|
Variations: Fisher Web Exclusive
V A R I A T I O N S
A M C E F F I C I E N C Y
W E B E X C L U S I V E
7 October 2004
Variations In The Longitudinal
Efficiency Of Academic
Medical Centers
Increased intensity of
care does not appear to be associated
with higher quality or to result in
better survival at AMCs.
By Elliott S. Fisher,
David E. Wennberg, Thérèse A. Stukel,
and Daniel J. Gottlieb
ABSTRACT:
Recent studies have revealed dramatic differences among
academic medical centers (AMCs) in the quantity of care provided to their
patients. The implications, however, depend upon whether the additional resources
provided by some centers lead to better results. This study describes the
content, quality, and outcomes of care across AMCs that differ by up to 60
percent in the overall intensity of medical services delivered to patients
with serious chronic illnesses. Efforts to reduce costs will require attention
to supply-sensitive services (the frequency of hospital stays, physician visits,
specialist consultations, diagnostic tests, and minor procedures) and should
include a focus on the longitudinal efficiency of hospitals and medical staffs.
Academic medical centers (AMCs) have
primary responsibility for the scientific basis of clinical medicine and are
thus widely viewed as setting the standard for high-quality care. Their important
mission and perceived quality advantages have been used to justify higher
costs and ensure their inclusion in preferred networks of providers, which
has threatened, at least in theory, current market-based cost containment
efforts.1 The
preeminence of these institutions, however, is difficult to reconcile with
the growing body of research documenting dramatic differences in practices
among them, particularly in the use of “supply-sensitive” services:
the use of the hospital as a site of care, the frequency of physician visits
and specialist consultations, and the use of diagnostic tests and minor procedures.
Studies from the early 1990s found that patients with chronic conditions cared
for at Boston University Medical Center received twice as much hospital-based
care over three years of follow-up than similar patients cared for by Yale–New
Haven hospital.2 More
recent analyses have focused on patients at the end of life: Compared with
similar patients at Stanford University Medical Center, for example, Medicare
enrollees who were cared for at New York University (NYU) Medical Center spent
more than twice as many days in the hospital (twenty-seven versus ten) and
had more than three times as many physician visits (seventy-six versus twenty-three)
during their last six months of life.3
These differences in the amount of care provided to similar patients raise
important questions about longitudinal efficiency—that is, whether some
hospitals may be able to achieve comparable outcomes in caring for a defined
population at much lower cost than other hospitals.4 Recent
population-based studies compared the process and outcomes of care for Medicare
beneficiaries residing in regions with nearly twofold differences in per capita
spending and found that the additional spending was almost entirely devoted
to greater use of supply-sensitive services and that higher spending did not
result in higher-quality care, lower mortality, better functional outcomes,
or greater satisfaction.5 The implications
of these findings for AMCs, however, are uncertain: Findings at the regional
level may not apply uniformly to the populations served by AMCs located within
them.
This study focuses on Medicare enrollees with chronic medical conditions who
received most of their care from AMCs. We compare the patterns of practice,
quality of care, and health outcomes across hospitals that differ by up to
60 percent in the overall amount of hospital and physician services provided
to similar patients.
Methods
Design overview. We
previously published a comprehensive study of the process and outcomes of
care for Medicare enrollees residing in U.S. regions that differed dramatically
in per capita Medicare spending. The study adopted a “natural randomization” approach
to assign patients to one of five different “treatment groups” (quintiles
of practice intensity) that corresponded closely to regional differences in
price and to illness-adjusted Medicare spending.6 To
confirm the natural randomization, we showed that patients in each quintile
were similar in health status at the time they entered the study but that
subsequent overall health care use (measured as price-adjusted spending on
hospital and physician services) differed dramatically. We then described
the process and outcomes of care for these cohorts across intensity (and thus
Medicare spending) quintiles.
This study reports on the subset of these patients who received their initial
hospitalization and most of their subsequent care in AMCs. Patients were assigned
to levels of intensity based on the location of the hospital where they received
their initial care. By assigning patients based upon the practice intensity
of their region (where intensity was determined based on the experience of
other Medicare beneficiaries), we maintained the essential features of the
natural randomization design. We then show that patients are similar at baseline
across hospital intensity levels (Exhibit
1) but that they are treated differently
(Exhibit
2). We then examine the content, quality, and outcomes of care across
intensity levels.
Study cohorts.
The three hospital-based cohorts were defined on the basis of an initial hospitalization
for acute myocardial infarction, colorectal cancer, or hip fracture among
fee-for-service Medicare enrollees, ages 65–99, that occurred during
the period 1993–1995.7 We
chose these conditions for three reasons. First, even if there are some differences
across regions or hospitals in the underlying health status of the populations
they serve (such as in smoking, obesity, or frailty), those differences will
be reflected in the incidence of disease; populations with lower rates of
smoking will have lower rates of heart attack, for example, but incident cases
of heart attack are likely to be relatively similar, reflecting the joint
operation of relevant risk factors. Second, follow-up of these cohorts would
provide insight into the quality of care and outcomes of three different clinical
care systems: cardiovascular disease, oncology, and orthopedics. Third, any
patterns of practice found to be consistent across these three conditions
would provide insights into the longitudinal care of patients with chronic
disease, regardless of the specific diagnosis that brought them to the hospital.
This study focuses on those patients whose initial hospitalization was at
one of the 299 hospitals that are members of the Council of Teaching Hospitals
(COTH).8
Measures.
Baseline characteristics—including
age, sex, race, socioeconomic status, illness severity, and comorbidities—were
defined using data obtained from the Medicare Denominator and Medicare Provider
Analysis and Review (MEDPAR) files (all three cohorts), chart review (acute
myocardial infarction, or AMI, cohort), and the U.S. census, based upon ZIP
code of residence (all three cohorts). To characterize the content and quality
of care during and after the index admission, we relied on chart review data
(for the AMI cohort only) and on analyses of Medicare physician and hospital
claims files. Our summary measures of spending for the cohorts were calculated
using standardized national prices for both hospital and physician services.9 Mortality
was ascertained from the Medicare Denominator file. Attributes of the regions
(supply of hospital beds and physicians, Medicare spending) were obtained
from the Dartmouth Atlas of Health Care.10
Analyses.
To determine whether average illness levels varied across hospitals of differing
intensity levels, we used logistic regression to determine each patient’s
predicted risk of death at one year, based on demographic, socioeconomic,
and clinical characteristics. We then compared the average predicted risk
of death for each cohort across intensity levels. We compared the use
of services across intensity levels using both unadjusted and adjusted rates.
Because results were similar, we present unadjusted rates in the exhibits.
To test for differences in survival across intensity levels, we used Cox
survival analyses, adjusting for patients’ demographics, illness severity,
comorbidity, socioeconomic characteristics, and U.S. census region.
Study Results
Characteristics
of the study populations. Key
characteristics of the regions, major teaching hospitals, and study cohorts
for the current study, grouped according to the assigned intensity levels
of patients’ AMC, are presented in Exhibit
1. Hospitals in the lowest-intensity
group (Q1) include St. Mary’s Hospital of Rochester, Minnesota (affiliated
with the Mayo Clinic); Strong Memorial Hospital of Rochester, New York; and
Richland Memorial Hospital of Columbia, South Carolina. Hospitals in the highest-intensity
group (Q5) include NYU Medical Center in Manhattan; Cedars-Sinai Medical Center
in Los Angeles; and Jackson Memorial Hospital in Miami.11 There
are more major teaching hospitals (and thus more patients) in the higher-intensity
groups, reflecting the underlying geographic distribution of teaching hospitals:
more are located in high spending regions.12 Patients
with AMI had a higher predicted risk of death (more than 30 percent) than
patients with hip fracture (about 24 percent) or colorectal cancer (about
20 percent), but there was little difference in illness levels across hospitals
when they were grouped according to intensity levels.
