discussion question 1

At the Intersection of Health, Health Care and Policy

doi: 10.1377/hlthaff.26.4.1104 26, no.4 (2007):1104-1110Health Affairs

Hospital Quality Alliance Measures The Inverse Relationship Between Mortality Rates And Performance In The

Ashish K. Jha, E. John Orav, Zhonghe Li and Arnold M. Epstein Cite this article as:

http://content.healthaffairs.org/content/26/4/1104 available at:

The online version of this article, along with updated information and services, is

Permissions : For Reprints, Links &

http://content.healthaffairs.org/1340_reprints.php

Email Alertings : http://content.healthaffairs.org/subscriptions/etoc.dtl

To Subscribe : https://fulfillment.healthaffairs.org

without prior written permission from the Publisher. All rights reserved. mechanical, including photocopying or by information storage or retrieval systems, may be reproduced, displayed, or transmitted in any form or by any means, electronic or States copyright law (Title 17, U.S. Code), no part of by Project HOPE - The People-to-People Health Foundation. As provided by United Suite 600, Bethesda, MD 20814-6133. Copyright ©

is published monthly by Project HOPE at 7500 Old Georgetown Road,Health Affairs

Not for commercial use or unauthorized distribution

o n N

o ve

m b e r 6

, 2 0 1 6 b

y H W

T e a m

H e a lth

A ffa

irs b

y h ttp

://co n te

n t.h

e a lth

a ffa

irs.o rg

/ D

o w

n lo

a d e d fro

m

o n N

o ve

m b e r 6

, 2 0 1 6 b

y H W

T e a m

H e a lth

A ffa

irs b

y h ttp

://co n te

n t.h

e a lth

a ffa

irs.o rg

/ D

o w

n lo

a d e d fro

m

M a r k e t Wat c h

The Inverse Relationship Between Mortality Rates And Performance In The Hospital Quality Alliance Measures These findings show that the HQA measures are likely to play a central role in monitoring and improving hospital care.

by Ashish K. Jha, E. John Orav, Zhonghe Li, and Arnold M. Epstein

ABSTRACT: The Hospital Quality Alliance (HQA) program gives us the opportunity to sys- tematically monitor the quality of hospital care nationwide. To gauge the importance of the HQA indicators, we examined the relationship between hospitals’ performance on HQA quality indicators and mortality for Medicare enrollees admitted for acute myocardial infarc- tion, congestive heart failure, and pneumonia. We found that higher condition-specific per- formance on this national quality reporting program is associated with lower risk-adjusted mortality for each of the three conditions. The relationship between high HQA performance and lower risk-adjusted mortality is an important validation for this national hospital quality rating program. [Health Affairs 26, no. 4 (2007): 1104–1110; 10.1377/hlthaff.26.4.1104]

T h e h o s p i t a l Q u a l i t y A l l i a n c e (HQA) is the first national program to publicly report on detailed process-

based quality of care provided by hospitals.1

Based on provisions in the Medicare Prescrip- tion Drug, Improvement, and Modernization Act (MMA) of 2003, the U.S. Department of Health and Human Services (HHS) insti- tuted a program to collect data on key mea- sures of hospitals’ management of three com- mon medical conditions—acute myocardial infarction (AMI), congestive heart failure (CHF), and pneumonia—and to withhold up to 0.4 percent of the Medicare fee schedule update for hospitals that choose not to partic- ipate. The HQA data provide hospitals with performance benchmarks and can be used to guide quality improvement. Moreover, these

indicators give us a potentially important tool for monitoring the quality of care nation- wide.

Although there exist other programs for rating hospitals, the HQA has quickly become the largest and most comprehensive program, with near-universal participation. The three conditions assessed in the HQA are common (they constitute more than 15 percent of Medi- care hospital medical and surgical admissions) and have important impacts on patients’ mor- bidity and mortality. Initial evaluation of the program has shown that hospitals vary greatly in quality performance.2

The HQA performance indicators are de- rived from a broad consensus of experts and are usually based on strong evidence of effi- cacy. But how these quality metrics perform in

1 1 0 4 J u l y / A u g u s t 2 0 0 7

H e a l t h T r a c k i n g

DOI 10.1377/hlthaff.26.4.1104 ©2007 Project HOPE–The People-to-People Health Foundation, Inc.

