The EHR-derived model predicted 30-day ischemic stroke with AUCs of 0.85 in derivation and 0.83 in validation, highlighting significant risk factors like prior stroke/TIA (aOR 4.53).
An EHR-derived model integrating procedural categories and cerebrovascular disease markers provides strong discrimination for predicting 30-day perioperative ischemic stroke risk.
Absolute Event Rate: 0% vs 0%
Background and Purpose: Perioperative ischemic stroke is uncommon overall but frequent in cardiac, major vascular, and neurosurgical procedures. Existing risk tools often exclude these contexts and omit cerebrovascular imaging markers. We developed and externally validated an EHR-derived 30-day stroke prediction model that incorporates both. Methods: Adults undergoing procedures at 3 hospitals (January 2016–June 2024) were included. The model was derived at a tertiary center and validated at two affiliated hospitals. Candidate predictors included demographics, comorbidities (e.g., prior stroke/TIA, atrial fibrillation, hypertension, diabetes), carotid stenosis and intracranial atherosclerosis, procedure setting (ambulatory vs inpatient/emergency), and procedural service. The primary outcome was 30-day ischemic stroke. Discrimination (AUC) and calibration were evaluated with bootstrap internal validation and external validation. Results: Strokes occurred in 1,235/255,850 derivation procedures (0.48%) and 418/189,095 validation procedures (0.22%). Independent predictors were older age (aOR 1.02 per year), prior stroke/TIA (aOR 4.53), carotid stenosis (aOR 2.13), intracranial atherosclerosis (aOR 2.93), atrial fibrillation (aOR 1.30), diabetes (aOR 1.23), hypertension (aOR 1.49), inpatient/emergency setting (aOR 4.40 vs ambulatory), and cardiovascular or neurosurgical procedures (aOR ~3–4 vs general surgery). Discrimination was high (derivation AUC 0.85, 95% CI 0.84–0.86; validation AUC 0.83, 95% CI 0.81–0.85) with good calibration. Predicted risk strata (5%) separated observed risk in both cohorts (derivation 0.23%, 2.19%, 6.45%; validation 0.13%, 1.05%, 3.09%). Conclusions: An EHR-derived perioperative stroke model that integrates high-risk procedural categories and cerebrovascular disease achieved strong discrimination and calibration across diverse procedures and may inform preoperative counseling and targeted prevention. A web-based calculator accompanies this work and access details will be provided in the Poster.
Shu et al. (Thu,) reported a other. The EHR-derived model predicted 30-day ischemic stroke with AUCs of 0.85 in derivation and 0.83 in validation, highlighting significant risk factors like prior stroke/TIA (aOR 4.53).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: