Project Lead: Rachel Prusynski, UW Medicine

Data Science Lead: Curtis Atkisson

Over a million Medicare beneficiaries receive nursing and rehabilitation therapy – physical therapy (PT), occupational therapy (OT), and speech language pathology (SLP) services annually in skilled nursing facilities (SNFs) after hospitalization. The goal of post-acute SNF care is to facilitate medical and functional recovery so patients can safely return to the community. The intensity of therapy services, measured as minutes of therapy per day, that patients receive during their SNF stays has declined substantially in recent years due to Medicare payment reforms and the COVID-19 pandemic. In 2019, Medicare payment reforms intended to make therapy delivery in SNFs more patient centered and responsive to patient characteristics rather than financial incentives. However, little is known about what factors predict therapy intensity in SNFs under these reforms.

This project uses Medicare administrative data from over 1.5 million skilled SNF stays from 2019-2021 to identify and rank the predictors of therapy intensity. Predictors will include a rich set of patient demographic and clinical data alongside facility characteristics to help determine whether patient or facility factors are stronger predictors of care delivery in SNFs. Machine learning methods will be applied to impute missing data on facility characteristics and help rank therapy intensity predictors.