Call for Panels
Call for Panels
Investigating the Factors that Influence Hospitalization for
Chronic Medical Conditions
Chair: Nancy McCall, Sc.D.
Sunday, June 6 • 9:30 a.m.- 11:00 a.m.
• Are Changing Rates of Admission for Chronic Medical
Conditions Simply a Reflection of Changes in the
Demographics, Health Status, and Geographic Migration
Patterns of the Elderly?
Nancy McCall, Sc.D., Lee Mobley, Ph.D., Erica Brody, MPH,
Sujha Subramanian, Ph.D., Mary Kapp
Presented by: Nancy McCall, Sc.D., Principal Scientist,
Division of Health Economics Research, RTI International,
1615 M Street NW, Suite 740, Washington, DC 20036; Tel:
202.728.1968; Fax: 202.728.2095; E-mail: nmccall@rti.org
Research Objective: Previous research has shown that
beneficiary factors such as age, race, Medicaid status, health
status, and geographic area are related to cross-sectional
variation in rates of admission for medical conditions
commonly referred to as Ambulatory Care Sensitive
Conditions (ACSCs). This paper examines the influence of
changes in sociodemographic, health status, and geographic
location characteristics among elderly Medicare beneficiaries
on the rate of growth of admissions for two chronic disease
ACSCs, congestive heart failure (CHF) and chronic obstructive
pulmonary disease (COPD).
Study Design: Annual rates of inpatient admissions and rates
of ER/Observation Bed use for the selected conditions and
median health status of the Medicare FFS population as
measured by the Principal Inpatient Diagnostic Cost Group
(PIP-DCG) were constructed using 1992-2000 Medicare
claims data files. Annual estimates of the proportion of the
Medicare population with specific attributes of interest (e.g.,
% Black, % died during the year) and residing in geographic
areas were estimated from the 1992-2000 Medicare
Denominator Files. All data were aggregated to urban and
rural places within states yielding two observations per state.
A dummy variable representing each study year was included
in the models to capture the time trend. Pooled crosssectional time series multivariate regression models were
estimated with SYSTAT to produce heteroskedastic-consistent
standard errors.
Population Studied: Medicare FFS beneficiaries age 65 and
older, enrolled in Medicare Part A and B during the full year.
Beneficiaries who died during the year but would have
otherwise qualified were included in the analysis. The total
number of study participants was 27 million in 1992 and
decreased to 25 million in 2000, reflecting a shift from FFS to
managed care within the Medicare program.
Principal Findings: In 1992, the rate of admission was
22.7/1,000 Medicare FFS beneficiaries for CHF and 7.6/1,000
for COPD. The hospitalization rate remained essentially
unchanged between 1992 and 2000 for CHF. In contrast, the
rate of admission for COPD increased 52% by 2000.
Significant evidence of different trend processes in urban and
rural areas were found necessitating separate urban/rural
modeling. Including explanatory variables such as age
distribution and other sociodemographic characteristics and
health status measures over time in the models eliminated
time trends observed in the raw rates. However, after
controlling for changes in beneficiary characteristics
significant unexplained spatial variation in ACSC rates
remained.
Conclusions: Much of the variation observed in the rate of
admission for CHF and COPD over time can be explained by
changes in sociodemographic, health status, and geographic
location characteristics among elderly Medicare beneficiaries.
However, the models also illustrate the importance of
accounting for spatial variation.
Implications for Policy, Delivery or Practice: Although
beneficiary characteristics appear to significantly influence the
rate of admissions observed over time, the remaining spatial
variation suggests that other factors not examined in this
paper play an important role in admission rates and warrant
further investigation.
Primary Funding Source: CMS
• Does Access to Usual Source of Care and Supplemental
Insurance Prevent Hospitalization for Chronic Medical
Conditions among the Elderly?
Sujha Subramanian, Ph.D., Nancy McCall, Sc.D., Lee Mobley,
Ph.D., Erica Brody, MPH
Presented by: Sujha Subramanian, Ph.D., Health Economist,
Division of Health Economics Research, RTI International, 411
Waverley Oaks Road, Waltham, MA 02452; Tel: 781.788.8100
x149; Fax: 781.788.8101; E-mail: SSubramanian@rti.org
Research Objective: To investigate whether access to usual
source of care and supplemental insurance prevent
hospitalization for chronic conditions among Medicare FeeFor-Service beneficiaries.
Study Design: We obtained information on patient
characteristics from the 1999 Medicare Current Beneficiary
Survey (MCBS) and derived admissions related to Ambulatory
Care Sensitive Conditions (ACSCs) from 2000 Medicare
inpatient claims. Since the information on access to care and
usual source of care information was collected in the last
quarter in 1999 in the MCBS, these variables should serve as
an indicator for ACSC admissions in 2000. We examined two
frequently occurring chronic conditions among the elderly,
congestive heart failure (CHF) and chronic lung disease
(Asthma/COPD), which were identified using ICD-9 codes.
Logistic regressions were estimated separately for each
condition and the independent variables included were
sociodemographics, health status, insurance status (duals,
supplemental insurance, prescription drug coverage), length
of association with usual source of care, propensity to seek
care, unmet need, urban/rural residence, and prior year (1999)
admission for specific ACSCs. All analyses were performed in
SUDAAN and odds ratios at the 5% or lower level were
considered as statistically significant results.
Population Studied: All 65 years or older community
residents who were continuously enrolled in fee-for-service
Medicare in 2000 were identified from the 1999 MCBS Access
to Care file. Among the sample of 8,240 beneficiaries
analyzed we observed that 1.9% were admitted for CHF and
1.1% for chronic lung disease.
Principal Findings: Length of association with usual source
of care and unmet need were not significant in both sets of
regressions. Propensity to seek care for CHF, and
supplemental insurance for chronic lung disease both tended
towards being significant. Beneficiaries with lower selfreported health status were 3 to 4 times more likely to have an
admission. In addition, beneficiaries with larger number of
chronic conditions were more likely to have ACSC admissions.
Prior hospitalization for an ACSC in 1999 increased the
likelihood of another admission in 2000 by about 11 times for
CHF and 37 times for chronic lung disease.
Conclusions: Previous admission for ACSC is the strongest
predictor of future ACSC admissions for the conditions
studied. Surprisingly, we did not identify any consistent
pattern to conclude that access to a usual source of care and
supplemental insurance, including prescription drug coverage,
lowers the probability of a preventable hospitalization.
Implications for Policy, Delivery or Practice: Overall, the
analyses presented raises important questions concerning the
lack of impact of usual source of care in reducing admissions
for chronic conditions and the role of prior admissions in
predicting future ones. Targeting hospitalized individuals for
close monitoring such as via disease management programs
after discharge may help reduce the rate of subsequent
admissions.
Primary Funding Source: CMS
• Do Ambulatory Care Sensitive Conditions Affect
Beneficiaries’ Experience and Satisfaction with Health
Care?
Bernard Shula L., Ph.D., R.N., Brody Erica, M.P.H.
Presented by: Brody Erica, MPH, Health Policy Analyst,
Division of Health Economics Research, RTI International,
3040 Cornwallis Road, Cox 176, Research Triangle Park, NC
27709; Tel: 919.541.2788; E-mail: ebrody@rti.org
Research Objective: To examine whether the incidence of
ambulatory care sensitive conditions are associated with self
reports of poor access to care or dissatisfaction with health
care services.
Study Design: We used data from the 2000 National
Medicare Fee-for-Service (MFFS) CAHPS survey, conducted
for the Centers for Medicare and Medicaid Services (CMS).
The sample of beneficiaries, drawn from a sampling frame
constructed from the CMS Enrollment Data Base (EDB),
resulted in 103,551 (64%) completed surveys. MFFS survey
data were merged with 1999-2001 Medicare claims data.
Using ICD-9 codes, and data from the twelve months
preceding the survey response date, we constructed a
measure of Ambulatory Care Sensitive Conditions (ACSC). In
addition, we estimated case-mix adjusted means for 5 CAHPS
measures using the CAHPS Macro version 3.4. The case-mix
adjusted means were stratified by the ACSC indicator and twosample tests for mean differences were performed. Due to
large sample sizes the Gaussian distribution was assumed.
Population Studied: Beneficiaries enrolled in the Medicare
Fee-for-Service program in 2000.
Principal Findings: We found that Medicare beneficiaries
who had a claim for an inpatient, observation stay or
emergency room visit for an ACSC were more likely to report
problems with getting needed medical care as measured by
the CAHPS Needed Care Composite. We did not find any
relationship between ACSC and the likelihood of problems
with the Getting Care Quickly or the Communication
Composites. There was no relationship between having an
ACSC and ratings of satisfaction with Medicare or health care.
Conclusions: Our findings suggest that Medicare
beneficiaries who report access problems with getting needed
care also experience higher rates of ACSC. The lack of a
similar finding for the Getting Care Quickly composite is
puzzling. In addition, beneficiaries who had an ACSC do not
differ in their reports of satisfaction from beneficiaries who did
not have an ACSC, suggesting that ACSC do not impact
negatively on more attitudinal measures of quality or that
beneficiaries do not attribute an ACSC to the plan
performance.
Implications for Policy, Delivery or Practice: The findings
suggest that patients reporting barriers to obtaining needed
care do in fact have a higher rate of adverse medical events.
Further, this analysis suggests that while some reports are a
direct result of the quality of care received and may be used to
monitor clinical quality, satisfaction with care is not associated
with this clinical measure of quality and access.
Primary Funding Source: CMS
• Spatial Analysis of Healthcare Markets: Separating the
Signal from the Noise in Ambulatory Care Sensitive
Condition Admission Rates
Mobley Lee, Ph.D., Elisabeth Root, M.P.H., Nancy McCall,
Sc.D., Sujha Subramanian, Ph.D.
Presented by: Mobley Lee, Ph.D., Health Economist, Division
of Health Economics Research, RTI International, 3040
Cornwallis Road, Cox Building, Research Triangle Park, NC
27709; Tel: 919.541.7195; E-mail: lmobley@rti.org
Research Objective: For decades, geographic approaches
and spatial analytic methods have been used in health
services research. Building on that approach and a unique GIS
database, we critically examine the influence of geographic or
market-level supply and demand factors on rates of
Ambulatory Care Sensitive Conditions (ACSCs) among
Medicare Fee-for-Service (FFS) beneficiaries. But more
critically, our varying-parameter spatial model allows us to
directly examine the influence of two recent policy initiatives,
implementation of home health and skilled nursing facility
(SNF) payment reform, on the supply of home health and
SNF services and the rate of ACSC admissions.
Study Design: Multivariate analyses are being conducted to
examine beneficiary characteristics, demand, supply and policy
factors that may influence the probability of being admitted for
one of the 3 selected chronic disease ACSCs: Chronic
Obstructive Pulmonary Disease, Congestive Heart Failure and
Peripheral Vascular Disease. ACSCs were aggregated by
Hospital Referral Regions (HRRs) using Geographic
Information Systems (GIS). The HRR is a more relevant
market than state or county for analysis of ACSCs because it
better reflects actual flows of patients within markets (which
often cross state and county boundaries). Defining the unit of
observation this way in a multivariate analysis explicitly
recognizes the patterns of health care utilization have an
explicit geographic configuration, thereby defining relevant
markets. Data on Medicare beneficiaries (using CMS claims
and enrollment files) are combined with information on
general population characteristics, rural vs. urban
designations, provider supply, competitive market factors,
managed care market penetration and provider
characteristics. The use of spatial modeling allows for a more
in-depth examination of the nature of the changes in ACSC
hospitalization rates than has previously been conducted.
Linear regression results from two cross sections, before and
after implementation of two payment reform initiatives in
1997, are compared with a pooled sample regression to
determine whether parameter changes across time are
meaningful and/or statistically significant.
Population Studied: The Medicare population continuously
enrolled in Medicare FFS Part A and B for one year between
1995 and 2000.
Principal Findings: Preliminary findings show substantial
variation in ACSC rates and their determinants across
geographic regions. Analyses of the influence of two key
Medicare payment policy reforms on ACSC rates of admission
are ongoing but will be completed within the next month and
will be available for reporting at Academy Health.
Conclusions: Preliminary findings suggest that rates of
admission for ACSCs may well be influenced by factors
beyond the control of practitioners within the clinical setting.
Implications for Policy, Delivery or Practice: If evidence that
market factor impacts vary with admission rates, then a
broader health policy approach may be necessary to effect a
reduction in admissions for chronic diseases such as those
examined in this paper.
Primary Funding Source: CMS
Call for Panels
Impacts of Incremental Public Health Insurance
Expansions
Chair: Barbara Lyons, Ph.D.
Sunday, June 6 • 3:00 p.m.- 4:30 p.m.
• Chip Shots: Association between the State Children’s
Health Insurance Programs (SCHIP) and Immunization
Rates
Ted Joyce, Ph.D., Andrew Racine, M.D., Ph.D.
Presented by: Ted Joyce, Ph.D., National Bureau of Economic
Research, Inc., 365 Fifth Ave., 5th Floor, New York, NY 10016;
Tel: 212.817.7960; Fax: 212.817.1597; E-mail:
ted_joyce@earthlink.net
Research Objective: To test whether SCHIP is associated
with differential gains in age-appropriate immunization rates
among poor and near-poor children relative to their non-poor
counterparts and whether the uptake of new vaccines among
poor and near-poor children was faster than would have been
observed in the absence of SCHIP.
Study Design: We use eight years of data (1995-2002) from
the recently released National Immunization Survey (NIS) to
associate changes in immunization with implementation of
SCHIP. The NIS has information on vaccine receipt,
characteristics of the providers as well as social and economic
measures of the household. Specifically, we compare changes
in up-to-date immunization status before and after
implementation of SCHIP in all 50 states and the District of
Columbia stratified by poor, near-poor and non-poor children.
If SCHIP has improved access to, and the quality of pediatric
care, then we would expect to observe relative improvements
in up-to-date immunization status among poor and near-poor
children relative to non-poor.
Population Studied: National Immunization Survey (NIS) is
a population-based survey of more than 30,000 households
per year with children between 19 and 35 months of age.
The NIS is conducted in all 50 states and the District of
Columbia.
Principal Findings: We show that the probability that a poor
or near-poor child is up to date for the 4:3:1:3:3 vaccine series
increased approximately 10 percentage points after SCHIP.
However, we observe a similar increase for non-poor children.
As to new vaccines, we demonstrate that uptake of the
varicella vaccine increases between 8 and 19 percentage
points more among poor and near-poor relative to non-poor
children after implementation of SCHIP. However, we present
evidence that the differential gains by poor and near-poor
children are related to the diffusion of new vaccines and not
specifically to SCHIP.
Conclusions: Our results suggest that the availability of
publicly provided health insurance for poor and near-poor
children may be a necessary but not a sufficient condition to
narrow the income gradient for immunizations.
Implications for Policy, Delivery or Practice: A fullyimmunized child now receives upwards of 19 shots over eight
visits in the first two years of life. To insure that poor and
near-poor children do not fall behind, we need to go beyond
policies directed at ability to pay and to concentrate perhaps
on the continuity of care, computerized registries and parent
reminder systems.
• The Effect of SCHIP Expansions on Health Insurance
Decisions by Employers
Thomas Buchmueller, Ph.D., Kosali Simon, Ph.D., Philip
Cooper, Ph.D., Jessica Vistnes, Ph.D.
Presented by: Philip Cooper, Ph.D., Senior Economist, Center
for Cost and Financing Studies, Agency for Healthcare
Research & Quality, 2101 East Jefferson Street, Suite 500,
Rockville, MD 20852; Tel: 301.594.2011; Fax: 301.594.2166; Email: pcooper@ahrq.gov
Research Objective: This research tests whether the State
Children's Health Insurance Program of 1997, a program that
extended free or heavily subsidized health insurance to
children in families under approximately 200% of the Federal
Poverty Level, changed private insurance outcomes at the
employer level. The outcomes of interest are: whether
employers continue to offer health insurance in low-wage
firms, whether employees continue to take-up coverage that is
offered, and whether employers change the generosity of
health insurance in any way.
Study Design: Specifically, our work looks at the effect of
eligibility expansions on
1) Employer decisions a) to offer health insurance to workers
in the low-wage sector b) to require cost sharing on the part of
employees in employee contributions
2) Worker decisions to a) take-up offered employer coverage
b) switch between family and single health insurance
coverage.
We do this by examining changes in insurance outcomes for
firms that we expect to be affected by SCHIP (those with a
substantial share of low-wage workers) relative to other firms.
Since there is variation in timing of SCHIP adoption and level
of generosity across states, we use this as a second source of
identification.
Population Studied: We examine this question using a data
set on employers that covers the 1996-2001 time period. The
data come from the Medical Expenditure Survey Insurance
Component List Sample (MEPS IC List). This is a large
nationally representative annual survey of employers (roughly
25,000 employers a year) which is conducted by the Census
Bureau and funded by the Agency for Healthcare Research and
Quality (AHRQ). This survey collects information about health
insurance plans offered by the employer, rich details such as
the premiums, the employee contributions, plan
characteristics such as cost sharing, and employer
characteristics. The survey is fielded to employers who offer
health insurance as well as those who do not.
Principal Findings: Our preliminary results are consistent
with the hypothesis that employers are encouraging workers
that are likely to be eligible for public subsidies to enroll
dependents in SCHIP rather than employer coverage.
However, the magnitude of the effect is small.
Conclusions: Our initial analysis indicates that employers are
not dropping coverage in response to SCHIP expansions,
however, they may be increasing premium cost-sharing.
Implications for Policy, Delivery or Practice: Policymakers
should be concerned that if states cut-back on SCHIP
expansions due to budget pressures and employers have
increased their cost-sharing provisions for health insurance in
response to previous levels of SCHIP generosity, low wage
workers could be left worse off.
• How Much Can Really Be Saved by Rolling Back SCHIP?