Overall treatment
intensity and spending. Overall
treatment intensity (and thus price-adjusted spending) differed greatly across
the five intensity groups for each clinical cohort (Exhibit
2). Because all
patients experienced at least one hospitalization, differences in total spending
during the first six months are relatively modest. After the acute episode,
however, per patient spending on hospital and physician services was between
47 percent (AMI) and 58 percent (hip fracture) higher in the highest-intensity
teaching hospitals than in the lowest-intensity hospitals. The proportion
of care provided at the original hospital, which we define as “loyalty,” was
high in each intensity group. Over the full five years of follow-up, more
than 80 percent of the hospital experience of these cohorts was provided by
their index hospital.13
The content of care. Exhibit
3 provides data on the use of physician services during the first six months
of follow-up for each of the cohorts across hospitals in the differing intensity
groups, classified according to the major categories defined by the Berenson-Eggers
Type of Service (BETOS) system. Spending on major surgical procedures was
similar across intensity groups (only 1–4 percent higher in the highest-intensity
hospitals than in the lowest). For all other services (minor procedures, diagnostic
tests, imaging, and evaluation and management services), however, resource
use increased linearly and substantially across the different intensity groups.
In the highest- compared with the lowest-intensity major teaching hospitals,
use of evaluation and management services, for example, is 56–82 percent
higher; imaging, 20–26 percent higher; and diagnostic testing, 73–94
percent higher. The consistency of the relative increase across intensity
groups, even when the underlying utilization rates differ across clinical
conditions, is notable.
A more detailed examination of specific physician services within each category
reveals a generally similar pattern.14 All
evaluation and management services were performed at higher rates among
patients cared for in the highest- compared with the lowest-intensity
hospitals, with rates of office visits and outpatient consultations
about 1.2 to 1.3 times higher, and rates of inpatient visits and new
inpatient specialist consultations about twice as high. Rates of imaging
services ranged from 1.3 times higher (chest x-ray) to 1.6 times higher
(MRI), with the notable exception of mammography, which was performed
at similar rates across intensity levels. The same pattern held for
diagnostic tests and minor procedures—all of which were performed
at increasing rates in higher-intensity hospitals, with the exception
of certain procedures that are less likely to be discretionary: repair
of laceration, excision of malignant skin lesions, and influenza and
pneumococcal vaccinations. Patients at the high-intensity hospitals
were also treated much more aggressively at the end of life.
The use of the hospital as a site of care was also much greater for patients
in higher-intensity hospitals. During the first six months of follow-up (when
all were hospitalized at least once), total inpatient days were 16 percent
(for colorectal cancer patients) to 38 percent (for hip fracture patients)
higher in the highest- than in the lowest-intensity hospital groups. After
the first six months, inpatient use was 61 percent higher for colorectal cancer
patients, 68 percent higher for heart attack patients, and 84 percent higher
for hip fracture patients.
Quality and access.
Our quality-of-care measures were drawn from the chart reviews completed by
the Cooperative Cardiovascular Project (Exhibit
4). The data show that among
major AMCs, higher-intensity practice was associated with either no difference
or a decrement in quality as intensity increased. Moreover, there was no greater
use of major cardiovascular procedures or other major elective procedures
in the higher-intensity hospitals. A somewhat higher proportion of patients
with AMI saw a cardiologist within the first thirty days of their hospitalization.
Long-term survival.
We also compared long-term mortality rates for patients cared for in hospitals
of differing intensity levels. As in the original study, in which comparisons
were based on regional differences in practice, there was no association between
the increased intensity across these major medical centers and improved survival.
Among hip fracture patients, there were no significant differences in mortality
across intensity groups. For the other two cohorts, there was a small, statistically
significant increase in the risk of death as intensity increased.15
Discussion
Major U.S. AMCs differ dramatically in the overall intensity of services they
provide to similar patients. The increased intensity does not appear to be
associated with higher quality of care or to result in better survival. Patients
in the higher-intensity hospitals simply spend more time in the hospital and
intensive care unit (ICU); have more frequent physician visits (especially
in the inpatient setting); have more specialists involved in their care; and
receive more imaging services, diagnostic testing, and minor (but not major)
procedures. The similar results achieved with markedly different levels of
resource inputs imply large differences in the longitudinal efficiency of
chronic disease care across these hospitals.
