Abstract Body: Introduction: Delays in endovascular therapy for acute large vessel occlusion (LVO) stroke can contribute significantly to disability following successful recanalization. The implementation of an automated intelligence LVO detection and interdisciplinary communication platform can shorten times to treatment.
Methods: We conducted a multicenter retrospective observational cohort study of consecutive adults with acute occlusion of the internal carotid, proximal middle cerebral, or basilar artery. Hub-and-spoke networks implementing Viz.ai queried electronic medical records 6 months prior to and 6 months following implementation of Viz.ai. Patients were included if they had a National Institutes of Stroke Scale (NIHSS) score ≥6, pre-stroke modified Rankin Scale 0-1, and presented within 24 hours of last known well (or unknown). The primary outcome was time from initial hospital contact to arterial puncture, which was compared between study periods using descriptive statistics, regression with robust standard errors clustered by site, and adjusted inverse probability of treatment weighting (IPTW) in which probability weights were used to reduce imbalance between study periods in a causal inference model. The model was adjusted for age, NIHSS, sex, comorbidities, overnight arrival, hub versus spoke arrival, academic quarter, and pre-stroke modified Rankin Scale which was imputed when missing using chained equations as an ordinal covariate.
Results: Of the 474 included patients across 7 sites (n=215 post-Viz, 45.4%), the median age was 67 years (interquartile range [IQR] 57-77) and median NIHSS was 17 (IQR 11-22). Using descriptive statistics, there was a trend toward a shorter time from hospital contact to puncture during the post-Viz period (median 103min, IQR 68-146, vs. 106min, IQR 76-169, p=0.10). In unadjusted regression with robust errors, clustered by site, the trend persisted (β -26.3, 95% confidence interval [CI], -53.7 to 1.3, p=0.058). In the adjusted IPTW model, arrival during the post-Viz period was associated with a shorter adjusted average treatment effect (time difference) of 31 minutes (95% CI, 14 to 48 minutes, p<0.001) when compared to arrival during the pre-Viz period.
Conclusions: Implementation of the Viz.ai platform led to a significant decrease in time to arterial puncture for patients with acute LVO. The degree to which these changes contributed to better clinical outcomes is being explored in subsequent analyses.