Skip main navigation
×

Health System Characteristics and Rates of Readmission After Acute Myocardial Infarction in the United States

Originally publishedhttps://doi.org/10.1161/JAHA.113.000714Journal of the American Heart Association. 2014;3:e000714

    Abstract

    Background

    Interventions to reduce early readmissions have focused on patient characteristics and the importance of early follow‐up; however, less is known about the characteristics of health systems, including quality, capacity, and intensity, and their influence on readmission rates in the United States. Therefore, we examined the association of hospital patterns of medical care with rates of 30‐day readmission.

    Methods and Results

    Medicare beneficiaries hospitalized for an AMI (n=188 611) between 2008 and 2009 in 1088 hospitals in the United States were included in our cohort. We tested the association between hospital patterns of medical care quality (discharge planning care quality), capacity (hospital size measured as the number of beds, hospital‐level Medicare all medical admission rates, supply of primary care physicians and cardiologists), and intensity (measures of care during the last 6 months of life) on CMS risk‐adjusted rates of 30‐day readmission using Poisson multilevel mixed‐effects models adjusting for patient‐ and hospital‐level covariates. There were 38 350 readmissions at 30‐days (20.3%) AMI discharges. Controlling for patient characteristics, measures of hospital care associated with higher rates of readmission included higher hospital‐level rates for all medical admissions, per capita primary care physicians and cardiologists, and last 6 months of life care intensity measures including increased number of hospital days, number of ICU days, number of physician visits, and 10 or more different physicians seen during the last 6 months of life. Better discharge quality and larger hospitals were associated with lower rates of readmission.

    Conclusions

    In addition to quality of care, high 30‐day readmission rates are associated with hospital‐level measures of capacity and intensity. Efforts to reduce readmission rates may need to address these broader patterns of medical care.

    Introduction

    One in 5 older adults in the United States are re‐hospitalized within 30 days of a medical or surgical hospital stay, costing Medicare over $17 billion dollars annually.1 In an effort to engage hospitals to reduce readmissions, the Affordable Care Act directed the Centers for Medicare and Medicaid Services (CMS) to institute the hospital readmissions reduction program (HRRP) with Medicare reimbursement penalties for hospitals with higher than expected rates or readmission beginning in fiscal year 2013. Currently, 61% of US hospitals have reimbursement penalties that average 0.24% with 10% of hospitals facing the highest penalty of 1%.2

    Clinicians and health systems have focused efforts on identifying patients at a high risk of readmission and improving the quality of inpatient clinical processes of care, care transitions, and post‐discharge continuity of outpatient care.3, 4 Recent evidence suggests that non‐clinical factors of regional quality and supply of health care may also influence readmission rates. Non‐clinical factors include regional baseline admission rates and bed supply.5 However, other novel measures of health care intensity and capacity have not been fully evaluated with rates of readmission. Furthermore, these measures have not been evaluated at the patient and hospital level. In addition, current investigations have not incorporated patient‐level risk adjustment methods used by CMS for calculating readmission penalties.

    In this paper, we test the hypothesis that hospital‐specific patterns of medical care quality, capacity, and intensity of health care at the end of life, are important factors in predicting 30‐day readmission rates for acute myocardial infarction (AMI).

    Methods

    Setting and Participants

    We first identified 2008 and 2009 100% fee‐for‐service Medicare beneficiaries age 66 and over with Part A and Part B coverage who had an AMI hospitalization based on principal discharge diagnosis ICD‐9 codes 410.00, 410.01, 410.10, 410.11, 410. 20, 410.21, 410.30, 410.31, 410. 40, 410.41, 410.50, 410.51, 410.60, 410.61, 410.70, 410.71, 410. 80, 410.81, 410.90, and 410.91, excluding 410.×2 (AMI, subsequent episode of care) using the Medicare Denominator and Medicare Provider Analysis and Review (MEDPAR) Files. AMI admissions were excluded if the patient was hospitalized for any reasons within 3 months prior to the AMI admission. For patients with more than one AMI admission during the study period, we randomly selected only one of the AMI admissions per beneficiary for inclusion in the cohort. The study cohort was further limited to hospitals with complete data on all measures of quality, capacity, and intensity and with at least 50 AMI discharges to assure statistical stability. In total, there were 188 611 index AMI admissions across 1088 hospitals. Patients were also assigned to the Hospital Service Area (HSA) of residence for hospital service area measures. HSAs are defined geographic service areas representing hospital use of care previously described by the Dartmouth Atlas of Healthcare.6

