Life Expectancy after Myocardial Infarction, According to Hospital Performance

Transcription

1 The new england journal of medicine Original Article Life Expectancy after Myocardial Infarction, According to Hospital Performance Emily M. Bucholz, M.D., Ph.D., M.P.H., Neel M. Butala, M.D., Shuangge Ma, Ph.D., Sharon Lise T. Normand, Ph.D., and Harlan M. Krumholz, M.D. ABSTRACT From the Department of Medicine, Boston Children s Hospital (E.M.B.), the Department of Internal Medicine, Massachusetts General Hospital (N.M.B.), the Department of Health Care Policy, Harvard Medical School (S.-L.T.N.), and the Department of Biostatistics, Harvard T.H. Chan School of Public Health (S.-L.T.N.) all in Boston; and the Departments of Biostatistics (S.M.) and Health Policy and Management (H.M.K.), Yale School of Public Health, the Section of Cardiovascular Medicine, Department of Internal Medicine, and Robert Wood Johnson Clinical Scholars Program, Yale School of Medicine (H.M.K.), and Center for Outcomes Research and Evaluation, Yale New Haven Hospital (H.M.K.) all in New Haven, CT. Address reprint requests to Dr. Krumholz at the Department of Internal Medicine, Yale School of Medicine, 1 Church St., Suite 200, New Haven, CT 06510, or at harlan. krumholz@ yale. edu. This article was updated on October 6, 2016, at NEJM.org. N Engl J Med 2016;375: DOI: /NEJMoa Copyright 2016 Massachusetts Medical Society. BACKGROUND Thirty-day risk-standardized mortality rates after acute myocardial infarction are commonly used to evaluate and compare hospital performance. However, it is not known whether differences among hospitals in the early survival of patients with acute myocardial infarction are associated with differences in long-term survival. METHODS We analyzed data from the Cooperative Cardiovascular Project, a study of Medicare beneficiaries who were hospitalized for acute myocardial infarction between 1994 and 1996 and who had 17 years of follow-up. We grouped hospitals into five strata that were based on case-mix severity. Within each case-mix stratum, we compared life expectancy among patients admitted to high-performing hospitals with life expectancy among patients admitted to low-performing hospitals. Hospital performance was defined by quintiles of 30-day risk-standardized mortality rates. Cox proportional-hazards models were used to calculate life expectancy. RESULTS The study sample included 119,735 patients with acute myocardial infarction who were admitted to 1824 hospitals. Within each case-mix stratum, survival curves of the patients admitted to hospitals in each risk-standardized mortality rate quintile separated within the first 30 days and then remained parallel over 17 years of follow-up. Estimated life expectancy declined as hospital risk-standardized mortality rate quintile increased. On average, patients treated at high-performing hospitals lived between 0.74 and 1.14 years longer, depending on hospital case mix, than patients treated at low-performing hospitals. When 30-day survivors were examined separately, there was no significant difference in unadjusted or adjusted life expectancy across hospital risk-standardized mortality rate quintiles. CONCLUSIONS In this study, patients admitted to high-performing hospitals after acute myocardial infarction had longer life expectancies than patients treated in low-performing hospitals. This survival benefit occurred in the first 30 days and persisted over the long term. (Funded by the National Heart, Lung, and Blood Institute and the National Institute of General Medical Sciences Medical Scientist Training Program.) 1332

2 Public reporting has become a mainstay of national efforts to improve the quality of care delivered in U.S. hospitals. 1 Increasingly, risk-standardized mortality rates are used to benchmark quality and gauge hospital performance because they reflect meaningful and widely interpretable results of hospital care. 2,3 Since 2007, the Centers for Medicare and Medicaid Services (CMS) has reported hospital-specific 30-day risk-standardized mortality rates for several common conditions, and more recently, riskstandardized mortality rates have been incorporated into payment policies. 4-6 Although several studies have evaluated the association of condition-specific risk-standardized mortality rates with other short-term quality metrics, 7-16 it is not known whether patients admitted to hospitals that are associated with better short-term outcomes have improved longterm survival. The short-term survival benefit conferred at high-performing hospitals may dissipate over time, which would lend support to the theory that these hospitals discharge more patients alive but with higher subsequent mortality. Alternatively, patients treated at high-performing hospitals may have a survival benefit that persists over time, which would suggest that differences in the quality of care delivered produce an early benefit that endures over time. Accordingly, we used data from the Cooperative Cardiovascular Project (CCP), a large, nationally representative cohort study of Medicare beneficiaries who had been hospitalized with acute myocardial infarction and had more than 17 years of follow-up, to evaluate the association between hospital 30-day risk-standardized mortality rates and life expectancy after acute myocardial