2016 Conference Presentation
Objective: As hospitals are increasingly held accountable for patients’ post-discharge outcomes under new payment models, hospitals may choose to acquire skilled nursing facilities (SNFs) to better manage these outcomes. This raises the question of whether patients discharged to hospital-based SNFs have better outcomes.
Data and methods: Answering the research question of the effect of SNF ownership on outcomes and costs is complicated by selection. Clearly, whether a hospital discharges a patient to its’ own SNF is not random. Simple comparisons of discharge outcomes across hospital-based and freestanding SNFs, controlling for observable characteristics, will not yield causal estimates of the effect of SNF hospital-based status on discharge outcomes. To address this issue, we instrument for choice of a hospital-based SNF using differential distance from the patient’s home to the nearest hospital with and without a SNF. The identifying assumption is that the instrument will be correlated with selection of a hospital with a SNF but independent of patient-specific health issues that would determine selection. With this instrument, we mimic randomization of residents into hospital-based SNFs when estimating the effects of hospital-based status on SNF discharge outcomes. Using national Minimum Data Set assessments linked with Medicare claims, we study a national cohort of residents who were newly admitted to SNFs from a hospital in 2009.
Results: After instrumenting for hospital-based status, we found that hospital-based SNF patients spent roughly 5 more days in the community and 6 fewer days in the SNF in the 180 days following their original hospital discharge with no significant effect on mortality or hospital readmission. Medicare spent almost $2,900 less on a hospital-based SNF patient in the 30 days following their original hospital discharge.
Policy implications: Our paper is suggestive of the idea that hospital-based SNFs generate some savings for Medicare over the 6-month hospital discharge period. These savings are largely achieved via lower SNF spending. As health care systems and payers increasingly take on risk for Medicare hospital discharge episodes, these results provide some support for vertical integration of hospitals and SNFs. Clearly, we must be careful in over-interpreting these results. The marginal patient in our model is someone who received treatment in a hospital-based SNF due to their prior residence being differentially closer to a hospital with a SNF. The experience of these individuals may not generalize to the universe of SNF patients. However, at least for these patients at the margin, we do observe some potential savings via lower SNF utilization.