    Design Overview

    We studied the factors associated with 30‐day readmissions in elderly Medicare beneficiaries with an initial admission for AMI. The unit of analysis was the patient with models incorporating system factors at the hospital or HSA level in addition to patient level covariates. We examined 3 categories of medical care: quality, capacity, and intensity. Medical care quality was assessed from the 2007 Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) on answering yes to both of these questions involving discharge: “During this hospital stay, did doctors, nurse or other hospital staff talk with you about whether you would have the help you needed when you left the hospital?” and “During the hospital stay, did you get information in writing about what symptoms or health problems to look out for after you left the hospital?” We also included a quality measure of hospital care transitions defined by whether or not a patient visited a primary care physician within 14 days of discharge.7 All measures were calculated at the hospital level except for all medical admission rate, primary care provider supply and cardiologist supply, which were calculated at the HSA level to represent broader system capacity in the area.6, 7, 8, 9 For these 3 measures, patients were assigned to their HSA of residence. Medical care capacity was assessed using the hospital or HSA level for all medical admission rates, hospital size as a measure of the number of beds, and per capita physician workforce supply measured by the Dartmouth Atlas of Healthcare (HSA level).7, 8 Medical care intensity at the hospital level was assessed using measures of intensity of care in the last 6 months of life: total hospital days, total intensive care unit days, number of physician visits, and 10 or more different physicians seen by the Dartmouth Atlas (all measured at the hospital level).9 This study was approved the Institutional Review Board.

    Outcomes and Follow‐Up

    The outcome of interest was all‐cause 30‐day readmissions from the AMI discharge to any acute care hospital,7 excluding readmissions for elective percutaneous coronary intervention or coronary artery bypass graft surgery.

    Patient‐level risk factors defined in the CMS AMI risk model were abstracted from MEDPAR (in‐patient), Carrier (physicians), and Outpatient files (health centers) during the year preceding and including the index AMI.10 All patterns of medical care quality, capacity, and intensity were calculated by the Dartmouth Atlas of Healthcare.6, 7, 8, 9 All measures were calculated at the hospital level except for all medical admission rates, primary care provider supply, and cardiologist supply, which were calculated at the HSA level to represent broader system capacity in the area.6, 7, 8, 9 For these 3 measures, patients were assigned to their HSA of residence. Discharge quality was assessed from the 2007 Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) as a yes answer to both of these questions involving discharge: “During this hospital stay, did doctors, nurse or other hospital staff talk with you about whether you would have the help you needed when you left the hospital?” and “During the hospital stay, did you get information in writing about what symptoms or health problems to look out for after you left the hospital?” We also determined if each patient visited a primary care physician within 14 days after discharge from the Medicare Carrier and Outpatient data files. We identified PCP vistis for each patient as follows: we selected ambulatory claims from Medicare Part B using CPT, place of service, and physician specialty codes. We abstracted Rural Health Clinic and Federally Qualifying Heath Clinic claims from the Medicare Outpatient file. Visits that occurred on the same day as the AMI index date were not counted; otherwise visits occurring up to 14 days from the discharge date were counted as a PCP visit. The all‐medical admission rate was calculated at the HSA levels for 2008. Hospital size was measured as the number of hospital beds for 2008.8 The 2006 primary care physician and cardiologist workforce supply at the HSA level were obtained from the Dartmouth Atlas project.8 The rate of patients visiting a primary care physician within 14 days after discharge was calculated at the hospital level. As covariates in our models, we also used hospital‐level covariates that the Dartmouth Atlas project identified as medical care intensity measures (2007): last 6 months of life measures of hospitals days, intensive care unit days, and seeing more than 10 different physicians.9 We conducted sensitivity analyses removing deceased patients within the first 30 days where the date of death from the Medicare denominator file is through year 2010.

    Statistical Analysis

    Using univariate poisson regression we tested the association between quartiles of hospital‐level measures of quality, capacity, and intensity with 30‐day readmission. We then used hierarchical binary regression models, with random effect for hospital to test the association of measures of quality, capacity, and intensity of medical care on 30‐day readmission adjusting for patient‐ and system‐level covariates. Poisson regression was used in order to yield risk rates. The sandwich variance was used to determine the standard error of rates. Two modeling approaches were conducted: simple and full. Simple models: Using hierarchical Poisson regression, the statistical significance of each factor was evaluated comparing each quartile with the 1st quartile (lowest, referent). These simple models also included patient‐level probability of 30‐day readmission using patient‐level risk factors of readmission derived from the CMS AMI readmission model.10Full model: We then included all variables in a Poisson multivariable model with the patient predicted probability of readmission using the CMS AMI readmission model.10 To determine if the effects of these measures on 30‐day readmission were consistent among all AMI patients alive at 30 days, we conducted a sensitivity analysis that excluded all patients deceased within 30 days of discharge.

