A Novel Algorithm for Prehospital Stroke Triage: A Population-Based In Silico Trial

Monday, March 23, 2026

9:00 AM - 10:00 AM Central Time

Location: Connections Central - RST 06

Nicholas Mohr, MD, MS, FCCM

University of Iowa Hospital and Clinics, Iowa

Disclosure(s): No relevant financial relationship(s) to disclose.
GB

Grant Brown, PhD

Associate Professor
University of Iowa

Disclosure information not submitted.
JS

Jacob Seedorf, N/A

Student
University of Iowa

Disclosure information not submitted.
FH

Ferney Henao-Ceballos, N/A

Graduate Student
University of iowa

Disclosure information not submitted.
AZ

Anne Zepeski, PharmD

Assistant Professor
University of Iowa

Disclosure information not submitted.
JC

Jorge Cespedes Segura, MD

Research Coordinator
University of Iowa

Disclosure information not submitted.
EK

Eric Kontowicz, PhD

Assistant Research Scientist
University of Iowa

Disclosure information not submitted.
DS

Dan Shane, PhD

Associate Professor
University of Iowa

Disclosure information not submitted.
NP

Natalia Perez de la Ossa, MD

Physician
N/A

Disclosure information not submitted.
MR

Marc Ribo, MD

Physician
N/A

Disclosure information not submitted.
JS

Jeffrey Saver, MD

Professor
University of California - Los Angeles

Disclosure information not submitted.
SO

Santiago Ortega-Gutierrez, MD, MS

Associate Professor
University of Iowa

Disclosure information not submitted.

Introduction: Effective prehospital stroke triage is essential for optimizing access to time-sensitive therapies. The objective of this study was to evaluate the clinical impact of a novel personalized prehospital stroke triage algorithm based on Bayesian predictive modeling to improve stroke outcomes.

Methods: To develop the Modeling Ambulance-based Prehospital Stroke Triage to Optimize Recovery (MAP-STROKE) destination selection algorithm, we used clinical trials data to model time-varying treatment efficacy of thrombolytic treatment and thrombectomy. Then we derived a diagnostic model (outcome=stroke subtype) and a treatment model (outcome=modified Rankin Scale [mRS]) to predict 3-month neurologic outcome for an individual patient. To estimate the impact of triage scenarios, we simulated stroke events in the U.S. over a 10-year period, then we predicted neurologic outcomes for this population under different scenarios: MAP-STROKE triage, AHA guideline, and nearest hospital routing.

Results: In our simulated population of 145 million stroke alerts across 20 replications, MAP-STROKE recommended the same hospital as the AHA guidelines in 39.6% of cases. Including all patients with a prehospital stroke alert, the MAP-STROKE algorithm improved the number of patients with good neurologic outcome by 0.5% (95% credible interval [Cr-I] 0.4-0.5%) over AHA guidelines and 0.7% (95%Cr-I 0.7-0.8%) over nearest hospital routing. In the subgroup of patients with large vessel occlusion (LVO), though, the neurologic improvement over AHA guidelines was greater (4.8% increase, 95% Cr-I 4.7-4.8%), which was even more pronounced for rural LVO patients (8.3% increase, 95%Cr-I 8.0-8.5%). Much of this change was attributable to a 76.5 min (95%Cr-I 76.0-77.0) reduction in time to thrombectomy compared with AHA guideline-adherent routing. These benefits were tempered by a 0.3% decrease in good neurologic outcome in non-LVO ischemic stroke patients, attributable to a 20.9 min delay in thrombolytic use and 2.7% fewer non-LVO patients who received thrombolytic treatment.

Conclusions: In prehospital stroke alert patients, use of the MAP-STROKE algorithm was estimated to improve good neurologic recovery through improved triage and destination selection. This improvement was most pronounced for LVO patients and those in rural areas.