Population-level impact of universal early AED implementation on out-of-hospital cardiac arrest outcomes: a machine learning analysis using synthetic patient data of 45-year projection study from Japan

BACKGROUND: Out-of-hospital cardiac arrest affects 130,000 individuals annually in Japan, with favorable neurological outcomes of <4 %. Despite widespread AED deployment (>670,000 devices), bystander utilization remains suboptimal at 5 % in witnessed cases. METHODS: We used machine learning-generated synthetic data from 2022 All-Japan Utstein Registry to project population-level impacts of universal early AED access through 2075. The synthetic dataset (n = 966,493) enabled population-level projections, while time-to-AED data were available only for witnessed cases. The intervention scenario modeled bystander AED application within 5 min for all witnessed arrests, compared with current patterns. Classification and Regression Trees incorporated Japan's demographic projections. Primary outcomes were one-month survival and favorable neurological outcome (CPC 1or 2). Rate ratios with 95 % confidence intervals were estimated using modified Poisson regression. RESULTS: Universal early AED implementation showed peak effectiveness in 2055-2060: favorable neurological outcome RR 2.16 (95% CI, 2.07-2.25; 116% relative increase). Over the 45-year projection period (2030-2075), universal early AED implementation could prevent approximately 235,000 deaths (95% CI, 210,500-261,500) and result in 160,000 more survivors with good neurological recovery after OHCA (95% CI, 142,000-181,000). Analyses estimated a number needed to treat of 19-29 (median 24) witnessed OHCA patients receiving bystander AED within 5 min to achieve one additional favorable neurological outcome. CONCLUSIONS: Machine learning-based modeling projects that universal bystander AED application in Japan over a period of four and a half decades could prevent approximately 235,000 deaths and result in 160,000 more survivors with good neurological recovery after OHCA.