Hospital readmissions have gained substantial attention from policymakers and health care providers because of their high frequency and tremendous costs (1). In particular, reducing readmission after heart failure (HF), acute myocardial infarction (AMI), and pneumonia has become a leading priority for many health systems since the Centers for Medicare & Medicaid Services (CMS) initiated public reporting in 2009 and financial penalties for readmissions in 2013 (2, 3). Performance on these condition-specific readmission measures determines whether hospitals receive across-the-board payment reductions for all Part A Medicare admissions for these conditions (4). As a result, research has focused largely on hospital readmissions after hospitalizations for HF, AMI, and pneumonia among the Medicare population (516). However, these publicly reported conditions account for only a small fraction of all hospital admissions (1719). Because Medicare readmission rates for specific conditions are used in global hospital quality metrics, such as those of the consumer-oriented Hospital Compare Web site (3), it is important to know whether readmission rates for these conditions reflect broader hospital-wide performance among all conditions and all payers (20, 21). This knowledge, however, currently does not exist. Likewise, whether the relationship between readmission rates for publicly reported and unreported conditions varies according to hospital characteristics has not been investigated. This knowledge might guide future policy regarding public reporting or reimbursement and inform hospital quality improvement strategies. To that end, this study investigated whether 30-day risk-adjusted readmission measures for publicly reported conditions among Medicare patients reflect readmission rates for 2 reference groups: Medicare patients hospitalized for unreported conditions and non-Medicare patients hospitalized with HF, AMI, or pneumonia. Methods Data Source and Study Participants Data from the Healthcare Cost and Utilization Project's Nationwide Readmissions Database (NRD) for 2013 and 2014 were used for this study. The NRD contains verified patient linkage numbers to track patient encounters across hospitals within a state while adhering to strict privacy guidelines and has been used previously for research (19, 2224). This database is a nationwide all-payer hospital inpatient data bank that includes patients of all ages, accounting for 49% of all U.S. hospitalizations each year. The 2013 NRD contains approximately 14 million discharges from 2006 hospitals across 21 states (weighted to estimate around 36 million discharges), and the 2014 database contains approximately 15 million discharges from 2048 hospitals across 22 states (weighted to estimate around 35 million discharges). Although the NRD does not comprise a random sample of states, it is designed to be geographically representative and to generate national estimates of readmissions when sampling weights are used. Hospitals with a minimum of 24 index admissions with a primary discharge diagnosis of a reported condition tied to financial penalties among Medicare patients (at least 8 each for HF, AMI, and pneumonia) were included. Similar to algorithms used by CMS (2527), admissions with an in-hospital death or a discharge against medical advice were excluded. Hospitalizations resulting in transfer to another acute care facility also were excluded, because such episodes of care would be accounted for by including the last hospitalization in the transfer chain. Only the first readmission within 30 days of an index hospitalization was considered. All subsequent admissions after 30 days from discharge were evaluated as another index hospitalization. Readmission hospitalizations within 30 days of a prior discharge were not included as index admissions. Variables For each eligible admission, the discharge diagnosis was ascertained and characterized as reported if it was HF, AMI, or pneumonia and as unreported otherwise. This analysis focused on publicly reported conditions tied to financial penalties during the study period; therefore, admissions with discharge diagnoses of chronic obstructive lung disease, hip or knee arthroplasty, or coronary artery bypass grafting were excluded from the unreported group, because readmissions after these conditions began to be penalized after 2014. The primary insurance payer for each admission was identified and characterized as Medicare or non-Medicare. Hospital characteristics examined included size (defined by number of beds), teaching status (metropolitan teaching, metropolitan nonteaching, or nonmetropolitan), and ownership status (public, private, or nonprofit). Metropolitan categorization included large and small metropolitan areas on the basis of a simplified adaptation of the 2003 version of the Urban Influence Codes. A hospital was classified as teaching if it had an American Medical Associationapproved residency program, was a member of the Council of Teaching Hospitals, or had a ratio of full-time equivalent interns and residents to beds of 0.25 or higher. Ownership status was obtained from the American Hospital Association Annual Survey of Hospitals. Whether a readmission was planned or unplanned was determined by using the algorithm for the 2013 condition-specific readmission measures (28), which has been applied elsewhere in all-payer contexts (29). Statistical Analysis For each hospital, 30-day all-cause risk-standardized unplanned readmission rates (RSRRs) and excess readmission ratios (ERRs) were estimated for 3 groups of patients: Medicare beneficiaries admitted with HF, AMI, or pneumonia (Medicare reported group); Medicare beneficiaries admitted for all other conditions (Medicare unreported group); and non-Medicare beneficiaries admitted for HF, AMI, or pneumonia (non-Medicare group). Patients younger than 65 years were excluded from the Medicare groups because they represented a