The Marginal Cost of Public Health Insurance for Children
Thomas Selden, Ph.D., Julie L. Hudson, Ph.D.
Presented by: Thomas Selden, Ph.D., Economist, Center for
Financing, Access, and Cost Trends, Agency for Healthcare
Research and Quality, 540 Gaither Road, Rockville, MD 20850;
Tel: 301.427.1677; Fax: 301.427.1277; E-mail: tselden@ahrq.gov
Research Objective: The State Children’s Health Insurance
Program (SCHIP) has helped to increase insurance coverage,
increase access to care, and reduce the financial burdens
facing low-income families. Recently, however, state budgets
have come under intense pressure, raising concerns that the
current generosity of state rules may be unsustainable.
Already, some states have modified benefit packages, while
others have slowed or halted new enrollment in SCHIP.
The objective of this paper is to present estimates of how
much can really be saved by rolling back SCHIP. The highest
cost children covered through SCHIP would likely have been
eligible for public coverage even had these expansions not
taken place – either through Medicaid spend-down provisions
or other public programs. Assuming such safety-net
programs would remain in place, rolling back SCHIP would
only shift those expenditures back to the safety net programs.
Reductions in eligibility might also lead to increased
uncompensated care, as well as increased tax expenditure
should private insurance increase. The net savings may well
be substantially smaller than one might conclude based on a
naive analysis of budgetary data.
Study Design: We use data from the 2000 Medical
Expenditure Panel Survey. We begin by identifying SCHIP
enrollees in our data and then simulate coverage and
expenditures by source of payment given a contraction in
eligibility for SCHIP. Our simulation accounts for changes in
public coverage, private coverage, medical care utilization,
expenditures by source of payment (including uncompensated
care), and the associated impact on public and private
transfers at the federal and state levels. Whenever possible,
our analysis is driven by relationships estimated from earlier
years of MEPS data.
Population Studied: Nationally-representative sample of
children aged 18 and under in the civilian noninstitutionalized
population.
Principal Findings: Our simulations suggest that in the
average state the actual savings from cutting SCHIP would be
only a half to a third what one would surmise based solely on
budgetary data. This is because children denied SCHIP would
cause offsetting increases elsewhere in state and federal
budgets. We consider three such offsets in our paper:
increased Medicaid medically needy spending, increased tax
subsidies to private insurance, and increased costs associated
with uncompensated care.
Conclusions: Our simulations suggest that in the average
state the actual savings from cutting SCHIP would be only a
half to a third what one would surmise based solely on
budgetary data. This is because children denied SCHIP would
cause offsetting increases elsewhere in state and federal
budgets. We consider three such offsets in our paper:
increased Medicaid medically needy spending, increased tax
subsidies to private insurance, and increased costs associated
with uncompensated care.
Implications for Policy, Delivery or Practice: Sound policy
decisions regarding children’s insurance must be based on
reliable cost estimates. Our analysis suggests that decisions
to roll back SCHIP, if based solely on budgetary data regarding
outlays per enrollee, may substantially overstate the cost
savings that states and the federal government might actually
hope to realize.
Primary Funding Source: AHRQ
• Effects of the State Children’s Health Insurance Program
(SCHIP) on Insurance Coverage of Low-Income Children
Lisa Dubay, Sc.M., Genevieve Kenney, Ph.D.
Presented by: Genevieve Kenney, Ph.D., Principal Research
Associate, Health Policy Center, The Urban Institute, 2100 M
Street, NW, Washington, DC 20037; Tel: 202.261.5568; Fax:
202.223.1149; E-mail: jkenney@ui.urban.org
Research Objective: To estimate the effects of SCHIP
eligibility expansions and concomitant investments that were
made in enrollment systems and outreach on insurance
coverage of both Medicaid and SCHIP eligible children.
Study Design: We used a pre-post design, with multiple
comparison groups, comparing changes in insurance
coverage before and after SCHIP implementation. The two
treatment groups consisted of newly SCHIP eligible children
and children eligible for Medicaid under the poverty-related
expansions. The comparison groups included children in the
same state with incomes slightly above the eligibility
thresholds for SCHIP and parents in the same income group.
Estimated difference-in-difference regression models
controlled for observable child, family, and area
characteristics. Data were from the 1996, 1999, and 2002
National Survey of America’s Families (NSAF). Coverage
categories included: Medicaid and SCHIP, employersponsored coverage, and no coverage. Treatment and
comparison groups were identified using an algorithm that
incorporates federal and state specific SCHIP eligibility
requirements to replicate the eligibility determination process.
Other NSAF data used in this study derive from questions
asked about awareness and perceptions of Medicaid and
SCHIP programs.
Population Studied: Children aged 0-17 years.
Principal Findings: Following the implementation of SCHIP
expansions, coverage under Medicaid and SCHIP grew by
between 11.0 and 13.1 percentage points for Medicaid-eligible
children and by between 12.9 and 17.9 percentage points for
SCHIP-eligible children. SCHIP-eligible children experienced
reductions in uninsurance between 4.9 and 9.7 percentage
points while reductions for Medicaid eligible children were
between 4.9 and 6.5 percentage points. The estimated effects
on employer-sponsored coverage varied from 5.9 and 8.2
percentage points. Despite the reductions in uninsured rates,
more than one in ten children eligible for either Medicaid or
SCHIP remained uninsured in 2002. Other information
collected on the NSAF suggests that over 80 percent of these
uninsured children have parents who say that they would
enroll their child in Medicaid or SCHIP if they were told that
their child was eligible.
Conclusions: Eligibility expansions and related changes in
SCHIP and Medicaid programs reduced uninsurance among
low-income children. Coverage gains occurred both for
children made newly eligible under SCHIP and for children
who were already eligible for Medicaid. The eligibility
expansions also appeared to lead to a substitution of public
for private coverage, implying substitution rates of between 50
and 60 percent.
Implications for Policy, Delivery or Practice: Further
progress in improving coverage among SCHIP and Medicaid
eligible children may require additional investments in
outreach to reach and enroll more uninsured children.
However, additional resources are unlikely given the current
fiscal environment, which has led some states to cut back on
outreach and to establish enrollment caps and waiting lists.
Thus, the improvements in coverage documented in this
study may be reversed.
Primary Funding Source: RWJF
Call for Panels
Using Report Cards to Drive Consumer Choice
Chair: David Howard, Ph.D.
Sunday, June 6 • 3:00 p.m.- 4:30 p.m.
• The Effect of Quality information on Consumer Choice of
Health Plans: Evidence from the Buyers Health Care
Action Group
Jean Abraham, Ph.D., Roger Feldman, Ph.D., Caroline Carlin,
Jon Christianson, Ph.D.
Presented by: Roger Feldman, Ph.D., Professor, Division of
Health Services Research and Policy, University of Minnesota,
MMC 729, 420 Delaware Street S.E., Minneapolis, MN 55455;
Tel: 612.624.5669; Fax: 612.624.2196; E-mail:
feldm002@umn.edu
Research Objective: This study had two objectives: first, to
understand what factors lead employees to search for
information on health plan quality; second, to determine
whether such information affects the employee’s decision to
switch health plans.
Study Design: The study design is based on the expected
utility model of health plan choice, which hypothesizes that
utility depends on quality, premiums, and other features of a
plan. However, unlike other studies we take seriously the
notion that the choice is made based on imperfect
information about health plan quality. Individuals can search
for information and receive different “messages” about health
plan quality. Based on the content of these messages, the
individual may revise her rankings of the plans’ expected
utilities and switch at the next open enrollment opportunity.
Because employers are increasingly the source for information
about the quality of health plans, we specifically want to know
why some employees are aware of this information while
others are not, and whether the employees who are aware of
quality information are more likely to switch plans than those
who are not. Therefore, we develop a two-equation of model
of information awareness and switching. The model is
estimated both by two-stage least squares and by bi-variate
probit methods to deal with endogenous information
awareness in the switching equation
Population Studied: The model is estimated with primary
data collected in 2002 though a survey of 651 single
employees who work for member firms of the Buyers Health
Care Action Group (BHCAG) in Minneapolis. BCHAG is an
employer health care purchasing and reform coalition that
sponsors two forms of quality information: annual awards for
excellence in quality, and a performance results book that
rates health care provider systems on various aspects of
service quality and enrollee satisfaction.
Principal Findings: Employee awareness of quality
information is much higher if the employer distributes
booklets to employees, compared with more passive
approaches to disseminating the quality information.
Demographics also matter, with more highly-educated
employees, females, and those with longer job tenure being
more aware of the quality information. However, we found no
effect of information awareness in the switching equation.
Employees were more likely to switch if the premium for their
current provider system increased relative to those of
alternative systems. Loyalty to a personal physician decreased
the probability of switching, but employees who were at risk of
losing their personal physician were more likely to switch.
Conclusions: Employees may be aware of the quality
information, but they do not appear to use it in making health
plan choices.
Implications for Policy, Delivery or Practice: Employers
need to question their methods for measuring quality. The
current BCHAG rating system rating does not appear to tap
the dimensions of quality that employees care about. Our
results also cast some doubt on the idea that providers should
be divided in exclusive systems. This appears to create a high
degree of loyalty to physicians which inhibits price
competition.
Primary Funding Source: RWJF
• Short and Long-term Effects of a Public Performance
Report on Hospital Reputation
Judith Hibbard, Ph.D., Jean Stockard, Ph.D., Martin Tusler,
M.S.
Presented by: Judith Hibbard, PhD, Professor, Department of
Planning, Public Policy & Management, University of Oregon,
1209 University of Oregon, Eugene, OR 97403; Tel:
541.346.3364; Fax: 541.346.2040; E-mail:
jhibbard@uoregon.edu
Research Objective: In a previous investigation we observed
that a public report on hospital performance stimulated
quality improvement efforts primarily as a result of the
hospitals’ concern about the impact of the report on their
public image and reputation. In this investigation we assess
the impact of the same public report on consumers’ views
about the relative quality of the hospitals in their community.
We explore whether a well designed, “evaluable” public report
actually did affect the reputation of hospitals included in the
report. Both short-term impacts and longer-term impacts on
hospital reputation are assessed.
Study Design: The study design includes a consumer
telephone survey prior to the release of the report, another
survey 2 months after the release, and a third interview 2 years
post report. In the first two surveys a panel design is used
with those respondents included in the pre-report survey reinterviewed in the post-report period (N=680). The third
interview was conducted on a random digit dial sample in the
community (N=800). Because the report was designed to be
highly “evaluable,” with the data displayed in a way that
consumers could quickly and easily discern high and low
performing hospitals, it has the potential to create a “viral
effect.” A viral effect would occur if viewers of the report came
away with an impression of which were better and worse
hospitals, if they retained those impressions and shared them
with others. We looked for evidence of a viral effect in the
study.
Population Studied: The study population includes a
consumer telephone survey conducted in the Madison, WI
metro area in the fall of 2001, just prior to the release of the
QualityCounts Hospital report, another survey 2 months after
the release, and a third interview 2 years post report.
Principal Findings: The results indicate that the public
performance report did have an impact on how consumers
viewed the hospitals. Evidence for a viral effect was also
observed. Consumer views on better and worse hospitals
shifted after the release of the report and the ability to recall
which hospitals were higher or lower performers was
observed. Consumers gained an impression and remembered
those impressions in the short term. We are currently
analyzing the data from the third survey, 2 years post report,
and the preliminary data suggest that the shift observed
shortly after the report was released is maintained.
Conclusions: Public reports, if designed to be evaluable, can
affect hospital reputations.
Implications for Policy, Delivery or Practice: A public report
that is evaluable and widely distributed apparently has the
potential to significantly affect the public image of hospitals
included in that report. Protecting their public image was
found to be an important motivator for stimulating hospitals
to put effort into quality improvement activities. This
suggests that, in addition to making informed choices, an
important consumer role in stimulating quality improvement,
is to simply attend to and form an opinion of hospital quality.
Primary Funding Source: RWJF
• Cards and Consumer Choice in Kidney Transplantation
David Howard, Ph.D., Bruce Kaplan, Ph.D.
Presented by: David Howard, Ph.D., Assistant Professor,
Department of Health Policy and Management, Emory
University, 1518 Clifton Road NE, Atlanta, GA 30322; Tel:
404.727.3907; Fax: 404.727.9198; E-mail:
dhhowar@emory.edu
Research Objective: This study was undertaken to examine
quality the responsiveness of consumers’ choice of kidney
transplant center to quality using a unique dataset consisting
of the universe of registrants for cadaveric kidney
transplantation. There are three principal study questions: 1)
Does quality influence patients’ choice of transplant center? 2)
Do report cards for transplant centers influence choice? 3)
Does the mode of report card dissemination (Internet versus
hardcopy) matter?
Study Design: Patient choice was modeling using a
conditional logit model with random coefficients. Transplant
center attributes – quality (as measured by graft failure rates
one year post-transplant), distance from patients’ home,
lagged market share, and hospital bed size – were interacted
with patient characteristics, including age, primary diagnosis,
education, and insurance type. Coefficients from the model
were used to compute predicted choice probabilities, and
standard errors computed via the delta method. The impact of
report cards was assessed by comparing observed to expected
ratios of the number of patients choosing hospitals that
experienced an improvement in outcomes from one report
card to the next versus hospitals that experienced a decline in
outcomes versus those that experienced no change.
Population Studied: The universe of patients registering for a
cadveric kidney transplant in the continental United States
between January 1, 1998 and June 30, 2001 (N = 47,125).
Principal Findings: The conditional logit model shows that
quality matters; a center that experienced an increase in its
actual one-year graft failure rate of one standard deviation
(0.05) from the sample average (0.09 to 0.14) could expect a 5
percent decline in patient registrations. Transplant centers
with large improvements in outcomes between report cards
experience larger than expected gains in patient registrations.
The converse is true for hospitals that experience a decline in
performance. The placement of report cards on the Internet in
September of 1999 was not associated with a shift in choice
patterns.
Conclusions: 1) At least some patients and their physicians
consider quality when choosing transplant centers. Transplant
centers with high survival rates will attract more patients. 2)
Choice patterns are responsive to the information contained
in report cards. 3) The placement of report cards on the
Internet had no detectable impact on choice patterns.
Implications for Policy, Delivery or Practice: While, is
impossible to gauge from the estimates whether demand is
sufficiently responsive to differences in outcomes to induce
hospitals to make socially optimal investments in quality,
policymakers should not assume that patient welfare is
diminished by increased competition in health care and use of
selective contracting by managed care. Report cards lead
patients to choose hospitals with lower graft failure rates, but
dissemination of report cards on the Internet only may be
insufficient for reaching the types of patients with end stage
renal failure.
• The Effects of Health Plan Performance Measurement
and Reporting on Quality of Care for Medicare
Beneficiaries
M. Kate Bundorf, Ph.D., Laurence Baker, Ph.D.
Presented by: M. Kate Bundorf, Ph.D., Assistant Professor,
Health Research and Policy, Stanford University, HRP T152,
Stanford, CA 94305; Tel: 650.725.0067; Fax: 650.725.6951; Email: bundorf@standford.edu
Research Objective: The Medicare program, which provides
health insurance coverage for elderly and disabled Americans,
has been a key driver of initiatives to provide consumers with
information to compare the quality of health care providers.
In 1997, the Centers for Medicaid and Medicare Services
(CMS, formerly known as HCFA) required all health plans
participating in the Medicare+Choice program to participate in
standardized performance measurement and reporting. The
objective of this research is to examine the effect of mandated
health plan quality measurement and reporting on the use of
measured services by Medicare beneficiaries. We examine the
effects on both beneficiaries enrolled in health plans
participating in performance measurement and those enrolled
in traditional Medicare.
Study Design: The primary data source for our analysis is the
Medicare Current Beneficiary Survey, an annual survey of
Medicare beneficiaries conducted by CMS. Our project uses
data from 1994-2000. We estimate multivariate regression
models to examine the utilization of measured services
among elderly Medicare beneficiaries enrolled in
Medicare+Choice plans before and after the implementation
of mandated participation in HEDIS performance
measurement and reporting and to compare changes among
these beneficiaries to those enrolled in traditional Medicare.
We also examine the effect of market level enrollment in
Medicare+Choice plans participating in quality measurement
and reporting on utilization of measured services among
individuals enrolled in traditional Medicare to identify the
spillover effects of health plan participating in performance
measurement.
Population Studied: A nationally representative sample of
Medicare beneficiaries aged 65 and older.
Principal Findings: Our findings demonstrate the impact of
mandatory health plan participation in performance
measurement and reporting on the utilization of measured
services for Medicare beneficiaries enrolled in both
Medicare+Choice plans and traditional Medicare.
Conclusions: Health plan performance measurement and
reporting may affect both individuals enrolled in participating
plans and those with other types of insurance coverage. Our
results provide evidence on the extent to which provider report
cards directed at consumers improve quality of care both
directly and through spillovers to individuals with other types
of coverage.
Implications for Policy, Delivery or Practice: This project
provides key information about the value of health plan
performance measurement and reporting both generally and
among elderly Medicare beneficiaries. This information is
essential for both large purchasers, such as the Medicare
program, and employers or employer coalitions in deciding
whether to invest in health plan performance measurement as
a tool to improve quality of care. It is also essential to policy
makers making similar decisions.
Call for Panels
Does Participation in Collaborative Quality Improvement
Programs Improve Care for Patients with Chronic Illness?
Chair: Emmett Keeler, Ph.D.
Sunday, June 6 • 3:00 p.m.- 4:30 p.m.
• Do Collaborative Quality Improvement Programs Reduce
Cardiovascular Risk for Persons with Diabetes?
Roberto Vargas, M.D., M.P.H., Carol Mangione, M.D. M.Sc.,
Joan Keesey, B.A., Steve Asch, M.D., Matthias Schonlau,
Ph.D., Emmett Keeler, Ph.D.
