Institutional Post-Acute Care Use After Hospitalization by Medicare Type, 2017-2022

Abstract

Objectives To compare national discharge to inpatient rehabilitation facilities (IRF) and skilled nursing facilities (SNF) after short-stay hospitalization among Medicare Advantage (MA) and Traditional Medicare (TM) beneficiaries. Design Retrospective cohort study using publicly available aggregate Medicare utilization data. Setting and Participants Participants included United States MA and TM beneficiaries from 2017 to 2022. Methods: Primary outcomes were discharge rates to IRFs and SNFs per 1000 acute hospital discharges and in-facility mortality. Secondary outcomes included community discharge and 30-day readmission among TM beneficiaries. Year-adjusted Poisson regression with acute discharges as an offset estimated incidence rate ratios (IRRs) comparing MA and TM utilization. Discharge-weighted logistic regression estimated odds ratios (ORs) for mortality, community discharge, and readmission. Trends in IRF vs SNF utilization were evaluated using year-adjusted logistic regression. Results Compared with TM beneficiaries, MA beneficiaries had lower discharge rates to IRF (IRR, 0.397; 95% CI, 0.395-0.398), with a smaller difference for SNF (IRR, 0.904; 95% CI, 0.903-0.905). Within IRF, in-facility mortality odds were higher for MA beneficiaries (OR, 14.69; 95% CI, 14.21-15.18). Within SNF, in-facility mortality odds were lower for MA beneficiaries (OR, 0.447; 95% CI, 0.444-0.450). Across both payers combined, mortality was lower in IRF than SNF (0.71% vs 2.46%; OR, 0.28; 95% CI, 0.278-0.286; P < .001). Among TM beneficiaries, discharge to IRF was associated with lower odds of 30-day readmission (OR, 0.76; 95% CI, 0.75-0.76) and higher odds of community discharge (OR, 1.84; 95% CI, 1.83-1.85). Conclusions and Implications MA beneficiaries had lower downstream use of IRF after hospitalization than TM beneficiaries. In-facility mortality differed among post-acute settings and payer. Patient-level case-mix at the referral source is needed to evaluate whether these differences persist after case mix adjustment.