Cost-Effectiveness and Clinical Utility of a Retinal Cardiovascular Artificial-Intelligence-Guided Strategy for...

To evaluate the clinical utility and long-term cost-effectiveness of artificial intelligence (AI)-guided retinal cardiovascular risk assessment, compared with the standard of care, for guiding statin therapy in primary cardiovascular disease prevention in South Korea.

In this multicenter study, 274 prospectively enrolled AI-guided participants were compared with 274 propensity score-matched retrospective standard of care controls; guideline concordance was adjudicated by an independent evaluator. Long-term cost-effectiveness was assessed from a healthcare-system perspective using a decision-tree and lifetime Markov microsimulation of 10 000 hypothetical individuals aged 63 years. Costs and quality-adjusted life-years (QALYs) were discounted at 4.5%; deterministic and probabilistic sensitivity analyses included 10 000 Monte Carlo iterations.

AI-guided care was associated with higher guideline compliance (82.8% vs 61.3%) and appropriate statin initiation (53.3% vs 4.7%) and lower inappropriate noninitiation (11.7% vs 35.8%; all P < .001), which were directly observed decision-level outcomes. In the model, the AI-guided strategy was projected to be dominant, yielding more QALYs (12.57 vs 11.39) at a lower lifetime cost ($123 271 vs $140 865) and an incremental cost-effectiveness ratio of -$14 972/QALY. Incremental net monetary benefit was $43 389 at 30 000 000 Korean won/QALY ($21 952), and the strategy was cost-effective in every probabilistic iteration and all examined sensitivity analyses, including sex-stratified analyses.

Linking retinal cardiovascular AI outputs to guideline-based preventive care was associated with more guideline-concordant statin decisions and, in a lifetime model, was projected to be economically dominant over usual care. Long-term outcomes were projected, not observed, and require confirmation in other settings.