Predictors and Temporal Trends of Withdrawal of Life-Sustaining Therapy After Acute Stroke in the Florida Stroke Registry

OBJECTIVES: Temporal trends and factors associated with the withdrawal of life-sustaining therapy (WLST) after acute stroke are not well determined. DESIGN: Observational study (2008-2021). SETTING: Florida Stroke Registry (152 hospitals). PATIENTS: Acute ischemic stroke (AIS), intracerebral hemorrhage (ICH), and subarachnoid hemorrhage (SAH) patients. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Importance plots were performed to generate the most predictive factors of WLST. Area under the curve (AUC) for the receiver operating curve were generated for the performance of logistic regression (LR) and random forest (RF) models. Regression analysis was applied to evaluate temporal trends. Among 309,393 AIS patients, 47,485 ICH patients, and 16,694 SAH patients; 9%, 28%, and 19% subsequently had WLST. Patients who had WLST were older (77 vs 70 yr), more women (57% vs 49%), White (76% vs 67%), with greater stroke severity on the National Institutes of Health Stroke Scale greater than or equal to 5 (29% vs 19%), more likely hospitalized in comprehensive stroke centers (52% vs 44%), had Medicare insurance (53% vs 44%), and more likely to have impaired level of consciousness (38% vs 12%). Most predictors associated with the decision to WLST in AIS were age, stroke severity, region, insurance status, center type, race, and level of consciousness (RF AUC of 0.93 and LR AUC of 0.85). Predictors in ICH included age, impaired level of consciousness, region, race, insurance status, center type, and prestroke ambulation status (RF AUC of 0.76 and LR AUC of 0.71). Factors in SAH included age, impaired level of consciousness, region, insurance status, race, and stroke center type (RF AUC of 0.82 and LR AUC of 0.72). Despite a decrease in the rates of early WLST (< 2 d) and mortality, the overall rates of WLST remained stable. CONCLUSIONS: In acute hospitalized stroke patients in Florida, factors other than brain injury alone contribute to the decision to WLST. Potential predictors not measured in this study include education, culture, faith and beliefs, and patient/family and physician preferences. The overall rates of WLST have not changed in the last 2 decades.