, 1576 Sherman
Street, Suite 300, Denver, CO 80203, Phone:
(303) 831-4200, Fax: (303) 831-4237,
Email: bontragerj@coloradohealthinstitute.org
Research Objective: To assess the capacity of
Colorado’s health care safety net system to meet
the basic medical, dental and mental health care
needs of Colorado’s most vulnerable
populations and to provide relevant, reliable and
timely information on which sound policy
decisions can be based.
Study Design: This project employs both
quantitative and qualitative methods over a
multi-year timeframe. Initial steps included
establishing operational definitions of the safety
net and vulnerable populations who use safety
net services. With the guidance of an advisory
committee of influential stakeholders, data gaps
were identified and criteria for selection of an
initial set of indicators were established. Data
sharing agreements were secured with various
safety net providers (e.g., federally qualified
health centers, rural health clinics, non-profit and
faith-based clinics, community mental health
centers and hospitals). A set of interactive Webbased tools is currently under development to
uniformly collect, analyze, and report safety net
indicators. Whenever possible, county and subcounty data were identified, although state-level
data were utilized when sub-state data were not
available. Concurrently, agreements were
secured within select Colorado communities to
participate in community case studies.
Population Studied: For this project, vulnerable
populations were defined as Medicaid and
SCHIP enrollees, as well as uninsured
Coloradans with family incomes below 250
percent of the federal poverty level.
Principal Findings: The first year of data
collection has revealed a complex array of safety
net providers distributed unevenly across the
state. Although these providers recognize the
importance of uniformly collected data about
safety net services and the populations that use
them, their ability to collect, analyze and report
patient-level data varies widely. Development of
a user-friendly Web-based data portal holds great
promise for this data collection effort. Building a
uniform database comprising patient
demographic information, access variables and
health status indicators is a critical starting point
for a safety net monitoring system. Preliminary
estimates indicate that increases in safety net
capacity have been roughly matched by growth
in the size of vulnerable populations. Other early
analysis that will be presented is the relationship
between regional variations in the incidence of
ambulatory care sensitive conditions and safety
net capacity.
Conclusions: Striking a balance between
qualitative and quantitative approaches, state
and local data and short-term and long-term
reporting strategies is essential to building
Colorado’s safety net monitoring system.
Establishing collaborative and mutually beneficial
data sharing agreements with safety net
providers has also assisted in addressing
development and implementation challenges.
Sub-state safety net data are most in demand by
policymakers and stakeholders but continues to
be the least available on a uniform basis.
Implications for Policy, Delivery or Practice:
With increasing interest among Colorado
policymakers for shoring up the state’s safety
net, the necessity of collecting, analyzing and
translating reliable data into policy-relevant
information is key among the project’s
objectives. Data on the characteristics of the
uninsured and those enrolled in state coverage
programs have already assisted policymakers to
better target policy options. The project will
continue to build a portfolio of communications
strategies for the dissemination of information
that can be utilized for informed policy decisionmaking.
Funding Source(s): Other Foundation,
Colorado Health Foundation
ƒ Monitoring the Safety Net at the Local
Level: Oregon’s Health Indicators Project
(HIP)
Heidi Allan, M.S.W.
Presented by: Heidi Allan, M.S.W., OHREC
Project Director, Office for Oregon Health Policy
& Research, State of Oregon, 255 Capitol Street
NE, 5th floor, Salem, OR 97310, Phone: (503)
378-2422, Fax: (503) 378-5511,
Email: heidi.allen@state.or.us
Research Objective: In Oregon, there is no
single source where safety net data is collected
or analyzed. Information on the safety net is
difficult to assemble and incomplete. Local
safety net initiatives struggle to describe gaps in
services, to measure the success or failure of
community interventions, and to compare
themselves with other communities or state
averages on measures of access. The Health
Indicators Project was developed to provide local
safety-net collaboratives with a common unit of
analysis for purposes of comparison and
monitoring, and to identify shared indicators of
access that communities could use utilizing
existing sources of data.
Study Design: Safety-net collaborations in
Oregon’s urban areas were contacted to
participate in focus group discussions with the
goal of achieving agreement on defining
common-sense geographic units of analysis and
a set of indicators measuring health care access.
A series of focus groups were held in five urban
areas.
Population Studied: Participants were
individuals who had participated in 100% safetynet access collaborations and represented a
variety of public and private stakeholder
organizations.