    Results

    Overall there were 188 611 Medicare AMI patients across 1088 hospitals during the study period with 38 350 readmissions at 30‐days (20.3%) (Table 1). There was an average of 116 AMI discharges per hospital. The mean age was 78.4 years with 51.6% male. Measures of hospital quality (discharge planning and proportion of patients visiting a primary care provider within 14 days of the AMI discharge), capacity (physician supply, hospital size, and all medical admissions), and intensity measures (last 6 months of life total hospital days, total ICU days, number of physicians seen, and ≥10 different physicians seen) are reported in Table 1.

    Table 1. Hospital and Patient Characteristics and Measures of Readmission

    Frequency or Mean (SD)
    Patient characteristics
    Number of AMI patients, n188 611
    Number of AMI patients per hospital116 (125)
    Mean age, y78.4 (7.7)
    Male sex, %51.6 (50.0)
    Black race, %6.5 (24.6)
    Health system characteristics
    Number of hospitals, n=10881088
    Academic hospitals, %8.5 (27.9)
    Hospital outcomes
    30‐day readmission rate, %20.3 (6.4)
    Number of readmissions, n38 350
    Case mix: predicted probability of readmission, %21.6 (7.1)
    30‐day death rate, %3.6 (18.6)
    Measures of quality, capacity, and intensity
    Quality measures
    Discharge planning: HCAHPS (%, hospital level)79.2 (4.6)
    PCP visit within 14‐days (%, hospital level)60.7 (11.6)
    Capacity measures
    PCPs per 100 000 population (HSA level)72.2 (18.8)
    Cardiologists per 100 000 population (HSA level)6.6 (2.9)
    Hospital size (average number of beds, hospital level)396 (264)
    All medical admission rates per 1000 beneficiaries (HSA level)244 (41)
    Intensity measures
    Last 6 months of life hospital days (hospital level)13.4 (3.1)
    Last 6 months of life ICU days (hospital level)4.5 (2.4)
    Last 6 months of life number of physician visits (hospital level)36.8 (11.9)
    Last 6 months of life 10 or more physicians seen (hospital level)10.1 (2.0)

    The value is the percentage of patients responding affirmatively to 2 questions about discharge planning on the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey: “During this hospital stay, did doctors, nurse or other hospital staff talk with you about whether you would have the help you needed when you left the hospital?” and “During the hospital stay, did you get information in writing about what symptoms or health problems to look out for after you left the hospital?” AMI indicates acute myocardial infarction; HSA, hospital service area; ICU, intensive care unit; PCP, primary care physician.

    We characterize the association of these measures of quality, capacity, and intensity by quartiles of the study population (AMI discharges) with 30‐day readmission in Table 2. Better discharge planning quality and larger hospital size as measured by the number of hospital beds were significantly associated with lower 30‐day readmission after AMI (Table 2). Higher area rates of Medicare all medical admissions, supply of primary care physicians and cardiologists were significantly associated with higher rates of 30‐day readmission. All measures of hospital care intensity during the last 6 months of life (total hospital days, intensive care unit days, number of physician visits, and ≥10 different physicians seen) were significantly associated with higher 30‐day readmission. However, the rate of patients visiting a primary care physician within 14 days of AMI discharge, a measure of hospital care transitions, was not associated with significantly lower 30‐day readmission except for the highest quartile.