Ashish Jha ([email protected]) is an assistant professor in the Harvard School of Public Health in Boston, Massachusetts. John Orav is an associate professor at Brigham and Women’s Hospital, also in Boston. Zhonghe Li is a statistical programmer and Arnold Epstein, a professor, at the Harvard School of Public Health.

o n N

o ve

m b e r 6

, 2 0 1 6 b

y H W

T e a m

H e a lth

A ffa

irs b

y h ttp

://co n te

n t.h

e a lth

a ffa

irs.o rg

/ D

o w

n lo

a d e d fro

m

the real world in identifying hospitals with better outcomes, such as lower risk-adjusted mortality across a number of clinical condi- tions, is largely unknown. Since the real goal of health care is to improve outcomes, it is of great interest to determine whether hospitals that perform well on the HQA measures also have better outcomes. The existence of such a relationship would bolster evidence of the va- lidity of the HQA measures. In this study, we sought to determine whether higher perfor- mance on the HQA measures was associated with lower risk-adjusted mortality for the three important medical conditions.

Study Data And Methods � Data. We used data from the 1 Decem-

ber 2005 release of the HQA program. These data report performance scores for 4,048 acute care hospitals for care provided from 1 April 2004 through 31 March 2005. Although HQA now has data on twenty process measures, we limited our analyses to the ten that are known as the “starter set” because MMA provides fi- nancial incentives for reporting only these ten measures. Therefore, while most hospitals re- port data on these ten measures, only a smaller number of hospitals also report data on the other ten. The recently passed Deficit Reduc- tion Act (DRA) of 2005 increased the financial incentive to hospitals for reporting and allows the HHS secretary to expand the number of measures required for the incentives. (The HQA data can be accessed at http://www .hospitalcompare.hhs.gov.)

We linked the HQA data with data from the American Hospital Association (AHA) an- nual survey to obtain information about hos- pital characteristics. We also linked each of these data sets with the 2003 Medicare Pro- vider and Analysis Review (MedPAR) Part A data set (100 percent file), which has dis- charge-level data for hospitalizations of all fee- for-service (FFS) Medicare beneficiaries. The MedPAR data set provided patient demo- graphic and clinical information through Inter- national Classification of Diseases, Ninth Revision (ICD-9), diagnoses for severity adjustment, as well as our primary outcome of mortality.

� Hospital quality metrics. For each hos- pital, we used ten HQA performance indica- tors to calculate a summary performance score for each of the three clinical conditions: AMI, CHF, and pneumonia. There were five perfor- mance indicators for AMI: aspirin at arrival, aspirin at discharge, beta-blocker at arrival, beta-blocker at discharge, and angiotensin- converting enzyme (ACE) inhibitor for left ventricular systolic (LVS) dysfunction; two in- dicators for CHF: left ventricular function as- sessment and ACE inhibitor for LVS dysfunc- tion; and three indicators for pneumonia: initial antibiotic timing (antibiotics provided in four hours or less), pneumococcal vaccina- tion, and oxygenation assessment.

For each hospital, we created condition- specific summary scores if the hospital re- ported a sample size of at least twenty-five for all of the indicators for that condition. The summary score is the simple weighted average of all measures for that condition. For exam- ple, if a hospital’s performance on the two CHF indicators was 90 percent (for indicator 1) and 96 percent (for indicator 2), the CHF summary score would be 93 percent. Because of the high level of correlation among the five AMI measures (Cronbach’s alpha 0.85), we also calculated summary scores for hospitals that had adequate sample sizes (twenty-five or more) for four out of five AMI measures.3 Fi- nally, to ensure that our results were robust to different methods for calculating summary scores, we also calculated summary scores us- ing alternative methods and found comparable results.4

� Outcomes. Three subgroups of patients were chosen from the MedPAR database, and we limited our analyses to those age sixty-five or older. The first subgroup consisted of pa- tients discharged with primary diagnosis of AMI (using ICD-9 codes 410.XX); the second included patients discharged with the primary diagnosis of CHF (ICD-9 codes 398.91 and 428.0–428.9); and the third subgroup con- sisted of patients discharged with the primary diagnosis of pneumonia (ICD-9 codes 480– 486). We excluded patients who were admit- ted as a result of transfer from other inpatient

M a r k e t W a t c h

H E A L T H A F F A I R S ~ V o l u m e 2 6 , N u m b e r 4 1 1 0 5

o n N

o ve

m b e r 6

, 2 0 1 6 b

y H W

T e a m

H e a lth

A ffa

irs b

y h ttp

://co n te

n t.h

e a lth

a ffa

irs.o rg

/ D

o w

n lo

a d e d fro

m

acute care facilities or were transferred to an- other acute care facility. Inpatient mortality was used as our primary measure of clinical outcome. The MedPAR database also provided patients’ age, race, sex, and comorbidities so that risk adjustment could be performed in our regression models using the Elixhauser comorbidity adjustment scheme.5 This tech- nique is now widely used and supported by the Agency for Healthcare Research and Qual- ity (AHRQ) as the risk-adjustment method of choice; it accounts for both demographic in- formation and the presence or absence of im- portant comorbidities.

