Impact of RapidAI® Implementation on Workflow Metrics and Access to Reperfusion Therapies in Acute Ischemic Stroke: A 10-Year Retrospective Analysis
Cerebrovascular Diseases, 2026
Abstract
Background and Purpose
Artificial intelligence‑assisted perfusion imaging can support selection and coordination in acute ischemic stroke (AIS). This study assessed institutional changes in access to reperfusion therapies and workflow metrics before and after RapidAI® implementation.
Methods
We retrospectively analyzed consecutive AIS admissions (July 2014–October 2024) at a single comprehensive stroke center, comparing pre‑implementation and post‑implementation periods (RapidAI® introduced June 2018). Primary outcomes were utilization of intravenous thrombolysis (IVT) and mechanical thrombectomy (MT), mRS 0–1 at discharge, and decompressive craniectomy (DC). Secondary outcomes were 90‑day mRS, door‑to‑needle (DTN) time, and hospital length of stay (LOS). Analyses were adjusted for key confounders and sensitivity analyses were made.
Results
Among 2,144 patients, adjusted analyses showed no significant differences in IVT (adjusted OR 0.93; 95% CI 0.52–1.69; p = 0.822) or MT (adjusted OR 2.73; 95% CI 0.70–10.61; p = 0.147) between periods. Rescue decompressive craniectomy was lower in the post period (adjusted OR 0.15; 95% CI 0.03–0.82; p = 0.029) and hospital length of stay was reduced (adjusted B −2.33 days; 95% CI −3.62 to −1.03; p < 0.001). No significant adjusted differences were observed for door to needle time (B −0.56 minutes; 95% CI −32.44 to 31.32; p = 0.972), discharge mRS 0–1 (adjusted OR 1.47; 95% CI 0.90–2.39; p = 0.126), discharge mRS ordinal (B −0.29; 95% CI −0.62 to 0.05; p = 0.091), or 90 day mRS ordinal (B −0.26; 95% CI −0.64 to 0.12; p = 0.174). Sensitivity analyses after multiple imputation produced consistent results.
Conclusions
RapidAI® implementation coincided with improved metrics in hospital length of stay and the need for decompressive craniectomy. Given low software utilization and potential secular trends, findings should be interpreted cautiously.
