Abstract WP65: Impact of Artificial Intelligence on Acute Ischemic Stroke Treatment in a Large Academic Healthcare System

Abstract

Introduction: We aimed to evaluate the impact of the implementation of a stroke triage artificial intelligence (AI) software on a large academic healthcare system. Methods: A retrospective study was conducted in our spoke and hub network comparing equal corresponding periods of pre- and post- Viz.ai implementation between January 2021 and December 2022. Door to needle, door to puncture (DTP), successful reperfusion rates, transfer rates, and clinical outcomes at discharge were compared. Results: The analysis encompassed a total of 5,456 patients including 2313 pre- and 2275 post-Viz.ai patients across 11 spokes and 416 pre- and 452 post-Viz.ai patients across 4 hubs. Among 868 patients undergoing EVT at the 4 hubs, there was no differences in baseline data except for a higher baseline NIHSS score in the post Viz.ai group. We found an overall absolute reduction of transfers by 60 cases equating to a statistically significant decrease of 2.3% (P=0.04). The results showed an overall decrease by 13.5 minutes (8.1%) which was not statistically significant (P=0.3). There was an overall statistically significant decrease of 6 minutes (9.5%) in DTP times from 63 (23-98) minutes to 57 (23-79) minutes (P=0.01). No other differences in outcomes were noted between both groups. Conclusion: The automatic AI software was able to show a notable impact in our hub and spoke mature network driven by decrease of unnecessary transfers, DIDO and DTP. These data support the role of AI software as an effective triaging tool in the stroke and hub networks. More studies are needed to confirm these results.