� Statistical analyses. We used analyses of variance and chi-square tests, as appropri- ate, to compare hospitals that reported data to the HQA with a sample size adequate for us to calculate a summary score for each condition and those that reported no data. We then built separate multivariable logistic regression models to determine if high performance on the condition-specific HQA metric was asso- ciated with lower condition-specific inpatient mortality. Separate models were built for AMI patients, CHF patients, and pneumonia pa- tients, with each discharge as the unit of analy- sis and dead (versus alive) at discharge as the primary outcome. HQA scores were examined both as continuous variables and by quartiles of performance. In our first set of models, we created predicted probabilities of death in hos- pitals, stratified by that hospital’s performance on HQA (by quartiles), for the person with av- erage values for each of the covariates (age, race, sex, and each of the comorbidities in the Elixhauser scheme) in the entire data set. Next, we used generalized estimating equa- tions to account for clustering of patients at the hospital level and again adjusted for pa- tient characteristics including age, sex, race, and the absence or presence of each of thirty comorbidities. We then further adjusted for the characteristics of the hospital where the patient was treated, including bed size (in three categories), region (by four census cate- gories), teaching status (member of the Coun- cil of Teaching Hospitals versus not), profit status (for-profit versus nonprofit), location

(urban versus rural), and the presence or ab- sence of a coronary or medical intensive care unit (ICU).

To gauge the potential clinical importance of our findings, we calculated the potential number of deaths avoided if patients in the hospitals with the lowest quartile of HQA quality performance had mortality rates of pa- tients from hospitals with the highest quartile. We used our results on risk-adjusted mortality by quartile of HQA performance and the prev- alence of admitted patients with the three conditions as evident in 2003 MedPAR data to perform the appropriate calculation. All analy- ses were conducted using SAS 9.1.

Study Results Of the 4,648 hospitals listed in the AHA

survey as hospitals that care for medical and surgical patients, 4,140 reported some data to the HQA program, and 3,720 of these reported adequate data to the HQA program to allow us to generate at least one summary score, al- though the number of hospitals for which we had summary scores varied by condition. Hos- pitals that choose not to report to the HQA are generally small and often critical-access hospi- tals. Characteristics of reporting hospitals and all U.S. hospitals (reporting and not) are de- scribed in Exhibit 1. Hospitals that reported adequate data to allow for the creation of sum- mary scores varied in profit status, location (urban versus rural), and presence of ICUs.

� Association of HQA and risk-adjusted mortality. Hospitals in the bottom quartile of HQA performance had a predicted mortality of 10.8 percent for AMI, 5.0 percent mortality for CHF, and 7.9 percent mortality for pneu- monia (Exhibit 2). Higher performance was consistently associated with lower adjusted mortality rates, as hospitals in the top quartile had nearly 1 percent lower mortality among patients with AMI, 0.4 percent among pa- tients with CHF, and 0.8 percent among pa- tients with pneumonia.

In our models that accounted for clustering, higher HQA performance was associated with lower mortality for each of the three condi- tions. For example, patients who were dis-

1 1 0 6 J u l y / A u g u s t 2 0 0 7

H e a l t h T r a c k i n g

o n N

o ve

m b e r 6

, 2 0 1 6 b

y H W

T e a m

H e a lth

A ffa

irs b

y h ttp

://co n te

n t.h

e a lth

a ffa

irs.o rg

/ D

o w

n lo

a d e d fro

m

charged from the top quartile of HQA perfor- mance in AMI had 11 percent lower odds of dying than patients who were discharged from hospitals in the bottom quartile of HQA AMI performance (odds ratio 0.89, 95 percent con-

fidence interval 0.85, 0.94). Similarly, patients admitted to hospitals in the top quartile of CHF performance had 7 percent lower odds of death, and patients admitted to hospitals in the top quartile of pneumonia performance