These findings raise three major questions. What is causing the differences
in utilization? Why don’t the additional resources lead to better quality?
What is AMCs’ responsibility for these differences in practice and costs?
Differences in utilization.
Several possible explanations for the differences in utilization deserve consideration.
Unmeasured differences in case-mix and comorbidities are often
invoked as a possible explanation for variations in use or outcomes. The natural-experiment
design on which this study is based minimizes this possibility. Patients’ preferences
also may matter. For example, patients wanting more aggressive management
of their illnesses may select AMCs as their place of care. We would therefore
expect patients’ preferences to differ relatively little on this dimension
across these hospitals. Data from other studies also argue against patients’ preferences
as a major determinant of the differences in practice we observed across hospitals.
Differences within cities in overall intensity have been well documented.16 Also,
the SUPPORT study, which had direct measures of patients’ preferences
for site of care at the end of life, found that the marked differences in
practice across the five participating AMCs were unrelated to patients’ preferences.17
Characteristics of the institutions themselves (or the physicians practicing
in them) provide the major alternative explanation for the differences in
intensity we found. One factor could be the size of the teaching programs.
Prior studies have shown that as the size of these programs increases, so
does the cost of care.18 In our population-based
longitudinal analysis, however, we found no association between larger teaching
programs and per patient use of physician services.19
Another potential explanation is the capacity of each hospital’s care
system relative to the size of the population it serves. At the regional
level, the population at risk and local capacity can be measured independently,
and a direct relationship has long been recognized between supply and utilization.
Exhibit
5 examines these associations at the regional level more closely.
In population-based analyses of the entire fee-for-service Medicare
population, regions with higher-intensity practice patterns (and higher per
capita Medicare spending) had a much higher per capita supply of hospital beds
and physicians than the more conservative regions.20 In
Exhibit
5, hospital referral regions (HRRs) are grouped into quintiles
according to the local supply of medical specialists (including internists)
and hospital beds. The data indicate that the influence of an increase in the
supply of physicians varies according to the supply of beds. In regions in
the lowest quintile of hospital bed supply, an increase from the lowest to
the highest quintile of internist/medical specialist supply is associated with
an 18 percent increase in the intensity of care. In areas that have the largest
per capita supply of beds, however, the same increase in physician supply is
instead associated with a 34 percent increase in per capita intensity.21
The association between system-specific per capita supply of resources
and intensity of care is consistent with the following behavioral
interpretation. Given the current culture and financial incentives
of U.S. medicine, which favors intervention among the seriously
ill (even in situations where the scientific evidence is weak),
physicians in a given care system are likely to draw upon all available
resources to care for their patients. Moreover, it is often easier
to manage patients’ care in the inpatient setting;
the outpatient setting imposes a greater burden on the physician
(for example, to arrange telephone, visiting nurse, or office visit
follow-up to ensure that the patient is responding to therapy).
The inpatient setting also lowers the threshold for further intervention:
It is easier to obtain tests, perform minor discretionary procedures,
or request a consultation.22 Studies
of ICU bed availability reveal the same phenomenon: When more beds are available,
the average severity of the patients cared for in the ICU is lower and the
average length-of-stay is longer.23 These
data are consistent with the theory that hospitals with the highest-intensity
practice patterns have developed this pattern of practice because of a relatively
higher per capita supply of beds and physicians.24
Relation of spending
and quality.
