Predictors of 30-Day Hospital Readmission Following Acute Stroke (S32.006)

OBJECTIVE: To determine the clinical factors associated with 30-day readmission in patients with acute stroke. BACKGROUND: 30-day hospital readmission will soon be used to measure quality of care. Whether characteristics of patients at high risk of readmission can be identified prior to discharge is uncertain. DESIGN/METHODS: We designed a retrospective case-control study. Patients with acute ischemic or hemorrhagic strokes readmitted to an academic center within 30 days of discharge over an 18 month period were identified (cases) and compared to diagnosis-matched controls, not readmitted over this period. Demographic and clinical information (e.g. number of prior hospitalizations, admission NIH stroke scale) were collected. Descriptive analyses were performed using Chi-square and Wilcoxon tests. Multivariate stepwise logistic regression was used to identify variables associated with 30-day readmission. RESULTS: The cohort included 198 patients (93 cases, 105 controls). Age, gender, and race did not differ between groups. Median time to readmission was 10 days (IQR 5 to 17). Readmission diagnoses included ischemic stroke (21.5%), infection (11%), and cardiac dysfunction (10%). Readmitted patients were more likely to have been hospitalized 2 or more times in the year prior to stroke (22.6% vs 4.8% in controls, p=0.0007). History of congestive heart failure (CHF) and pneumonia during the incident admission were associated with readmission (p=0.0097 and 0.033, respectively). The multivariate model showed that admission NIHSS (OR 1.056, 95% CI 1.010-1.103 per 1 point increase; p=0.015), prior hospitalizations (OR 1.754, 1.258-2.446 per additional admission; p=0.0009), and absence of hyperlipidemia (OR 0.501, 0.267-0.940; p=0.031) were independently associated with readmission. CONCLUSIONS: Admission NIHSS and frequent hospitalizations prior to a stroke are associated with subsequent 30-day readmission following acute ischemic or hemorrhagic stroke. Although this model requires validation, the clinical characteristics identified in this study can be used to target high-risk patients and reduce readmission. Disclosure: Dr. Stiles has nothing to disclose. Dr. Strowd has nothing to disclose. Dr. Bishop has nothing to disclose. Dr. Umesi has nothing to disclose. Dr. Craig has nothing to disclose. Dr. Lefkowitz has nothing to disclose. Dr. Reynolds has nothing to disclose. Dr. Arnan has nothing to disclose. Dr. Bushnell has received research support from Genentech.