infarction. We selected acute myocardial infarction because it was one of the first conditions for which 30-day risk-standardized mortality rates were developed and because there is considerable heterogeneity in risk-standardized mortality rates across hospitals. We used life expectancy to measure long-term survival because it is an easily interpretable metric that is meaningful to patients and can be used to calculate the years of life saved by treatment at high-performing versus low-performing hospitals. Methods Study Design and Conduct The study was designed by all the authors and was approved by the institutional review board at Yale University. Data were obtained through an agreement with Qualidigm, a quality-improvement organization that contracts with Medicare for data analysis and release. Qualidigm was not involved in any of the data analyses or in the preparation of the manuscript. Funding was provided by the National Heart, Lung, and Blood Institute and the Medical Scientist Training Program of the National Institute of General Medical Sciences. The first author had full access to all the data in the study and assumes responsibility for the completeness and integrity of the data and the accuracy of the data analyses. Study Population We analyzed data from the CCP, a quality-improvement initiative from the Health Care Financing Administration (now CMS) for patients with acute myocardial infarction, which has been described previously. 17,18 The CCP included fee-forservice Medicare beneficiaries hospitalized with a principal discharge diagnosis of acute myocardial infarction (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 410), except acute myocardial infarction readmissions (ICD-9-CM code 410.x2). Each hospital was sampled for an 8-month period during the interval from January 1994 through February Trained personnel at centralized dataabstraction centers abstracted patient records for information on demographic characteristics, medical history, clinical presentation, laboratory and electrocardiographic data, and treatment. In our analyses, we used inclusion criteria that were similar to those used for established risk-standardized mortality rate measures. 19,20 Specifically, we limited our sample to patients 65 years of age or older who had a clinically confirmed acute myocardial infarction. When patients were hospitalized more than once during the study period, we used only the first admission. We excluded patients who were admitted directly from the ambulatory surgery department, 1333

3 The new england journal of medicine patients who were transferred from other acute care hospitals, and patients who left the hospital against medical advice. Finally, we limited the analysis to hospitals that had at least 30 patients with a discharge diagnosis of acute myocardial infarction during the 8-month sampling period. Outcome Variable We used data from the Medicare denominator files for 1994 through 2012 to ascertain survival over 17 years of follow-up. The denominator files contain demographic and enrollment information on all Medicare fee-for-service and Medicare Advantage beneficiaries in a given year, including dates of death. Time to death was defined as the number of days from admission to the date of death, with data censored at 17 years of follow-up. Calculation of Risk-Standardized Mortality Rates We calculated risk-standardized mortality rates using a medical-record model described by Krumholz and colleagues 19,20 (see also the Methods section in the Supplementary Appendix, available with the full text of this article at NEJM.org). In brief, we used hierarchical logistic regression to model the log odds of death within 30 days after admission as a function of patient demographic and clinical variables as well as a random hospital-specific effect. Riskstandardized mortality rates were calculated as the ratio of predicted to expected mortality at a given hospital, multiplied by the national observed mortality rate. 20 For each hospital, the denominator ( expected mortality) is the number of deaths expected within 30 days on the basis of national mortality data for the case mix of that hospital. It was calculated with the use of a common intercept for all hospitals and thus can be viewed as a measure of hospital case mix. The numerator ( predicted mortality) is the number of predicted deaths on the basis of the performance of that specific hospital. It was calculated with the use of a hospital-specific intercept and is analogous to observed mortality. Conceptually, the ratio of predicted to expected mortality compares the performance of a particular hospital with that of an average hospital with the same case mix. Sample Stratification To ensure that we were comparing life-expectancy estimates among patients admitted to hospitals with a similar case mix, we stratified hospitals into quintiles that were based on their expected mortality (i.e., the denominator of the 30-day risk-standardized mortality rate model), which is a marker of case mix or patient risk (Fig. 1). Life-expectancy analyses were performed separately for each case-mix stratum to permit comparisons of patients admitted to hospitals with a similar case mix. Within each case-mix stratum, we ranked hospitals by risk-standardized mortality rate and grouped them into quintiles (Fig. 