Principal Findings: Defining the geographic unit
of analysis required several meetings in each
community to problem-solve issues related to
changing demographics. Ultimately, the
participants developed Primary Care Service
Areas (PCSA) that represented the five urban
areas and aligned with the rural PCSAs to create
a state-wide unit of analysis representing
contemporary and common-sense boundaries
for local safety-net systems. Multiple shared
indicators measuring access were identified from
existing data sources and agreed upon by
participants. Utilizing existing data sources
presented several challenges: 1) it is difficult to
define access within the limitations of the
indicators available for analyses, there is rarely a
denominator including individuals who are not
utilizing services; 2) there is little data granular
enough for analysis at the PCSA level; 3) there
were no good proxies of access to outpatient
mental health and preventative dental services;
and 4) race data is not widely available.
Conclusions: The Health Indicators Project has
provided safety-net communities with a common
unit of analysis for comparisons and a set of
shared indicators of health care access. Ongoing monitoring of the indicators could assist
safety-net collaboratives in using limited financial
and data capacity resources in an efficient
progression towards identified objectives by:
o Accurately defining and quantifying the
problem at the level of concern (PCSA, local,
regional, or state).
o Utilizing baseline measures to evaluate the
relative success of interventions.
o Applying for funding to address unmet needs.
Implications for Policy, Delivery or Practice: In
order to truly maximize the benefits of the
Health Indicator Project, limitations in the
available data sources to measure health access
should be addressed, particularly related to
identifying demand (including foregone health
care) and disparities. Developing a coordinated
and targeted approach to understanding health
care access at the local, regional, and state level
will require continued collaboration between
state and safety-net community collaboratives.
Funding Source(s): HRSA
Call for Panels
Patient Outcomes and Costs Related to Nurse
Staffing
Chair: Chuan-Fen Liu, Ph.D., M.P.H.
Tuesday, June 5 • 1:00 p.m.-2:30 p.m.
Panel Overview: In this panel, we present
results of recent, large-scale studies of nurse
staffing, patient outcomes and costs of care at
124 medical centers in the Veterans Health
Administration, Department of Veterans Affairs
(VA). Previous studies of nurse staffing and
patient outcomes have focused primarily on
facility-level analyses, in part due to lack of data
at the unit level. Using detailed data from VA
administrative databases, we were able to
examine associations between nurse staffing,
nursing skill mix and patient mortality at the unit
level as well as the facility level. Using survey
data on reorganization of VA hospitals during
the past decade, we examined the effect of
nursing service organization (nurse executive
line authority for nurse staffing) on patient care
costs. Our results suggest that nurse staffing
and skill mix have different effects on patient
mortality for patients with ICU stays versus those
without any ICU stay. While nurse staffing levels
and market variables affect patient care costs,
nursing service organization measured as line
authority for nurse staffing does not appear to be
associated with patient care costs. We interpret
these results in the context of a wide range of
studies on the relationships between nurse
staffing, skill mix, nursing work environment,
and patient outcomes such as length of stay and
hospital-acquired complications, with attention
to data and methods used in previous research.
ƒ The Association Between Nursing Factors &
Patient Morality in the Veterans Health
Administration: The View from the Nursing
Unit Level
Presented By: Anne Sales, Ph.D., M.S.N.,
Associate Professor, University of Alberta, 3-114
Clinical Sciences Building, Edmonton, Alberta
T6G 2G3, Canada, Phone: (780) 492-8432, Fax:
(780) 492-2551, Email: anne.sales@ualberta.ca
Abstract: Nurse staffing is not constant across
an entire hospital. Most previous studies of the
association of nurse staffing with in-hospital
patient mortality have used data aggregated to
the facility level, due to limitations of the
available data. In this study, we present findings
from the first large-scale study of in-hospital
patient mortality and its association with
registered nurse staffing and skill mix at the unit
level. We included patient-level data from all
inpatient admissions to acute care (intensive
care and non-intensive care) units in 124
Veterans Health Administration Medical Centers
within the United States Department of Veterans
Affairs (VA). Staffing and skill mix variables at
the unit level came from VA national accounting
and payroll databases. We obtained patient inhospital mortality outcomes and other
characteristics for risk adjustment from VA’s
National Patient Care Databases. We used
hierarchical multi-level multiple regression
models to adjust for patient, unit, and hospital
characteristics. Estimating results at the facility
level, even with adjustment for facility level
clustering, yielded more highly significant results
for both RN hours per patient day and RN skill
mix than when results were estimated at the unit
level, with clustering accounted for at both unit
and facility level. Splitting the sample among
patients with any and no ICU stay during their
hospitalization resulted in differing effects of RN
staffing and skill mix across these two patient
populations. Our findings suggest that the
association between RN staffing and skill mix
and in-hospital patient mortality is not as clear
as it appears when facility-level data are used.