    Table 2. Univariate Models for Quality, Capacity and Intensity Measures

    QuartileRR95% CIP Value
    Quality measures
    Discharge planning: HCAHPS (%, hospital level)Q11.00
    Q20.93(0.90 to 0.96)<0.001
    Q30.85(0.82 to 0.88)<0.001
    Q40.86(0.84 to 0.89)<0.001
    PCP visit within 14‐days (%, hospital level)Q11.00
    Q21.01(0.97 to 1.04)0.717
    Q31.01(0.98 to 1.04)0.471
    Q41.07(1.04 to 1.11)<0.001
    Capacity measures
    PCPs per 100 000 population (HSA level)Q11.00
    Q21.07(1.04 to 1.10)<0.001
    Q31.06(1.03 to 1.09)<0.001
    Q41.14(1.11 to 1.18)<0.001
    Cardiologists per 100 000 population (HSA level)Q11.00
    Q21.04(1.01 to 1.07)0.022
    Q31.10(1.06 to 1.13)<0.001
    Q41.21(1.17 to 1.25)<0.001
    Average number of beds (hospital level)Q11.00
    Q20.86(0.83 to 0.89)<0.001
    Q30.82(0.79 to 0.84)<0.001
    Q40.80(0.78 to 0.83)<0.001
    All medical admission rates per 1000 beneficiaries (HSA level)Q11.00
    Q21.13(1.09 to 1.16)<0.001
    Q31.21(1.18 to 1.25)<0.001
    Q41.30(1.26 to 1.35)<0.001
    Intensity measures
    Last 6 months of life hospital days (hospital level)Q11.00
    Q21.11(1.07 to 1.15)<0.001
    Q31.19(1.16 to 1.23)<0.001
    Q41.41(1.36 to 1.45)<0.001
    Last 6 months of life ICU days (hospital level)Q11.00
    Q21.02(0.99 to 1.06)0.141
    Q31.06(1.03 to 1.09)<0.001
    Q41.16(1.13 to 1.20)<0.001
    Last 6 months of life number of physician visits (hospital level)Q11.00
    Q21.09(1.06 to 1.13)<0.001
    Q31.16(1.12 to 1.20)<0.001
    Q41.31(1.27 to 1.35)<0.001
    Last 6 months of life 10 or more different physicians seen (hospital level)Q11.00
    Q21.08(1.05 to 1.12)<0.001
    Q31.13(1.09 to 1.17)<0.001
    Q41.28(1.24 to 1.32)<0.001

    All estimates are calculated from univariate Poisson models comparing the highest quartile measure to the lowest. HCAHPS: The value is the percentage of patients responding affirmatively to 2 questions about discharge planning on the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey: “During this hospital stay, did doctors, nurse or other hospital staff talk with you about whether you would have the help you needed when you left the hospital?” and “During the hospital stay, did you get information in writing about what symptoms or health problems to look out for after you left the hospital?” HSA indicates hospital service area; ICU, intensive care unit; PCP, primary care physician; RR, relative risk.

    In multivariable hierarchical models, controlling for patient characteristics, significant associations with 30‐day readmission at the patient and hospital level were observed and reported in Table 3. Table 3 stratifies these models into simple (evaluating each measure separately while adjusting for the CMS Readmission model factors) and full models (combining all measures into 1 model while adjusting for the CMS Readmission model factors). Quality measures: the third and fourth quartiles of discharge planning were significantly associated with lower 30‐day readmission, however seeing a primary care physician within 14 days was not significantly protective against 30‐day readmission. Capacity measures: all quartiles of hospital size (number of beds) were associated with significantly lower 30‐day readmission, while all quartiles of the rates of Medicare all medical admission was associated with significantly higher 30‐day readmission. In addition, only the highest quartile of per capita rates of primary care physician supply and cardiologist supper were associated with significantly higher 30‐day readmission, however this effect was not preserved in both the simple and full models. Intensity measures: the last 6 months of life intensity measures significantly associated with higher 30‐day readmission including total hospital days (all quartiles), total ICU days (only the highest quartile), number of physicians seen (quartiles 3 and 4), and ≥10 different physicians seen (all quartiles). All of the intensity measures were confirmed using both the simple and full model approaches with the exception of total ICU days, which lost statistical significance in the full model. In addition, the direction of the association changed for number of physician visits from a positive association with 30‐day readmission in the simple model to a statistically significant negative association with 30‐day readmission in the full model. We report the results for all hospitals without a denominator restriction in Table S1.