1). These quintiles reflect hospital performance on 30-day mortality among hospitals with a similar case mix and were the primary unit of analysis for all life-expectancy calculations. We used the terms high-performing and low-performing to refer to hospitals in the lowest and highest 30-day mortality quintile, respectively, within each case-mix stratum. Life-Expectancy Calculations Life expectancy was defined as mean survival and was calculated as the area under the survival curve. We used a three-step process to calculate life expectancy (see the Methods section and Fig. S1 in the Supplementary Appendix). First, we fit a separate marginal Cox proportionalhazards model within each case-mix stratum. The model included dummy variables for riskstandardized mortality rate quintile and patient age. Second, we plotted the expected survival curves separately for patients within each riskstandardized mortality rate quintile and extrapolated the curves to age 100 using exponential models. We selected exponential models because we lacked information on the shape of the survival curves and thus chose a model with a constant hazard that does not make assumptions about changes to the hazard function over time. Third, we calculated mean life-expectancy estimates by summing the areas under the individual Cox and exponential survival functions. We then calculated the years of life saved as the difference in life expectancy between patients treated at hospitals in the lowest riskstandardized mortality rate quintile and those treated at hospitals in the highest risk-standard- 1334

4 234,771 Patients were included in the initial study cohort 115,036 Were excluded 15,307 Were <65 yr of age 22,891 Had unconfirmed acute myocardial infarction 139 Were hospitalized outside the United States 27,500 Had repeat hospitalizations 19,379 Were admitted from ambulatory surgery or other acute hospital stays 6,818 Were transferred from other acute care hospitals 237 Left hospital against medical advice 27,694 Were admitted to hospitals with <30 patients with discharge diagnosis of acute myocardial infarction during the study period 119,735 Patients in 1824 hospitals were included in the study Step 1: Hospitals were stratified into quintiles according to expected mortality rates to compare hospitals with a similar case mix Healthiest Patients Sickest Patients Case-mix stratum 1: 21,054 patients in 364 hospitals Case-mix stratum 2: 24,891 patients in 365 hospitals Case-mix stratum 3: 26,228 patients in 365 hospitals Case-mix stratum 4: 24,618 patients in 365 hospitals Case-mix stratum 5: 22,944 patients in 365 hospitals Step 2: Within each case-mix stratum, hospitals were stratified into quintiles according to risk-standardized mortality rates to evaluate hospital performance among hospitals with a similar case mix (lowest quintile is defined as high-performing; highest quintile is defined as low-performing) Step 3: Separate models were performed for each case-mix stratum to calculate and compare life expectancy for patients admitted to hospitals in each performance quintile Figure 1. Patient Cohort, Case-Mix Strata, and Hospital-Performance Quintiles. The total cohort included 119,735 patients with acute myocardial infarction who were distributed across 1824 hospitals (for excluded patients, more than one criterion may apply). Hospitals were rank-ordered by expected mortality rate (a measure of case mix) and stratified into quintiles ranging from those admitting more-healthy patients to those admitting less-healthy patients. Within each case-mix stratum, hospitals were then ranked by risk-standardized mortality rate (a measure of hospital performance) and stratified into quintiles ranging from higher-quality to lower-quality hospitals. The association between risk-standardized mortality rate quintile and life expectancy after myocardial infarction was analyzed separately for each case-mix stratum. ized mortality rate quintile within each case-mix stratum. To determine whether differences in life expectancy across risk-standardized mortality rate quintiles could be explained by differences in patient characteristics and treatment among hospitals, we repeated the life-expectancy calculations above, adjusting first for clinical characteristics and then for treatment characteristics. The clinical model included 35 characteristics related to sociodemographic factors, medical history, frailty, and clinical presentation. The treatment model was adjusted for the same variables as those in the clinical model, in addition to reperfusion therapies, aspirin, and betablockers on admission. (For details of these 1335