M a r k e t W a t c h

H E A L T H A F F A I R S ~ V o l u m e 2 6 , N u m b e r 4 1 1 0 7

EXHIBIT 1 Characteristics Of Hospitals In The Hospital Quality Alliance (HQA) Program

Hospitals with AMI summary scores (n = 1,965)

Hospitals with CHF summary scores (n = 2,394)

Hospitals with pneumonia summary scores (n = 3,270)

All hospitals (n = 4,648)

Hospital characteristics Number Percent Number Percent Number Percent Number Percent

Bed size <100 100–400 >400

138 1,408

419

7.0 71.7 21.3

321 1,653

420

13.4 69.1 17.5

1,010 1,843

417

30.9 56.4 12.8

2,178 2,048

422

46.9 44.1 9.1

Region West Midwest South Northeast

376 441 721 404

19.4 22.7 37.1 20.8

401 564 953 457

16.9 23.7 40.1 19.2

550 848

1,300 543

17.0 26.2 40.1 16.8

883 1,369 1,737

612

19.0 29.5 37.4 13.2

For-profit Teaching Urban

302 265

1,654

15.4 13.5 84.2

382 273

1,882

16.0 11.4 78.6

522 270

2,107

16.0 8.3

64.4

688 277

2,446

14.8 6.0

52.6

Presence of cardiac ICU Presence of medical ICU

1,168 1,815

60.8 94.4

1,285 2,194

54.8 93.6

1,460 2,776

45.7 86.9

1,528 2,995

32.9 64.4

SOURCE: Authors’ analysis of HQA data, December 2005 release. NOTES: ICU is intensive care unit. AMI is acute myocardial infarction. CHF is congestive heart failure.

EXHIBIT 2 Adjusted Mortality Rates, Stratified By Hospital Performance On Hospital Quality Alliance (HQA) Summary Scores

Predicted mortality rate (95% confidence interval)

HQA performance AMI CHF Pneumonia

1st quartile 10.0% (9.7, 10.4)

4.6% (4.4, 4.8)

7.1% (6.9, 7.4)

2d quartile 10.2% (10.0, 10.5)

4.9% (4.8, 5.1)

7.4% (7.2, 7.6)

3d quartile 10.6% (10.3, 10.9)

5.0% (4.8, 5.2)

7.5% (7.2, 7.7)

4th quartile 10.8% (10.5, 11.2)

5.0% (4.8, 5.1)

7.9% (7.6, 8.1)

p value for trend <0.001 0.005 <0.001

SOURCE: Authors’ analysis of HQA data, December 2005 release, and Medicare data. NOTES: Adjusted for patient age, sex, race, and the presence or absence of each of thirty comorbidities. AMI is acute myocardial infarction. CHF is congestive heart failure.

o n N

o ve

m b e r 6

, 2 0 1 6 b

y H W

T e a m

H e a lth

A ffa

irs b

y h ttp

://co n te

n t.h

e a lth

a ffa

irs.o rg

/ D

o w

n lo

a d e d fro

m

had 15 percent lower odds of death (Exhibit 3). These findings persisted and changed little when we controlled for hospital characteris- tics including teaching status, region, bed size, and urban versus rural location (Exhibit 3).

� Clinical impact of differences in risk- adjusted mortality. We calculated the poten- tial number of deaths avoided if patients in the hospitals with the lowest quartile of HQA quality performance had the mortality rates of patients from hospitals with the highest quartile. For the three conditions, approxi- mately 2,200 deaths would have been avoided (474 for AMI, 627 for CHF, and 1,112 for pneu- monia).

Discussion A major goal of the HQA program is to

monitor hospital quality of care and drive quality improvement. The motivation for hos-

pitals to improve their quality performance depends on the face validity of the indicators and the evidence that better performance on them is associated with better outcomes for patients. Thus, our findings—that higher per- formance on the AMI, CHF, and pneumonia HQA indicators was each associated with lower risk-adjusted mortality—provide im- portant data that are likely to bolster the HQA’s impact.

