Abstract MP056: A Discrete Event Simulation for the Assessment of Emergency Medical Services Routing Policies for Large Vessel Occlusion Stroke: A Comparison Between Two US Counties
Introduction: Proposed Emergency Medical Services (EMS) routing policies permit additional transport time for suspected large vessel occlusion acute ischemic stroke (LVO) patients if the nearest hospital is not an endovascular center (EC). The effectiveness of these policies may depend on the region.
Methods: We created a discrete event simulation of EMS-screened suspected stroke patients over a year, assuming that 40% were strokes and 20% of strokes were LVO. We used hospital locations and demographic data of Mecklenburg County, NC (2 ECs, 6 non-ECs, pop. 990k, 546 mi2) and King County, WA (4 ECs, 10 non-ECs, pop. 2.4M, 2300 mi2). Patients were assigned to census tracts using estimated annual strokes per tract calculated from published incidence rates. A patient’s location within the tract was randomized. We retrieved real road travel times to estimate transport time to hospitals. Last known well time (LKW) was probabilistically assigned. A patient was EC-routed if they positively screened for LVO, had LKW ≤6 hours, were within the permitted transport time, and the closest hospital was not an EC. We simulated policies varying stroke severity screen (Los Angeles Motor Scale ≥ 4 (LAMS): 81% sensitivity, 89% specificity; Cincinnati Stroke Triage Assessment Tool ≥ 2 (C-STAT): 83% sensitivity, 40% specificity) and the permitted additional transport time to EC (10, 20, and 30 minutes). Each policy was replicated 20 times. Number needed to route (NNR) is the number of patients EC-routed for one LVO to be routed to an EC.
Results: EMS screened on average 3102 patients in Mecklenburg and 5178 in King County. In Mecklenburg, 67%, 99%, and 100% of LVOs were within a respective 10, 20, and 30 minutes of additional transport time to an EC; an EC was not the closest hospital for 57%, 71%, and 71% of these. In King, fewer LVOs met the same transport time criteria (43%, 59%, and 68%) and fewer of these were closest to a non-EC (37%, 55%, and 62%). EC-routing added a mean 6, 10, and 10 minutes (Mecklenburg) and 6, 10, and 13 minutes (King) for 10, 20, and 30 minute policies, respectively. EC-routed LVOs totaled 54, 98, and 98 (Mecklenburg) and 55, 110, and 143 (King) for 10, 20, and 30 minute policies using LAMS. With LAMS, 62% of EC-routed patients were non-LVO, totaling 83, 148, and 158 annually (Mecklenburg) and 89, 177, and 218 (King) for 10, 20, and 30 minute polices. Using LAMS, NNR was 2.54, 2.51, and 2.61 (Mecklenburg) and 2.62, 2.61, and 2.52 (King) for 10, 20, and 30 minute policies. For C-STAT, 89% were non-LVO with NNR>8 for all policies.
Conclusions: Our customizable simulation evaluates EC routing policies. Low specificity of a screening tool coupled with a large permitted EC routing time may contribute to congested ECs. Our results suggest that with respect to NNR, Mecklenburg may prefer to permit 20 additional minutes for EC transport, but King may prefer 30 minutes. We demonstrate that one policy may not be optimal for all regions.
Author Disclosures: B.M. Bogle: None. A. Asimos: None. W.D. Rosamond: None.
- © 2017 by American Heart Association, Inc.