Presented by: Roberto Vargas, M.D., M.P.H., Natural
Scientist, Health, RAND, 1700 Main St, Santa Monica, CA
90407; Tel: 310.393.0411; Fax: 310.393.4818; E-mail:
roberto_vargas@RAND.org
Research Objective: In patients with diabetes, cardiovascular
morbidity and mortality is a significant and potentially
preventable complication. There is a need to identify
successful interventions to improve clinical outcomes for
these patients that can be reproduced in real world settings.
We aim to examine the impact of implementing the Chronic
Care Model (CCM) through Breakthrough Series
collaboratives on the reduction of cardiovascular disease risk
in patients with diabetes. associated with success in four
collaboratives.
Study Design: The pre and post periods were one year before
and after the intervention. Patients were eligible for inclusion if
they were over 25 years old, had 5 out of the 6 variables
needed to calculate UKPDS cardiovascular risk score (78 % of
sample). We used hierarchical regression to compare the
cardiovascular risk of patients from intervention and control
sites pre and post participation in the collaborative. This
model adjusted for patient level variation in age, medical comorbidities, and severity of diabetes as well as clustering by
health care organization site.
Population Studied: Representatives from 40 healthcare
organizations participated in a collaborative instructing them
on how to implement organizational changes to improve
quality of care for diabetes in an intervention group of their
patients. Of these, 17 organizations were large enough to
provide matched diabetic patients from non-intervention
clinical sites as controls, and 13 agreed to participate. We
identified patients with diabetes by chart review from both
intervention and within organization control sites and
calculated the risk for fatal or non-fatal myocardial infarction
or sudden death using a modified United Kingdom
Prospective Diabetes Study (UKPDS) risk engine based on
glycemia, blood pressure, and cholesterol level
Principal Findings: 564 patients from intervention sites and
606 patients from control sites met the inclusion criteria. The
mean age was 64 years, 39% were female, and the mean 1
year risk of myocardial infarction or cardiovascular death as
estimated by the UKPDS score pre-intervention was 3.1% for
the intervention group and 3.0% for the control group.
Although the UKPDS risk score improved in both groups, the
intervention group had significantly greater risk reduction
when compared to the control group (p <.0001)
Conclusions: Over a one year interval, implementation of the
CCM through the Breakthrough Series significantly lowered
the cardiovascular risk of patients with diabetes who were
cared for in the participating organization´s settings.
Implications for Policy, Delivery or Practice: Effective
collaborative interventions to improve quality of care are can
be implemented in practice based settings.
Primary Funding Source: RWJF
13 and 26 percent in number of changes made; and between
20 and 42 percent in depth of changes made.
Conclusions: Organizational commitment to measuring and
pursuing patient satisfaction, the presence of team
champions, and involvement of physicians on teams all
enhanced team effectiveness. More effective teams, in turn,
made more changes to improve care for patients with chronic
illnesses, and also made changes that were expected to be
more effective in improving patient health.
Implications for Policy, Delivery or Practice: The data
support recommendations that clinical and managerial
leaders should develop effective teams, show support for
focusing on patient satisfaction, identify and support team
champions, and involve physicians to improve the quality of
care for patients with chronic illness.
Primary Funding Source: RWJF
• The Role of Team Effectiveness in Improving Chronic
Illness Care
Stephen Shortell, Ph.D., Jill Marsteller, Ph.D., M.P.P., Michael
Lin, M.S.P.H., Marjorie Pearson, Ph.D., Shin-Yi Wu, Ph.D.,
Peter Mendel, Ph.D.
• Improvements in Communication, Education, and SelfManagement through Implementation of the Chronic Care
Model for Patients with Heart Failure
David Baker, M.D., M.P.H., F.A.C.P., Asch, Steven, M.D.,
Brown, Julie, M.A., Dracup, Kathy, R.N., Ph.D., Chan, Kitty,
Ph.D., Keeler, Emmett, Ph.D.
Presented by: Jill Marsteller, Ph.D., M.P.P.,
AcademyHealth/NHCS Fellow, Hospital Care Statistics
Branch, National Center for Health Statistics/CDC, 3311
Toledo Rd., Hyattsville, MD 20872; Tel: 301.458.4098;E-mail:
ble2@cdc.gov
Research Objective: The importance of teams for improving
quality of care has received increased attention. We examine
both the correlates of self-assessed team effectiveness and the
consequences of this effectiveness for actually making
changes to improve care for people with chronic illness.
Study Design: Survey data were obtained from individual
providers on 40 teams participating in the national Evaluation
of the Improving Chronic Illness Care Program. Based on
current theory and literature, measures were derived of
organizational culture, a focus on patient satisfaction, team
composition, team effectiveness and the presence of a team
champion. The actual number and depth of changes made to
improve chronic illness care were assessed using monthly
reports from teams to faculty of the quality improvement
program.
Population Studied: 40 multidisciplinary teams from
hospitals, physician groups, health plans and Bureau of
Primary Health Care clinics. Teams participated in one of
three IHI/MacColl Institute Improving Chronic Illness Care
collaboratives on asthma, diabetes, congestive heart failure or
depression. Teams comprised from 3 to 14 nurses, physicians,
pharmacists, case managers, administrators, and other staff.
Principal Findings: A focus on patient satisfaction, the
presence of a team champion, and the involvement of the
physicians on the team were each consistently and positively
associated with greater team effectiveness. Maintaining a
balance among culture values of participation, achievement,
openness to innovation, and adherence to rules and
accountability also appear to be important. Team
effectiveness, in turn, was consistently associated with both a
greater number and depth of changes made to improve
chronic illness care. The various measures explain between 24
and 40 percent of the variance in team effectiveness; between
Presented by: David Baker, MD, MPH, FACP, Chief, Division
of General Internal Medicine, , Feinberg School of Medicine of
Northwestern University, 676 N St. Clair St, Suite 200,
Chicago, IL 60611, IL 60611; Tel: 312.695.0917; Fax:
312.695.0951; E-mail: dwbaker@northwestern.edu
Research Objective: We sought to determine whether
participation in the Institute of Healthcare Improvement
Breakthrough Series (IHI BTS) led to better communication,
education, knowledge, and self-management skills based
upon direct reporting from patients.
Study Design: We conducted a quasi-experiment in 4
organizations participating in the IHI BTS for CHF
(participants) in 1999-2000 and 4 comparable control
organizations (controls). Participating organizations sent
teams to three IHI meetings designed to facilitate major, rapid
changes in CHF care based on the Chronic Care Model.
Participating organizations provided a registry of patients;
patients were interviewed by telephone approximately one year
after the IHI BTS regarding provider-patient communication,
education received, objective knowledge of CHF and selfmanagement, self-management behaviors, satisfaction, and
quality of life. The statistical significance of differences
between groups was determined with multivariate methods,
adjusting for demographics, socioeconomic status, coronary
artery disease, previous revascularization, physician specialty,
and study site.
Population Studied: Of the 781 patients who completed the
survey 62% were age 65 or older, 52% were female, 71% were
white, and 66% had a history of coronary artery disease.
Principal Findings: Patients in the participant group were
more likely to report that their doctor and nurse 1) gave them
choices and options about their treatment, 2) gave them
confidence in their ability to make changes in their life to
control their heart failure, and 3) regularly reviewed how they
were doing managing all aspects of their condition (p < 0.01
for all). Patients in the participant group were more likely to
report having received education about lifestyle modifications
and self-monitoring; e.g., 86.5% of patients in the participant
group reported being told to weigh themselves daily and
record their weight compared to 34.0% of controls (p < 0.01).
Patients in the participant group were approximately twice as
likely as control patients to say that someone with heart failure
should check his weight at least several times per week (81.9%
versus 43.5%, p < 0.01), and they were more likely to know
that swelling of the legs and ankles, waking up at night short
of breath, and weight gain were signs of worsening heart
failure (p < 0.01). Participant patients were more likely than
controls to have a scale at home (adjusted percentages 92.1
versus 77.7%, respectively; p = 0.002), and 66.3% of patients
in the participant group said they monitored their weight every
day compared to 32.9% of control patients (p < 0.001).
However, there were no significant differences in quality of life
between the two groups.
Conclusions: Participation in the IHI BTS was associated with
better communication, education, knowledge, and selfmanagement skills. However, heart-failure related quality of
life was not higher for patients in participating groups
compared to patients in control organizations.
Implications for Policy, Delivery or Practice: Expanding the
use of collaboratives may be an effective strategy for
promoting rapid organizational change. However, further
research is needed to understand how chronic disease
management programs can improve patients’ quality of life.
The quality of patient-centered care measures can be
monitored through patient surveys.
Primary Funding Source: RWJF
Call for Panels
The Effects of the Balanced Budget Act of 1997 on
Hospital Finances, Uncompensated Care Provision &
Quality of Care
Chair: Kevin Volpp, M.D., Ph.D.
Sunday, June 6 • 5:00 p.m.- 6:30 p.m.
• The Influence of Health Policy and Market Factors on the
Hospital Safety Net
Gloria Bazzoli, Ph.D., Richard Lindrooth, Ph.D., Ray Kang, Jack
Needleman, Ph.D., Romana Hasnain-Wynia, Ph.D.
Presented by: Gloria Bazzoli, Ph.D., Professor, Health
Administration Department, Virginia Commonwealth
University, P.O. Box 980203, 1008 E. Clay Street, Richmond,
VA 23298-0203; Tel: 804.828.5223; Fax: 804.828.1894; E-mail:
gbazzoli@vcu.edu
Research Objective: US hospitals play an important role in
serving indigent patients, providing uncompensated care that
totaled $21.5 billion in 2001. The late 1990s presented several
challenges to hospitals given the effects of the 1997 Balanced
Budget Act (BBA) on Medicare and Medicaid payments
coupled with increased private payer pressures. Earlier
research suggests that voluntary non-profit hospitals likely
reduced their uncompensated care in response to these
pressures, but as a group only to a small degree. Such
findings may allay policymaker concern about potential
adverse payment policy effects on the safety net. However,
substantial variation exists in voluntary non-profit hospital
involvement in indigent care. We focus on a subset of
hospitals with a demonstrated commitment to indigent care
to assess how BBA and other pressures affected their
uncompensated care provision.
Study Design: Following the Institute of Medicine’s 2000
report on the hospital safety net, we identify core and
voluntary safety net hospitals based on their demonstrated
commitment to indigent care. Core safety net hospitals have
a legal mandate or strong commitment to maintain an open
door to indigent patients. Voluntary safety net hospitals treat a
significant number of indigent patients but to a lesser degree
than core institutions. Generally, prior research has equated
voluntary safety net hospitals with voluntary non-profit
hospitals. We, however, use pre-BBA data on actual hospital
uncompensated care provision to identify hospitals with a
commitment to indigent care. We analyzed data for 1996 to
2000, using estimation approaches that control for fixed
hospital and market characteristics.
Population Studied: All urban US hospitals in operation
between 1996 and 2000, which totaled 2,446 institutions.
Principal Findings: Our results suggest that voluntary safety
net hospitals we identified were very responsive to Medicare
net revenue changes when providing uncompensated care
and that this response varied by market. Voluntary safety net
hospitals in markets that lacked substantial hospital
competition or high HMO market share kept pace with other
hospitals in their uncompensated care provision. Voluntary
safety net hospitals in harsher markets, however, had slower
growth in uncompensated care relative to other hospitals.
Further, voluntary safety net hospitals we identified were more
responsive to financial pressures in their uncompensated care
provision than were voluntary non-profit hospitals overall.
Conclusions: Our findings are consistent with a growing body
of research that suggests not all voluntary non-profit hospitals
are alike in their objectives and actions. Further, our results
raise concern that institutions that once were key participants
in the health care safety net reduced their indigent care
involvement during the late 1990s, especially if they operated
in markets with substantial hospital competition or high HMO
market share.
Implications for Policy, Delivery or Practice: Because BBA
was intended to reduce federal deficits rather than affect the
hospital safety net, our analysis indicates that BBA had
unintended effects that require redress. In addition, the
findings support the continued targeting of public support to
those hospitals that demonstrate a commitment to indigent
care provision. Currently, Medicare and Medicaid
disproportionate share hospital payments and some state
uncompensated care pools provide such support but these
may need to be strengthened if indigent care is to be broadly
distributed across institutions rather than concentrated in a
few.
Primary Funding Source: RWJF
• The Influence of Health Policy and Market Factors on
Quality in the Hospital Safety Net
Jack Needleman, Ph.D., Gloria Bazzoli, Ph.D., Richard
Lindrooth, Ph.D., Ray Kang,Romana Hasnain-Wynia, Ph.D.
Presented by: Jack Needleman, Ph.D., Associate Professor,
Department of Health Services, University of California Los
Angeles School of Public Health, 650 Charles Young Dr. S.,
Room 31-236B CHS, Los Angeles, CA 90095-1772; Tel:
310.267.2706; Fax: 310.825.3317; E-mail: needlema@ucla.edu
Research Objective: US hospitals play an important role in
serving indigent patients, providing uncompensated care that
totaled $21.5 billion in 2001. Substantial variation exists in
voluntary non-profit hospital involvement in indigent care. We
focus on a subset of hospitals with a demonstrated
commitment to indigent care. The late 1990s presented
several challenges to hospitals given the effects of the 1997
Balanced Budget Act (BBA) on Medicare and Medicaid
payments coupled with increased private payer pressures, and
the core safety net hospitals may have been especially
affected. One response to financial pressure is to cut
expenses in ways that threatens quality of care. We examine
the extent to which quality of care in core safety net hospitals
declined relative to other voluntary hospitals.
Study Design: Following the Institute of Medicine's 2000
report on the hospital safety net, we identify core and
voluntary safety net hospitals based on their demonstrated
commitment to indigent care, based on pre-BBA data on
actual hospital uncompensated care provision reported to the
American Hospital Association. We construct Agency for
Healthcare Research and Quality patient safety indicators
using state hospital discharge data and regress adverse events
on measures of safety net status, implementation of the BBA,
the patient risk profile in the hospital and other hospital
characteristics.
Population Studied: Urban US hospitals in operation
between 1996 and 2000 in seven states: California, Florida,
New Jersey, New York, Washington and Wisconsin.
Principal Findings: Findings to be presented at the meeting.
Conclusions: Conclusions to be presented at the meeting.
Implications for Policy, Delivery or Practice: To the extent
that the BBA has contributed to a decline in relative quality in
core safety net hospitals, it will have the unintended
consequence of placing the population dependent on these
providers at risk. If we find that relative quality has declined,
Medicare and Medicaid disproportionate share hospital
payments and some state uncompensated care pools that
provide support but for safety net providers may need target
additional resources to these hospitals assure they can
provide quality care.
Primary Funding Source: RWJF
Call for Panels
Does Hospital Financial Condition Affect Patient Care &
Safety?
Chair: Gloria Bazzoli, Ph.D.
Monday, June 7 • 8:30 a.m.- 10:00 a.m.
• How Much of the Variation in Hospital Financial
Performance is Explained by Service Mix?
Richard C. Lindrooth, PhD, Gloria J. Bazzoli, PhD, Jan P.
Clement, PhD, Mei Zhao
Presented by: Richard C. Lindrooth, PhD, Assistant Professor,
Department of Health Administration and Policy, Medical
University of South Carolina, 19 Hagood Avenue, 408
Harborview Tower, Charleston, SC 29425; Tel: 843.792.2192;
Fax: 843.792.3327; E-mail: lindrorc@MUSC.EDU
Research Objective: To quantify the amount of variation in
hospital operating margins that is due service mix, payer mix,
operational decisions, market characteristics, and
ownership/mission.
Study Design: We regress hospital operating margins from
the Medicare Cost Reports on the share of hospital inpatients
in the 100 most common DRGS and MDC category; payer
mix; ownership indicator variables; market characteristics (e.g.
per capita income); and operational decisions (e.g. staffing
ratios). Next we calculate the partial (and adjusted) R-squared
for each variable category. Finally, we test for differences in
service mix between the hospitals with margins in the top
quartile and hospitals in the bottom quartile.
Population Studied: All nonfederal, general short-term
hospitals operation between 1995-2000 in AZ, CA, CO, FL, IA,
MA, MD, NJ, NY, WA, and WY. We base our service and
payer mix variables using all inpatient admissions report in
the HCUP-SID data for these states. The measures are
combined with operating margins from the Medicare cost
reports and hospital characteristics from the AHA Annual
Survey, and market characteristics from the Area Resource file.
Principal Findings: Based on our analysis of all hospitals in
the FL data (this will be expanded to all states for the
presentation), we find that of the variation in operating
margins 13% is explained by service mix; 9.6% by ownership
characteristics; 5.9% by payer mix; 4.2% by market
characteristics; and less than 1% by operating characteristics.
The results for only investor-owned hospitals are more
striking: 72% (34.4% using adjusted partial r-squared) of the
variation is explained by service mix; 1.7% by payer mix; 7.7%
by market characteristics; and less than 7.1% by operating
characteristics. The financially successful hospitals are
significantly likely to have a higher share of inpatients in DRGs
associated with cardiac care such as (104-107, 110).
Conclusions: Hospitals are increasingly faced with entry by
specialty hospitals and ambulatory care clinics that specialize
in profitable services. Losing patients that are profitable will
hurt the financial margins of general hospitals, which also
supply less profitable services. If the trend continues we
would expect to increasingly see closures or reductions in the
provision of unprofitable services. This could lead to access
problems for patients with certain unprofitable diagnoses or
quality problems if hospital volume in these services
diminishes sharply.
Implications for Policy, Delivery or Practice: It is well known
that Medicare is very generous in its reimbursement for some
DRGs and less generous in its reimbursement for other
DRGs. While it is extremely difficult for CMS to observe the
true cost of providing a service and set the appropriate rate,
policy makers should be cognizant of the fact such distortions
can lead to hospital access or quality problems.
