Abstract TP146: Prediction of 30-Day Subsequent Stroke Hospitalization After an Emergency Department Dizziness Visit: A Cohort Study

Introduction: Stroke after ED dizziness visits is higher than in comparison groups. However, studies evaluating the ability to predict subsequent stroke using readily available clinical data are lacking. We assessed the prediction of subsequent stroke after ED dizziness visits using electronic medical record data. Methods: We performed a retrospective cohort study at a multicenter health system in Southern California (January 2017-December 2020). Adults presenting to the ED for dizziness who were discharged home were included. Extracted variables included demographics, vascular risk factors, dizziness-relevant prior medical history (prior dizziness, migraine, anxiety/depression, Elixhauser comorbidities), index visit diagnosis (ICD10: R42.x, H81.x, vs all others), and index visit neuroimaging. The primary outcome was stroke hospitalization within 30 days. Cox proportional hazards model was used to estimate associations between the baseline factors and stroke risk. Model performance was evaluated using Harrell’s C for discrimination and calibration across risk strata. Results: Among 77,315 patients discharged after an ED dizziness visit, 94 (0.12%, 1 in 833) had stroke hospitalization within 30 days. Of these, 43% (40/94) were on antiplatelets/anticoagulants and 40% (38/94) had index visit neuroimaging. Higher ASCVD score (HR, 1.16, 95%CI 1.08-1.26) and prior stroke (HR, 2.70; 95%CI, 1.33-5.52) were associated with subsequent stroke, whereas other prior medical history, index visit diagnosis, and neuroimaging were not. Model discrimination was modest (Harrell’s C, 0.67) with good calibration across risk quantiles. The highest risk quantile had a mean predicted stroke risk of 0.2% (1 in 500), approximately twice that of the lower quantiles, captured 50% of the subsequent strokes (47/94), and showed baseline antiplatelet/anticoagulant use of 55% (10638/19328) and index visit imaging of 54% (10429/19328). Conclusions: Baseline vascular risk assessment may help identify ED dizziness patients at modestly higher short-term stroke risk, though absolute risk remains low even in the highest risk group. Other prior medical history, ED diagnosis and neuroimaging at index visit did not impact subsequent stroke probability. Attention to standard short-term vascular risk reduction (e.g., blood pressure, antiplatelets) may warrant further consideration prior to discharge of ED dizziness visits.