ABSTRACT
Objective
To compare outcomes between patients admitted to different clinicians within skilled nursing facilities for post-acute care, leveraging the plausibly random distribution of patients to admitting clinicians in the case of clinicians who specialize in nursing facility care (SNFists). We also compare patient outcomes between SNFists who are physicians versus advanced practice providers (APPs).
Study Setting and Design
We used multi-level modeling to estimate within-SNF variation in the characteristics and outcomes of patients admitted to different SNFists and linear regression to compare patient characteristics and outcomes between physician and APP SNFists. Our main outcomes were 30-day hospitalizations, 30-day mortality, and antipsychotic use.
Data Sources and Analytic Sample
We analyzed claims data for a 20% sample of traditional Medicare beneficiaries admitted to a SNF for post-acute care from 2016 to 2019.
Principal Findings
The sample included 81,789 post-acute patients seen by 6273 SNFists at 1479 facilities between 2016 and 2019. Within-facility variation in patient characteristics across admitting SNFists was modest and substantially greater across admitting clinicians who were not SNFists, consistent with our key assumption that patients are distributed in a more balanced fashion across admitting clinicians who are SNFists. With patient-level confounding limited by this focus on SNFists, there was minimal to modest variation in the rates of mortality (adjusted standard deviation: −0.14), hospitalization (0.40), and antipsychotic use (1.10) across admitting clinicians. Outcomes also did not differ between APP and physician admitting SNFists (mortality: 0.001 [95% CI: −0.001, 0.003]; hospitalization: 0.004 [95% CI: −0.001, 0.010], antipsychotic use: −0.001 [95% CI: −0.006, 0.003]). In contrast, outcomes varied substantially across admitting clinicians who were not SNFists.
Conclusions
Quasi-experimental assignment of patients to clinicians in SNFs reveals that the admitting clinician appears to have little influence on key outcomes in the post-acute setting, in contrast with similar research conducted in other care settings. An analysis of non-SNFists might falsely conclude that the impact of clinician factors is large because of evident non-random sorting of patients to non-SNFist clinicians in SNFs.