Primary Funding Source: AHRQ
• Hospital Financial Distress and Patient Outcomes: A
Panel Study
Mei Zhao, MHA, Gloria J. Bazzoli, PhD, Henry J. Carretta,
MPH, Askar Chukmaitov, MPH
Presented by: Mei Zhao, MHA, Research
Associate/Instructor, Department of Health Administration,
Virginia Commonwealth University, P.O. Box 980203, 1008 E.
Clay Street, Richmond, VA 23298-0203; Tel: 804.828.5223; Fax:
804.828.1894; E-mail: mzhao@hsc.vcu.edu
Research Objective: The hospital industry was confronted
with intense financial pressure during the late 1990s, resulting
in a considerable number of hospitals that were in financial
distress. In particular, government reports indicated that
nearly 1/3 of hospitals had negative total margins. At the
same time, the Institute of Medicine raised concern about a
substantial number of patients receiving substandard care.
Earlier research suggests that hospital financial condition may
have an impact on patient quality of care, but these studies
typically only used one or two mortality measures as their
quality indicators. The objective of the current study is to
examine what differences exist in the quality of care and
patient safety between financially weak (distressed) and strong
(non-distressed) hospitals using multiple patient outcome
measures.
Study Design: A panel study design is applied to data from
1995-2000 to compare hospital patients’ outcomes in the
lowest tercile of operating margin with those in the highest
tercile operating margin. Five mortality rates (IQIs) and nine
patient safety indicators (PSIs) are used to represent patient
outcomes. To date, descriptive analysis and t-tests have been
conducted to examine the difference between these two
groups.
Population Studied: All nonfederal short term general
medical-surgical hospitals in operation between 1995 and
2000 from 11 states that participated in the AHRQ’s HCUP
State Inpatient Databases (SID) program. This totaled 1,611
institutions.
Principal Findings: Overall, we found that distressed
hospitals consistently had significantly poorer IQIs and PSIs in
each study year, although the gap between distressed and
non-distressed hospitals narrowed somewhat between 1998
and 2000. Mortality rates for both distressed hospitals and
non-distressed hospitals fell between 1995 and 1998 but then
increased from 1998 to 2000. The incidence of most medical
complications as measured by the PSIs, on the other hand,
increased throughout the period for both the distressed and
non-distressed groups.
Conclusions: The findings of this study are consistent with
existing literature that suggests hospital quality of patient care
is related to hospital financial condition. Hospitals with better
financial performance have better patient outcomes than
those in poorer financial condition. Although this is the case,
it is interesting that the gap between distressed and nondistressed hospitals narrowed for the mortality indicators
beginning in 1998. This coincides with the implementation of
major payment changes of the 1997 Balanced Budget Act,
which affected all hospitals nationwide.
Implications for Policy, Delivery or Practice: Both public
policies and private payer strategy implemented in the 1990s
were intended to reduce costs without compromising quality
of patient care. However, these policies created considerable
financial pressure on hospitals. Our analysis indicates that
these polices may have had unintended negative effects on
patient outcomes. Hospitals experiencing financial distress
may have fewer resources to invest in the quality of their
services, which in turn affected patient outcomes. In addition,
Medicare payment policies implemented in the late 1990s
may have had an immediate adverse effect on patient
outcomes for both distressed and non-distressed hospitals.
Primary Funding Source: AHRQ
• A Profile of Inpatient Care and Safety in Hospitals with
Differing Case-Mix and Financial Condition
Sema K. Aydede, Ph.D., Gloria J. Bazzoli, Ph.D., Jerod Loeb,
Ph.D., Barbara Braun, Ph.D.
Presented by: Sema K. Aydede, PhD, Associate Scholar,
Institute for Child Health Policy, University of Florida, 5700
SW 34th Street, Suite 323, Gainsville, FL 32608-5367; Tel:
352.265.7220 x86335; E-mail: ska@ichp.edu
Research Objective: It has long been recognized that some
hospitals treat disproportionately larger share of patients with
complex conditions. On the other hand, during mid to late
1990’s, all hospitals have faced increased pressures to contain
expenditures. This study explores patient safety and the
quality of inpatient care in hospitals with differing case-mix
and financial condition.
Study Design: Hospitals in 11 geographically dispersed states
with mandatory participation in AHRQ’s HCUP State Inpatient
Data (SID) program are grouped into those treating patients
with more/less severely ill and those that were financially
weak/strong. All-Patient Refined Diagnosis-Related Groups
(APR-DRG) software was applied to SID discharge data for
severity of illness groupings. Hospitals treating patients more
severely ill were defined as those with above average percent
of APR-DRG major and extreme cases for at least 5
consecutive years during 1995-2000. Financially weak
hospitals are those with a negative 3 year (1993-1995) average
operating margin. Seven AHRQ Patient Safety Indicators
(PSI) (that involve the largest number of patients who are at
risk for adverse outcomes) and 5 AHRQ Inpatient Quality
Indicators (IQI) (where majority of hospitals treat a sufficient
number of patients) are considered.
Population Studied: 1363 nonfederal, acute care hospitals.
Principal Findings: When all hospitals (n=1363) were
considered: (1) rates for 4 (two sentinel and two nursingrelated) out of 5 PSI were higher; and (2) rates for all 3 inhospital mortality IQI and one technical complication PSI were
lower for hospitals serving more severely ill patients than they
are for hospitals serving less severely ill patients in 1996.
These results persisted in 2000. When the subgroup of
financially strong hospitals (n=610) were considered, quality
comparisons across hospitals serving more/less severely ill
patients gave the same results as above for both 1996 and
2000. When the focus was on financially weak hospitals
(n=753): (1) the same result as above persisted for two
sentinel and two nursing-related PSIs both in 1996 and 2000;
but (2) rates for 3 IQI and one PSI were not different across
hospitals serving more/less severely ill patients in 1996; and
(3) rates for 3 IQI and one PSI were lower for hospitals serving
more severely ill patients than they were for hospitals serving
less severely ill patients in 2000.
Conclusions: Generally, these results suggest that hospitals
treating a more severely ill patient mix have better outcomes
for selected mortality and technical complication measures
when compared to hospitals treating less severely ill patient
populations. These findings are consistent with theories of
learning by doing, but it is also clear that this does not carry
through to all PSIs. Hospital financial distress may erode this
advantage, however, especially in the short-run.
Implications for Policy, Delivery or Practice: Although
hospitals that treat a more complex patient mix may be able to
overcome the adverse effects of financial distress over time,
they may experience detrimental changes in patient outcomes
in the short-run. As such, these hospitals may require onetime subsidies or targeted financial assistance in the short-run
to help avert unfavorable patient outcomes.
Primary Funding Source: AHRQ
• Disparities in Quality and Safety Outcomes, 1995-2000
Jan P. Clement, Ph.D., Askar Chukmaitov, M.P.H.
Presented by: Jan P. Clement, PhD, Professor, Department of
Health Administration, Virginia Commonwealth University,
P.O. Box 980203, 1008 E. Clay Street, Richmond, VA 232980203; Tel: 804.828.1886; Fax: 804.828.1894; E-mail:
jclement@hsc.vcu.edu
Research Objective: Although researchers have examined
disparities in access to care for population groups, there is
less research on potential disparities in hospital care
delivered. This study assesses whether there are disparities in
the quality of hospital care received by Medicaid and self-pay
patients compared to patients covered by more generous
private insurance plans and whether any differences have
become more or less pronounced over time.
Study Design: Rates for AHRQ inpatient quality and patient
safety indicators (IQIs and PSIs) are determined for hospitals
for each year from 1995 through 2000. The measures are risk
adjusted for the patient’s age, gender and APR-DRG and
smoothed to eliminate random variation in rates. We use
multiple measures, choosing the measures where a large
number of patients would be at risk of the adverse outcome.
Thus, we selected five IQIs because a large share of US
hospitals treat a sufficient number of patients with diagnoses
relevant to the IQI. However, because PSIs are not condition
specific, we selected the eight PSIs that involve the largest
number of patients who are at risk of adverse outcomes.
Population Studied: All short term general medical-surgical
(or acute care) hospitals from 11 states (AZ, CA, CO, FL, IA,
MD, MA, NJ, NY, WA, and WI) that participated in the
AHRQ’s HCUP State Inpatient Databases (SID) program.
Principal Findings: Preliminary findings suggest that relative
to private pay hospital mortality rates (IQIs), rates for
Medicaid and self pay patients are significantly higher in 1995
and 1998 even though the average Medicaid case-mix is lower
than the average private pay case-mix. The average self-pay
case mix is similar to the private pay case-mix. In 2000, the
findings are somewhat different. Only one hospital Medicaid
IQI rate is significantly higher than the hospital private pay
rate. Three of the five hospital self pay IQI rates are still
significantly higher than the private pay rates. In contrast, for
seven of the eight hospital patient safety indicator rates (PSIs)
both the hospital Medicaid and self-pay rates are consistently
significantly higher than the private pay rates in 1995, 1998
and 2000. Over time, self pay and private pay IQI and PSI
rates have consistently increased while those for Medicaid
have either increased or decreased.
Conclusions: Mortality and patient safety rates differ by payer
and to have changed at different rates over time.
Implications for Policy, Delivery or Practice: Although
preliminary findings indicate Medicaid and self-pay patients
have worse outcomes on average than private pay patients,
there appear to be improvements in some of the hospital
mortality rates over time but not in the patient safety
indicators. In addition to seeking to eliminate disparities in
access to care, researchers and policy-makers need to examine
differences in the underlying processes of hospital care for
disadvantaged populations.
Primary Funding Source: AHRQ
Call for Panels
Medical Debt: Causes, Consequences & Policy
Implications
Chair: Robert Seifert, M.P.A.
Monday, June 7 • 10:30 a.m.- 12:00 p.m.
• How Medical Debt Threatens Economic Security and
Access to Care: Findings from the Commonwealth Fund’s
2003 Survey of Health Insurance and Access
Jennifer Edwards, Dr.P.H., Sara Collins, Ph.D., Michelle Doty,
Ph.D.
Presented by: Jennifer Edwards, Dr.P.H., Deputy Director,
Task Force on the Future of Health Insurance, The
Commonwealth Fund, One E. 75th Street, New York, NY
10021; Tel: 212.606.3835; E-mail: je@cmwf.org
Research Objective: A combination of increases in the
number of people uninsured and higher levels of patient cost
sharing for the insured make it likely that medical debt will
become a wider concern for the under-65 population. Past
studies have found medical bills contribute to nearly half of all
personal bankruptcies. More recently, media reports have put
a negative spotlight on aggressive efforts of hospitals to
collect from patients unable to pay and hospitals indicate
concerns with rising levels of bad debt and charity care. To
date, these accounts have mainly focused on local markets or
court records. The goal of this study is to provide a national,
2003 estimate of the prevalence of medical bill and medical
debt problems and to examine the associated consequences
for family savings and access to medical care. The study uses
a household survey to document the insurance status and
demographic characteristics of those in debt, and tracks
changes since a 2001 survey.
Study Design: We surveyed a nationally representative
sample of adults 19-64 (n= 3,293) between July and November
2003. Telephone interviews in English and Spanish asked each
adult about access to health care services, insurance coverage,
experiences with their provider and the health system
generally, and an array of questions about medical debt.
Population Studied: Adults age 19-64
Principal Findings: Nearly one in four non-elderly Americans
reports they or a family member had difficulty paying medical
bills in the last twelve months. Twenty-one percent of adults
were contacted by a collection agency; and sixteen percent
reported having to change their way of life significantly to pay
medical bills. Most of these medical debts were incurred when
the patient was insured. A startling 41% of the insured who
reported problems paying their bills -- 18% of all non-elderly
Americans with insurance -- used up all or most of their
savings to pay medical bills.
Additional analysis will allow us to answer the following
questions: What are the characteristics of people in medical
debt (age, income, health status, etc.)? How does being
underinsured correlate with medical debt? Did patients expect
their insurance would cover the care they received? As a result
of medical debt, how have patients or their families changed
their way of life? Have concerns about health care costs kept
people in medical debt from seeking further care? Are people
in medical debt less likely to receive recommended preventive
care or go without recommended tests or treatment? Are
people with chronic illness or disability more likely to be in
debt? How much have these problems grown since 2001?
Conclusions: Although medical debt problems are greatest
for the uninsured, the problem is clearly spreading to those
with private insurance. Findings indicate medical debt is
undermining long term economic security for low and middle
income working families. In addition, access to care has been
affected, which has the potential to impact future health,
especially for those with chronic illnesses needed regular
attention.
Implications for Policy, Delivery or Practice: Short term
solutions that have been proposed include clarifying federal
requirements that appear to compel providers’ collection
practices; setting guidelines for free or discounted care; and
offering low-income people rates more similar to what
Medicaid pays. In the long run, however, providing health
insurance that meets patients’ needs for financial security may
be the only solution that doesn’t continue to burden providers
with uncompensated care costs.
Primary Funding Source: CWF
• Hospital Practices and Medical Debt
Jeffrey Prottas, Ph.D., Deborah Gurewich, Ph.D., Robert
Seifert, MPA
Presented by: Jeffrey Prottas, Ph.D., Professor, Heller
Graduate School, Brandeis University, 415 South St., Waltham,
MA 02254; Tel: 781.736.3955; Fax: 781.736.3950; E-mail:
prottas@brandeis.edu
Research Objective: Examine how hospital policies and
practices dealing with self-pay patients (uninsured patients
and insured patients with cost-sharing responsibilities) affect
patients' indebtedness
Study Design: Structured comparative case studies in
hospitals with different tax statuses. Studies conducted in
several states with different uncompensated care
reimbursement policies
Population Studied: Non-profit, for-profit and public
hospitals in three states. Data collected about hospital policies
and processes concerning eligibility for charity care programs,
billing and collection of payment from patients who are
uninsured or with insurance that leaves them with unpaid
portions of the hospital bills.
Principal Findings: Hospital policies show wide variations.
Bill discounting is common but often ad hoc and
unsystematic. Eligibility for free and/or discounted care is
often poorly communicated to patients. Collection policies are
as variable as free care and discounting policies.
Conclusions: Hospitals have great discretion in billing and
collection policies, despite Federal regulations. Policies can be
crafted that improve the accessibility to free or discounted
care for indigent patients.
Implications for Policy, Delivery or Practice: More
transparent hospital policies would increase the accessibility
of free and discounted care. They would also insure equitable
access to existing programs. Some forms of health insurance
seem to provide only minimal financial protection to patients,
which may raise regulatory issues. Free care programs that
exclude the low-income insured make little sense and expose a
large population to serious debt problems following illness.
• Medical Bankruptcy Incidence and its Legal and Practical
Limits
Melissa Jacoby, J.D.
Presented by: Melissa Jacoby, J.D., Associate Professor of
Law, Temple University, School of Law, 1719 N. Broad Street,
Philadelphia, PA 19122; Tel: 215.204.4489; E-mail:
Melissa.jacoby@temple.edu
Research Objective: The study of bankruptcy filers originally
was initiated to explore the characteristics and problems of
individuals who seek relief from the federal bankruptcy
system. After a high incidence of medical-related bankruptcy
was observed, I undertook further analysis of the data and
studied the relevant laws to evaluate the value of bankruptcy
as a method of health care finance.
Study Design: For data, this paper draws primarily on a
weighted sample of anonymous surveys, containing
approximately 28 questions with mostly pre-coded answers,
collected during first quarter of 1999 in eight federal judicial
districts. To the extent the data are available, this paper also
will incorporate medical-related information from the 2001
Consumer Bankruptcy Project, which involves case file
information, initial surveys, and follow up interviews with
bankruptcy filers in five states. The study also involves an
analysis of bankruptcy law and broader debtor-creditor laws.
Population Studied: Individual bankruptcy filers.
Principal Findings: Findings from data analysis: Nearly half
of all bankruptcy filers either have a substantial medical debt
or indicate illness or injury as a reason they filed for
bankruptcy. Among joint (married) bankruptcy filers who both
have health insurance, almost 40% nonetheless list a provider
of health goods or services as a creditor. Like filers who did
not indicate medical-related bankruptcy, about 80% of
medical-related bankruptcy filers indicated having some
insurance (and the 2001 Consumer Bankruptcy Project will
shed light on lapses in insurance). Of those indicating illness
or injury as a reason for filing, 20.5% identified the possibility
of losing their homes as a reason for filing, and were slightly
more likely to indicate home loss than those who did not
indicate illness or injury as a reason for filing. Findings from
legal analysis: bankruptcy may prevent medical providers from
collecting medical debt and may result in the discharge of that
debt under some circumstances. This may be helpful to a
debtor/patient because medical providers who extend credit
for their services have a wide range of informal and formal
collection mechanisms at their disposal in the debtor-creditor
system. However, the discharge of medical debt may hamper
the receipt of some medical services in the future. Bankruptcy
also does not help with future repeated medical costs or the
income-related effects of illness. In addition, if the debtor has
converted medical-debt into a debt secured by her home, she
probably will remain liable for that debt after bankruptcy and
risks losing her home. Finally, legislation pending in the U.S.
Congress over the past several years would hinder the
accessibility of bankruptcy, which would have implications for
medical-related filers.
Conclusions: The bankruptcy data not only reveal further
cracks in the health care finance system (including people who
arguably fall within the middle class), but also support
concerns that having some health insurance does not
insulate individuals from medical-related financial distress.
Bankruptcy is an understandable response to unexpected
medical-related financial problems, but may not be a
particularly helpful approach for people with chronic health
problems, and may become less accessible to ill and injured
people in financial distress if pending legislation becomes law.
