A TWO-PART MODEL FOR THE IMPACT OF HOSPITAL AND COMMUNITY FACTORS ON MEDICARE READMISSION PENALTIES

Purpose: The Hospital Readmissions Reduction Program (HRRP) lowers Medicare reimbursements for hospitals with excessive 30-day unplanned readmissions for select conditions. Our objectives are to determine hospital and community characteristics associated with 1) whether a hospital is penalized or not under HRRP and 2) if penalized, the influence of these factors on the magnitude of the penalty.

Method: Currently, four fiscal years (FYs) of HRRP penalty data are available; in addition, we synthesized data from 8 different sources, such as Medicare claims and Nursing Home/Home Health Compare to capture hospital and county characteristics. The two-part model employs a logistic regression to model the propensity for a hospital to be penalized, followed by a conditional ordinary least squares regression with a log-transformed outcome to model the magnitude of the fine.

Result: The intra-class correlations from a 2-level null model, which represent the percent of variance in HRRP penalties attributable to counties, were 0.37, 0.52, 0.26, and 0.19 from FYs 2013-2016, respectively. Preliminary results suggest hospitals in the top quartile of percent Medicare inpatient days, Medicaid discharges, and dual-eligible patients were associated with significantly increased odds of being penalized and, if penalized, higher amounts compared to hospitals in the lower quartiles. Conversely, hospitals located in counties in the top quartile of per capita hospice access, full-time medication aides in nursing homes, 5-star nursing home ratings, and percent patients who received a flu shot (as determined by a home health team) were associated with significantly decreased odds of being penalized and, if penalized, reduced amounts compared to hospitals in the lower quartiles.

Conclusion: Community factors at the county level account for anywhere between 20 to 50% of the variation in HRRP penalties, depending on the FY. In addition, geographical characteristics related to socioeconomic status, nursing home quality, and home health quality were found to be significantly associated with the probability and magnitude of HRRP penalties, which suggests penalizing hospitals for unplanned readmissions may be inappropriate.