Related Experiment Video
Updated: Jun 10, 2026
Implementation of a Real-Time Psychosis Risk Detection and Alerting System Based on Electronic Health Records using CogStack
Published on: May 15, 2020
Development and External Validation of an Electronic Health Record-Derived Model for 30-Day Periprocedural Ischemic
Liqi Shu1, Karen L Furie1, Ana-Lucia Garcia Guarniz1
1Department of Neurology The Alpert Medical School of Brown University Providence RI USA.
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.