5 The new england journal of medicine models, see the Methods section in the Supplementary Appendix.) Finally, to determine whether the survival benefit associated with admission to a higherperforming hospital occurred exclusively in the first 30 days or continued to increase after 30 days, we repeated unadjusted and adjusted lifeexpectancy calculations among 30-day survivors only. The same methods and covariates were applied to calculate life expectancy and years of life saved in all the patients and in 30-day survivors. Life expectancy was calculated from the time of admission for the overall cohort and from 30 days for 30-day survivors. All statistical analyses were performed with the use of SAS software, version 9.3 (SAS Institute). Results Patient Characteristics The final study sample included 119,735 patients with acute myocardial infarction who were admitted to 1824 hospitals (Fig. 1). Chart-abstracted baseline characteristics of patients in each casemix stratum are shown in Table 1, and clinicalpresentation and treatment characteristics of patients in each case-mix stratum are shown in Table S1 in the Supplementary Appendix. Patients admitted to hospitals in the highest casemix stratum (i.e., highest expected mortality) were older on average, had a higher prevalence of diabetes, and were less likely to have undergone previous coronary revascularization procedures than patients admitted to hospitals in the lowest case-mix stratum (i.e., lowest expected mortality). Patients admitted to hospitals in the highest case-mix stratum had higher rates of shock and heart failure on presentation than patients in the lowest case-mix stratum. In addition, they were less likely to have received reperfusion therapy or aspirin on admission. Life Expectancy Unadjusted survival curves for patients in each risk-standardized mortality rate quintile were plotted for each of the five case-mix strata (Fig. S2 through S6 in the Supplementary Appendix). Within each case-mix stratum, patients in the lowest risk-standardized mortality rate quintile (i.e., those admitted to high-performing hospitals) had the highest cumulative probability of survival, whereas patients in the highest riskstandardized mortality rate quintile (i.e., those admitted to low-performing hospitals) had the lowest cumulative probability of survival over all 17 years of follow-up. Unadjusted life-expectancy estimates after acute myocardial infarction showed similar patterns according to risk-standardized mortality rate quintile and case-mix stratum (Fig. S7 and Table S2 in the Supplementary Appendix). As casemix severity increased (moving from case-mix stratum 1 to stratum 5), life expectancy decreased. Similarly, within each case-mix stratum, life expectancy decreased as risk-standardized mortality rate quintile increased, which resulted in significant differences in life expectancy between the high-performing hospitals and the low-performing hospitals (Table 2). For example, in the lowest case-mix stratum, patients treated at high-performing hospitals lived, on average, 1.07 years (95% confidence interval [CI], 0.75 to 1.39) longer than patients treated at low-performing hospitals. Similarly, in the highest casemix stratum, the difference in life expectancy among patients treated at high-performing versus low-performing hospitals was 0.83 years (95% CI, 0.63 to 1.13). These differences varied slightly by case mix but were significant in all five case-mix strata (P<0.001 for all comparisons). Differences in patient life expectancy between high-performing and low-performing hospitals persisted after adjustment for patient sociodemographic and clinical characteristics in the clinical model (Table 2, and Fig. S8 and Table S2 in the Supplementary Appendix) and after further adjustment for treatment in the full model (Table 2 and Fig. 2, and Table S2 in the Supplementary Appendix). After full adjustment, patients treated at high-performing hospitals lived an average of 1.14 years (95% CI, 0.84 to 1.44) longer than patients treated at low-performing hospitals in the lowest case-mix stratum and 0.84 years (95% CI, 0.60 to 1.09) longer in the highest case-mix stratum. Differences in life expectancy across risk-standardized mortality rate quintiles remained significant in all five casemix strata (P<0.001 for all comparisons). Life Expectancy among 30-Day Survivors When 30-day survivors were examined separately, survival curves for patients in the five risk-standardized mortality rate quintiles were nearly identical in each of the five case-mix strata (Fig. S9 1336

6 Table 1. Patient Characteristics According to Hospital Case-Mix Stratum.* Variable Overall (N=119,735) Stratum 1 (N = 21,054) Stratum 2 (N = 24,891) Stratum 3 (N = 26,228) Stratum 4 (N = 24,618) Stratum 5 (N = 22,944) Demographic characteristics Age yr 76.5± ± ± ± ± ±7.6 <0.001 Female sex no. (%) 58,453 (48.8) 9,681 (46.0) 11,855 (47.6) 12,817 (48.9) 12,268 (49.8) 11,832 (51.6) <0.001 Nonwhite race no. (%) 10,502 (8.8) 2,140 (10.2) 2,025 (8.1) 2,170 (8.3) 2,109 (8.6) 2,058 (9.0) <0.001 Median ZIP Code level household 49.5± ± ± ± ± ±29.1 <0.001 income percentile Missing data no. (%) 4,879 (4.1) 1,118 (5.3) 1,263 (5.1) 1002 (3.8) 876 (3.6) 620 (2.7) <0.001 Medical history Diabetes no. (%) 36,690 (30.6) 6,162 (29.3) 7,568 (30.4) 8,157 (31.1) 7,588 (30.8) 7,215 (31.4) <0.001 Hypertension no. (%) 74,531 (62.2) 13,131 (62.4) 15,490 (62.2) 16,452 (62.7) 15,231 (61.9) 14,227 (62.0) 0.31 History of AMI no. (%) 35,668 (29.8) 6,160 (29.3) 7,510 (30.2) 7,978 (30.4) 7,256 (29.5) 6,764 (29.5) 0.02 History of CABG no. (%) 15,484 (12.9) 3,007 (14.3) 3,334 (13.4) 3,427 (13.1) 2,982 (12.1) 2,734 (11.9) <0.001 History of PCI no. (%) 8,294 (6.9) 1,753 (8.3) 1,858 (7.5) 1,822 (6.9) 1,559 (6.3) 1,302 (5.7) <0.001 History of stroke or TIA no. (%) 16,701 (13.9) 2,745 (13.0) 3,301 (13.3) 3,763 (14.3) 3,433 (13.9) 3,459 (15.1) <0.001 COPD no. (%) 24,167 (20.2) 4,173 (19.8) 5,064 (20.3) 5,417 (20.7) 4,921 (20.0) 4,592 (20.0) 0.15 Current smoker no. (%) 17,371 (14.5) 3,310 (15.7) 3,789 (15.2) 3,860 (14.7) 3,441 (14.0) 2,971 (12.9) <0.001 Obesity no./total no. (%) 18,824/100,857 (18.7) 3,438/18,030 (19.1) 4,115/21,290 (19.3) 4,268/22,024 (19.4) 3,758/20,692 (18.2) 3,245/18,821 (17.2) <0.001 Missing data no. (%) 18,878 (15.8) 3,024 (14.4) 3,601 (14.5) 4,204 (16.0) 3,926 (15.9) 4,123 (18.0) <0.001 CHF no. (%) 25,610 (21.4) 3,753 (17.8) 4,901 (19.7) 5,788 (22.1) 5,630 (22.9) 5,538 (24.1) <0.001 Chronic kidney disease no. (%) 5,817 (4.9) 836 (4.0) 1,155 (4.6) 1,182 (4.5) 1,301 (5.3) 1,343 (5.9) <0.001 Cancer no. (%) 2,770 (2.3) 446 (2.1) 573 (2.3) 592 (2.3) 598 (2.4) 561 (2.4) 0.13 Dementia no. (%) 7,116 (5.9) 1,069 (5.1) 1,268 (5.1) 1,544 (5.9) 1,556 (6.3) 1,679 (7.3) <0.001 Anemia no. (%) 8,394 (7.0) 1,367 (6.5) 1,644 (6.6) 1,825 (7.0) 1,801 (7.3) 1,757 (7.7) <0.001 Frailty indexes Admission from skilled nursing 7,661 (6.4) 947 (4.5) 1,368 (5.5) 1,624 (6.2) 1,693 (6.9) 2,029 (8.8) <0.001 facility no. (%) Immobile on admission no./ 22,028/116,654 (18.9) 3,256/20,566 (15.8) 4,310/24,317 (17.7) 4,842/25,610 (18.9) 4,724/23,963 (19.7) 4,896/22,198 (22.1) <0.001 total no. (%) Missing data no. (%) 3,081 (2.6) 488 (2.3) 574 (2.3) 618 (2.4) 655 (2.7) 746 (3.3) <0.001 Incontinent on admission no./ 8,505/117,105 (7.3) 1,131/20,703 (5.5) 1,586/24,403 (6.5) 1,891/25,651 (7.4) 1,847/24,059 (7.7) 2,050/22,289 (9.2) <0.001 total no. (%) Missing data no. (%) 2,630 (2.2) 351 (1.7) 488 (2.0) 577 (2.2) 559 (2.3) 655 (2.9) <0.001 P Value for Trend * Plus minus values are means ±SD. A higher case-mix stratum signifies hospitals that admitted sicker patients with a higher expected mortality rate. Thus, hospitals in stratum 1 admitted the healthiest patients with the lowest expected mortality rate, and hospitals in stratum 5 admitted the sickest patients with the highest expected mortality rate. AMI denotes acute myocardial infarction, CABG coronary-artery bypass grafting, CHF congestive heart failure, COPD chronic obstructive pulmonary disease, PCI percutaneous coronary intervention, and TIA transient ischemic attack. P values are for the trend across all five case-mix strata. 1337

7 The new england journal of medicine Population and Hospital Case-Mix Stratum 30-Day Expected Mortality Rate Mean Years of Live Saved in Patients Treated at High-Performing versus Low-Performing Hospitals (95% CI) Adjusted for Clinical Unadjusted Characteristics Adjusted for Treatment percent All patients Stratum 1: healthiest patients 13.2± (0.75 to 1.39) 1.20 (0.89 to 1.51) 1.14 (0.84 to 1.44) Stratum ± (0.82 to 1.25) 0.89 (0.62 to 1.15) 0.74 (0.48 to 1.01) Stratum ± (0.67 to 1.23) 0.94 (0.71 to 1.18) 0.80 (0.57 to 1.03) Stratum ± (0.64 to 1.26) 0.85 (0.60 to 1.11) 0.77 (0.52 to 1.01) Stratum 5: sickest patients 20.3± (0.63 to 1.13) 0.93 (0.68 to 1.18) 0.84 (0.60 to 1.09) 30-Day survivors only** Stratum 1: healthiest patients 13.2± ( 0.15 to 0.57) 0.41 (0.07 to 0.74) 0.40 (0.07 to 0.72) Stratum ± ( 0.18 to 0.48) 0.11 ( 0.18 to 0.41) 0.06 ( 0.24 to 0.35) Stratum ± ( 0.22 to 0.40) 0.24 ( 0.03 to 0.51) 0.20 ( 0.06 to 0.46) Stratum ± ( 0.31 to 0.43) 0.03 ( 0.25 to 0.31) 0.02 ( 0.26 to 0.29) Stratum 5: sickest patients 20.3± ( 0.39 to 0.23) 0.08 ( 0.22 to 0.37) 0.06 ( 0.22 to 0.35) * Plus minus values are means ±SE. Hospitals were grouped into case-mix strata with the use of the expected mortality estimates generated from the risk-standardized mortality rate models as a marker of case-mix severity. Mean expected mortality estimates for hospitals in each stratum are provided here to show the range of case-mix severity across strata. High-performing hospitals were defined as those in the lowest quintile of risk-stratified mortality rates, and low-performing hospitals were defined as those in the highest quintile of risk-stratified mortality rates. The clinical model was adjusted for patient sociodemographic factors (age, sex, race, and socioeconomic status), medical history (diabetes, hypertension, history of acute myocardial infarction, history of coronary-artery bypass grafting, history of percutaneous coronary intervention, history of stroke or transient ischemic attack, chronic obstructive pulmonary disease, congestive heart failure, chronic kidney disease, cancer, dementia or Alzheimer s disease, anemia, current smoking, and obesity), frailty (admission from a skilled nursing facility, immobility on admission, and incontinence on admission), and clinical presentation (heart rate on presentation, initial systolic blood pressure, shock within the first 48 hours of admission, heart failure within the first 48 hours of admission, duration of chest pain before presentation, blood urea nitrogen level, creatinine level, white-cell count, ST-segment elevation myocardial infarction, left or right bundle-branch block, heart block, and anterior or lateral acute myocardial infarction). The treatment model was adjusted for all the variables in the clinical model in addition to revascularization (percutaneous coronary intervention or coronary-artery bypass grafting) within 30 days after admission, fibrinolytic therapy during hospitalization, aspirin within 48 hours after admission, and beta-blockers within 48 hours after admission. Table 2. Mean Years of Life Saved among Patients Treated at High-Performing versus Low-Performing Hospitals, According to Hospital Case-Mix Stratum.