Implications for Policy, Delivery or Practice: Implications
for health policy: The data and analysis suggest that the
insured/uninsured dichotomy is no longer particularly helpful
or even descriptive taxonomy. The quality of health insurance
programs must be evaluated. Implications for bankruptcy
policy: Even though bankruptcy is not the ideal approach to
health care finance, it may be an important stop-gap measure
for some households, and thus Congress should proceed very
carefully before making changes to the bankruptcy system that
will hamper access to and the benefits of bankruptcy. Yet, it
probably is not possible to make bankruptcy more hospitable
to the needs of medical-related filers due to important
structural limitations within the system. Because of these
inherent limitations, lawyers should be very cautious in
advising financially distressed debtors and ideally should
pursue other options before recommending bankruptcy.
Primary Funding Source: For 1999 Consumer Bankruptcy
Project: Harvard Law School. For legal analysis and follow-up
data analysis: Temple University.
• Medical Debt: Causes and Responses in Des Moines,
Iowa
Becky Miles-Polka, M.S., C.N.M., Anne Baber Wallis, Ph.D.
leaders and residents about the issue and to develop means
of changing the system at the policy and practitioner levels.
Population Studied: Residents of designated low-income
Census Tracts in Des Moines, and residents of Polk County.
Qualitative data collection was conducted with a purposive
sample of residents in low-income Des Moines
neighborhoods.
Principal Findings: Survey findings indicate that medical debt
is a significant problem for Des Moines residents. Thirty-four
percent of respondents reported debt related to medical
expenses. Working families with incomes between $15,000
and $25,000 are hardest hit (40.3%) as are people with jobs
(37.3%), families with children (42.4%), and AfricanAmericans (41.0%). Qualitative research is investigating
amounts of medical debt and results and sources of medical
debt. One preliminary finding of the qualitative research is that
affordable housing advocates report medical debt as one of
the most common barriers to home ownership.
Conclusions: Medical debt is revealed in local survey data as
an important factor related to health care access and family
economic success. Community-level data help to refine
understanding of the problem and to craft local responses to
service, information and organizational needs.
Implications for Policy, Delivery or Practice: Work in Des
Moines, Iowa, with community providers and city- and statelevel policy makers shows the link between research and
practice. Data from future surveys will be used in the
community as an indicator to monitor changes in health care
access.
Primary Funding Source: Annie E. Casey Foundation
Presented by: Becky Miles-Polka, M.S., C.N.M., President,
Within Reach Consulting, 3525 Witmer Parkway, Des Moines,
IA 50310; Tel: (515) 274-1330; Fax: 515.274.1330; E-mail:
b.milespolka@mchsi.com
Research Objective: The purpose of this research is to study
the problem of medical debt in Des Moines, IA, and to use
research data to inform community activity around policy and
institutional improvements to reduce medical debt and lessen
its impact on low-income populations.
Study Design: We conducted about 700 door-to-door surveys
and about 700 telephone surveys in the city of Des Moines
and surrounding Polk County. The survey was part of the
Annie E. Casey Foundation's Making Connections initiative
and was conducted by the Urban Institute and NORC. Data
on medical debt were analyzed at the University of Iowa
College of Public Health and supplemented with additional
qualitative data collection in low-income Des Moines
neighborhoods. Research findings were used to educate local
Presented by: Chapin White, Ph.D., Post-doctoral Fellow,
National Bureau of Economic Research, 1050 Mass. Ave.,
Cambridge, MA 02138; Tel: 617.588.0358; Fax: 617.868.2742;
E-mail: chapin_white@post.harvard.edu
Research Objective: Medicare implemented a new
prospective payment system (PPS) for skilled nursing facilities
(SNFs) beginning in 1998. The new PPS changed payment
levels ("average payment effect") and eliminated cost
reimbursement ("marginal payment effect"). The objective of
this research is to measure the effects of the new payment
system on payments, nurse staffing levels and quality of care
in the freestanding SNF setting.
Study Design: Nurse staffing and quality of care were
measured at the SNF level in 1997 (pre-PPS) and 2001 (postPPS). Changes in nurse staffing and quality of care are
regressed on SNF-level payment effects using a long
differences OLS model. The average payment effect is
measured at the SNF level using the change in revenues per
resident per day attributable to the SNF PPS. The marginal
Call for Panels
Informing Medicare Policy on Post-Acute Care
Chair: Joseph Newhouse, Ph.D.
Monday, June 7 • 10:30 a.m.- 12:00 p.m.
• Medicare's New Prospective Payment System for Skilled
Nursing Facilities: Effects on Staffing and Quality of Care
Chapin White, Ph.D.
payment effect is measured by the "Medicare resident
fraction," i.e. the fraction of residents whose stays were
covered by Medicare in 1997; this represents the fraction of a
SNF's residents for whom cost reimbursement was
eliminated. Quality of care is measured using two processbased measures (survey deficiencies and fraction of residents
in restraints) and two outcomes-based measures (incidence of
pressure sores and incidence of unplanned weight gain or
loss). Nurse staffing and quality of care are adjusted for
resident casemix.
Population Studied: All freestanding SNFs in the U.S. were
included if they were open in 1997 and 2001 and had nonmissing data (n=9067).
Principal Findings: Payment levels increased for most SNFs,
and this increase was associated with a positive but small
change in nurse staffing. Real Medicare payments per
Medicare resident per day increased by $40.51; this is
estimated to increase spending on nurse staff by only $0.40
per resident per day. The marginal payment effect, on the
other hand, is associated with a sharp drop in nurse staffing
and a deterioration in the process-based quality of care
measures. The marginal payment effect (evaluated at the
mean Medicare resident fraction) is associated with a
decrease in spending on nurse staff of $2.10 per resident per
day, and an increase of 0.23 in the number of deficiencies.
Conclusions: The new SNF PPS changed both average
payments and marginal payments, and these payment effects
varied widely across facilities. The new PPS had partially
offsetting effects on nurse staffing in freestanding SNFs. The
increase in payment levels appears to have increased nurse
staffing somewhat, but this effect was overshadowed by the
marginal payment effect, which was associated with large
decreases in nurse staffing and a deterioration in some
measures of quality of care.
Implications for Policy, Delivery or Practice: Medicare
policymakers are concerned that quality might be deteriorating
in SNFs post-PPS. This research underscores that concern,
and shows that raising payment levels may not address the
problem. Under the new PPS, SNFs have a strong incentive to
reduce costs, regardless of the level of payments they receive.
Improving the quality of care in the SNF setting may,
therefore, require other measures such as reintroduction of
some element of cost reimbursement, enhanced reporting of
quality measures, and/or increased regulatory oversight.
Primary Funding Source: NIA
• Informing Medicare Policy on Post-acute Care
Sharon Bee Cheng, M.S., Christopher Hogan, Ph.D.
Presented by: Sharon Bee Cheng, M.S., Analyst, , Medicare
Payment Advisory Commission, 601 New Jersey Avenue,
N.W., Suite 9000, Washington, DC 20001; Tel: 202.220.3712;
Fax: 202.220.3759; E-mail: scheng@medpac.gov
Research Objective: To determine whether certain vulnerable
types of eligible Medicare beneficiaries were excluded from the
benefit during the period of declining use.
Between 1996 and 2000, the number of beneficiaries who
used home health services fell by one million. Over the period
of decline, the Congress and the administration sought to
reduce fraud and abuse, change eligibility for the benefit, and
implement two substantial changes to the payment system.
While an overall reduction in use was the intended
consequence of the changes they made, they did not intend to
exclude eligible beneficiaries who would benefit from receiving
home health services. This study provides a national picture of
the situation, compares current users to those before the
decline, and determines whether patient access has changed.
Study Design: Changes in service use were assessed with
claims data from before and after the decline, in 1996 and
2001. Users’ characteristics were measured in two ways: for
those who were discharged from an inpatient hospital, the
diagnosis related group (DRG) was used; for those who were
admitted to home health directly from the community, the
(DXG) was used. Users were grouped according to the rate of
use of home health in 1996 among beneficiaries with that
DRG or DXG. Finally, we compared the rate of use by the
groups in 2001 to the rate in 1996 to determine whether any
group experienced a disproportionate decline. We also
grouped users by state of residence and demographic
characteristics and made similar comparisons.
Population Studied: Medicare beneficiaries enrolled in
traditional, fee-for-service Medicare.
Principal Findings: Some proportion of every type of
beneficiary used home health in 1996 and in 2001; no type
was wholly excluded in 2001. However, the types did
experience different rates of decline. The type of beneficiaries
who had the highest rate of use in 1996 experienced the
smallest decline in use. Those with the lowest rate of use
experienced a disproportionate decline in use. Very high use
states had greater declines than low use states. Rates of
decline by demographic characteristics were proportionate
among high users and low users.
Conclusions: Because all types of beneficiaries used the
benefit in 1996 and still used the benefit in 2001, we can
conclude that no particular type of beneficiary was excluded
from the benefit. Those with lowest rate of use, thus the least
clear need for the benefit, experienced greater declines than
those with clearer needs. The higher rate of decline in higheruse states suggests that some of the unexplained state-bystate variation in use had diminished.
Implications for Policy, Delivery or Practice: Payment policy
may be one factor in the disproportionate decline in use for
some types of Medicare beneficiaries. To the extent that the
decline was caused by the reduction in fraud and abuse, we
can conclude that eligible beneficiaries have not lost access to
the benefit. However, if structures of the payment system
have caused access problems, then structural change is
necessary.
Primary Funding Source: Medicare Payment Advisory
Commission
• Long-Term Care Hospitals’ Role In Medicare Post-Acute
Care
Sally Kaplan, Ph.D., Chapin White, Ph.D.
Presented by: Sally Kaplan, Ph.D., Research Director,
Medicare Payment Advisory Commission, 601 New Jersey
Avenue, N.W., Suite 9000, Washington, DC 20001; Tel:
202.220.3700; Fax: 202.220.3759; E-mail:
skaplan@medpac.gov
Research Objective: To determine what role in Medicare
long-term care hospitals (LTCHs) play and where clinically
similar patients are treated in areas without them. LTCHs are
the most expensive and least frequently used post-acute care
setting and are not available in many areas. As the number of
LTCHs has doubled since 1993 and spending quintupled from
1993 to 2001, policymakers have asked more questions. This
study was designed to add to policymakers’ knowledge and
inform policy.
Study Design: Medicare administrative data were used to
construct over 5 million episodes of care that began in the first
six months of 2001. Episodes begin with admission to an
acute hospital and end with death, readmission, or no postacute care for 61 days. Two-stage regressions compare
Medicare spending and outcomes. A qualitative component
adds context and color to the quantitative analyses.
Population Studied: Beneficiaries enrolled in fee-for-service
Medicare.
Principal Findings: One of the most persistent questions
from policymakers is where clinically similar patients are
treated in areas without LTCHs. Some patients stay in the
acute hospital longer, moving from an intensive care unit
(ICU) to a medical bed or to a unit stepped down from ICUlevel care. A few are treated in inpatient rehabilitation facilities
and some are treated in skilled nursing facilities equipped to
handle patients with multiple complex illnesses or on
ventilators.
LTCH patients are stable but medically complex and some of
their underlying medical conditions may not have been
resolved before transfer to the LTCH. The probability of using
an LTCH increases with clinical severity and proximity to these
facilities. The strongest predictor of LTCH use is a diagnosis
of tracheostomy. Regardless of diagnosis, patients with the
highest level of severity are four times more likely to use an
LTCH. As proximity increases, the probability of using an
LTCH increases. Being admitted to an acute hospital that has
a co-located LTCH quadruples the probability that a patient
will use LTCH care.
Preliminary findings show that for all patients discharged alive
from the acute hospital in 2001, Medicare payments averaged
$12,000 for acute and post-acute care. For patients with the
greatest probability of using an LTCH, payments averaged
$30,000.
Conclusions: Medicare defines LTCHs exclusively by average
length of stay. The presence of an LTCH in an area or a
hospital strongly predicts the setting in which beneficiaries are
treated. The rapid growth in LTCHs and the potential for
overuse of this setting indicate that policymakers need to
better define these facilities and the patients appropriate for
care in them.
Implications for Policy, Delivery or Practice: Little has been
known about LTCHs. This research provides policymakers
with more information on the role of LTCHs in Medicare. It
also provides information on how Medicare spending and
outcomes compare for LTCHs patients and similar patients
treated in alternative settings.
Primary Funding Source: Medicare Payment Advisory
Commission
Call for Panels
Caring of the Elderly near the End-of-Life:
Studies of Hospice Care and Informal Care
Chair: David Grabowski, Ph.D.
Monday, June 7 • 2:00 p.m.- 3:30 p.m.
• Do Non-Profit and For-Profit Hospices Behave
Differently?
Richard Lindrooth, Ph.D
Presented by: Richard Lindrooth, Ph.D.
Research Objective: To test whether non-profit and for-profit
hospices respond differently to the financial incentives of
Medicare payment policy.
Study Design: A hospice can maximize profit maximizing
patient length of stay (LOS), because while daily costs are
high several days after admission and several days before
impending death, payment is fixed per day and does not vary
be diagnosis. Thus by maximizing LOS, the length of interim
period of relatively low costs, and higher reimbursement, is
also maximized. We posit that a hospice can influence LOS,
i.e. patient mix, in the following ways: it can offer attributes
that are attractive to patients with a diagnosis associated with
a long hospice LOS; market itself to physicians/ discharge
planners who treat patients with diagnoses associated with
long LOS; and encourage physicians to admit patients earlier
in the progression of the disease than they otherwise would.
We test these predictions using a dataset of all Medicare
hospice admissions in 1993. We examine the shares of each
diagnosis in for-profit and non-profit hospices, and test
whether there are more ex-ante profitable patients at for-profit
hospices. In addition, we test whether there are differences in
LOS within each diagnosis.
Population Studied: We use a dataset that consists of all
Medicare reimbursed hospice admissions in 1993 from the
Medicare SAF files. Hospice information was obtained from
the MEDPAR files. We have information from a 720 look back
window as well as a 2.5 year follow-up.
Principal Findings: Most of the variability in expected LOS is
explained by diagnosis, and thus diagnosis is an easily
observed patient trait that can be used to cream-skim. Of the
diagnoses with relatively high LOS, 12 out of 13 have higher
shares at for-profit hospices 10 of these differences are
significantly less than zero. Of the diagnoses with relatively
low expected LOS, 8 out of 14 have lower shares at for-profit
hospices, but 3 out of 14 also had a significantly higher share.
The results of the analysis by diagnosis are mixed and no
significant pattern emerged.
Conclusions: The results show that for-profit hospices
maximized profit and did attract a higher share of patients
with long expected LOS. The results within diagnosis reveal
that for-profit hospices were not able to differentially draw
relatively profitable patients within each diagnosis; rather they
attracted patients with relatively profitable diagnoses. This is
probably due to the fact that it is notoriously difficult to predict
LOS within a given diagnosis. The results tie directly to the
financial incentive inherent in Medicare reimbursement for
hospice care.
Implications for Policy, Delivery, and Practice: Medicare
reimbursement for hospices may cause access problems for
patients with diagnoses with relatively short lengths of stay
because they are unprofitable. We find that if Medicare would
reimburse hospices relatively high upon admissions and again
relatively high upon death, hospices will not have the incentive
to attract certain types of patients. Such a reimbursement
schedule would mirror costs more closely and eliminate the
incentive to admit only those patients with a long expected
LOS.
Primary Funding Source: Robert Wood Johnson Foundation
• Do Selection or Treatment Effects Explain Differences in
Medicare End-of-Life Costs among Hospice and Usual
Care Decedents?
Donald Taylor, Jr.
Presented by: Donald Taylor, Jr.,
Research Objective: To determine whether lower end-of-life
Medicare costs observed among decedents are due to
selection or treatment effects.
Study Design: The use of hospice has increased steadily
since it became a covered Medicare benefit in 1982, with
nearly 20 percent of Medicare decedents using hospice care in
1999. Determining whether the hospice benefit truly reduces
total Medicare costs for end-of-life care compared to usual
care has been difficult because it requires disentangling
treatment effects from selection effects. Selection implies that
persons choosing hospice are systematically different utilizers
of health care than usual care decedents, while treatment
implies that hospice truly reduces total Medicare costs of endof-life care after controlling for selection. We use hazard
models to control for the different propensities to use hospice
care among all decedents, and create a matched control
sample based on daily predicted probabilities of hospice care
prior to death between hospice users and non-users. We then
compared costs per day from hospice entry to death to costs
per day over the same period of time for the matched
controls. We also explored whether the results differed by the
following diagnoses: cancer, heart disease, COPD, and
neurological degenerative disorders, including Alzheimer’s
disease.
Population Studied: We analyzed 9,127 members of a
screening sample for the National Long Term Care Survey
who died between July 1, 1991 and December 31, 1999. We
used Medicare claims records of all types, including inpatient,
outpatient, home health, physician supplier, SNF, DME, and
hospice, to determine total Medicare-financed costs among
decedents from January 1, 1991 until death or the censor date.
At a minimum we had costs for six months prior to a person’s
death, and data over 8.5 years.
Principal Findings: Bivariate findings show that costs during
the last 30 days of life were significantly lower for those using
hospice, $4,498, compared to usual care, $6,227, a pattern
that remained when considering the last 60 days of life,
$8,320 versus $9,577. P-values were less than 0.02 in both
cases. Costs during the last 90 days of life did not differ by
hospice use, $11,421 versus $11,708, p-value equal to 0.19.
When comparing costs among hospice users and usual care
decedents using our matched approach that controls for
selection effects, we found that hospice costs were 6% lower
per day compared to controls. The p-value was less than 0.01.
Furthermore, differences were larger for cancer decedents
than for other persons using hospice.
Conclusions: The results suggest that hospice has a true
treatment effect on total Medicare costs after the point of
entry into hospice; differences are not only due to selection.
Implications for Policy, Delivery, and Practice: Expanding
hospice care to a larger proportion of Medicare beneficiaries
needing end-of-life care may reduce rather than increase the
total costs to the Medicare program.
Primary Funding Source: Robert Wood Johnson Foundation
Call for Panels
New Research on Health Literacy
Chair: David Howard, Ph.D.
