Cost-effectiveness of artificial intelligence model for glaucoma screening in Thailand

Objective This study evaluated the cost-effectiveness of an artificial intelligence (AI)-based glaucoma screening strategy compared with standard care in Thailand. Design Markov model-based cost-utility analysis. Participants A hypothetical cohort of 1,000 asymptomatic Thai individuals aged 50 years was modeled. Methods From a societal perspective, a Markov model-based economic evaluation was conducted using local and international data to estimate the costs and quality-adjusted life-years (QALYs) associated with glaucoma screening strategies. The AI screening strategy simulated annual fundus photography with AI assessment, whereas the standard-care strategy involved opportunistic detection through routine healthcare visits. Incremental cost-effectiveness ratios (ICERs) and net monetary benefits (NMBs) were calculated. Extensive sensitivity analyses were performed. Results Over a 30-year time horizon, annual AI-based glaucoma screening was associated with a total expected cost of 14.077 million Thai Baht (THB) or 423,108 United States dollars (USD), compared with 11.088 million THB or 333,257 USD under standard care. The expected health outcomes were 14,710 QALYs for the AI-based strategy and 14,684 QALYs for standard care. The ICER was 115,298 THB (3,466 USD) per QALY gained with an incremental NMB of 1.159 million THB (34,835 USD), which is considered cost-effective at the Thai willingness-to-pay threshold. The annual AI-based screening strategy remained cost-effective when the total screening cost was below 336 THB (10 USD) per test or when the AI specificity exceeded 75%. Given that routine screening visits and fundus photography costs were assumed to be constant, the cost of the AI software should not exceed 183 THB (6 USD) per screening to maintain cost-effectiveness. Conclusions AI-based glaucoma screening is likely to be a cost-effective strategy in Thailand, particularly when AI implementation costs are controlled and the diagnostic specificity is sufficiently high. Therefore, it represents a promising strategy for early detection, with the potential to reduce the burden of visual impairment attributable to glaucoma.