Although strong scientific evidence under- pins many of the HQA performance measures, much of this evidence is based on efficacy studies and clinical trials. Thus, it was not ob- vious that hospitals with high performance on process indicators for a specific condition would have better outcomes for that condition in real-world operations, especially given the technical limitations of outcomes measure- ment and the differences between assessing ef-

1 1 0 8 J u l y / A u g u s t 2 0 0 7

H e a l t h T r a c k i n g

EXHIBIT 3 Adjusted Odds Ratio Of In-Hospital Mortality, By Hospital Quality Alliance (HQA) Performance Quartile

Adjusted odds ratio (95% confidence interval)

HQA performancea AMI CHF Pneumonia

1st quartile 0.89 (0.85, 0.94)

0.93 (0.88, 0.98)

0.85 (0.81, 0.89)

2d quartile 0.93 (0.89, 0.98)

0.99 (0.94, 1.04)

0.87 (0.83, 0.92)

3d quartile 0.96 (0.91, 1.01)

1.01 (0.96, 1.07)

0.90 (0.86, 0.95)

p value for trend <0.001 0.005 <0.001

HQA performanceb

1st quartile 0.91 (0.86, 0.96)

0.92 (0.88, 0.98)

0.90 (0.86, 0.95)

2d quartile 0.95 (0.90, 1.00)

0.98 (0.93, 1.04)

0.91 (0.87, 0.96)

3d quartile 0.97 (0.93, 1.02)

1.00 (0.94, 1.05)

0.92 (0.88, 0.97)

p value for trend 0.0002 0.0056 <.0001

SOURCE: Authors’ analysis of HQA data, December 2005 release, and Medicare data. NOTES: AMI is acute myocardial infarction. CHF is congestive heart failure. a Adjusted for patient age, sex, race, and the presence or absence of each of thirty comorbidities. b Adjusted as in Note a, as well as for teaching status, bed size, for-profit status, and region of the country.

o n N

o ve

m b e r 6

, 2 0 1 6 b

y H W

T e a m

H e a lth

A ffa

irs b

y h ttp

://co n te

n t.h

e a lth

a ffa

irs.o rg

/ D

o w

n lo

a d e d fro

m

fectiveness and efficacy. Moreover, some of the measures are unlikely to have a direct impact on inpatient mortality. This is true for pneu- mococcal vaccinations for pneumonia patients and true in part for beta-blockers at the time of discharge, depending on whether beta- blockers were prescribed only at discharge or earlier. Thus, the strength and consistency of the associations we found, especially for AMI and pneumonia care, suggest that these mea- sures identify hospitals with programs that ensure the provision of high- quality care that likely ex- tends beyond the specific processes measured in the HQA program. This level of effect of the HQA perfor- mance should help alleviate concerns that hospitals might focus only on the indicators being measured to the detri- ment of the patient’s well- being.6

� HQA progress and prospects. The HQA is an important development in the ef- fort to provide publicly available data on hos- pital performance. The effort so far has focused primarily on three common medical condi- tions. Recently, the HQA program has added quality indicators, provided now by only a small number of hospitals, on surgical infec- tion prevention. The DRA gave the Centers for Medicare and Medicaid Services (CMS) the option of expanding the set of “required” indi- cators even further, and the CMS has an- nounced plans to require an expanded set of indicators going from the current number of ten to more than twenty. This effort to monitor quality of care nationally is likely to continue and be expanded to other areas of clinical care. We expect this system to play a central role in U.S. efforts to monitor and improve the quality of hospital care nationally.

� Comparison with a previous study. We are aware of only one previous examina- tion of the relationship between HQA quality measures and risk-adjusted mortality at the hospital level. Rachel Werner and Ed Bradlow examined relationships between individual

HQA measures and risk-adjusted mortality rates and had broadly similar findings.7 They focused on differences between “high perform- ers” (hospitals that performed at the seventy- fifth percentile for all metrics) and “low per- formers” (hospitals that performed at the twenty-fifth percentile for all metrics) and found significant differences in mortality rates. We were able to examine the relation- ship between the HQA and outcomes across the entire spectrum of performance, and our

findings suggest that the rela- tionship is consistent: that across all four quartiles of HQA performance, mortality rates improve for each of the three conditions. Further- more, we performed two sets of analyses: the first without adjusting for hospital charac- teristics and the second, with adjustment for hospital char- acteristics. By demonstrating

that the results do not change meaningfully, our findings suggest a robust relationship be- tween HQA measures and outcomes that is less likely to be attributable to unmeasured confounders.

� Clinical importance of our study. Al- though the differences in mortality rates across the four quartiles were statistically sig- nificant, one might wonder whether they were clinically meaningful or important. We believe that they are. First, the mortality benefit of be- ing a patient in a hospital in the top (versus bottom) quartile is comparable to that of re- ceiving a beta-blocker after an AMI versus not receiving one—roughly a 14 percent diminu- tion in the rate of mortality.8 Further, we calcu- lated that if hospitals in the bottom quartile had mortality rates that were comparable to those in the top quartile, 2,200 fewer elderly Americans would die each year from the three conditions studied.