Monday, June 7 • 2:00 p.m.- 3:30 p.m.
• Health Literacy Research and the IOM Report
David Kindig, M.D.
Presented by: David Kindig, M.D., Professor, Population
Health Sciences, University of Wisconsin, 610 Walnut Street,
Madison, WI 53726; Tel: 608.263.6294; Fax: 608.262.6404; Email: dakindig@facstaff.wisc.edu
Research Objective: The forthcoming Institute of Medicine
(IOM) report provides an up-to-date summary of all the
research that has been conducted on the subject of health
literacy. However, many unanswered questions remain. The
purpose of this presentation is to describe the unfinished
research agenda on health literacy, focusing on the topics on
which IOM panel members believe more research is necessary
for advancing knowledge in the field.
Study Design: Review of the literature and testimony before
the IOM panel.
Principal Findings: The following topics were identified as
important areas for future research.
1. What interventions can successfully improve health literacy
or reduce barriers to persons with low health literacy?
2. Does low health literacy explain racial and socioeconomic
disparities?
3. What are the fiscal implications of policies to improve
health literacy?
4. What are the implications of low health literacy for
Medicaid?
5. What are the implications of low health literacy for public
health communication?
6. Is aging associated with deteriorating health literacy?
7. What is the relationship between prior educational
attainment and health literacy?
8. How should interventions to improve health literacy be
incorporated into the health system? Who should pay for
them?
Conclusions: The presentation will give an overview of each
topic and why it is important for formulating policies to
address low health literacy. It will conclude with a description
of resources for research in the field of health literacy.
Primary Funding Source: Institute of Medicine, NAS
• The Prevalence of Low Health Literacy
Michael Paasche-Orlow, M.D., M.P.H., Julie A Gazmararian,
Ph.D., Ruth Parker, M.D.
Presented by: Michael Paasche-Orlow, MD, MPH, Assistant
Professor of Medicine, Department of Medicine, Boston
University, 91 East Concord Street, Boston, MA 02118; Tel:
(617) 414-5877; Fax: 617.414.4676; E-mail:
Michael.PaascheOrlow@bmc.org
Research Objective: The Institute of Medicine has identified
health literacy along with self-management as a national
priority area for the promotion of health care quality. No
study, however, has systematically evaluated the prevalence of
low health literacy reported in the medical literature or the
methods used for this research.
Study Design: We conducted a review of the literature to
summarize the methods and findings of studies that examine
the prevalence of low health literacy conducted in the United
States and to synthesize these findings by comparing the
characteristics of studies with similar and dissimilar results.
Articles and abstracts pertaining to health literacy were
identified by experts active in the field and through MEDLINE,
CINAHL, PsychInfo, and Sociological Abstracts database
searches for the period 1963 through October 2003. Of 1376
abstracts reviewed, 81 studies met specified criteria:
presentation of primary data on the prevalence of health
literacy; description of the study population, identification of a
measurement instrument, and data collection methods; study
conducted in the United States with >25 adult subjects. Low
literacy was defined as the rate of subjects scoring at an
Inadequate level on TOFHLA or 6th and below on other
measures. Weighted analysis of variance was used to compare
the mean rates of low literacy according to quartiles of
demographic characteristics.
Principal Findings: The studies reviewed include data on
30,351 subjects, and report a weighted mean prevalence of low
health literacy of 25%, 95% CI 22% to 29% (range, 0 to 54).
Subjects in 34/81studies were excluded if they did not speak
English; vision and cognitive function were mentioned as
selection criteria in 15 and 7 studies, respectively. While 16
different instruments were used to evaluate literacy, 68%
(55/81) of the studies used either the REALM or the TOFHLA
and had similar rates of low literacy. Literacy was not
associated with the rate of female (P=0.14) or Caucasian
subjects (P=0.26). Low literacy was associated with higher
rates of failure to complete high school (P=0.02), African
American subjects (P=0.04), and older subjects (P=0.0002).
Data for Spanish language testing was separately reported for
5% (1,504/30,351) of subjects and revealed a higher rate of low
literacy than for those tested in English (44% versus 25%,
P=0.002).
Conclusions: One in four subjects tested were found to be
low literate. The instruments used to measure literacy,
populations sampled, and study methods varied across the
reviewed studies and influence the prevalence estimates
presented. Despite significant methodological differences,
low literacy is consistently associated with level of education,
ethnicity, and age
Implications for Policy, Delivery or Practice: Low literacy is
pervasive. Patient education and efforts to simplify the
healthcare system must be advanced.
• Impact of Low Health Literacy on Medical Costs
David Howard, Ph.D., Julie Gazmararian, Ph.D.
Presented by: David Howard, Ph.D., Assistant Professor,
Department of Health Policy and Management, Emory
University, 1518 Clifton Road NE, Atlanta, GA 30322; Tel: (404)
727-3907; Fax: 404.727.9198; E-mail: dhhowar@emory.edu
Research Objective: Despite the prevalence of low health
literacy and the number of studies documenting its impact on
health outcomes, there is a paucity of information on the
implications for medical costs. The objective of this study was
to examine the impact of low health literacy on medical costs
and resource use.
Study Design: Costs incurred over a one year period were
compared between persons with low and adequate health
literacy using propensity scores. Propensity scores were based
on age, race, income, education, and chronic conditions.
Population Studied: A total of 3,260 non-institutionalized
elderly persons enrolling in one of four Medicare managed
health care plans (Cleveland, Ohio; Houston, Texas; South
Florida; Tampa, Florida) in 1997.
Principal Findings: Controlling for observed covariates,
persons with low health literacy incur expenditures for
emergency room care that are $86 (66% of the sample
average emergency room expenditures) higher (1997 U.S.
dollars) (P<0.01) than persons with adequate health literacy.
Costs for outpatient care and prescription drugs are $137 (7%)
lower (P = 0.42) and $27 (4%) higher (P = 0.60), respectively.
The cost of inpatient care was $979 (18%) higher (P = 0.25)
but statistically insignificant, possibly due to the extreme
skewness of the inpatient spending distribution. When
persons incurring inpatient expenditures above the 99th
percentile were excluded, the difference was $1,017 (19%) and
significant at the 5% level.
Conclusions: Persons with low health literacy incur higher
expenditures for emergency room care. Comparisons of the
likelihood of using inpatient care, inpatient admissions, and
costs with outliers excluded indicate that persons with low
health literacy use more inpatient resources as well. Future
research should seek data sources with larger sample sizes,
examine the sensitivity of results to alternative models for
comparing health spending, and determine whether results
are consistent across different demographic groups.
Implications for Policy, Delivery or Practice: Programs to
improve health literacy and simplify communication of health
information have the potential to reduce health care costs.
Primary Funding Source: Pfizer Foundation
• Chronic Disease Management Mitigates the Relationship
Between Literacy and Health Outcomes
Darren DeWalt, M.D., Michael Pignone, M.D., M.P.H., Russell
Rothman, M.D., M.P.P., Morris Weinberger, Ph.D.
Presented by: Darren DeWalt, M.D., Robert Wood Johnson
Clinical Scholar, Internal Medicine and Pediatrics, University
of North Carolina, 5038 Old Clinic Building, CB#7105, Chapel
Hill, NC 27599; Tel: 919.966.3712; Fax: 919.843.9237; E-mail:
darren_dewalt@med.unc.edu
Research Objective: To evaluate the role of patient literacy in
the effectiveness of chronic disease management programs
for diabetes and heart failure.
Study Design: We have performed two randomized
controlled trials of disease management interventions for
patients with diabetes and heart failure in our university based
general internal medicine clinic. The interventions included a
one-on-one education session with a clinical pharmacist or
health educator using easy to read educational materials
focused on basic self-care skills. Patients then received several
follow-up phone calls and regular communication with the
care coordinator who helped to reduce barriers to care.
Control patients received usual care by their physician.
Population Studied: For the diabetes program, we enrolled
217 patients with poor glucose control and measured literacy
with the Rapid Estimate of Adult Literacy in Medicine
(REALM). The main study outcome was glucose control
measured as glycated hemoglobin (A1C). For the heart failure
program, we enrolled 129 patients with New York Heart
Association class II-IV symptoms and measured literacy with
the Test of Functional Health Literacy in Adults (TOFHLA).
The main study outcomes were HF related quality of life
(HFQOL) and a combined end-point of hospitalization or
death.
Principal Findings: After 12 months follow-up for patients
with diabetes, we found that intervention patients had greater
improvement in A1C than control patients(-2.7% vs –1.5%,
p=0.01). Among patient with literacy levels less than 7th
grade, improvement in A1C was 1.3 points more for
intervention than control patients (CI 0.6, 2.1), but in patients
with literacy levels at or above 7th grade, improvement in A1C
was similar between control and intervention patients (0.4 (CI
–0.4, 1.3). The intervention increased the odds of obtaining
goal A1C (<7.0) more for patients with low literacy (adjusted
OR 6.3 (CI 1.9, 21.1) than for those with higher literacy
(adjusted OR 1.0 (CI 0.5, 2.1). After 6 months follow-up for
patients with heart failure, we found that patients in the
intervention group had improvements in HF specific
knowledge and self-efficacy, but no differences in HFQOL
compared to the control group. After adjusting for baseline
differences between the groups, the incidence rate ratio (IRR)
for hospital admission or death was 0.68 (CI 0.36, 1.32). A
subgroup analysis of patients categorized as having
inadequate literacy on the TOFHLA (n=53) revealed an
adjusted IRR of 0.35 (CI 0.13, 0.94).
Conclusions: Disease management programs that address
literacy appear to have greater effect among patients with low
literacy and may provide a key strategy for mitigating the
relationship between low literacy and adverse health
outcomes.
Implications for Policy, Delivery or Practice: Health care
providers for vulnerable populations should consider disease
management as an effective strategy to improve health
outcomes for patients with low literacy.
Primary Funding Source: Pfizer Foundation
Call for Panels
Evaluating a New Breast Cancer Procedure:
Values in Conflict
Chair: Richard Rettig, Ph.D.
Monday, June 7 • 4:00 p.m.- 5:30 p.m.
• Values in Conflict: HDC/ABMT Legal Issues
Peter Jacobson, J.D., M.P.H.
Presented by: Peter Jacobson, J.D., M.P.H., Associate
Professor, School of Public Health, University of Michigan,
109 Observatory, Ann Arbor, MI 48109-2029; Tel: (734) 9360928; Fax: (734) 764-4338; E-mail: pdj@umich.edu
Research Objective: To understand: (1) Key strategies for
both plaintiffs’ and defense attorneys in HDC/ABMT litigation;
(2) judicial trends in HDC/ABMT decisions; and (3) the
implications for courts in resolving similar clinical scientific
controversies.
Study Design: We interviewed eight leading defense and
plaintiffs’ attorneys (four each) in the HDC/ABMT litigation.
Together they have handled hundreds of these cases,
including jury trials, judicial and arbitration hearings, and
settlement negotiations. We identified the attorneys by asking
knowledgeable experts at insurance companies and elsewhere
familiar with the litigation. Interviews were designed to elicit
discussion of the following topics: 1) HDC/ABMT litigation
experience; 2) litigation and negotiation strategies, including
changes over time; 3) respondents’ analysis of the reported
case law; and 4) lessons learned. We also collected
supportive documentary evidence when available (e.g., nonconfidential transcripts and legal briefs). We also analyzed
trends in HDC/ABMT litigation by reading and classifying all
reported judicial cases.
Population Studied: Attorneys.
Principal Findings: Each case involved four interrelated legal
issues: (1) Contract: plaintiffs’ attorneys were able to show
that contractual exclusions for HDC/ABMT were ambiguous
or inconsistent with insurers’ marketing information. (2)
Standard of care: plaintiffs’ attorneys demonstrated that the
procedure had diffused widely among academic and
community oncologists and transplanters; defense attorneys
relied on lack of proven effectiveness. (3) Bad faith (the key
strategic issue): plaintiffs’ used inconsistent insurer decisions
to argue bad faith; defense argued that coverage denial was
legitimate, hence, no bad faith. (4) Informed consent: defense
stressed informed consent to show that patients knew the
procedure was experimental; courts ignored this argument.
Important additional strategies included: (5) Expert witnesses:
defense relied on the insurers’ medical directors, who lacked
credibility, because other experts unwilling to testify; plaintiffs
used treating physicians and community oncologists. (6) The
clinical science: the absence of effectiveness data should have
been an advantage for defense; but plaintiffs successfully
argued that insurers covered other unproven procedures;
moreover, HDC/ABMT was not laetrile, i.e., a quack
treatment. (7) Sympathy and emotion: strongly favored
plaintiffs, as did seemingly insensitive insurer behavior.
Plaintiffs won approximately 50% of the litigated cases,
contrary to widespread perceptions. No data were available
on negotiated settlements.
Conclusions: Strategically, the cases were basically struggles
over whether insurers acted in good faith or unreasonably
denied appropriate care. Defense attorneys were
disadvantaged because of the procedure’s widespread
diffusion, the lack of expert witnesses, insurer arrogance, and
uncertainty about scientific benefit. Oddly, both sides were
correct: the procedure was indeed widespread, but lacked
justifying scientific data and was experimental. Courts should
have disregarded physician custom and ruled that the
procedure was scientifically unproven.
Implications for Policy, Delivery or Practice: First, courts are
not the right forum for resolving controversial clinical scientific
disputes. Policymakers need to focus on improving the
technology assessment process and funding randomized
control trials to determine a procedure’s effectiveness.
Second, like it or not, the legal system performed as it is
designed to do—to protect individual litigants. Insurers will
have difficulty denying similar procedures if solid evidence of
effectiveness is lacking; evidence of ineffectiveness will
strengthen their position. The nation’s culture of technology
is embedded too deeply to accept limits very easily.
Primary Funding Source: RWJF
HDC/ABMT. Since then a total of 14 RCTs have been
published or presented and no benefit has been shown in
overall survival. Treatment related deaths continue to be a
concern.
Conclusions: The widespread availability of HDC/ABMT in
the 1990s hampered the ability of RCTs to accrue patients.
The failure to allow an evidence-based evaluation of the
procedure to inform decision-making led to a decade of an
expensive and harmful treatment of a group of women who
had a limited life expectancy. This case study highlights the
problems of conducting RCTs when a procedure is widely
available and of making evidence-based coverage decisions
under such circumstances.
Primary Funding Source: RWJF, Commonwealth Fund
• Women with breast cancer & HDC/ABMT in the US
Cynthia Farquhar, M.B., Ch.B., M.D., M.P.H.
• Insurers' Perspectives on HDC/ABMT
Wade Aubry, M.D.
Presented by: Cynthia Farquhar, M.B., Ch.B., M.D., M.P.H.,
Associate Professor of Reproductive Medicine, Obstetrics and
Gynecology, University of Auckland, National Women's
Hospital, Auckland, 0 ; Tel: +64 9 630 9943 x3240; Fax: +64 9
630 9585; E-mail: c.farquhar@auckland.ac.nz
Research Objective: To determine the number of women in
the US who received HDC/ABMT for breast cancer during the
1990s.
Study Design: Analysis of AHRQ Health Care Utilization
Project (HCUP) inpatient data.
Population Studied: Women with metastatic and high-risk
breast cancer.
Principal Findings: HDC/ABMT was initially used to treat
patients with hematological malignancies. However, in the
late 1980s it began to be used to treat women with either
metastatic or high-risk breast cancer. The rationale for the
procedure was that if a woman responded to conventional
chemotherapy doses, then higher doses would result in
improved success rates; and these higher doses could be
made possible by harvesting and reinfusing a patient’s own
bone marrow. There is little consensus on when and how
access to experimental interventions should be made
available. Do small, single-site Phase 2 studies provide
adequate data or are large, multi-site Phase 3 randomized
clinical trials required to show effectiveness. In the private
sector, coverage decisions are made by health insurers and
health plans, usually after consideration of both the evidence
of effectiveness of the procedure and its medical necessity for
an individual patient. Most recipients of the HDC/ABMT
procedure were younger women, and not Medicare-eligible.
Thus, coverage requests that were first received in the late
1980s were addressed to private insurers; and these requests
were declined by most insurance companies. Eventually,
litigation, state and other mandates resulted in many health
insurance companies providing coverage. From 1993 until
2001 more than 20,000 women in the United States with the
diagnosis of breast cancer received either HDC/ABMT as
inpatients (HCUP database of the AHRQ). (A higher estimate
based on the North American Transplant Registry data puts
the number of women close to 40,000.) The inpatient
mortality rate fell from 4.7% in 1993 to 3.5% in 2001. Cost and
length of stay also declined over this time period. The results
of 5 RCTs (including 2 US trials) first presented in 1999, with
one exception, showed no overall survival benefit from
Presented by: Wade Aubry, M.D., Associate Clinical Professor
of Medicine, Institute for Health Policy Studies, University of
California San Francisco, 3333 California Street,Suite 265, San
Francisco, CA 94118; Tel: (415) 476-0615; Fax: (415) 476-0705;
E-mail: wmaubry@itsa.ucsf.edu
Research Objective: To describe health insurers'perspectives
on HDC/ABMT, both short- and long-term.
Study Design: Personal experience as former SrVP of Blue
Shield of California; former chairman of Medical Advisory
Panel to Blue Cross Blue Shield Association; interviews with
key participants; document analysis.
Principal Findings: The emergence and demand for
HDC/ABMT to treat breast cancer placed new pressures on
health insurers that contributed significantly to voluntary and
regulatory changes in how coverage decisions are made.
