Abstract P471: Predictors of Referral to Community-Based Rehabilitation Following Stroke

Abstract

Background: While several studies have examined care transitions following stroke, few have assessed use of rehabilitation. We explored factors predictive of referral to community-based rehabilitation following stroke. Setting: 40 NC acute care hospitals that participated in the COMprehensive Post-Acute Stroke Services study from 2016-2019. Participants: Adults discharged home following stroke or TIA (N=11,195) Methods: Guided by an a priori conceptual framework, 41 predictors of referral to community-based rehabilitation were grouped into 3 patient-level domains: demographics, stroke severity, and medical history (e.g., prior stroke); 2 hospital-level domains: structural characteristics (e.g., bed size) and stroke-specific characteristics (e.g., stroke center status); and 1 community-level domain (county-level therapist supply). We estimated predictors of referral using a logistic mixed model with a hospital-specific random effect. We used a hierarchical backward selection approach, first performing domain-specific model selection, then entering retained variables into a complete model. Results: The cohort was 49% female, 72% white, with a mean age of 66.8 (SD=14.0) years and overall low stroke severity [mean [SD] NIHSS 2.4 [3.8]). Thirty-six percent of patients were referred to rehabilitation. Referral varied across hospitals, ranging from 3-78%, with a median of 35%. In the final model, older age, female sex, non-white race, higher stroke severity, longer length of stay, previous stroke, current smoking, and heart failure were associated with higher odds of referral as was hospital-employed vs. external-contracted OT services (Table). Discussion: Approximately one-third of stroke survivors discharged home were referred to community-based rehabilitation. Women and non-white patients had higher odds of referral. Hospital and community factors were largely not independent predictors of referral. Unmeasured process measures may play a role in unexplained hospital variation.