    Table 3. Hierarchical Models for Patient and Hospital Level Analyses

    QuartileSimple ModelFull Model
    RR95% CIP ValueRR95% CIP Value
    Quality measures
    Discharge planning: HCAHPS (%, hospital level)Q11.001.00
    Q20.96(0.92 to 1.00)0.0770.99(0.95 to 1.03)0.738
    Q30.90(0.86 to 0.94)<0.0010.93(0.89 to 0.97)0.001
    Q40.92(0.88 to 0.96)<0.0010.95(0.91 to 0.99)0.028
    PCP visit within 14‐days (%, hospital level)Q11.001.00
    Q21.00(0.95 to 1.04)0.8400.99(0.95 to 1.04)0.756
    Q30.98(0.93 to 1.03)0.3550.96(0.92 to 1.00)0.039
    Q41.01(0.97 to 1.06)0.6640.98(0.94 to 1.02)0.358
    Capacity measures
    PCPs per 100 000 population (HSA level)Q11.001.00
    Q21.03(1.00 to 1.07)0.0431.03(0.99 to 1.06)0.105
    Q31.02(0.99 to 1.06)0.1811.01(0.98 to 1.05)0.447
    Q41.05(1.01 to 1.08)0.0101.03(0.99 to 1.06)0.128
    Cardiologists per 100 000 population (HSA level)Q11.001.00
    Q21.00(0.97 to 1.04)0.8451.00(0.97 to 1.04)0.870
    Q31.02(0.99 to 1.05)0.2250.99(0.96 to 1.03)0.669
    Q41.04(1.01 to 1.08)0.0150.98(0.95 to 1.02)0.412
    Average number of beds (hospital level)Q11.001.00
    Q20.90(0.87 to 0.94)<0.0010.88(0.85 to 0.92)<0.001
    Q30.87(0.83 to 0.90)<0.0010.83(0.80 to 0.86)<0.001
    Q40.85(0.81 to 0.89)<0.0010.79(0.75 to 0.82)<0.001
    All medical admission rates per 1000 beneficiaries (HSA level)Q11.001.00
    Q21.06(1.02 to 1.10)0.0031.04(1.00 to 1.08)0.032
    Q31.13(1.09 to 1.18)<0.0011.10(1.06 to 1.15)<0.001
    Q41.16(1.11 to 1.21)<0.0011.12(1.07 to 1.16)<0.001
    Intensity measures
    Last 6 months of life hospital days (hospital level)Q11.001.00
    Q21.07(1.03 to 1.12)0.0021.07(1.01 to 1.12)0.003
    Q31.11(1.06 to 1.16)<0.0011.10(1.05 to 1.16)<0.001
    Q41.23(1.18 to 1.29)<0.0011.23(1.16 to 1.31)<0.001
    Last 6 months of life ICU days (hospital level)Q11.001.00
    Q21.01(0.96 to 1.05)0.8310.99(0.95 to 1.03)0.500
    Q31.02(0.98 to 1.07)0.2981.00(0.96 to 1.04)0.996
    Q41.09(1.04 to 1.13)<0.0011.02(0.98 to 1.07)0.327
    Last 6 months of life number of physician visits (hospital level)Q11.001.00
    Q21.04(1.00 to 1.09)0.0650.96(0.92 to 1.01)0.087
    Q31.08(1.03 to 1.13)0.0010.93(0.88 to 0.98)0.010
    Q41.14(1.09 to 1.20)<0.0010.88(0.82 to 0.95)0.001
    Last 6 months of life 10 or more different physicians seen (hospital level)Q11.001.00
    Q21.05(1.01 to 1.10)0.0251.09(1.05 to 1.14)<0.001
    Q31.07(1.02 to 1.12)0.0041.11(1.05 to 1.16)<0.001
    Q41.14(1.09 to 1.19)<0.0011.15(1.08 to 1.23)<0.001

    All estimates are calculated from Poisson models comparing the highest quartile measure to the lowest. Simple models: test each measure independently while adjusting for patient factors used in the CMS AMI readmission models. Full model: tests the independent effect of all measures combined in 1 model while adjusting for patient factors used in the CMS AMI readmission models. In the hospital analysis, PCP, cardiologist supply and all medical admissions are per‐capita measures calculated at the hospital service area. HCAHPS: The value is the percentage of patients responding affirmatively to 2 questions about discharge planning on the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey: “During this hospital stay, did doctors, nurse or other hospital staff talk with you about whether you would have the help you needed when you left the hospital?” and “During the hospital stay, did you get information in writing about what symptoms or health problems to look out for after you left the hospital?” AMI indicates acute myocardial infarction; CMS, Centers for Medicare and Medicaid Services; HSA, hospital service area; ICU, intensive care unit; PCP, primary care physician; RR, relative risk.

    We conducted sensitivity analyses whereby we removed all discharged AMI patients who died within 30 days of discharge; all results were qualitatively similar to the hospital‐level analyses of all AMI patients (Table S2). Overall, discharge planning, hospital size, all medical admission rates, and total hospital days in the last 6 months of life care intensity measures were the only factors that were associated with readmission rates after hospitalization for AMI irrespective of the analytic strategy and sensitivity analyses.

