Objective: To investigate how an AI case-finding and clinical coaching intervention impacted mortality and how this impact varied by age, gender, and deprivation status. Design: Multi-site parallel prospective two-arm Randomised Controlled Trial led by Nuffield Trust and delivered by HN (Health Navigator Ltd). Patients were randomised on a 2:1 ratio to the intervention after consent and the automated and manual screening processes. Setting: Secondary care-based patient identification for a community-based intervention; Eight hospital sites across England were enrolled onto the study (York, Staffordshire, Essex, and Kent). Participants: Subjects aged 18 and over, who have experienced at least one emergency attendance in the preceding six months and identified as high-risk of unplanned hospitalisation via a prediction model. Subjects were also manually screened for their suitability to intervention. Intervention: One-to-one telephone-based health coaching, led by registered nurses or paramedics. Primary outcome measure: 24-month mortality. Results: The intervention was associated with reduced overall mortality (posterior probability: 92.2%), predominantly driven by the impact for males aged 75 and over (log-rank p-value: 0.0011, Hazard Ratio (HR) [95% CI]: 0.57 [0.37, 0.84], number needed to treat: 8). Excluding one site unable to adopt the prediction model indicated stronger impact (HR [95% CI]: 0.45 [0.26, 0.76]), suggesting a role of prediction in reducing mortality. Conclusions: Early mortality, specifically in elderly males, may be prevented by predicting individuals at risk of unplanned hospitalisation and supporting them with a clear outreach, out-of-hospital nurse-led, telephone-based coaching and care model. Trial registration: IRAS project ID: 173319; and clinicaltrials.gov ID: 2015-000810-23