Anemia indicates deficiency anemia; diabetes, diabetes, uncomplicated; fluid/electrolytes, fluid and electrolyte disorders; heart failure, congestive heart failure; hypertension, hypertension, complicated and uncomplicated; neurologic, other neurologic disorders; peripheral vascular, peripheral vascular disorders; pulmonary, chronic pulmonary disease; and renal, renal failure.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Ibrahim AM, Dimick JB, Sinha SS, Hollingsworth JM, Nuliyalu U, Ryan AM. Association of Coded Severity With Readmission Reduction After the Hospital Readmissions Reduction Program. JAMA Intern Med. 2018;178(2):290–292. doi:10.1001/jamainternmed.2017.6148
The Hospital Readmission Reduction Program (HRRP), established in 2010, placed significant financial penalties on hospitals with rates of readmission that were higher than expected for 3 targeted medical conditions.1 Under the program, a hospital’s readmission rate is adjusted based on patients’ coded severity of illness. Although these adjustments reflect an effort from the Centers for Medicare & Medicaid Services to avoid unfairly penalizing hospitals caring for patients with higher severity of illness, hospitals can improve their calculated rates of readmission by increasing their coded level of severity. It is unknown whether changes in coded severity of illness explain the previously described reductions in readmissions after implementation of the HRRP.1,2
Data for this study came from discharges from the Medicare Provider Analysis and Review file between January 1, 2008, and November 30, 2014. The post-HRRP period included admissions on or after April 1, 2010. Readmissions were defined as any inpatient hospitalization within 30 days of discharge from the index hospitalization. The study sample included Medicare beneficiaries admitted for 1 of 3 targeted medical conditions (acute myocardial infarction, heart failure, or pneumonia) to acute care hospitals exposed to the HRRP and to critical access hospitals that were exempt from the HRRP (controls). We performed a difference-in-differences analysis to compare changes in 30-day rates of readmission between exposed and control hospitals. We estimated the following 2 episode-level logistic models: the first adjusted for admitting diagnosis, demographics (age, sex, race/ethnicity [white or nonwhite]), and season of admission; the second also adjusted for coded severity of illness based on comorbidities as described by Elixhauser et al.3 Both models used inclusion criteria from the Centers for Medicare & Medicaid Services4 to identify the eligible sample of index admissions.
Standard errors were calculated to be robust to hospital-level clustering. P < .05 (2-sided) was considered statistically significant. Analyses were performed using Stata, version 14 (Stata Corp).
A total of 6 302 389 beneficiary episodes—spanning the interval between admission and 30 days after discharge—at 3259 HRRP-exposed hospitals and 1115 control hospitals were included. The mean number of comorbidities during index admissions as defined by Elixhauser et al3 per admission at control hospitals increased 19.6% (from 2.50 to 2.99) vs 38.8% (from 2.50 to 3.47) at hospitals exposed to the HRRP (Table). Exposed hospitals had larger increases in the frequency of coding each of the 10 most common comorbidities during the study period (Figure, A).
Trends in rates of readmission were parallel between control and exposed hospitals before implementation of the HRRP (Figure, B and C). Rates of readmission adjusted for patient demographics, admitting diagnosis, and season decreased from 20.71% to 19.94% at control hospitals and 20.44% to 19.30% at HRRP-exposed hospitals, with a resultant difference-in-differences estimate of –0.38% (95% CI, –0.72% to –0.04%) (Table). After adjustment for severity of illness, the difference-in-differences estimate was –1.03% (95% CI, –1.39% to –0.68%). Sensitivity analysis using hierarchical condition categories to assess severity and using linear models with hospital fixed effects confirmed our results.
Our findings raise concern that a substantial portion of estimated reductions in readmissions after implementation of the HRRP are the result of hospital documentation rather than underlying improvements in the delivery of care. A total of 63.1% of the reduction in the risk-adjusted rate of readmission after implementation of the HRRP was due to increases in coded severity ([–0.38% – (–1.03%)]/–1.03% = 63.1%). This finding is consistent with prior evidence that 40% of the reduction in risk-adjusted coronary artery bypass graft mortality rates after public reporting in New York State were the result of increases in coded severity of illness.5 Moreover, the finding extends concerns about other performance measures used by payers that are also influenced by coded severity of illness and exclusion criteria.6
Our study is limited by administrative data that could not be used to determine the appropriateness of the increases in coding severity of illness. It is possible that hospitals exposed to the HRRP coded illnesses as less severe prior to implementation of the HRRP, rather than coding illnesses as more severe after implementation of the program. Regardless of whether changes in coded severity of illness appropriately reflect clinical risk, our study demonstrates that increases in coding were responsible for a large share of the observed reduction in risk-adjusted rates of readmission.
Corresponding Author: Andrew M. Ryan, PhD, School of Public Health, University of Michigan, 1415 Washington Heights, School of Public Health II, Room M3124, Ann Arbor, MI 48109 (email@example.com).
Accepted for Publication: September 2, 2017.
Published Online: November 13, 2017. doi:10.1001/jamainternmed.2017.6148
Author Contributions: Drs Ibrahim and Ryan had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Ibrahim, Dimick, Sinha, Ryan.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Ibrahim, Ryan.
Critical revision of the manuscript for important intellectual content: Ibrahim, Dimick, Sinha, Hollingsworth, Ryan.
Statistical analysis: Nuliyalu, Ryan.
Obtained funding: Dimick, Sinha, Hollingsworth, Ryan.
Administrative, technical, or material support: Dimick, Ryan.
Study supervision: Ryan.
Conflict of Interest Disclosures: None reported.
Funding/Support: Dr Ibrahim reported receiving funding from the Robert Wood Johnson Foundation and the US Department of Veterans Affairs supporting his role as a Robert Wood Johnson Clinical Scholar. Dr Dimick reported receiving award R01AG039434-04 from the National Institute of Aging of the National Institutes of Health. Dr Dimick reported holding a financial interest in ArborMetrix Inc, which had no role in the analysis herein. Dr Sinha reported receiving grant T32-HL007853 from the National Institutes of Health. Dr Hollingsworth reported receiving grants R01 HS024525 01A1 and R01 HS024728 01 from the Agency for Healthcare Research and Quality. Dr Ryan reported receiving grant R01-AG-047932 from the National Institute on Aging.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.