    We repeated our analyses across 306 hospital referral regions (HRR, Tables S3). The HRR cohort included all 306 HRRs with complete data on all measures of quality, capacity, and intensity. Overall, the HRR level analysis confirmed our hospital level analysis with the following exceptions whereby the hospital‐level analysis demonstrated significant associations in both the simple and full models for bed supply, last 6 months of life number of physician visits, and ≥10 different physicians seen and the HRR level analysis only demonstrated significance in the simple models.

    Discussion

    We demonstrate an association between hospital measures of quality, capacity, and intensity of medical care and 30‐day readmission after AMI. These associations persisted even after including the CMS AMI risk model in our analyses to account for patient risk factors for readmission. After adjusting for patient‐level factors, patient‐reported quality of discharge communication was associated with lower 30‐day readmission. We found a consistent effect that hospitals with greater capacity for care, as manifest by a greater tendency to admit patients, also had higher 30‐day readmission. However, contrary to recent findings, we demonstrate that larger hospitals were more likely to have lower 30‐day readmission. In addition, hospital measures of greater intensity of care, more hospital days in the last 6 months of life, more physicians seen, and ≥10 different physicians seen, were associated with higher 30‐day readmission.

    Our findings are consistent with current evidence that one effective remedy for high readmission rates is better discharge planning.4, 11 Jha and colleagues11 originally reported no relationship between discharge planning and all‐cause 30‐day readmission for congestive heart failure and pneumonia using the Hospital Quality Alliance score. However, when the HCAHPS score was used in the analysis, Jha and colleagues reported a statistically significant inverse trend with the higher quartiles of HCAHPS score for discharge planning and 30‐day readmission rates. These previous findings by Jha and colleagues11 are supported in our hospital‐level analysis in a new cohort of AMI discharges showing higher HCAHPS scores for discharge planning are associated with lower rates of 30‐day readmission.

    Our paper also extends the previous study by Epstein and colleagues5 that regional hospital admission rates are associated with readmissions by incorporating measures of health care intensity at the end of life. Using multivariable models, in contrast to Epstein's univariate analyses, we did not observe an independent positive association of physician or hospital bed supply with 30‐day readmissions. In contrary, our patient‐ and hospital‐level analysis, we found larger hospitals as measured by number of beds was significantly associated with lower 30‐day readmission. Our finding suggests that analyses, when done at the aggregated population level as in the case of the Epstein and colleagues5 analysis at the HRR‐level may mask underlying phenomenon of larger hospitals. Larger hospitals may have more resources and are more likely to be academic hospitals, thus have more resources to target key penalty areas such as readmissions.

    Current efforts to reduce readmission rates fail to target more general health system effects. The strong association of the number of hospital days and ≥10 different physicians seen in the last 6 months of life and 30‐day readmission highlights the importance of local patterns of hospital use. We demonstrate patterns of medical care and readmissions at the hospital level, which have not been shown before, and suggest rates of readmission are driven partly by the intensity of health care delivery within a hospital.

    Reducing readmissions may also require health systems, hospitals, physicians, allied health professionals, as well as patients and families to work together towards reducing unnecessary end of life health care intensity. First, we recommend the use of advance care planning for patients at the end of life.12 Patient preferences for the level of intensity of their health care and preference for location of death as much as possible should be matched with the setting and care. Larochelle and colleagues well characterize the role of the physician in end of life treatment intensity and noted variation in communication and collaboration with the care team, patient, and family members.13 As such, we suggest targeted interventions are thus needed to reduce variation and improve the divide between physician and patient preferences. Second, high‐intensity areas at the end of life could incorporate successful models such as ENABLE14 to support patients and families at the end of life with care liasons to aid in managing pain, symptoms, and comfort while matching patient preferences for intensity of care and location of that care and death (eg, home, hospital or managed care facility).

    Our finding of lower 30‐day readmission with higher HCAHPS discharge quality scores is consistent with the benefits seen in intervention trials that improved patient education and information prior to discharge such as care transitions and medication reconciliation.4, 15, 16, 17 Improvements in the care patients receive after discharge including regular phone and office follow‐up has also been shown to be effective and is an area of active improvement.3, 4 However, these strategies may fall short of sustained change unless they are coupled to improvements in care that generally lower the threshold of hospital admission and raise the quality of ambulatory care.