These changes include independent medical review, support
for patient care costs of selected qualifying clinical trials, and
greater emphasis on scientific evidence, consistency of
decision making, and increasing transparency of processes to
ensure fairness to enrollees and limit legal risk and bad
publicity for insurers. From an insurer’s point of view,
HDC/ABMT was in a class by itself. No other single
technology came close to requiring as much medical director
time, to generating as much emotion, as much liability risk, or
as much impact on coverage decision-making. By the late
1980s, medical necessity clauses in health insurance contracts
and processes for reviewing new technologies and developing
coverage guidelines were standard practice in most health
insurance organizations (indemnity plans, HMOs, PPOs, and
self-funded plans). Although many technologies had been
subjected to review and policy development, disputes over
coverage did not merit much public attention or liability risk
for insurers. Attempts to treat HDC/ABMT the same as other
technologies were largely unsuccessful for many reasons,
including the lack of effective alternatives for desperate
patients seeking treatment, the high cost of the procedure in
an era of rising health care costs and managed care, the
efforts of advocacy groups, ambiguities in contracts such as
separate benefits for chemotherapy and bone marrow
transplantation, and the willingness of reputable physicians to
provide “off protocol” treatment and support lawsuits against
insurers.
Conclusions: Insurers responded in many ways and
ultimately helped support clinical research that demonstrated
the lack of benefit of HDC/ABMT in the treatment of breast
cancer. First and foremost, insurers conducted and
commissioned many technology assessments that
demonstrated repeatedly the lack of evidence of effectiveness
and the need to conduct randomized clinical trials of
HDC/ABMT for breast cancer. After the Fox v. Health Net
verdict in 1993 for $89m, many insurers paid for cases to
avoid the possibility of huge punitive damage awards by
sympathetic juries. Voluntary independent medical review
(IMR) programs for experimental therapies began in the early
1990s primarily to review these cases, later expanding to
reviews of a broader range of disputed emerging technologies.
Forty-one states and the District of Columbia have now
mandated external review programs for medical necessity
denials. Blue Cross Blue Shield Plans and other commercial
insurers developed programs to support the patient care costs
of clinical trials, which have subsequently led to state
mandates and Medicare regulations for coverage of patient
care costs of qualifying clinical trials.
Primary Funding Source: RWJF
• HDC/ABMT: Implications for Policy, Delivery and
Practice
Richard Rettig, Ph.D., Peter Jacobson, J.D., M.P.H., Cynthia
Farquhar, M.D., M.P.H., Wade Aubry, M.D.
Presented by: Richard Rettig, Ph.D., Senior Social Scientist,
DC/RAND Health, RAND, 1200 South Hayes Street,
Arlington, VA 22202-5050; Tel: 703.413.1100 x5299; Fax:
703.414.4717; E-mail: rettig@rand.org
Research Objective: To tell the story of high-dose
chemotherapy with autologous bone marrow transplantation
(HDC/ABMT) for treating metastatic and high-risk breast
cancer; and to draw lessons for the future.
Study Design: Multiple methodologies of interviews, legal
case analysis, analysis of HCUP inpatient data, document
analysis.
Population Studied: the participants in the HDC/ABMT story
– women patients, physicians, patient advocates, lawyers,
judges, juries, insurers, state legislatures, technology
assessors, clinical trialists, and the press.
Principal Findings: HDC/ABMT, as a medical procedure, was
not required to be evaluated by a Phase 3 randomized clinical
trial (RCT), as FDA requires of new drugs. Some in the
medical profession, some insurers, and the National Cancer
Institute, supported RCTs. Others in oncology legitimated the
procedure’s use on the basis of Phase 2 studies. Physician
legitimation, patient demands, patient advocacy, litigation,
state legislative mandates, drove utilization of the procedure
concurrent with its evaluation by RCTs. Accrual of patients to
clinical trials was severely hampered by the availability of the
procedure.
Conclusions: The basic conflict in the HDC/ABMT experience
is between the need to listen and respond to women with a
terminal diagnosis who wish access to an experimental
treatment and, on the other hand, the need to protect and
maintain the integrity of the evidence-based evaluation
process. Some institutions (courts, plaintiffs’ attorneys, the
press, patient advocates) respond well to individual patients;
others (insurers, technology assessors, clinical trialists)
respond well to the need for population-based data on
effectiveness; a few (independent medical review, providers of
information for patients, and some patient advocates) attempt
to respond to both.
Implications for Policy, Delivery or Practice: Effective policy
for dealing with new medical procedures must devise
transparent ways to respond to both of the conflicting values
of patient demands for access to experimental treatment AND
to the need to protect & maintain the evaluation process. At
the emergence of a new procedure, NCI, the medical
profession, health insurers, and patient advocates should
provide clear information about known risks and benefits to
prospective patients for the treatment in question on a far
more collaborative basis than in the past. Financing of Phase 3
trials of procedures should receive sustained attention.
Educational efforts about clinical trials should be directed to
patients, advocates, judges, and the media. Technology
assessments should be more available to the public. The
response to demands for early access to experimental
procedures must be active and evidence-based, coupled with
independent medical review.
Primary Funding Source: RWJF
Call for Panels
Will the New Medicare Drug Plan Lead to Lower Spending
on Medical Services?
Chair: Geoffrey Joyce, Ph.D.
Monday, June 7 • 4:00 p.m.- 5:30 p.m.
• How Much Would a Medicare Prescription Drug Benefit
Cost? Offsets in Medicare Part A Costs by Increased Drug
Use
Zhou Yang, Ph.D., Edward Norton, Ph.D.
Presented by: Zhou Yang, Ph.D., Assistant Professor,
Department of Medicine, Michigan State University, B412
Clinical Center, East Lansing, MI 48824; Tel: 517.432.8653; Fax:
517.432.9471; E-mail: zhou.yang@ht.msu.edu
Research Objective: To investigate the effect of outpatient
prescription drugs utilization on demand for inpatient care
and Medicare Part A cost and quantify the offsets in Part A
cost from more drug use among elderly Medicare
beneficiaries.
Study Design: This study estimates a set of dynamic
simultaneous demand equations of outpatient prescription
drugs and other Medicare covered health services based on
health capital and health production function theory to
quantify the effects of outpatient prescription drugs use on
subsequent demand for inpatient care and Medicare Part A
cost. The longitudinal Medicare Current Beneficiary Survey
data from 1992 to 1998 is used to estimate the empirical
model. Demographic features, insurance status, functional
status and health shocks are controlled as independent
variables in all the demand equations. Discrete random effect
is used to control for unobserved individual level factors that
may lead to biased estimation of the relationship between
drug use and other medical care cost.
Population Studied: Medicare beneficiaries age 65 or more,
from 1992 to 1998.
Principal Findings: Increase in outpatient prescription drugs
use leads to a minor, but significant offsets in subsequent
Medicare Part A cost at 4.4 percent on average for the entire
sample over a year. The magnitudes of the offsets differ
across different subgroups. The offset rates of Medicare Part
A cost from more drug use are higher among people without
disability (offset rate 4.9%), with chronic diseases (diabetes,
offset rate 10.9%), in lower income (less than $15,000 a year,
offset rate 7.3%) than people with disability, without chronic
diseases and in higher income level.
Conclusions: More outpatient prescription drugs use may
help the elderly Medicare beneficiaries to maintain better
health, and reduce their subsequent demand for inpatient care
as well as Medicare Part A cost. Increase in prescription
drugs utilization could be more influential among people in
low income or high demand for drugs, like chronically ill
patients to help them rely less on inpatient care.
Implications for Policy, Delivery or Practice: The future cost
of Medicare drug benefit may not be as much as the
predictions from previous research that did not consider
possible offsets in Medicare Part A cost from the increased
outpatient prescription drugs consumption stimulated by the
drug benefit. However, the magnitudes of the offsets are not
as optimistic as predicted by other existing research that
concluded the savings in hospital care cost could even exceed
the increase in drug cost. Further research consider both the
immediate effect of drug use on demand for inpatient care
and long term effect of drug use on general health, disability,
mortality and total demand for inpatient care at the
population level will inform the Medicare drug policy debate
even better.
Primary Funding Source: NIA
• Prescription Drug Coverage, Drug Vintage and Medical
Expenditure of Medicare Beneficiaries: Evidence from the
Medical Expenditure Panel Survey
Frank Lichtenberg, Ph.D.
Presented by: Frank Lichtenberg, Ph.D., Professor, 614 Uris
Hall, Columbia University, 3022 Broadway, New York, NY
10027; Tel: 212.854.4408; Fax: 212.316.9219; E-mail:
frl1@columbia.edu
Research Objective: To examine the impact of prescription
drug insurance coverage on the vintage of pharmaceuticals
used, on the quantity of non-drug medical services (e.g.
hospital admissions and emergency-room visits), and on
overall medical expenditure.
Study Design: Estimation of medical care utilization
equations, by type of medical care, person, and medical
condition, using 1996-1998 Medical Expenditure Panel Survey
data, linked to data on FDA approval dates of prescribed
medicines. Explanatory variable of primary interest is the
mean vintage (years since initial FDA approval) of the active
ingredients contained in the prescribed medicines used by an
individual to treat a given condition. Availability of data by
person by condition allows us to control for unobserved
person fixed effects as well as condition fixed effects.
Population Studied: Entire U.S. population, and the Medicare
population, during 1996-1998.
Principal Findings: (1) Medicare beneficiaries with private
prescription drug coverage during 1996-1998 used newer
drugs than those without such coverage. (2) Use of newer
drugs is associated with lower utilization of non-drug medical
services, especially hospital care, controlling for fixed person
and condition effects. (3) The use of newer drugs reduces
non-drug medical costs much more than it increases drug
costs.
Conclusions: Prescription drug coverage affects (increases)
the vintage, or quality, as well as the quantity, of prescription
drugs. Enactment of the Medicare drug benefit is expected to
encourage use of newer medicines, and thereby to reduce
Medicare's non-drug costs, especially hospital costs.
Implications for Policy, Delivery or Practice: Government
and private payers should account for the effects of
prescription drug coverage on the vintage, or quality, of drugs
utilized, and for the indirect effects of this coverage on other
medical costs.
Primary Funding Source: National Pharmaceutical Council
• Will the Substitution of Drug for Non-Drug Care under a
Voluntary Drug Benefit Lower Medicare Spending?
Evidence from a State Pharmacy Assistance Program
Boyd Gilman, Ph.D., Barbara Gage, Ph.D.
Presented by: Boyd Gilman, Ph.D., Senior Economist, , RTI
International, 411 Waverley Oaks Road, Suite 330, Waltham,
MA 02452; Tel: 781.788.8100; Fax: 781.788.8101; E-mail:
bgilman@rti.org
Research Objective: To evaluate the impact of outpatient
drug coverage for low-income elderly and disabled
beneficiaries on the use and cost of non-drug medical
services.
Study Design: Medicaid eligibility files were used to identify
enrollees in two pharmacy assistance programs for lowincome elderly and disabled beneficiaries between 1995 and
1999. Medicare claims for inpatient, outpatient, professional,
home health and DME services were extracted and merged for
enrollees and non-participating beneficiaries in the same state.
Medicare costs were summed by year and type of service for
each beneficiary. The study used a difference-in-differences
model to estimate the change in Part A and B spending
among state program participants before versus after
enrollment relative to the change in Medicare spending
among non-participating beneficiaries over the same period.
The two-stage utilization and expenditure models controlled
for observable between-group differences associated with
beneficiary age, sex, race, health status, disability, dual
eligibility and residence. Observable health status
characteristics were also controlled for by using a prospective
Hierarchical Coexisting Condition (HCC) index based on
beneficiary demographic characteristics and prior year
inpatient and Part B diagnoses.
Population Studied: Elderly and disabled Medicare
beneficiaries residing in Vermont between 1995 and 1999.
The ever-enrolled population included over 12,000
beneficiaries. The never-enrolled population totaled
approximately 100,000 Medicare beneficiaries.
Principal Findings: The study failed to reveal any conclusive
evidence of overall savings to the Medicare program
associated with the provision of outpatient prescription drug
coverage and a subsequent substitution of drug for non-drug
medical services. The total savings effect was dominated by
an initial year spike in expenditures, particularly for inpatient
services, suggesting that beneficiaries are most likely to apply
for public drug coverage only after they experience an acute
hospitalization and a subsequent need for outpatient
prescription medication. After the initial year spike in
Medicare payments, expenditures largely returned to trended
baseline levels.
Conclusions: The findings suggest that, when measured
across a broad sample of low-income Medicare beneficiaries,
drug coverage is unlikely to significantly reduce the rate and
intensity of average medical resource use and, thus, will not
result in any offsetting reduction in total Part A and B
expenditures. However, despite our best efforts, the savings
estimate may still be biased due to unobserved differences in
health status between those who choose to enroll in a
voluntary subsidized program and those who do not. Adverse
selection stemming from baseline differences will cause
savings to be underestimated. Selection emanating from a
precipitating acute care episode and subsequent regression to
the mean will cause savings to be overestimated.
Implications for Policy, Delivery or Practice: Based on the
results from our study, there is no compelling reason to date
to make an adjustment in the net cost projections of the
Medicare Part D drug benefit plan based on a substitution of
drug for non-drug services when measured over a nonstratified sample of beneficiaries. Further efforts to identify
the source of adverse selection and control for unobservable
differences in health status accordingly are warranted.
Primary Funding Source: CMS
suggest that Medicare beneficiaries with private drug coverage
have significantly lower non-drug spending than those without
drug coverage, although the savings vary across sources of
coverage and types of service.
Implications for Policy, Delivery or Practice: Perhaps the
most important takeaway is the strong evidence of
unobserved selection bias into drug coverage. This documents
that concerns about risk selection are indeed real and calls
into question the viability of standalone drug coverage. The
other important lesson from this study is the confirmation
that managed care matters. There is strong evidence that drug
coverage from Medicare HMOs offers the most significant
opportunity to realize any savings effect.
Primary Funding Source: Consortium of Pharmaceutical
Companies
• Selection Bias and the Effects of Prescription Drug
Coverage on Non-Drug Medical Spending: Evidence from
the 1994-1999 Medicare Current Beneficiary Survey
Michael Furukawa, M.S., ABD
• Measure, Learn, and Improve: Have Physicians Begun to
Engage in the Quality Improvement Cycle?
Anne-Marie Audet, M.D., M.Sc., S.M., Stephen Schoenbaum,
M.D., M.P.H., Michelle Doty, Ph.D., M.P.H., Jamil Shamasdin
Presented by: Michael Furukawa, M.S., ABD, The Wharton
School, University of Pennsylvania, 3641 Locust Walk,
Philadelphia, PA 19104; Tel: 215.575.6165;E-mail:
mifuruka@wharton.upenn.edu
Research Objective: To identify the presence of unobserved
adverse or favorable selection bias into private supplemental
plans with drug coverage and estimate the selectioncontrolled effect of drug coverage on total non-drug and types
of medical spending.
Study Design: Using the Medicare Current Beneficiary Survey
from 1994-1999, I estimate the effect of drug coverage from
private supplemental plans on total and type of non-drug
medical spending. I explicitly address self-selection into drug
coverage by specifying and estimating endogenous treatment
effect models using IV and control function estimators.
Population Studied: Elderly Medicare beneficiaries with
private supplemental insurance plans (employer-sponsored,
individually-purchased Medigap and Medicare HMO)
Principal Findings: I find strong evidence that naïve drug
coverage effects are biased by unobserved adverse selection
into employer and Medigap drug coverage and unobserved
favorable selection into Medicare HMO drug coverage. After
controlling for selection bias, I find evidence of significant
savings on total non-drug spending for employer and
Medicare HMO drug coverage. These effects are driven by
lower total ambulatory spending and lower probability of any
post-acute spending. There is no evidence of significant
savings on total acute or total post-acute spending.
Conclusions: Not controlling for these unobserved adverse or
favorable selection induces bias into the estimates of the drug
coverage savings effect. The selection-corrected estimates
Presented by: Anne-Marie Audet, M.D., M.Sc., S.M., Assistant
Vice President, Quality Improvement, Quality of Care, The
Commonwealth Fund, One East 75th Street, New York, NY
10021; E-mail: ama@cmwf.org
Research Objective: Payers and regulatory and oversight
organizations have shown interest in using quality
improvement principles and physician performance measures
to improve health care. But little is known about how
physicians themselves use data to monitor and improve the
care they deliver.
We conducted a national survey of physicians with the
following objectives: 1) to explore whether physicians have
adopted basic quality improvement principles (i.e., measure,
learn, improve); 2) to identify the quality-of-care data to which
physicians say they have access; 3) to describe physicians’
willingness to share these data; and 4) to determine whether
physicians engage in quality improvement activities.
Study Design: Mail survey completed by a national random
sample of 3,598 physicians caring for adult patients. The
survey was conducted between March and May 2003. The
survey sample was randomly selected from a national list the
American Medical Association (AMA) that includes both AMA
members and non-members. All physicians in the sample
were involved in the direct care of adult patients and had been
in practice at least three years post-residency.
Population Studied: National random sample of US
physicians caring for adult patients.
Principal Findings: The response rate was 52.8% (1,837
physicians). Forty-three percent of respondents said they have
easy access to data about their patients’ clinical profile. Seven
of eight find it difficult or impossible to identify patients who
Call for Panels
Measure, Learn & Improve: Is the Science of Quality
Improvement Applied by Physicians? What Can be Done
to Accelerate Adoption?
Chair: Anne-Marie Audet, M.D., M.Sc., S.M.
Tuesday, June 8 • 11:15 a.m.- 12:45 p.m.
have abnormal laboratory results (84%) or take specific
medications (85%). One-half of physicians do not have
access to any data about the quality of the care they deliver.
Physicians in group practices with more than 50 members are
much more likely than solo physicians to have access to these
data (adjusted OR=2.14, p<0.001). Specialists are less likely to
have data on their quality compared with primary care
physicians (OR = 0.38, p <0.05). Health plans are the most
common source of quality-of-care data (25%). Only 14% of
physicians generate their own data. One-third of physicians
participate in quality improvement efforts. Physicians in
groups larger than 50 are more likely to participate than solo
physicians (OR = 2.38, p <0.05). Thirteen percent “definitely
agree” that performance data should be shared with their own
patients, while 45% disagree. Seventy percent said these data
should “probably not” or “definitely not” be shared with the
public.
