Comparing a Care Program and Usual Care to Help Patients Transition Home from the Hospital after a Stroke -- The COMPASS Study [Internet]

Background: The Centers for Medicare & Medicaid Services (CMS) implemented transitional care management (TCM) reimbursement to support the uptake of hospital-to-home transitional care (TC) to improve outcomes and reduce readmissions for complex patients like stroke survivors. Given the high risk and unmet needs of patients with stroke and the wide variability in TC interventions, the scientific community saw an urgent need to evaluate the uptake and effectiveness of comprehensive TC.

Objectives: The COMprehensive Post-Acute Stroke Services (COMPASS) study compared a new postacute stroke TC model (COMPASS-TC) with usual care for 90-day functional status, caregiver strain, and other patient-reported and claims-based outcomes.

Methods: The COMPASS trial was a pragmatic cluster-randomized trial. Forty North Carolina hospital units were randomized 1:1 to the COMPASS-TC intervention or usual care, stratified by volume of patients with stroke and stroke center certification. After approximately 1 year, the hospitals entered phase 2. Hospitals randomized to usual care crossed over to implement COMPASS-TC, and hospitals randomized to intervention in phase 1 sustained COMPASS-TC. COMPASS-TC plans for secondary prevention, recovery, and referrals to services and community-based resources. COMPASS-TC was consistent with CMS TCM reimbursement requirements. Hospital staff received training, which included 2 days of training prior to launching the intervention, monthly ongoing training, monthly one-on-one meetings to review performance metrics (eg, the percentage of patients who attended a follow-up clinic visit, the % receiving 2-day calls; case ascertainment), and access to our study website with long-standing training materials. In phase 1, hospital staff enrolled 6024 adults discharged home after stroke or transient ischemic attack (TIA). Twenty-five hospitals (3771 participants) were included in the phase 2 analysis. The primary outcome was functional status (Stroke Impact Scale-16); secondary outcomes were mortality, disability, medication adherence, depression, cognition, self-reported general health, fatigue, care satisfaction, home blood pressure (BP) monitoring, falls, physical activity, and caregiver strain. Telephone interviewers, blinded to treatment assignment, assessed these outcomes at 90 days. Additional secondary outcomes were identified from claims for 2925 participants, including CMS Medicare fee-for-service (FFS; n = 2262), NC Medicaid (n = 382), and Blue Cross Blue Shield of North Carolina (n = 277). Claims-based outcomes included readmissions, continuity of care (as indicated by ambulatory care visits), TCM billing, skilled nursing and inpatient rehabilitation facility visits, and emergency department visits. In intention-to-treat analyses, we estimated treatment effects from generalized linear mixed models adjusted for age, race, stroke severity, stroke type, and history of stroke. Additional covariates were considered in sensitivity analyses. We addressed missing covariate data using multiple-imputation procedures, and inverse probability weights accounted for missing patient-reported outcomes. We used Bayesian hierarchical models to explore implementation sustainability and effectiveness.

Results: Eighty-eight percent of participants had mild stroke or TIA. Receipt of the intervention was low; 35% of patients received COMPASS-TC in phase 1. Only 58% of COMPASS hospitals delivered the intervention uninterrupted during phase 1. The primary outcome measure mean ± SD Stroke Impact Scale-16 scores were similar between intervention and usual care (80.6 ± 21.1 vs 79.9 ± 21.4). Home BP monitoring was self-reported by 72% in the intervention arm vs 64% in usual care (adjusted odds ratio [OR], 1.43; 95% CI, 1.21-1.70). No other secondary patient-reported outcomes differed by arm, nor was there effect heterogeneity across prespecified subgroups of age, race, sex, diagnosis type, insurance status, or stroke severity. Within the FFS population, COMPASS-TC was associated with significantly higher rates of ambulatory care visits (hazard ratio, 1.27; 95% CI, 1.14-1.41) and TCM billing (OR, 1.87; 95% CI, 1.54-2.27). No differences by arm were found for other claims-based outcomes. Receipt of ambulatory care visits was considerably lower for Medicaid enrollees (12%) than for those with FFS and Blue Cross Blue Shield of North Carolina (89% and 78%, respectively). Eleven of the 20 sites randomized to intervention in phase 1 sustained COMPASS-TC in phase 2. Functional status, home BP monitoring, and measures of implementation remained stable across phases, suggesting that the effectiveness of the intervention on BP monitoring in phase 1 was maintained when hospitals sustained the intervention without study support. Finally, home BP monitoring, ambulatory care, and TCM billing were significantly improved in phase 2 compared with phase 1 in hospitals that crossed over to implement COMPASS-TC in phase 2.

Conclusions: This study provided a realistic measure of TC uptake for stroke and TIA survivors in current practice. COMPASS-TC was not consistently incorporated into real-world care despite being designed with stakeholder input and being consistent with CMS TCM policies. The intervention did not improve 90-day functional status nor most other secondary outcomes except BP monitoring. The findings may reflect the low intervention uptake or indicate that a TC follow-up visit could not impact long-term outcomes. Within FFS enrollees, COMPASS-TC led to significantly higher rates of ambulatory care visits and TCM billing. Future research should focus on improving TC implementation and evaluating what interventions and policy changes are effective, valued, and feasible.

Limitations: The benefits of pragmatic trials also come with challenges. The low rate of intervention adherence and outcome ascertainment limited statistical power. Unmeasured differences between individuals with and without observed outcomes may limit the conclusiveness of the results.