* All P values for the comparison of years of life saved in all patients admitted to high-performing versus low-performing hospitals are less than ** P values for the comparison of years of life saved in 30-day survivors admitted to high-performing versus low-performing hospitals are greater than 0.05, unless otherwise noted. P = through S13 in the Supplementary Appendix). As was the case in the analyses that included all patients, life-expectancy estimates declined as case-mix severity increased, which reflected the fact that hospitals in these case-mix strata were caring for patients who were at higher risk. However, life-expectancy estimates were similar for 30-day survivors in all risk-standardized mortality rate quintiles within a given case-mix stratum (Fig. S14 and Table S3 in the Supplementary Appendix). Unadjusted differences in life expectancy between the highest-performing and lowest-performing hospitals were not significant for any case-mix stratum (P>0.10 for all comparisons). Life-expectancy estimates were similar for 30-day survivors treated at highperforming and those treated at low-performing hospitals even after adjustment for differences in clinical characteristics and treatment among hospitals (Table 2 and Fig. 3, and Table S3 and Fig. S15 in the Supplementary Appendix). 1338

8 Mean Life Expectancy after Acute Myocardial Infarction (yr) RSMR quintile 1 (high-performing) Case-Mix Stratum 1 (hospitals that admitted the healthiest patients) RSMR quintile 2 RSMR quintile 3 RSMR quintile 4 RSMR quintile 5 (low-performing) Case-Mix Stratum 2 Case-Mix Stratum 3 Case-Mix Stratum 4 Case-Mix Stratum 5 (hospitals that admitted the sickest patients) Figure 2. Life Expectancy after Acute Myocardial Infarction, According to Risk-Standardized Mortality Rate (RSMR) Quintile. Life expectancy after acute myocardial infarction was calculated from the time of admission, with adjustment for patient characteristics and treatment, and was estimated separately for each case-mix stratum. A higher case-mix stratum signifies hospitals that admitted sicker patients, who had a higher expected mortality rate. Within each case-mix stratum, a higher RSMR quintile indicates lower-performing hospitals (those with a higher-than-expected mortality rate). I bars indicate 95% confidence intervals. Discussion Using data from a large medical-record study involving patients with acute myocardial infarction, we found significant differences in life expectancy between patients admitted to hospitals with high performance on 30-day mortality quality measures and patients admitted to hospitals with low performance on those measures. After grouping hospitals with similar case mixes, we found that patients treated at high-performing hospitals (i.e., low 30-day risk-standardized mortality rates) lived, on average, between 0.74 and 1.14 years longer after acute myocardial infarction than those treated at low-performing hospitals (i.e., high 30-day risk-standardized mortality rates). These findings were consistent across case-mix strata, which indicates that the relationship between hospital performance and long-term patient outcomes is independent of hospital case mix. The survival advantage for patients treated at high-performing hospitals arose from differences in survival during the first 30 days after hospitalization and then persisted during the remainder of follow-up. Previous studies have similarly shown that the survival benefits associated with individual therapies occur largely in the first 30 days and then persist over time. In the Second International Study of Infarct Survival (ISIS-2) and the Gruppo Italiano per lo Studio della Sopravvivenza nell Infarto-1 (GISSI-1) trial, aspirin and reperfusion therapy were associated with significant survival gains in the first 30 days to 1 year, which then persisted over a period of 10 years. 21,22 Our study extends these findings to short-term hospital-outcome measures and shows that the early survival advantage achieved by high-performing hospitals is durable. If hospitals with low 30-day risk-standardized mortality rates achieved lower-than-expected mortality by forestalling death for the first 30 days, we might expect higher long-term mortality rates and thus shorter life expectancy in 30-day survivors. Instead, our findings showed that patients treated at high-performing hospitals who survived the 1339

9 The new england journal of medicine Mean Life Expectancy after Acute Myocardial Infarction (yr) RSMR quintile 1 (high-performing) Case-Mix Stratum 1 (hospitals that admitted the healthiest patients) RSMR quintile 2 RSMR quintile 3 RSMR quintile 4 RSMR quintile 5 (low-performing) Case-Mix Stratum 2 Case-Mix Stratum 3 Case-Mix Stratum 4 Case-Mix Stratum 5 (hospitals that admitted the sickest patients) Figure 3. Life Expectancy after Acute Myocardial Infarction, According to RSMR Quintile, among 30-Day Survivors. Life expectancy after acute myocardial infarction was adjusted for patient characteristics and treatment and was estimated separately for each case-mix stratum. A higher case-mix stratum signifies hospitals that admitted sicker patients, who had a higher expected mortality rate. Within each case-mix stratum, a higher RSMR quintile indicates lower-performing hospitals (those with a higher-than-expected mortality