Skilled Nursing Facility Use Under Hospital Controlled Bundled Payments

Abstract

Hospitals increasingly bear financial risk for health care spending after hospital discharge through payment reforms such as bundled payments and accountable care organizations. Under Model 2 of Medicare’s Bundled Payments for Care Improvement (BPCI) initiative, hospitals took on financial responsibility for health care use during an episode of care beginning at hospital admission and lasting up to 90 days after discharge. The financial success of Model 2 participants hinged on managing post-acute care use, including skilled nursing facility (SNFs) care. During BPCI, the primary drivers of SNF Medicare spending were length of stay and therapy intensity, which determined daily payment rates. SNF therapy intensity increased considerably in recent years, despite no significant changes in SNF patient frailty or outcomes. Reducing unnecessary overuse of SNF therapy would have lowered Medicare spending without decreasing quality of care. The objective of this study is to assess changes in SNF treatment intensity (length of stay, therapy intensity, payments) under BPCI Model 2, as well as changes in SNF referral patterns and the impact of SNF market power. I focus on lower extremity joint replacement (LEJR) episodes and compare the impact of Model 2 in a cohort of hospital participants that took on risk before it was mandatory (early adopters) to a cohort of hospitals that took on risk when it was mandatory (late adopters). I find that Model 2 hospital participation was associated with differential reductions in SNF use. During both early adopter and late adopter episodes, I found differential reductions in SNF days and Medicare payments. SNF therapy intensity declined for early adopter episodes only. Within SNF changes drove reductions in SNF length of stay and Medicare payments. However, SNF therapy intensity reductions in the early adopter population were driven by changes in SNF referral patterns. I find limited evidence for large changes in referral concentration or historical SNF quality and efficiency under BPCI Model 2 across both early and late adopter hospitals. In terms of the impact of SNF market power, I find that SNF treatment intensity reductions were greatest in SNF markets with the greatest excess capacity and competition when using counties to define markets across both the early and late adopter cohorts. However, differences were not typically significant due to low statistical power. Based on the results of a power analysis, more hospital participants, may be required to detect statistically significant differences, particularly when stratifying participants into groups. Alternative results using health service areas (HSAs) to define SNF markets were less conclusive.