Skilled Nursing and Inpatient Rehabilitation Facility Use by Medicare Fee-for-Service Beneficiaries Discharged Home After a Stroke: Findings From the COMPASS Trial

Objectives: To examine the effect of a comprehensive transitional care model on the use of skilled nursing facility (SNF) and inpatient rehabilitation facility (IRF) care in the 12 months after acute care discharge home following stroke; and to identify predictors of experiencing a SNF or IRF admission following discharge home after stroke. Design: Cluster randomized pragmatic trial Setting: Forty-one acute care hospitals in North Carolina. Participants: 2262 Medicare fee-for-service beneficiaries with transient ischemic attack or stroke discharged home. The sample was 80.3% White and 52.1% female, with a mean (SD) age of 74.9 (10.2) years and a mean ± SD National Institutes of Health stroke scale score of 2.3 (3.7). Intervention: Comprehensive transitional care model (COMPASS-TC), which consisted of a 2-day follow-up phone call from the postacute care coordinator and 14-day in-person visit with the postacute care coordinator and advanced practice provider. Main Outcome Measures: Time to first SNF or IRF and SNF or IRF admission (yes/no) in the 12 months following discharge home. All analyses utilized multivariable mixed models including a hospital-specific random effect to account for the non-independence of measures within hospital. Intent to treat analyses using Cox proportional hazards regression assessed the effect of COMPASS-TC on time to SNF/IRF admission. Logistic regression was used to identify clinical and non-clinical predictors of SNF/IRF admission. Results: Only 34% of patients in the intervention arm received COMPASS-TC per protocol. COMPASS-TC was not associated with a reduced hazard of a SNF/ IRF admission in the 12 months post-discharge (hazard ratio, 1.20, with a range of 0.95-1.52) compared to usual care. This estimate was robust to additional covariate adjustment (hazard ratio, 1.23) (0.93-1.64). Both clinical and non-clinical factors (ie, insurance, geography) were predictors of SNF/IRF use. Conclusions: COMPASS-TC was not consistently incorporated into real-world clinical practice. The use of a comprehensive transitional care model for patients discharged home after stroke was not associated with SNF or IRF admissions in a 12-month follow-up period. Non-clinical factors predictive of SNF/IRF use suggest potential issues with access to this type of care.