ƒ Nurse Staffing & Patient Outcomes in
Hospitals: Current Findings & Future
Directions
Presented By: Jack Needleman, Ph.D., Associate
Professor, University of California, Los Angeles,
650 Charles Young Drive S., Room 31-236B CHS,
Los Angeles, CA 90095-1772, Phone: (310) 2672706, Email: needlema@ucla.edu
Abstract: The Department of Veterans Affairs
(VA), a federal agency of the United States
government, has implemented extensive
organizational changes to improve efficiency of
inpatient care since the mid-1990s. One
reorganization strategy, implemented during the
period 1996-2000, was the use of
interdisciplinary patient care teams within
service line structures, which in turn affected
nursing supervision and reporting structures.
Although Service Lines were intended to provide
more efficient and economical patient care,
shifting control of nurse staffing decisions to an
interdisciplinary service line also had the
potential to decrease efficiency by creating new
barriers to cross-service planning for staffing and
to reassignment of nursing staff from low to high
census areas. This study examines the
relationship between nurse executive line
authority for nursing staff and patient care costs
between 02/2003 and 06/2003 at 124 acute care
VA hospitals. The unit of analysis is the hospital.
We assessed nurse executive line authority for
nurse staffing based on information from the VA
Office of Nursing Services and a survey of nurse
executives. Other data sources were VA
Decision Support System inpatient extracts
(patient care costs and nurse staffing),
administrative databases (patient characteristics
and health outcomes); and national databases
(market/health service area characteristics). The
dependent variable was the total patient care
cost per bed day of care for each hospital.
Among 125 hospitals, 87 had Nurse Executive
line authority for nurse staffing in 2003.
Compared to hospitals without Nurse Executive
line authority for nurse staffing, these 87
hospitals had lower patient care costs per bed
day of care ($258 versus $283, p=0.058), fewer
registered nurse hours per patient day (4.2 hours
versus 4.7 hours, p=0.083), and fewer total
nursing hours per patient day (7.2 hours versus
7.9 hours, p=0.033). After controlling for nurse
staffing and patient, facility, and market area
characteristics, the relationship between patient
care cost and line authority for nurse staffing was
not statistically significant. The factors
significantly associated with high patient care
costs were higher nurse staffing levels (RN and
non-RN hours) and the RN wage index,
comparing costs of nursing wages across
different markets. Our study results provide no
evidence that moving to a service line
organizational structure offers cost efficiency,
nor evidence that there are inefficiencies
associated with this organizational change.
ƒ The Relationship Between Line Authority for
Nurse Staffing & Patient Care Cost
Presented By: Chuan-Fen Liu, Ph.D., M.P.H.,
Research Health Scientist, VA Puget Sound
Health Care System, 1100 Olive Way, Seattle, WA
98101, Phone: (206) 764-2587, Fax: (206) 7685343, Email: fliu@u.washington.edu
Abstract: Current research on nurse staffing and
patient outcomes in hospitals, using a wide
range of data sets and methods, finds strong
evidence that nurse staffing levels and skill mix
influence a variety of outcomes: length of stay,
hospital-acquired infections and other
complications, pain management, and patient
satisfaction. There is strong but less conclusive
evidence for the association of nurse staffing and
mortality. Other studies examine the association
of other dimensions of the nurse work
environment, such as use of overtime and
professional autonomy, on patient outcomes.
Estimates of the costs of increasing nurse
staffing levels to those of the best staffed
hospitals fall within a range of 1-3% of hospital
costs, but estimates of the offsetting cost
savings from reduced length of stay and avoided
complications, reduced staffing turnover, and
other impacts vary more widely. We review the
literature on nurse staffing and patient
outcomes, with particular attention to the data
and methods used and the limitations and
challenges these impose on the ability to study
these associations; identify issues of estimating
the costs and cost offsets of increasing nurse
staffing; and discuss directions for future
research to resolve current conflicts.