A more difficult question is why higher spending—at both the regional
and hospital levels—doesn’t lead to higher-quality care (and might
make things worse).25 First,
the generally poor state of clinical science in dealing with populations of
chronically ill patients must be recognized. For patients with specific chronic
conditions, for example, the scientific literature is largely silent on the
appropriate interval between repeat visits or the value of additional diagnostic
tests, hospitalizations, and stays in the ICU. The current evidence suggests,
however, that improving both processes and outcomes requires changes in the
system of care, such as the development of data systems to promote adherence
to care guidelines and the implementation of team approaches to ensure coordination.26 Current
fee-for-service reimbursement does not pay for these services. Moreover, higher
spending across the AMCs examined here is largely devoted to having patients
spend more time in the hospital and having more physicians involved in a given
patient’s care (Exhibit
6). Hospitals can be dangerous places.27 Also,
having more physicians involved in a given patient’s care makes communication
among physicians more difficult and could increase the risk of errors.28
Study limitations. Several
limitations of these analyses must be acknowledged. First, we have no data
about functional outcomes or satisfaction with care among the patients studied
at these hospitals. We do know, however, that procedure rates and use of important
medications following AMI were similar—and that the rates of major procedures intended
to improve the quality of life (such as knee replacements or back surgery)
were not consistently higher in the higher-intensity hospitals and indeed
were lower for some procedures. Moreover, at the regional level, where sample
sizes were adequate to detect small differences in procedure rates, there
were no significant differences across the five quintiles of per capita spending
in the rates of major elective surgical procedures. Finally, analyses of a
representative sample of the Medicare population found no significant differences
in the rate of functional decline (or in satisfaction with care) across U.S.
regions that differed by up to twofold in per capita spending.
Second, we have not addressed the question of why the differences in intensity
observed across these teaching hospitals are associated with per capita Medicare
spending levels in their regions. Although variations in hospital-specific
practice patterns within HRRs are well documented, our data indicate that
on average the intensity of care provided by teaching hospitals is correlated
with the overall practice intensity of the region where they are located.
This could reflect either or both of the following: patterns of practice among
residents who decide to stay in the area where they trained, or regional differences
in overall capacity that lead to adaptive behavior on the part of both teaching
and nonteaching hospitals.
Finally, some may be concerned that our focus on the hospital is misplaced.
To borrow a phrase from the gun control debate: Hospitals don’t order
tests or put patients in the ICU; physicians do. The rationale for a focus
on the hospital and its affiliated medical staff rests on three observations.
First, in spite of efforts to reduce hospital use, hospital services are still
the single largest component of health care spending. Second, the hospital
provides both the physical and the social context within which many of the
most costly and least studied clinical decisions are made (whether to readmit
patients with congestive heart failure, whether to care for them in the ICU,
how many specialist referrals should be made). Any effort to address the challenge
of supply-sensitive services will entail engaging hospitals—and the
physicians who use them—in an effort to better align resource levels
(including hospital and ICU beds) with the size and needs of the specific
population served by a hospital and its medical staff. Moreover, as this and
other studies demonstrate, most hospitals serve clearly defined populations
(especially among those with chronic disease), even if they are not now able
to easily identify these populations and manage their care. Finally, most
physicians remain in private solo or small-group practices, even at some major
AMCs. Efforts to address the current well-recognized deficiencies in the quality
and efficiency of care will need to ensure that these physicians are able
to take advantage of advances in informatics, quality improvement, and care
coordination that may be beyond the reach of an individual physician practice.
The hospital and its affiliated medical staff may therefore be the most practical
and least disruptive organizational unit within which to approach the needed
integration of clinical practice.29 Such
integration could occur formally (where payments are made to an integrated
prepaid group practice) or informally (where payment continues to be to the
individual physician practices, but care is “virtually” integrated
through the use of shared electronic medical records and care protocols).
In either case, payers may want to consider fostering the integration of physician
practice at the level of the hospital and its affiliated medical staff.
Who can lead this
effort? The evidence
we present has serious implications for those institutions that bear responsibility
for the scientific basis of clinical medicine: AMCs and the federal research
agencies and nonprofit institutions that set priorities and fund clinical research.30 The
remarkable differences in utilization revealed in the data presented here and
in our earlier work at the regional level pose a challenge and an opportunity.
The challenge is to learn how to improve both the quality and efficiency of
care. The opportunity is for AMCs to reclaim their academic authority and,
perhaps, to secure their financial future. Given the widely held perception
that quality is superior at academic institutions but that costs (when measured
per discharge) are higher, academic institutions that demonstrate both high
quality and lower long-term costs will be more likely to compete successfully
in a health care marketplace that is increasingly concerned about longitudinal
efficiency.
Financial support was provided by a grant from the Robert Wood Johnson
Foundation and the National Institute on Aging (Grant no. P01 AG019783).
NOTES
1. See L.M. Nichols et al., “Are Market Forces Strong Enough
to Deliver Efficient Health Care Systems? Confidence Is Waning,” Health
Affairs 23, no. 2 (2004): 8–21; and J.C. Robinson, “Hospital
Tiers in Health Insurance: Balancing Consumer Choice with Financial Incentives,” Health
Affairs, 19 March 2003,
content.healthaffairs.org/cgi/content/abstract/hlthaff.w3.135 (28 July 2004).
2. See E.S. Fisher et al., “Hospital Readmission Rates for Cohorts
of Medicare Beneficiaries in Boston and New Haven,” New
England Journal of Medicine 331, no. 15 (1994): 989–995.
3. See J.E. Wennberg et al., “Use of Hospitals, Physician Visits,
and Hospice Care during the Last Six Months of Life among Cohorts Loyal to
Highly Respected Hospitals in the United States,” British
Medical Journal 328, no. 7440 (2004): 607.
4. To our knowledge, the term longitudinal efficiency was
introduced by Arnold Milstein, Pacific Business Group on Health, in testimony
before the U.S. Senate Committee on Health, Education, Labor, and Pensions,
28 January 2004. Milstein refers to the need for a focus on the longitudinal
experience of defined populations, because differences in unit price ignore
potential differences in volume or quality that can influence the aggregate
costs (and outcomes) of care for a population.
5. E.S. Fisher et al., “The Implications of Regional Variations
in Medicare Spending, Part 1: The Content, Quality, and Accessibility of Care,” Annals
of Internal Medicine 138, no. 4 (2003): 273–287, and “The
Implications of Regional Variations in Medicare Spending, Part 2: Health Outcomes
and Satisfaction with Care,” Annals of Internal Medicine 138,
no. 4 (2003): 288–298.
6. The original study used two different indices of local practice
intensity to assign the cohorts to quintiles of price and illness-adjusted
Medicare spending, one based on differences in practice intensity during the
last six months of life (the End-of-Life Expenditure Index, or EOL-EI) and
one based on differences in average practice intensity during the first six
months after an acute episode of illness. Both measures were calculated on
samples that were distinct from the study cohorts. Both were uncorrelated with
differences in illness levels among Medicare enrollees across regions but highly
correlated with differences in total per capita Medicare spending. We use the
EOL-EI in the current study. For a detailed description of the methods, see
ibid. and an appendix at www.annals.org/cgi/data/138/4/288/DC1/1 (28 July 2004,
available only to subscribers of Annals of Internal Medicine).
7. The hip fracture and colorectal cancer cohorts were defined based
upon an initial hospitalization identified in the MEDPAR hospital discharge
files that occurred in the full three-year period 1 January 1993–31 December
1995 using appropriate International Classification of
Diseases, Ninth
Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes. The
acute myocardial infarction (AMI) cohort comprised Medicare enrollees in the
Cooperative Cardiovascular Project with completed medical record reviews whose
confirmed AMI occurred between February 1994 and November 1995.
8. COTH membership was ascertained from the American Hospital Association
(AHA) annual survey. The cohorts were defined using Medicare hospital claims
records (MEDPAR file), and the linkage to the AHA file was based on the Medicare
provider number recorded on the claims.
9. The standardized national price for hospital services was based
upon the diagnosis-related group (DRG) for each discharge and the relative
value units (RVUs) for each physician service. Details are provided in the
appendix to the original study, www.annals.org/cgi/data/138/4/288/DC1/1.
10. J.E. Wennberg and M.M. Cooper, eds., The Quality
of Medical Care in the United States: A Report on the Medicare Program, The
Dartmouth Atlas of Health Care in the United States 1999 (Chicago:
AHA Press, 1999). Data on resources, spending, and intensity measures for each
of the 306 U.S. HRRs are available at Dartmouth Atlas of Health Care, “Is
More Actually Better?” www.dartmouthatlas.org/annals/fisher03.php (28
July 2004).
11. A list of the ten largest hospitals in each intensity group is
provided as Appendix Table 1 on the Health Affairs Web
site, content.healthaffairs.org/cgi/content/full/hlthaff.var.19/DC2.
12. In the lower-intensity regions, only 10 percent of the hospital
beds are in major teaching hospitals, compared with 31 percent in the higher-intensity
regions. This difference provides a second motivation for the current study:
whether the higher-intensity practice pattern in high-cost regions was largely
attributable to the predominance of teaching hospitals (or, asked another way,
might teaching hospitals in the lower-intensity regions be just like teaching
hospitals in the higher-intensity regions?).
13. Loyalty was measured as the average percentage of hospital days
experienced by each patient that were provided by the index hospital; it was
more than 95 percent for the first six months, 60–68 percent for the
period from two to five years, and more than 50 percent for years two through
five. The average over five years was more than 80 percent for each cohort.
As a sensitivity analysis, we repeated all utilization analyses excluding patients
who received any care outside their index hospital. The results were essentially
unchanged, with much higher utilization rates observed in the higher-intensity
hospitals, which indicates that the major findings of the study—marked
hospital-specific differences in use—are not sensitive to differences
in loyalty across the hospitals included in the current study.
14. See Appendix Table 2 at content.healthaffairs.org/cgi/content/full/hlthaff.var.19/DC2.
15. The survival models were run on the entire hip fracture, colorectal
cancer, and AMI cohorts from the original study (see www.annals.org/issues/v138n4/toc.html),
with interaction terms for the categories of teaching hospital and HRR-level
end-of-life intensity index measured as a continuous variable. The estimates
of the effect of a 10 percent increase in practice intensity on mortality were
as follows: hip fracture, 1.003 (95 percent confidence interval: 0.999, 1.007);
colorectal cancer, 1.007 (95 percent CI: 1.000, 1.013); and AMI, 1.012 (95
percent CI: 1.005, 1.020).
16. Although it is reasonable to imagine that patients’ preferences
may differ between Miami and Minneapolis, it seems less likely that patients
receiving care in major U.S. AMCs in the same city would choose hospitals on
the basis of their preferences—and if they did, one might expect those
with strong preferences for “more intensive” care to choose the
hospital with a stronger reputation for high-technology care. Longitudinal
analyses in Boston showed that patients cared for at Massachusetts General
Hospital had hospital utilization rates that were 25 percent lower than those
cared for at Boston University Medical Center. Fisher et al., “Hospital
Readmission Rates.”
17. R.S. Pritchard et al., “Influence of Patient Preferences
and Local Health System Characteristics on the Place of Death: SUPPORT Investigators,
Study to Understand Prognoses and Preferences for Risks and Outcomes of Treatment,” Journal
of the American Geriatrics Society 46, no. 10 (1998): 1242–1250.
18. Medicare uses the ratio of interns and residents to beds to determine
the level of supplemental payments for the indirect costs of medical education,
which are presumed to include greater intensity of testing. L. Koenig et al., “Estimating
the Mission-Related Costs of Teaching Hospitals,” Health
Affairs 22, no. 6 (2003): 112–122.
19. In our data, the ratio of interns and residents to beds was either
unassociated (AMI, hip fracture cohorts) or negatively associated (colorectal
cancer cohort, r = –.30,
p < .001)
with the intensity of physician services provided during the first six months
of follow-up.
20. The notion that supply and utilization are associated is not new;
numerous studies have documented the association between the local supply of
resources and utilization. Early examples include Roemer’s now classic
observation of a strong association between the local bed supply and hospitalization
rates. Since that time, studies have consistently revealed an association between
specific resources and utilization, including physician supply and the frequency
of physician visits; the local hospital bed supply and admission rates for
discretionary (but not for emergent or nondiscretionary) conditions; and the
use and availability of ICU beds.
21. A more formal way to look at these relationships is through linear
regression. These two factors explain 43 percent of the regional variation
in intensity.
