Using Discrete Event Simulation to Assess the Regionally-Specific Impact of the Severity-Based Stroke Triage Algorithm for EMS on Patient Outcomes and Overtriage

Introduction: We previously developed a discrete event simulation environment to evaluate the regional impact of the AHA/ASA’s Severity-Based Stroke Triage Algorithm for EMS. It now is capable of projecting patient outcomes. Methods: We simulated EMS-assessed suspected acute ischemic stroke (AIS) patients with last known well < 6 hours over a year in Mecklenburg County, NC. Patients were geographically distributed per projected census tract estimates. Suspected large vessel occlusion (LVO) AIS patients were routed to an endovascular stroke center (ESC) under varied screening tools and transport time rules. LVO patients initially transported to non-ESCs were transferred to an ESC after IV-tPA. Based on patient type (LVO, non-LVO AIS, non-AIS) and hospital type (ESC, non-ESC), we probabilistically assigned time to hospital arrival, imaging, IV-tPA, transfer, and endovascular thrombectomy. Patient survival and modified Rankin Score (mRS) were estimated as a function of time and type of treatment received. Overtriage was calculated as the percent of patients unnecessarily transported directly to ESCs due to false positive screens. Results: An average of 2172 patients (172 LVO, 556 non-LVO AIS, 1444 non-AIS) were screened annually. Results by stroke severity screen and transport time rule are in the Table. For example, using C-STAT and a 10 minute transport threshold: 559 (79%) of direct arrivals to ESCs were overtriaged; 58% of LVOs were directly transported to an ESC; and 54% of LVOs received endovascular thrombectomy within 6 hours. Of LVOs, 82% survived, 37% achieved an mRS ≤2. Of non-LVO AIS patients, 91% survived and 50% achieved mRS≤2. Conclusions: Our data suggest the burden of overtriage with some algorithm implementations may outweigh the marginal gains in survivors and patients with favorable neurological discharge, but this likely varies by region, demonstrating the value of leveraging such modeling environments for tailoring triage algorithms.