Integrating Cardiovascular Risk Screening into Routine Diabetic Retinopathy Screening: Cost-Effectiveness of a Retinal Imaging–Based Artificial Intelligence System (JARVIS) in Singapore
Background: People with diabetes have increased risk of major adverse cardiovascular events (MACE). Conventional cardiovascular risk tools, such as the Framingham Risk Score (FRS), may have limited accuracy in Asian populations. JARVIS (Joint Automated Retinal and Vascular Imaging System) is a retinal imaging-based artificial intelligence approach for cardiovascular risk stratification. We evaluated the cost-effectiveness of JARVIS versus FRS among adults with diabetes in Singapore. Methods: A five-state Markov model estimated costs and quality-adjusted life-years (QALYs) over 20 years from the Singapore healthcare system perspective. Adults with diabetes erwent screening with JARVIS or FRS. True-positive and false-positive individuals received preventive treatment. Inputs were derived from published literature, Singapore cost sources, and age-weighted estimates of high MACE risk. Parameter uncertainty was assessed using 1,000 Monte Carlo simulations. The willingness-to-pay threshold was S$45,000 per QALY gained. Findings: JARVIS had higher costs and greater health benefits than FRS with an an incremental cost-effectiveness ratio of S$4,536 per QALY gained. Its probability of being cost-effective was 88%. With confirmatory coronary artery calcium scoring for test-positive individuals, JARVIS remained cost-effective (S$13,293 per QALY gained). In 100,000 individuals, JARVIS was projected to prevent 4,727 MACE events and avert 887 deaths over 20 years. Interpretation: JARVIS is likely cost-effective for cardiovascular risk screening among adults with diabetes in Singapore. Integration within retinal screening pathways may improve early risk detection, reduce cardiovascular events, and provide good value for money.