Using propensity stratification to compare patient outcomes in hospital-based versus freestanding skilled-nursing facilities

Medicare skilled-nursing facility (SNF) patients exhibit differences in resource use and outcomes by whether the SNF is hospital-based or freestanding. Some of the differences may be attributable to patient selection rather than underlying institutional differences. This study adjusts for patient selection by stratifying Medicare SNF patients by their likelihood of hospital-based SNF referral. Three outcomes are analyzed to illustrate this approach: Medicare SNF length of stay, discharge to home within 30 days, and preventable hospital readmissions. The estimations use claims and patient-assessment data merged with facility and market characteristics. The results provide strong evidence that good candidates for faster recovery and discharge to the community are preferentially selected into hospital-based units. While the unstratified regression approach controls for much of the selection, stratified regressions provide further reductions in setting-specific differences. Remaining differences may be because of patterns of care or reflect residual bias from unobserved factors.