rate). I bars indicate 95% confidence intervals. period of acute illness did not lose that advantage. Alternatively, if high-performing hospitals admitted patients with lower risk than what was captured by the risk model, we would expect the survival curves to continue to diverge. The fact that the survival curves remain parallel after the first 30 days suggests that the association of early hospital performance with outcomes is the result of quality differences and not residual confounding. Our results suggest that investing in initiatives to improve short-term hospital performance may also improve patient outcomes over the long term. Many quality-improvement efforts have focused on improving process-of-care measures for acute myocardial infarction, 23,24 but these efforts, although important, have failed to resolve the variation in risk-standardized mortality rates among hospitals ,25,26 We found an inconsistent relationship between risk-standardized mortality rate quintile and acute myocardial infarction process measures (Table S4 in the Supplementary Appendix), a finding also seen in past studies. A likely explanation for this observation is that many patients are not eligible for the process measures, these measures represent only a narrow assay of quality, or the follow-up time is too short to capture the effect of discharge treatments. 12 Hospital culture, organizational structure, and collaboration across providers may also be contributing factors This study has several limitations. First, we applied several patient and hospital exclusion criteria in our calculation of risk-standardized mortality rates. As a result, our findings may not be generalizable to all patients or hospitals but do reflect current methods of estimating shortterm outcomes. Second, approximately 7% of the patients in the CCP were still alive after 17 years of follow-up, which required extrapolation of the expected survival curves to calculate life expectancy. Third, we lacked information on patients who lost Medicare eligibility; however, we estimate this number to be relatively small. Fourth, our study is based on observational data, and as such, unmeasured factors related to hospital quality and to hospital selection could have confounded our analyses. Fifth, the quality of care 1340

10 for patients with acute myocardial infarction has improved greatly since the time when the CCP was conducted. As a result, patients surviving to 30 days in the current era may have different features than those in the mid-1990s, with different risks of heart failure, reinfarction, and even death over the long term. 32,33 In conclusion, we found that patients treated at hospitals with lower 30-day risk-standardized mortality rates had significantly longer life expectancies after acute myocardial infarction than patients treated at hospitals with higher riskstandardized mortality rates. These differences in life expectancy were attributable to more patients surviving the first 30 days at high-performing hospitals than at low-performing hospitals and the persistence of that benefit over time. The content of this article does not reflect the views of Qualidigm or the Centers for Medicare and Medicaid Services (CMS), nor does mention of organizations imply endorsement by the U.S. Government. The authors assume full responsibility for the accuracy and completeness of the ideas presented. Supported by the National Heart, Lung, and Blood Institute (NHLBI) and the National Institute of General Medical Sciences (NIGMS) Medical Scientist Training Program. Dr. Bucholz is supported by F30 Training grant F30HL A1 from the NHLBI and by NIGMS Medical Scientist Training Program grant T32GM Dr. Krumholz was supported by grant U01 HL (Center for Cardiovascular Outcomes Research at Yale University) from the NHLBI during the time that the work was conducted. Disclosure forms provided by the authors are available with the full text of this article at NEJM.org. We thank personnel at Qualidigm and CMS for providing the data that made this research possible. References 1. Medicare.gov. Hospital Compare. Baltimore: Centers for Medicare and Medicaid Services, 2015 ( 2. Krumholz HM, Normand SL. Public reporting of 30-day mortality for patients hospitalized with acute myocardial infarction and heart failure. Circulation 2008; 118: Lindenauer PK, Remus D, Roman S, et al. Public reporting and pay for performance in hospital quality improvement. N Engl J Med 2007; 356: Quality Initiatives Patient Assessment Instruments. Outcome measures. Baltimore: Centers for Medicare and Medicaid Services, 2014 ( Medicare/ Quality-Initiatives-Patient -Assessment-Instruments/ HospitalQuality Inits/ OutcomeMeasures.html). 5. Yale New Haven Health Services Corporation, Center for Outcomes Research and Evaluation. Medicare hospital quality chartbook 2012: performance report on outcome measures ( / Medicare/ Quality-Initiatives-Patient -Assessment-Instruments/ HospitalQuality Inits/ Downloads/ MedicareHospitalQuality Chartbook2012.pdf). 6. Patient Protection and Affordable Care Act, Pub. L. No (2010) ( / fdsys/ pkg/ PLAW-111publ148/ pdf/ PLAW-111publ148.pdf). 