22. This could explain why increases in the per capita supply
of physicians have a greater impact in high- than in low-bedded areas: It is
easier for the additional physician in a high-bedded area to provide patient
care in the inpatient setting, where more can be done more quickly and—from
the physician perspective—more
efficiently.
23. See D.E. Singer et al., “Rationing Intensive Care—Physician
Responses to a Resource Shortage,” New England Journal
of Medicine 309, no. 19 (1983): 1155–1160; and M.J. Strauss
et al., “Rationing of Intensive Care Unit Services: An Everyday Occurrence,” Journal
of the American Medical Association 255, no. 9 (1986): 1143–1146.
24. The Medicare data can be used to estimate the actual inputs of
hospital beds and nurse and physician full-time equivalents (FTEs). See J.E.
Wennberg et al., “Use of Medicare Claims Data to Monitor Provider-Specific
Performance among Patients with Severe Chronic Illness,” Health
Affairs, 7 October 2004, content.healthaffairs.org/cgi/content/abstract/hlthaff.var.5.
25. The regional analyses showed that for four of six acute care measures
and three of four measures of the quality of preventive care, higher-intensity
regions provided significantly lower quality of care. This is also consistent
with recent analyses that find that states with higher per capita Medicare
spending—and more specialist physicians—have lower levels of quality
for Medicare enrollees. K. Baicker and A. Chandra, “Medicare Spending,
the Physician Workforce, and Beneficiaries’ Quality of Care,” Health
Affairs, 7 April 2004, content.healthaffairs.org/cgi/content/abstract/hlthaff.w4.184 (28 July 2004).
26. Evidence suggests that the chronic care model developed by Ed Wagner
and colleagues, when implemented, can lead to improvements in both the process
and outcome of care. T. Bodenheimer, E.H. Wagner, and K. Grumbach, “Improving
Primary Care for Patients with Chronic Illness: The Chronic Care Model, Part
2,” Journal of the American Medical Association 288,
no. 15 (2002): 1909–1914.
27. M.C. Creditor, “Hazards of Hospitalization of the Elderly,” Annals
of Internal Medicine 118, no. 3 (1993): 219–223.
28. E.S. Fisher and H.G. Welch, “Avoiding the Unintended Consequences
of Growth in Medical Care: How Might More Be Worse?” Journal
of the American Medical Association 281, no. 5 (1999): 446–453.
29. An early proposal by Pete Welch and Mark Miller focused on fostering
integration at the medical staff level for the physician services provided
during a single inpatient stay, whereas we suggest a longer time window. Their
analysis also indicated that physicians have few hospital affiliations (averaging
about 1.5 per physician) but admit almost all of their patients to a single
hospital (averaging 90 percent). See W.P. Welch and M.E. Miller, “Proposals
to Control High-Cost Hospital Medical Staffs,” Health
Affairs 13, no. 4 (1994): 42–57.
30. A possible approach to implementing a collaborative effort of AMCs,
the National Institutes of Health (NIH), and the Agency for Healthcare Research
and Quality (AHRQ) to address these issues is discussed in J.E. Wennberg, E.S.
Fisher, and J.S. Skinner, “Geography and the Debate over Medicare Reform,” Health
Affairs, 13 February 2002, content.healthaffairs.org/cgi/content/abstract/hlthaff.w2.96 (28 July 2004).
Elliott Fisher (elliott.s.fisher{at}dartmouth.edu)
is a professor of medicine and of community and family medicine at Dartmouth
Medical School in Hanover, New Hampshire, and is codirector of the Veterans
Affairs (VA) Outcomes Group at the White River Junction (Vermont) VA Medical
Center. David Wennberg directs the Center for Outcomes Resesarch and Evaluation,
Maine Medical Center, in Portland. Thérèse Stukel is a senior scientist and
research director at the Institute for Clinical Evaluative Sciences, Toronto,
Ontario. Dan Gottlieb is a research associate at the Center for the Evaluative
Clinical Sciences, Dartmouth Medical School.
DOI: 10.1377/hlthaff.var.19
©2004 Project HOPE–The People-to-People Health Foundation, Inc.
|