fundamentally different population that was not included in CMS penalty calculations (2527). Hospital ERRs were estimated as the ratio of the predicted readmission rate for each hospital to the expected rate for a hospital with a similar case mix by using hierarchical models to approximate the risk-adjustment methodology used in the CMS Hospital-Wide Readmission Measure (30). Details on the risk-adjustment methodology and statistical procedures used are provided in Supplement 1. Supplement. Technical Appendix A, B, and C Because CMS uses the point estimate of the ERR as the foundation for determining how much a hospital will be penalized, its precise values are relevant to health policy (Supplement 2). We calculated the range of within-hospital differences in ERRs between groups overall and among subgroups of hospitals. We then assessed the agreement between the ERRs for the Medicare reported group and comparator groups by using BlandAltman plots (Supplement 3) (31). We plotted each hospital according to the average ERR between the Medicare reported group and either the Medicare unreported or the non-Medicare group along the x-axis and the difference in ERR between the Medicare reported group and the comparison group along the y-axis. Limits of agreement in these plots were defined as0.1, which would represent a 10% difference in total dollar amount of a hospital's financial penalty (a financially meaningful difference), assuming that the hospital's penalty does not cross a penalty floor or ceiling. To understand the characteristics of hospitals with the largest differences in ERRs, we examined discordant hospitals, defined as hospitals for which the ERR difference between the Medicare reported and comparator groups was greater than 0.1. In addition, hospital penalty status based on the Medicare reported group and potential penalty status based on non-Medicare and Medicare unreported groups were determined by identifying hospitals with an ERR greater than 1.00, in accordance with the approach used by CMS to calculate penalties (Supplement 2). As a robustness check, all analyses were repeated with the inclusion of hospitals with a minimum of 15 index Medicare admissions for reported conditions (5 for each condition) and 75 such admissions (25 for each condition). We also repeated the analyses, stratifying by year and using a conventional statistical threshold of1.96 SD of the Medicare reported ERR to define limits of agreement in the BlandAltman plots. Medicare groups derived from the NRD differ slightly from those used by CMS in that the NRD includes Medicare managed care; therefore, we also used data from the CMS Medicare Provider and Analysis Review files for the fee-for-service population to calculate raw readmission rates for Medicare groups and compared the results with those from the NRD to validate its use. All analyses were performed by using SAS, version 9.4 (SAS Institute). Role of the Funding Source This study was funded by the Richard A. and Susan F. Smith Center for Outcomes Research in Cardiology at Beth Israel Deaconess Medical Center. No external funding source had any role in the design, conduct, or analysis of the study or in the decision to submit the manuscript for publication. Results Our final sample included 2101 hospitals (1044 hospitals from 2013 and 1057 from 2014). During the study period, Medicare patients had 953086 admissions for reported conditions (HF, AMI, and pneumonia), with a median annual hospital RSRR of 19.8% (interquartile range, 18.4% to 21.4%), compared with 7121223 admissions for unreported conditions, with a median annual hospital RSRR of 17.4% (interquartile range, 16.0% to 18.9%) (Table 1). Non-Medicare patients had 446250 admissions for HF, AMI, or pneumonia, with a median annual hospital RSRR of 14.6% (interquartile range, 13.4% to 16.3%). Table 1. Readmission Rates and Hospital Volume for Publicly Reported Versus Unreported Conditions and Payer Status Hospital Readmission Rates for Reported Versu
[1]
J. Mellor,et al.
Does It Pay to Penalize Hospitals for Excess Readmissions? Intended and Unintended Consequences of Medicare's Hospital Readmissions Reductions Program
,
2017,
Health economics.
[2]
Leora I. Horwitz,et al.
Trends in readmission rates for safety net hospitals and non-safety net hospitals in the era of the US Hospital Readmission Reduction Program: a retrospective time series analysis using Medicare administrative claims data from 2008 to 2015
,
2017,
BMJ Open.
[3]
Leora I. Horwitz,et al.
Hospital Characteristics Associated With Risk-standardized Readmission Rates
,
2017,
Medical care.
[4]
Yun Wang,et al.
Short-term rehospitalization across the spectrum of age and insurance types in the United States
,
2017,
PloS one.
[5]
S. Yende,et al.
Proportion and Cost of Unplanned 30-Day Readmissions After Sepsis Compared With Other Medical Conditions
,
2017,
JAMA.
[6]
P. Pierorazio,et al.
Causes, Timing, Hospital Costs and Perioperative Outcomes of Index vs Nonindex Hospital Readmissions after Radical Cystectomy: Implications for Regionalization of Care
,
2017,
The Journal of urology.
[7]
R. Dreyer,et al.
Relationship Between Age and Trajectories of Rehospitalization Risk in Older Adults
,
2017,
Journal of the American Geriatrics Society.
[8]
Russell E. Mardon,et al.
Hospital-Level Factors Related to 30-Day Readmission Rates
,
2017,
American journal of medical quality : the official journal of the American College of Medical Quality.
[9]
Corwin M Zigler,et al.
Readmission Rates After Passage of the Hospital Readmissions Reduction Program
,
2017,
Annals of Internal Medicine.
[10]
Bruce E Landon,et al.
All-Payer Claims Databases - Uses and Expanded Prospects after Gobeille.
,
2016,
The New England journal of medicine.
[11]
Deepak L. Bhatt,et al.
Thirty-Day Readmissions After Transcatheter Aortic Valve Replacement in the United States: Insights From the Nationwide Readmissions Database
,
2016,
Circulation. Cardiovascular interventions.
