Background: Hospitalists increasingly confront patient care and policy dilemmas due to financial incentives created by bundled payments, single payments that “bundle” acute and post-acute care for specific diagnoses. The most common and most successful bundled payment has been for total joint replacement, where cost savings have been achieved through discharging patients to home rather than a skilled nursing facility, which had been the default discharge destination. This raises the question: what are the clinical outcomes of patients who used to go to SNF and are now being sent home? We seek to answer this question addressing three key limitations of prior literature: 1) we use a contemporary sample (Medicare data is only available up to 2016); 2) we use all payers rather than just Medicare; and 3) we focus on high-risk subgroups, rather than the larger population of healthier patients.

Methods: We used Pennsylvania all-payer claims data, including all inpatient and ambulatory surgical procedures, hospital visits and readmissions, and mortality. Patient characteristics included demographics, comorbidities, and index hospital length of stay. Hospital characteristics included size, region, and facility type. We compared outcomes of elective total joint replacements (TJR) in a pre-bundle period (September 2011-August 2013) to post-bundle (September 2016-August 2018). We used a cross-temporal matching approach, which combines strengths of propensity matching (patients who went to SNF in the pre-period matched to patients who went home post-period) and a difference-in-difference analysis to adjust for temporal trends. We used mixed models with hospital random effects to account for clustering.

Results: There were 85,121 TJRs in the pre-bundle period and 104,828 post-bundle; 46 of 166 hospitals participated in bundled payments for TJR (27.7%). The proportion of patients discharged home increased significantly post-bundle (63.6% to 78.4%). In adjusted difference-in-difference analyses using the propensity-matched cohort of patients who went to SNF in the pre-period and now go home post-period, 30- and 90-day readmissions decreased in the post-period (30d: -2.9%, 95% CI -4.2 to -1.6; 90d: -3.9%, 95% CI -.8 to -2.0). Mortality rates did not change. There were no differences in any outcome comparing hospitals that participated and those that did not participate in bundled payments.

Conclusions: Surprisingly, changes in hospital discharge destination after THR extend all payers and hospitals (even those not participating in bundled payments), and there is no evidence of harm in terms of readmissions or mortality. This could be consistent with improved peri-operative care, a beneficial change in practice (these are patients that should have gone home anyways), or unmeasured confounding. These findings raise important questions about whether other patient groups managed by hospitalists and subject to bundled payments might show similar results.