    It should be noted that the patient populations used to derive the measures of hospital intensity differ from the AMI admission cohort and are from a different time period. Yet, the probability that a patient spends more time in the hospital in the last months of life strongly predicts the chances that patients will be readmitted after an AMI. These deeply embedded styles of care (specifically discharge quality, medical admission rates, and last 6 months of total hospital days) may need as much attention as improvement in care transitions.

    Our results extend previous studies by evaluating system measures of capacity and health care intensity in for all AMI Medicare discharges, a relatively homogenous population compared with congestive heart failure or pneumonia.5 We also report on effects at both the hospital and regional (HRR, see Supplemental Material) levels and demonstrate that the associations of these measures are similar at both a hospital and regional (HRR) level. We conducted patient‐level risk adjustment to align with the CMS penalty calculations by using the published CMS AMI readmission model.10 Our results are consistent with previous research by Epstein and colleagues, who demonstrated similar associations for congestive heart failure and pneumonia HRR‐level discharge quality scores from HCAHPS and admission rates were associated with rates of 30‐day readmission.5 However, as discussed above, our hospital level analysis demonstrates larger hospitals are associated with lower 30‐day readmission and not higher readmission as reported by Epstein and colleagues.5

    There are several limitations of our analyses. First, while our nation‐wide analysis of fee‐for‐service Medicare beneficiaries hospitalized for AMI captures a sizeable national sample of AMI hospitalizations across the United States, it may not be applicable to Medicare Advantage or to younger patients. Second, the HCAHPS survey used the answer to 2 questions as a metric of hospital discharge quality. These questions do not capture all of the important aspects of discharge quality. Third, while our outcome of 30‐day readmissions excluded planned readmissions common after an AMI, our analysis did not focus solely on avoidable readmissions. Currently there is a lack of consensus on defining avoidable readmissions and it is not possible to review each readmission with a clinical chart review. Fourth, we employed patient‐level risk adjustment using the CMS AMI readmission model, however this is limited to comorbidities and procedures that are captured in the CMS claims. Fifth, we restricted our cohort to hospitals with complete data for all of our measures. There are a total of 4315 hospitals caring for Medicare AMI patients, however only 1773 hospitals had complete data on the measures used in this study. Hospitals not included and missing this data from The Dartmouth Atlas were small hospitals that did not meet the criteria for public reporting. We further eliminated 685 hospitals and 16 902 patients to meet a minimum denominator requirement of 50 AMI discharges per hospital resulting in the final cohort of 188 611 and 1088 hospitals to have stable statistical estimates.

    In conclusion, we demonstrate that patterns of hospital or regional medical care utilization, unrelated to AMI hospitalization, are associated with rates of 30‐day AMI readmission. Our results suggest that efforts to reduce 30‐day readmissions through better discharge planning and care coordination will help but may need to be accompanied by efforts to reduce overall hospital admission rates and end of life care intensity. Our findings point to the need for clinicians and hospital administrators to look beyond care transitions clinical workflows and also redesign the larger context of care delivery in order to reduce the overall number of hospital admissions.

    Sources of Funding

    We would like to acknowledge the Dartmouth Atlas of Health Care and The Dartmouth Institute of Health Policy and Clinical Practice. This study was supported, in part, by the Robert Wood Johnson Foundation and the Agency for Healthcare Research and Quality. The funders had no role in the design or conduct of the study, collection, management, analysis, or interpretation of the data, or the preparation, review, approval of the manuscript, or the decision to submit the manuscript for publication. Drs Brown and Goodman had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

    Disclosures

    Dr. Brown received grant support from the Agency for Healthcare Research and Quality K01 HS018443. Dr. Goodman received grant support from the Robert Wood Johnson Foundation and personal fees from Intermountain Healthcare, Canadian Health Services Research Foundation, and the Massachusetts Health Data Consortium.

    Supplementary Information

    Table S1. Hierarchical models for hospital level analyses without hospital denominator restrictions.

    Table S2. Hospital measures of readmission among patients alive through 30‐days without hospital denominator restriction.

    Table S3. Hierarchical models for HRR level analyses.