Conclusions: The results of the survey suggest that
physicians are not making full clinical use of available data
about their own practice to guide their care. Although they no
doubt strive to provide care of the highest quality, many
physicians do not have adequate systems in place to ensure
that they consistently do so. Clinical practice in the United
States, for the most part, is not data-driven, nor is it
transparent even within the context of the physician-patient
compact.
Implications for Policy, Delivery or Practice: The
implications of these findings are important and deserve the
attention of physicians, professional organizations, and
policymakers alike. It is hard to imagine how, in the absence
of quality-of-care information, effective solutions to some of
the pervasive problems with health care quality can be
generated. Although quality improvement is an essential
component of professionalism, most practicing physicians do
not appear to be participating in it. Physicians should be
taught the knowledge and skills to participate in quality
improvement activities, and the acquisition and application of
these skills should be rewarded.
Primary Funding Source: CWF
• The Formation of Residents: Acquiring the Habit of
Quality Improvement
David Leach, M.D.
Presented by: David Leach, M.D., Executive Director,
ACGME, 515 N State St, Suite 2000, Chicago, IL 60610; Tel:
312.755.5007; E-mail: dcl@acgme.org
Research Objective: The Accreditation Council for Graduate
Medical Education (ACGME) sets the standards for and
accredits the nation’s 7800 allopathic residency programs. It
accredits 114 different specialties. In 1997 it endorsed the use
of educational outcome measures as an accreditation tool.
The objective of this intervention was to improve the
formation of resident physicians by focusing on their
demonstrated skills (educational outcomes) rather than
simply their potential skills (learning opportunities). After an
extensive vetting process six general competencies were
identified as relevant to all specialties. These competencies
are: patient care; medical knowledge; practice-based learning
and improvement; interpersonal and communication skills;
professionalism; and system-based practice.
In the context of the larger study, practice-based learning and
improvement and system-based practice are especially
relevant to the acquisition of quality improvement skills. The
objective of these competencies is to acquire the skills needed
to analyze and improve the individual resident’s practice and
to acquire the skills needed to analyze and design safer and
more effective health care systems.
Study Design: This is a large and complex study. Introducing
competencies, assessment methods, clinical and procedural
databases to 7800 residency programs, and using educational
outcome data to inform accreditation decisions made by 27
Residency review Committees (RRCs) has been a multiyear
process. By design the Initiative was introduced in four
phases: Forming the Initial Response; Sharpening the Focus
and Clarifying the Definition of the Six Competencies; Linking
Educational Outcomes with Clinical Outcomes; and
Benchmarking Excellent Performance. We are currently in the
second phase of the Initiative.
Population Studied: 100,000 residents in 7800 ACGME
accredited residency programs
Principal Findings: Quality improvement and other
competencies and skills can be viewed as clinical habits that
physicians need to acquire early on during residency training,
so that they persist throughout their professional lives.
Training can either facilitate or inhibit learning these skills.
Facilitators include: getting physicians in training to develop
portfolios of clinical experiences and procedures; getting
residents to obtain data about the clinical outcomes of their
care; just-in-time education about how they can better
understand the relationship between their work and the
microsystems of care in which they do this work, and the
effects of system design on clinical outcomes; and simulation
of clinical experiences. Assessment techniques that monitor
resident formation include: focused and direct observation of
the clinical skills of the resident; 360 degree evaluations of
residents including input from patients; portfolios and
analysis of portfolios; and cognitive tests. Assessment over
time and on multiple occasions can monitor the growth of the
resident’s demonstrated skills.
Conclusions: Both individual attributes and contextual
expectations are important in the development of quality
improvement skills. Educational imperatives can be used to
foster clinical changes. The prospect of 100,000 residents
trying to improve their practices may have an important effect
on the larger system.
Implications for Policy, Delivery or Practice:
Accreditation of educational programs is traditional viewed as
a “trailing-edge” phenomenon. It may also be used to
promote leading edge changes that promote good learning for
good health care.
Primary Funding Source: RWJF, ACGME
• What can Professional Organizations Do to Foster
Adoption of Quality Improvement Principles and Methods
by Practicing Physicians?
John Tooker, M.D., Vincenza Snow, M.D., F.A.C.P., Christel
Mottur-Pilson, Ph.D.
Presented by: John Tooker, M.D., Executive Vice President,
The American College of Physicians, 190 N Independence
Mall West, Philadelphia, PA 19106-1572; Tel: 215.351.2800;
E-mail: jtooker@acponline.org
Research Objective: There is clear evidence that the transfer
of scientific knowledge and research findings into practice is
slow and inconsistent. Thus, this applies to the science of
quality improvement and the practice of medicine. While
improving physician knowledge is necessary, it is not sufficient
to change practice. The American College of Physicians (ACP)
is aware of this fact and has begun to complement its
evidence-based products, such as guidelines, etc. with
practice-based quality improvement interventions and
educational activities. Each of the programs identified below
reflect different avenues to quality improvement. Taken as a
whole they point the way to greater transparency and data
driven quality improvements. Each of these programs focuses
on the clinical practice environment of the average physician
in the hope that these practice-based interventions will lead to
wider adoption and use.
Study Design: Four separate initiatives the ACP is engaged in
will be described.
1. Office CME and team based learning:
Closing the Gap: Partnering for Change. This 3 year grant
from the Agency for Healthcare Research and Quality
(AHRQ), targets systems change through a practice-based
CME intervention. The team-oriented CME intervention
utilizes Wagner’s Chronic Care Model and the Institute of
Healthcare Improvement’s (IHI) Breakthrough Series Model
for improvement. Patients with type 2 diabetes will serve as
the prototype medical condition for this study. Practice teams
of physicians, nurses, and administrators will receive training
in systems change and improvement strategies, implement
the intervention, and provide feedback and practice data. We
hypothesize that this approach will lead to improved quality
of care and patient outcomes. We are currently in year two of
the grant. During the pilot year, we trained 4 practices that
are now implementing what they learned in their practices and
reporting back on their progress. We are currently recruiting
32 practices for years 2 and 3.
2. Practice-based Diabetes QI project
ACPNet is a two-year grant from AHRQ with the aim of
improving the infrastructure of practice-based research while
at the same time testing an electronic diabetes educational
intervention aimed at areas identified as needing
improvement via baseline chart abstraction data. The
outcomes data of the intervention group will be compared
with the control group (without intervention) to judge the
success of the intervention. We tested this model at a
medium sized IPA and are applying the lessons learned to the
grant.
3. Dissemination via Train-the-Trainer
Patient Safety: The Other Side of the Quality Equation is a
three-year grant from AHRQ focused on developing and
disseminating an ambulatory care patient safety curriculum in
the seven key areas identified by the IOM as crucial for
professional societies. These are systems, medication errors,
idealized office design, electronics, cognition and
communication and the role of the patient. The curriculum
has been widely disseminated via the College Chapter
infrastructure. Through informal channels the curriculum has
been adopted in part or whole by medical societies, program
directors, and clinical institutions. The curriculum is built
around clinical cases exemplifying given patient safety
problem areas.
4. QI and accountability via performance measures
Performance Measures Sub-committee. The College has
established a performance measures subcommittee to assess
and guide its membership in the adoption and education
concerning valid and feasible performance measures.
Population Studied: US physicians
Implications for Policy, Delivery or Practice: It is imperative
that professional organizations, and therefore, that the
American College of Physicians be a player in the professionwide effort to establish, refine, and apply the science of quality
improvement. Through it's various projects and programs,
and through it's policies, the College is playing an important
role at every step of the physician's improvement cycle measure, learn and improve.
Primary Funding Source: AHRQ, ACP
Call for Panels
Monitoring Outcomes of Medicare-Funded Health Care
with Administrative Data: The Medicare Quality
Monitoring System
Chairs: Lein Han, Ph.D., Myles Maxfield, Ph.D.
Tuesday, June 8 • 11:15 a.m.- 12:45 p.m.
• Quality of Care for Medicare Claimants with Diabetes:
1992 and 2001
Sylvia Kuo, Ph.D., Barbara Flemming, M.D., Ph.D., Lien Han,
Ph.D., Neil Gittings, M.A., Shiela Roman, M.D., M.P.H., Linda
Geiss, M.A.
Presented by: Angela Merrill, Ph.D., Researcher, Mathematica
Policy Research, Inc., 955 Massachusetts Avenue, Suite 801,
Cambridge, MA 02139; Tel: 617.491.7900 X 237; Fax:
617.491.8044; E-mail: amerrill@mathematica-mpr.com
Research Objective: To describe rates of complications and
receipt of associated preventive services among Medicare
diabetic claimants and to document how these trends have
varied over time and across different demographic subgroups.
This effort is part of the Medicare Quality Monitoring System
designed to monitor the quality of care received by Medicare
beneficiaries.
Study Design: Using the Medicare 5% Standard Analytic Files,
rates on 24 quality indicators associated with diabetes were
reported for Medicare diabetic claimants. These measures
were identified using claims with specific ICD-9, CPT/HCPSC
and Revenue Center codes. Rates were generated for a 5%
cross-sectional sample for 1992 and 2001, and were stratified
by demographic and enrollment characteristics of the sample.
Rates were reported as a percent of Medicare diabetic
claimants with any claim during the year associated with that
measure. The rates and standard errors were age-sex
standardized to control for variations in the age-sex
distribution across the two years.
Population Studied: The sample includes full-year fee-forservice Medicare diabetic claimants without end stage renal
disease from 1992 through 2001. Diabetic claimants are
defined as those who had either an inpatient or emergency
department visit or two ambulatory encounters coded for
diabetes during the year. Beneficiaries who died during the
year were excluded, as were individuals who had gestational
diabetes. The resulting sample sizes were 148,864 claimants
for 1992 and 227,541 for 2001.
Principal Findings: The use of preventive services (HbA1c
tests, lipid tests, eye exams, and self-monitoring of glucose
levels) increased greatly from 1992 to 2001. Additionally, rates
for many of the short-term and serious long-term
complications to diabetes (e.g. lower limb amputations and
most cardiovascular conditions) declined from 1992 to 2001.
However, other types of complications increased slightly
during the period studied (e.g. osteomyelitis and most eye
complications). Further, subgroup analyses demonstrated
consistently higher complication rates and lower use of
preventive services among non-whites and Medicare
beneficiaries who were also enrolled in Medicaid.
Conclusions: These analyses demonstrate significant
progress over time in some key areas, such as in the receipt of
preventive services and in some serious adverse diabetic
outcomes. More research is needed to determine whether the
increase in other types of complications reflect quality
improvements through delayed disease progression or
presage future prevalence of more serious outcomes.
Additionally, the results suggest possible disparities in the
quality of care for non-whites and Medicare beneficiaries who
were also enrolled in Medicaid.
Implications for Policy, Delivery or Practice: These results
provide an early warning sign to policymakers to focus
attention on the diabetes care provided to some vulnerable
subgroups, but also suggest some successes, such as in
achieving reductions in lower limb amputations.
Primary Funding Source: CMS
• Preventable Hospitalizations Among Medicare Fee-ForService Beneficiaries, 1995 to 2001
Tim Lake, Ph.D., Sheila Roman, M.D., M.P.H.
Presented by: Sheila Roman, M.D., M.P.H., Medical Officer,
Quality Measurement and Assessment Group, Quality
Measurement and Health Assessment Group, Centers for
Medicare & Medicaid Services, 7500 Security Blvd., Baltimore,
MD 21244; Tel: 410.786.6004; E-mail: roman@CMS.HHS.gov
Research Objective: To track trends in preventable
hospitalization rates for Medicare beneficiaries, and to
compare rates according to beneficiary characteristics and
geographic region. Preventable hospitalization rates are
considered an indicator of access to ambulatory care.
Study Design: We used Medicare hospital claims data on all
Medicare fee-for-service beneficiaries to calculate annual
hospitalization rates for ten ambulatory-care sensitive
conditions (ACSCs) from 1995 to 2001. The ACSCs included
congestive heart failure, bacterial pneumonia, chronic
obstructive pulmonary disease, urinary tract infection,
dehydration, angina without procedure, long-term
complications of diabetes, leg amputation for diabetes, shortterm complications of diabetes and uncontrolled diabetes.
The rates were calculated as the number of hospital
discharges per 100,000 Medicare fee-for-service beneficiaries.
We generated results for all beneficiaries and for subgroups
defined in terms of age, gender, race, reason for Medicare
eligibility, dual enrollment in Medicaid, urban/rural location,
and geographic region. All rates were adjusted for age/sex
differences.
Population Studied: Beneficiaries enrolled in Medicare feefor-service in 1995, 1996, 1997, 1998, 1999, 2000, or 2001.
Principal Findings: From 1995 to 2001, preventable
hospitalization rates increased by 10 to 15 percent for five of
the 10 measures studied: bacterial pneumonia, chronic
obstructive pulmonary disease (COPD), urinary tract infection,
dehydration, and long-term complications of diabetes.
Hospitalization rates for congestive heart failure, leg
amputation for diabetes, and short-term complications of
diabetes were largely unchanged, while rates for angina
without procedure and uncontrolled diabetes decreased
substantially during the period. The 2001 hospitalization rate
for all 10 ACSC measures combined was about 6,900
discharges per 100,000 beneficiaries, representing an overall
increase of nearly 3 percent from 1995. The oldest old (80
years or older), African Americans, those in the South, those
dually enrolled in Medicaid, and those eligible for Medicare
due to end-stage renal disease had higher rates of
hospitalization than other beneficiaries in 2001. These groups
also experienced relatively large increases in rates from 1995
to 2001.
Conclusions: Our findings indicate substantial growth in the
nationwide rate of preventable hospitalizations for selected
conditions in recent years, and may suggest disparities among
certain groups of Medicare beneficiaries. Because our results
were adjusted for age (and sex) differences, these trends are
unlikely to be explained by the aging of the Medicare
population. However, we do not know the extent to which
these trends were driven by differences in the rates of
underlying prevalence of disease, or by the severity of illness.
Implications for Policy, Delivery or Practice: Efforts to
improve access to care should aim to reduce or slow the rise
in preventable hospitalization rates over time—potentially
improving quality of life for Medicare beneficiaries while
reducing the cost of care. Reducing the disparity in rates
between certain subgroups is also an important goal. Despite
the uncertainty regarding appropriateness of rates in absolute
terms, these national findings can provide benchmarks for
those seeking to improve access to care for all beneficiaries or
to target services to particularly vulnerable groups.
Primary Funding Source: CMS
• Patient Safety Among Medicare Beneficiaries
Arnold Chen, M.D., David Hunt, M.D., F.A.C.S.
Presented by: Arnold Chen, M.D., Senior Clinical Researcher,
Health Research Department, Mathematica Policy Research,
Inc. P.O. Box 2393, 600 Alexander Park, Princeton, NJ 085432393; Tel: 609.275.2336; Fax: 609.799.0005; ; E-mail:
achen@mathematica-mpr.com
Research Objective: To determine the rates of potentially
preventable adverse events among hospitalized fee-for-service
Medicare beneficiaries in 2000 and 2001. Rates are computed
by demographic group and geographic area. The study uses
Medicare claims data and the MQMS Patient Safety Measures
(PSMs), a modified subset of AHRQ’s recently developed
Patient Safety Indicators (PSIs). Finally, the study
decomposes the PSMs into component complications to help
identify specific remediations hospitals could take.
Study Design: Cross-sectional study of Medicare beneficiaries
hospitalized in 2000 and 2001. Age and sex-adjusted rates of
PSMs, cross tabulations of PSM rates with beneficiary
characteristics, and frequencies of specific diagnoses and
DRGs within specific PSMs.
Population Studied: For each of the two study years, all
Medicare beneficiaries: eligible for Medicare in January of that
year, continuously enrolled in Medicare Part A and fee-forservice Medicare throughout the year or until death, and
hospitalized during the year in a short-stay acute care hospital.
Sample sizes varied from slightly over 1 million to roughly 22
million beneficiaries (the various PSMs have different
eligibility criteria).
Principal Findings: The three most common PSMs were
decubitus ulcers (27 events per 1,000 discharges, about
270,000 events), postoperative septicemia (13.4 events per
1,000 discharges, about 14,000 events), and postoperative
respiratory failure (8 events per 1,000 discharges, about
18,000 events). For decubitus ulcers, there was a three-fold
differential in rates between the youngest and oldest age
groups, and rates for black or dually eligible beneficiaries were
roughly double those for white and non-dually eligible
beneficiaries. Urban dwelling and ESRD beneficiaries also had
higher rates of decubitus ulcers. Rates of postoperative
septicemia were about two-fold higher among the youngest
and oldest age groups compared to the intermediate groups.
Male, dually eligible, and ESRD beneficiaries also had rates of
postoperative septicemia two or three times those of their
counterparts. Rates of postoperative respiratory failure were
about 1.5 to 2 times higher among male and dually eligible
beneficiaries.
Conclusions: Among hospitalized Medicare beneficiaries, the
serious and potentially preventable complications of decubitus
ulcers, postoperative septicemia, and postoperative
respiratory failure occur relatively commonly and affect large
numbers of people. Age, dual eligibility, black ethnicity, and
ESRD define a group of beneficiaries that are particularly
vulnerable to patient safety events.
Implications for Policy, Delivery or Practice: These results
suggest that Medicare policymakers and health care providers
focus further research and quality improvement efforts on
decubitus ulcers, postoperative septicemia, and postoperative
respiratory failure, particularly for beneficiaries who are elderly,
dually eligible, black, and have ESRD. The MQMS PSMs
provide a new, inexpensive way for the Medicare program,
which covers some 35 million fee-for-service beneficiaries
nationwide, to track the quality of care received by its
beneficiaries.