[12]
R. Dreyer,et al.
Trajectories of Risk for Specific Readmission Diagnoses after Hospitalization for Heart Failure, Acute Myocardial Infarction, or Pneumonia
,
2016,
PloS one.
[13]
E John Orav,et al.
Readmissions, Observation, and the Hospital Readmissions Reduction Program.
,
2016,
The New England journal of medicine.
[14]
K. Itani,et al.
Does Use of a Hospital-wide Readmission Measure Versus Condition-specific Readmission Measures Make a Difference for Hospital Profiling and Payment Penalties?
,
2016,
Medical care.
[15]
T. Gjørup,et al.
Can municipality-based post-discharge follow-up visits including a general practitioner reduce early readmission among the fragile elderly (65+ years old)? A randomized controlled trial
,
2015,
Scandinavian journal of primary health care.
[16]
Tami L. Remington,et al.
A multidisciplinary intervention for reducing readmissions among older adults in a patient-centered medical home.
,
2015,
The American journal of managed care.
[17]
Harlan M Krumholz,et al.
Trajectories of risk after hospitalization for heart failure, acute myocardial infarction, or pneumonia: retrospective cohort study
,
2015,
BMJ : British Medical Journal.
[18]
Harlan M Krumholz,et al.
Strategies to Reduce 30-Day Readmissions in Older Patients Hospitalized with Heart Failure and Acute Myocardial Infarction
,
2014,
Current Geriatrics Reports.
[19]
Leora I. Horwitz,et al.
Development and Use of an Administrative Claims Measure for Profiling Hospital-wide Performance on 30-Day Unplanned Readmission
,
2014,
Annals of Internal Medicine.
[20]
Ashish K. Jha,et al.
Characteristics of hospitals receiving penalties under the Hospital Readmissions Reduction Program.
,
2013,
JAMA.
[21]
Leora I. Horwitz,et al.
Diagnoses and timing of 30-day readmissions after hospitalization for heart failure, acute myocardial infarction, or pneumonia.
,
2013,
JAMA.
[22]
Matthew D. McHugh,et al.
Hospital Nursing and 30-Day Readmissions Among Medicare Patients With Heart Failure, Acute Myocardial Infarction, and Pneumonia
,
2013,
Medical care.
[23]
Leora I. Horwitz,et al.
Correlations among risk-standardized mortality rates and among risk-standardized readmission rates within hospitals.
,
2012,
Journal of hospital medicine.
[24]
A. Sarría-Santamera,et al.
Interventions to reduce hospital readmissions in the elderly: in-hospital or home care. A systematic review.
,
2011,
Journal of evaluation in clinical practice.
[25]
J. Thorpe,et al.
Medicare Hospital Readmissions Reduction Program
,
2011
.
[26]
Sara J Singer,et al.
Perceptions of hospital safety climate and incidence of readmission.
,
2011,
Health services research.
[27]
Harlan M Krumholz,et al.
Development, validation, and results of a measure of 30-day readmission following hospitalization for pneumonia.
,
2011,
Journal of hospital medicine.
[28]
Harlan M. Krumholz,et al.
An Administrative Claims Measure Suitable for Profiling Hospital Performance Based on 30-Day All-Cause Readmission Rates Among Patients With Acute Myocardial Infarction
,
2011,
Circulation. Cardiovascular quality and outcomes.
[29]
H. Krumholz,et al.
State-sponsored public reporting of hospital quality: results are hard to find and lack uniformity.
,
2010,
Health affairs.
[30]
Harlan M Krumholz,et al.
The performance of US hospitals as reflected in risk-standardized 30-day mortality and readmission rates for medicare beneficiaries with pneumonia.
,
2010,
Journal of hospital medicine.
[31]
S. Normand,et al.
Trends in length of stay and short-term outcomes among Medicare patients hospitalized for heart failure, 1993-2006.
,
2010,
JAMA.
[32]
S. Normand,et al.
Patterns of Hospital Performance in Acute Myocardial Infarction and Heart Failure 30-Day Mortality and Readmission
,
2009,
Circulation. Cardiovascular quality and outcomes.
[33]
B. A. Cohen,et al.
Reduction of 30-day postdischarge hospital readmission or emergency department (ED) visit rates in high-risk elderly medical patients through delivery of a targeted care bundle.
,
2009,
Journal of hospital medicine.
[34]
Sharon-Lise T. Normand,et al.
An Administrative Claims Measure Suitable for Profiling Hospital Performance on the Basis of 30-Day All-Cause Readmission Rates Among Patients With Heart Failure
,
2008,
Circulation. Cardiovascular quality and outcomes.
[35]
C. Steiner,et al.
Conditions With the Largest Number of Adult Hospital Readmissions by Payer, 2011
,
2006
.
[36]
D. Altman,et al.
STATISTICAL METHODS FOR ASSESSING AGREEMENT BETWEEN TWO METHODS OF CLINICAL MEASUREMENT
,
1986,
The Lancet.
[37]
E. Rackow.
Rehospitalizations among patients in the Medicare fee-for-service program.
,
2009,
The New England journal of medicine.