    Footnotes

    *Correspondence to: Jeremiah R. Brown, PhD, MS, HB 7505 DHMC One Medical Center Drive, Lebanon, NH. E‐mail:

    Accompanying Tables S1 through S3 are available at http://jaha.ahajournals.org/content/3/3/e000714/suppl/DC1

    References

    • 1 Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee‐for‐service program. N Engl J Med. 2009; 360:1418–1428.CrossrefMedlineGoogle Scholar
    • 2 Medicare Payment Advisory Commission (U.S.) . Report to the Congress: Medicare and the Health Care Delivery System. Washington, DC: MedPAC; 2013.Google Scholar
    • 3 Hansen LO, Young RS, Hinami K, Leung A, Williams MV. Interventions to reduce 30‐day rehospitalization: a systematic review. Ann Intern Med. 2011; 155:520–528.CrossrefMedlineGoogle Scholar
    • 4 Naylor MD, Brooten D, Campbell R, Jacobsen BS, Mezey MD, Pauly MV, Schwartz JS. Comprehensive discharge planning and home follow‐up of hospitalized elders: a randomized clinical trial. JAMA. 1999; 281:613–620.CrossrefMedlineGoogle Scholar
    • 5 Epstein AM, Jha AK, Orav EJ. The relationship between hospital admission rates and rehospitalizations. N Engl J Med. 2011; 365:2287–2295.CrossrefMedlineGoogle Scholar
    • 6 The Dartmouth Atlas of Health Care, 1999. Chicago IL: American Hospital Publishing; 1999.Google Scholar
    • 7 After hospitalization: a Dartmouth atlas report on post‐acute care for Medicare beneficiaries. The Dartmouth Atlas of Healthcare; 2011. Available at: http://www.dartmouthatlas.org/downloads/reports/Post_discharge_events_092811.pdf. Accessed September 28, 2011.Google Scholar
    • 8 Hospital and physician capacity update. The Dartmouth Atlas of Healthcare; 2009. Available at: http://www.dartmouthatlas.org/downloads/reports/Capacity_Report_2009.pdf. Accessed April 16, 2013.Google Scholar
    • 9 Trends and variation in end‐of‐life care for Medicare beneficiaries with severe chronic illness. The Dartmouth Atlas of Healthcare; 2011. Available at: http://www.dartmouthatlas.org/downloads/reports/EOL_Trend_Report_0411.pdf. Accessed December 5, 2012.Google Scholar
    • 10 Krumholz HM, Lin Z, Drye EE, Desai MM, Han LF, Rapp MT, Mattera JA, Normand SL. An administrative claims measure suitable for profiling hospital performance based on 30‐day all‐cause readmission rates among patients with acute myocardial infarction. Circ Cardiovasc Qual Outcomes. 2011; 4:243–252.LinkGoogle Scholar
    • 11 Jha AK, Orav EJ, Epstein AM. Public reporting of discharge planning and rates of readmissions. N Engl J Med. 2009; 361:2637–2645.CrossrefMedlineGoogle Scholar
    • 12 Curtis JR, Engelberg RA, Bensink ME, Ramsey SD. End‐of‐life care in the intensive care unit: can we simultaneously increase quality and reduce costs?Am J Respir Crit Care Med. 2012; 186:587–592.CrossrefMedlineGoogle Scholar
    • 13 Larochelle MR, Rodriguez KL, Arnold RM, Barnato AE. Hospital staff attributions of the causes of physician variation in end‐of‐life treatment intensity. Palliat Med. 2009; 23:460–470.CrossrefMedlineGoogle Scholar
    • 14 Bakitas M, Stevens M, Ahles T, Kirn M, Skalla K, Kane N, Greenberg ER; Project Enable C‐I . Project ENABLE: a palliative care demonstration project for advanced cancer patients in three settings. J Palliat Med. 2004; 7:363–372.CrossrefMedlineGoogle Scholar
    • 15 Brock J, Mitchell J, Irby K, Stevens B, Archibald T, Goroski A, Lynn J. Association between quality improvement for care transitions in communities and rehospitalizations among Medicare beneficiaries. JAMA. 2013; 309:381–391.CrossrefMedlineGoogle Scholar
    • 16 Gleason KM, Brake H, Agramonte V, Perfetti C. Medications at Transitions and Clinical Handoffs (MATCH) toolkit for medication reconciliation. (Prepared by the Island Peer Review Organization, Inc, under Contract No. HHSA2902009000 13C.) AHRQ Publication No. 11(12)‐ 0059. Rockville, MD: Agency for Healthcare Research and Quality; 2012 Revised August 2012.Google Scholar
    • 17 Gleason KM, McDaniel MR, Feinglass J, Baker DW, Lindquist L, Liss D, Noskin GA. Results of the Medications At Transitions and Clinical Handoffs (MATCH) study: an analysis of medication reconciliation errors and risk factors at hospital admission. J Gen Intern Med. 2010; 25:441–447.CrossrefMedlineGoogle Scholar