Key result
EHR-derived model predicts 30-day periprocedural ischemic stroke with ~0.86 AUC.
Why the study?
Existing risk calculators often exclude high-risk cardiac, major vascular, and neurosurgical procedures and rarely incorporate cerebrovascular disease markers available in routine electronic health records.
Cohort (n=444,945)
Yes
Effect estimate: AUC 0.86 (95% CI 0.84-0.87)
An EHR-derived predictive model integrating procedure type and markers of cerebrovascular disease accurately estimates 30-day periprocedural ischemic stroke risk across diverse surgical populations.
May support EHR-based periprocedural stroke risk stratification; leaves open prospective validation and outcome impact.
BACKGROUND: Perioperative ischemic stroke is uncommon overall but frequent in cardiac, major vascular, and neurosurgical procedures. Existing calculators often exclude these settings and rarely incorporate cerebrovascular disease markers available in routine electronic health record data. METHODS: We assembled a retrospective cohort of adults undergoing procedures at 3 hospitals (January 2016-June 2024). The model was derived at Rhode Island Hospital (255 850 procedures) and externally validated at 2 affiliated hospitals (The Miriam Hospital/Newport Hospital; 189 095 procedures). Candidate predictors included age, vascular comorbidities, documented carotid stenosis and intracranial atherosclerosis, procedure setting (ambulatory versus inpatient/emergency), and procedural service (including vascular versus nonvascular neurosurgery and open versus interventional cardiovascular procedures). We fit a multivariable logistic regression model with internal validation using 1000 bootstrap resamples and assessed calibration by observed versus predicted risk across deciles. External validation applied locked derivation coefficients without refitting. RESULTS: Strokes occurred in 1235/255 850 derivation procedures (0.48%) and 418/189 095 validation procedures (0.22%). Independent predictors included older age, prior stroke or transient ischemic attack (adjusted odds ratio [aOR], 6.66), inpatient/emergency setting (aOR, 4.25 versus ambulatory), vascular neurosurgery (aOR, 6.70 versus general surgery), and open cardiovascular procedures (aOR, 4.14). Discrimination was high (derivation area under the curve, 0.87 [95% CI, 0.87-0.88]; optimism-corrected area under the curve, 0.87; validation area under the curve, 0.86 [95% CI, 0.84-0.87]) with good calibration. Prespecified risk strata (<1%, 1-5%, >5%) separated observed event rates in both cohorts. CONCLUSION: A pragmatic, electronic health record-derived model integrating procedural category and cerebrovascular disease accurately predicts 30-day periprocedural ischemic stroke across diverse procedures and is available as a web-based calculator to support counseling and targeted prevention.
No takes yet. Share an insight, caveat, or question.
Shu et al. (2026) conducted a cohort in Periprocedural Ischemic Stroke (n=444,945). Electronic health record-derived prediction model was evaluated on 30-day periprocedural ischemic stroke (model discrimination in external validation cohort) (AUC 0.86, 95% CI 0.84-0.87). An electronic health record-derived model integrating procedural category and cerebrovascular disease accurately predicted 30-day periprocedural ischemic stroke with an AUC of 0.86.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: