Abstract WP349: Sex-Based Claims Models for the Prediction of 1-Year Mortality After Ischemic and Hemorrhagic Stroke

Abstract

Background: Claims-based models have been developed to predict 30-day mortality after ischemic and hemorrhagic stroke, yet few models are available for predicting 1-year post-stroke mortality among discharged patients. Moreover, there are limited data on whether factors associated with 1-year mortality differ for women and men. We developed claims-based models to predict 1-year mortality after hospital discharge for ischemic and hemorrhagic stroke stratified by sex. Methods: We identified fee-for-service Medicare beneficiaries aged ≥65y who were discharged alive with a principal diagnosis of ischemic (ICD-9 433, 434, 436) or hemorrhagic (ICD-9 430, 431) stroke in 2014. We fit Cox regression models with Markov-chain Monte Carlo simulation to identify risk factors for 1-year all-cause mortality; variables with a posterior probability ≥0.95 of being associated with 1-year mortality were considered significant and included in the models for the overall samples and for each sex. Results: The study included 235,737 ischemic (53% women; mean age 80.3y [women] and 77.2y [men]) and 43,084 hemorrhagic (52% women; mean age 80.0y [women] and 77.8y [men]) stroke patients. One-year mortality was 27.9% (95% CI 27.7%-28.2%) for women and 22.0% (21.8%-22.3%) for men who had an ischemic stroke and was 50.8% (50.2%-51.5%) for women and 46.4% (45.7%-47.1%) for men who had a hemorrhagic stroke. The c statistics for the models in the overall samples were 0.79 for ischemic stroke and 0.76 for hemorrhagic stroke (Figure). Although there were small differences in the variables included in the models by sex, the c statistics were comparable. Conclusions: One-year claims-based models can be used by hospitals to predict the risk of death for patients discharged after stroke. The models identified many risk factors common to women and men that can inform strategies to improve the clinical management of stroke patients during the first year of recovery.