Abstract WP41: Should They Stay Or Should They Go? The Impact Of AI-based Care Coordination On Stroke Patient Transfers Across A Hospital Network

Abstract

Introduction: A key decision facing non-thrombectomy capable hospitals is whether to transfer a suspected large vessel occlusion (LVO) patient to a thrombectomy-capable stroke center (TSC). We report our experience with a commercially available AI-based care coordination platform (Viz.ai) designed to detect and alert LVOs on imaging (CTA) and improve communication among spoke and hub providers. Methods: All patients presenting across our community hospital network (consisting of two primary stroke centers, one stroke-ready hospital, and two freestanding emergency departments) with the diagnosis of acute stroke were included. We investigated the effect of the Viz.ai software platform on the number of stroke patients transferred out of our health system to the TSC and CTA utilization by comparing transfer patterns in a cohort of patients before (pre-Viz, 12/2018 - 10/2020) and after (post-Viz, 10/2020 - 08/2022) implementation. Results: A total of 1778 pre-Viz and 1779 post-Viz patients were compared. The post-Viz transfers significantly decreased by 33% (158 post-Viz vs 236 pre-Viz, p<0.0001) and the post-Viz patients receiving a CTA within our system increased by 64% (1167 post-Viz vs 710 pre-Viz, p<0.0001). The proportion of patients receiving a thrombolytic (tPA or TNK) who were cared for at the spokes was significantly higher post-Viz (74% vs 63%, p=0.02). The baseline clinical characteristics and study outcomes are summarized in the Table. Conclusions: In a consecutive cohort of stroke patients across a system of community hospitals that are not thrombectomy-capable, the implementation of the Viz.ai software platform was associated with reduced transfers to the thrombectomy center and an increased use of in-house CTA for LVO screening. There was an increase in the proportion of patients treated with intravenous thrombolysis who did not undergo transfer and received care locally suggesting improved identification of patients who benefit from transfer.