� Study limitations. The HQA program only examines process measures across three conditions, and although these conditions are common and a source of major morbidity and mortality, they make up only 15 percent of hos-

M a r k e t W a t c h

H E A L T H A F F A I R S ~ V o l u m e 2 6 , N u m b e r 4 1 1 0 9

“If hospitals in the bottom quartile had mortality rates that were comparable to

those in the top quartile, 2,200 fewer

elderly Americans would die each year.”

o n N

o ve

m b e r 6

, 2 0 1 6 b

y H W

T e a m

H e a lth

A ffa

irs b

y h ttp

://co n te

n t.h

e a lth

a ffa

irs.o rg

/ D

o w

n lo

a d e d fro

m

pital admissions. Also, we used administrative data to perform risk adjustment for in-hospital mortality; thus, our ability to fully account for differences in underlying risk among patients was limited.

Next, as described above, some of the met- rics we used in calculating our summary scores have demonstrated long-term benefits but might be unlikely to improve mortality in the short term. Therefore, by including those metrics, we might have underestimated the strength of the relationship between HQA performance and mortality rates. Given that the HQA program does not differentiate be- tween metrics that have short- versus long- term benefits and that previous investigators have generally included all of the HQA metrics that are part of the “starter set,” we chose to not exclude any of the individual metrics.

Finally, despite the fact that our quality metrics focus on care for patients of all ages, our outcomes data were obtained from Medi- care files and only assessed outcomes for the elderly. However, at least one previous study suggests that outcomes for the elderly closely track those for all patients in most hospitals.9

I n t h i s s t u d y we found that high perfor- mance on the HQA was associated with 7–15 percent lower odds of death for each

of three clinical conditions. These data docu- ment the importance of the HQA indicators for patient outcomes and are likely to in- crease the impact of the CMS effort in quality reporting. With the recent passage of the DRA, it is clear that the United States has em- barked on a continuing and expanding initia- tive to monitor the quality of hospital care. Our findings underscore the potential of this effort for improving quality of care and changing patient outcomes.

The authors are grateful to Sheila Roman at the Centers for Medicare and Medicaid Services for her invaluable insights about the Hospital Quality Alliance data. This paper was presented at the 2006 AcademyHealth meeting in Seattle, Washington. The authors acknowledge the support of the Commonwealth Fund.

NOTES 1. A.K. Jha et al., “Care in U.S. Hospitals—The Hos-

pital Quality Alliance Program,” New England Jour- nal of Medicine 353, no. 3 (2005): 265–274; and S.C. Williams et al., “Quality of Care in U.S. Hospitals as Reflected by Standardized Measures, 2002– 2004,” New England Journal of Medicine 353, no. 3 (2005): 255–264.

2. Jha et al., “Care in U.S. Hospitals.”

3. Ibid.

4. C.N. Kahn III et al., “Snapshot of Hospital Qual- ity Reporting and Pay-for-Performance under Medicare,” Health Affairs 25, no. 1 (2006): 148–162.

5. A. Elixhauser et al., “Comorbidity Measures for Use with Administrative Data,” Medical Care 36, no. 1 (1998): 8–27.

6. R.M. Werner and D.A. Asch, “The Unintended Consequences of Publicly Reporting Quality In- formation,” Journal of the American Medical Associa- tion 293, no. 10 (2005): 1239–1244.

7. R.M. Werner and E.T. Bradlow, “Relationship be- tween Medicare’s Hospital Compare Perfor- mance Measures and Mortality Rates,” Journal of the American Medical Association 296, no. 22 (2006): 2694–2702.

8. Miami Trial Research Group, “Metoprolol in Acute Myocardial Infarction: Patient Popula- tion,” American Journal of Cardiology 56, no. 14 (1985): 10G–14G.

9. J. Needleman et al., “Measuring Hospital Quality: Can Medicare Data Substitute for All-Payer Data?” Health Services Research 38, no. 6, Part 1 (2003): 1487–1508.

1 1 1 0 J u l y / A u g u s t 2 0 0 7

H e a l t h T r a c k i n g

o n N

o ve

m b e r 6

, 2 0 1 6 b

y H W

T e a m

H e a lth

A ffa

irs b

y h ttp

://co n te

n t.h

e a lth

a ffa

irs.o rg

/ D

o w

n lo

a d e d fro

m