AI-driven digital health history system for perioperative cardiac risk assessment in non-cardiac surgery

Introduction: This prospective, multicenter, non-randomized study evaluated the association between an AI-assisted Digital Health History Device—the Comprehensive AI-Assisted Preoperative Evaluation (CAPE) software—and the quality and clinical outcomes of perioperative cardiology consultations in non-cardiac surgery. Materials and methods: A total of 405 patients were enrolled across three hospitals. Clinic-level pathway allocation was determined by local IT infrastructure readiness, assigning 200 patients to the CAPE software pathway and 205 to standard physician-led care. CAPE integrated ChatGPT (OpenAI) under physician supervision to convert structured data into guideline-driven consultation notes. The primary outcome was the 30-day composite rate of major perioperative cardiovascular and cerebrovascular complications. Results: The primary composite endpoint was significantly lower in the CAPE group compared to controls (7.0% vs. 19.5%; unadjusted OR: 0.31, 95% CI: 0.16–0.59, p < 0.001; and adjusted OR: 0.34, 95% CI: 0.17–0.68, and p = 0.002). Among secondary outcomes, hypertensive crises occurred less frequently in CAPE (2.0% vs. 7.8%, p = 0.0075; Bonferroni threshold α = 0.0083), whereas differences in hypotension, arrhythmias, and major bleeding were not statistically significant after adjustment. Turnaround time was shorter with CAPE (1.53 ± 1.12 vs. 6.38 ± 8.50 days, p < 0.001), and blinded raters scored CAPE-generated notes higher across all quality domains ( p < 0.001). Conclusions: Under physician oversight, CAPE software was associated with improved consultation quality, greater efficiency, and lower observed composite complication rates. The principal limitation is the non-randomized design, which precludes establishing direct causality and leaves potential for unmeasured confounding.