<![CDATA[BackgroundThe emergency medical services (EMS) sector presents substantial hurdles in providing excellent treatment while operating within limited resources. Artificial intelligence (AI) has significant potential in prehospital EMS, where rapid and accurate decision-making is essential. However, a comprehensive synthesis of AI’s effectiveness in this setting remains limited. The objective of this systematic review and meta-analysis was to evaluate AI’s predictive accuracy, compare its performance with human-based triage, and assess its feasibility in prehospital care. MethodsFollowing PRISMA guidelines, PubMed, Cochrane Library, and Web of Science were searched up to February 20, 2025. Fourteen peer-reviewed studies reporting quantitative outcomes were included. Study quality was assessed using the Newcastle-Ottawa Scale (median score = 7). Random-effects meta-analyses in R were conducted to synthesize area under the curve (AUC), AI-human AUC differences, and feasibility outcomes. ResultsAcross 14 studies involving 9,107,906 patients, AI demonstrated strong predictive performance with a pooled AUC of 0.874 (95% CI: 0.843–0.905). Individual study AUCs ranged from 0.805 to 0.997, with moderate heterogeneity (I²=52.3%). In eight studies, AI significantly outperformed human triage, showing a pooled AUC improvement of 0.074 (95% CI: 0.045–0.103; p<0.001), despite high heterogeneity due to varying human benchmarks. Feasibility analysis showed high clinical utility, with a pooled proportion of 0.929 and no heterogeneity (I²=0%). Funnel plots suggested slight asymmetry, but Egger’s tests indicated no significant publication bias. ConclusionArtificial intelligence demonstrated high predictive accuracy, slight superiority over human triage, and strong feasibility in prehospital care, supporting its integration into EMS with further validation and standardized reporting.]]>