7. Drye EE, Normand SL, Wang Y, et al. Comparison of hospital risk-standardized mortality rates calculated by using inhospital and 30-day models: an observational study with implications for hospital profiling. Ann Intern Med 2012; 156: McCrum ML, Joynt KE, Orav EJ, Gawande AA, Jha AK. Mortality for publicly reported conditions and overall hospital mortality rates. JAMA Intern Med 2013; 173: Horwitz LI, Wang Y, Desai MM, et al. Correlations among risk-standardized mortality rates and among risk-standardized readmission rates within hospitals. J Hosp Med 2012; 7: Girotra S, Cram P, Popescu I. Patient satisfaction at America s lowest performing hospitals. Circ Cardiovasc Qual Outcomes 2012; 5: Jha AK, Orav EJ, Li Z, Epstein AM. The inverse relationship between mortality rates and performance in the Hospital Quality Alliance measures. Health Aff (Millwood) 2007; 26: Werner RM, Bradlow ET. Relationship between Medicare s hospital compare performance measures and mortality rates. JAMA 2006; 296: Peterson ED, Roe MT, Mulgund J, et al. Association between hospital process performance and outcomes among patients with acute coronary syndromes. JAMA 2006; 295: Kontos MC, Rennyson SL, Chen AY, Alexander KP, Peterson ED, Roe MT. The association of myocardial infarction process of care measures and in-hospital mortality: a report from the NCDR. Am Heart J 2014; 168: Krumholz HM, Lin Z, Keenan PS, et al. Relationship between hospital readmission and mortality rates for patients hospitalized with acute myocardial infarction, heart failure, or pneumonia. JAMA 2013; 309: Krumholz HM, Merrill AR, Schone EM, et al. Patterns of hospital performance in acute myocardial infarction and heart failure 30-day mortality and readmission. Circ Cardiovasc Qual Outcomes 2009; 2: Marciniak TA, Ellerbeck EF, Radford MJ, et al. Improving the quality of care for Medicare patients with acute myocardial infarction: results from the Cooperative Cardiovascular Project. JAMA 1998; 279: Ramunno LD, Dodds TA, Traven ND. Cooperative Cardiovascular Project (CCP) quality improvement in Maine, New Hampshire, and Vermont. Eval Health Prof 1998; 21: Krumholz HM, Wang Y, Mattera JA, et al. An administrative claims model suitable for profiling hospital performance based on 30-day mortality rates among patients with an acute myocardial infarction. Circulation 2006; 113: Yale New Haven Health Services Corporation, Center for Outcomes Research and Evaluation for the Centers for Medicare and Medicaid Services Condition-specific measures updates and specifications report hospital-level: 30-day risk standardized mortality measures. Baltimore: Centers for Medicare and Medicaid Services, March ISIS-2 (Second International Study of Infarct Survival) Collaborative Group. Randomised trial of intravenous streptokinase, oral aspirin, both, or neither among 17,187 cases of suspected acute myocardial infarction: ISIS-2. Lancet 1988; 2: Franzosi MG, Santoro E, De Vita C, et al. Ten-year follow-up of the first megatrial testing thrombolytic therapy in patients with acute myocardial infarction: results of the Gruppo Italiano per lo Studio della Sopravvivenza nell Infarto-1 study. Circulation 1998; 98: Lewis WR, Peterson ED, Cannon CP, et al. An organized approach to improvement in guideline adherence for acute myocardial infarction: results with the Get With The Guidelines quality improvement program. Arch Intern Med 2008; 168: Process of care measures. Baltimore: Centers for Medicare and Medicaid Services, 2013 ( / Medicare/ 1341

11 Quality-Initiatives-Patient-Assessment -Instruments/ HospitalQualityInits/ HospitalProcessOfCareMeasures.html). 25. Joynt KE, Blumenthal DM, Orav EJ, Resnic FS, Jha AK. Association of public reporting for percutaneous coronary intervention with utilization and outcomes among Medicare beneficiaries with acute myocardial infarction. JAMA 2012; 308: Ryan AM, Nallamothu BK, Dimick JB. Medicare s public reporting initiative on hospital quality had modest or no impact on mortality from three key conditions. Health Aff (Millwood) 2012; 31: Bradley EH, Curry LA, Spatz ES, et al. Hospital strategies for reducing risk-standardized mortality rates in acute myocardial infarction. Ann Intern Med 2012; 156: Curry LA, Spatz E, Cherlin E, et al. What distinguishes top-performing hospitals in acute myocardial infarction mortality rates? A qualitative study. Ann Intern Med 2011; 154: Landman AB, Spatz ES, Cherlin EJ, Krumholz HM, Bradley EH, Curry LA. Hospital collaboration with emergency medical services in the care of patients with acute myocardial infarction: perspectives from key hospital staff. Ann Emerg Med 2013; 61: Curry LA, Linnander EL, Brewster AL, Ting H, Krumholz HM, Bradley EH. Organizational culture change in U.S. hospitals: a mixed methods longitudinal intervention study. Implement Sci 2015; 10: CMS.gov. Hospital engagement networks. Baltimore: Centers for Medicare and Medicaid Services ( / partnership forpatients.cms.gov/ about-the-partnership/ hospital-engagement-networks/ thehospital engagementnetworks.html). 32. Yeh RW, Sidney S, Chandra M, Sorel M, Selby JV, Go AS. Population trends in the incidence and outcomes of acute myocardial infarction. N Engl J Med 2010; 362: Masoudi FA, Foody JM, Havranek EP, et al. Trends in acute myocardial infarction in 4 US states between 1992 and 2001: clinical characteristics, quality of care, and outcomes. Circulation 2006; 114: Copyright 2016 Massachusetts Medical Society. Lynx, Bayerischer Wald National Park, Germany Christian Kaes, M.D. 1342