MODEL-BASED COST-UTILITY ANALYSIS OF ARTIFICIAL INTELLIGENCE-ASSISTED GENERAL PRACTITIONER-LED FOLLOW-UP MANAGEMENT FOR HYPERLIPIDEMIA IN SHANGHAI, CHINA
Author(s)
Yihan Hu, Master’s Candidate1, Ping Zhou, PhD1, Huajie Jin, PhD2, Kang Wang, MSc2, Tiantian Zhang, PhD1.
1Fudan University, Shanghai, China, 2King's College London, London, United Kingdom.
1Fudan University, Shanghai, China, 2King's College London, London, United Kingdom.
OBJECTIVES: Local policies in Shanghai support integrating lipid management into general practitioner (GP) services, but artificial intelligence (AI)-assisted models remain immature. This study assessed the cost-utility of AI-assisted GP management for primary prevention of atherosclerotic cardiovascular disease (ASCVD) among community adults with hyperlipidemia.
METHODS: A sex-stratified Markov cohort model simulated adults aged 20-59 years without ASCVD from a Chinese healthcare system perspective. Strategies were no intervention, defined as usual care outside the GP management pathway and relying on routine healthcare use and self-management; GP management; and AI-assisted GP management (GP+AI). The model used annual cycles until age 85 across hyperlipidemia states, first and recurrent myocardial infarction (MI), ischemic stroke (IS), and death. In primary prevention, management effects were modeled as improvements in statin use among eligible individuals and lifestyle modification. MI, IS, and mortality rates were adjusted using the resulting low-density lipoprotein cholesterol (LDL-C) reductions and established LDL-C-cardiovascular risk relationships. Annual per-person AI and GP costs were CNY200 and CNY50. Costs and QALYs were discounted at 3%; the willingness-to-pay threshold was CNY99,665/QALY. Sensitivity and scenario analyses assessed uncertainty.
RESULTS: No intervention cost CNY21,388 and yielded 17.78 QALYs; GP cost CNY21,006 and yielded 17.89 QALYs; GP+AI cost CNY23,844 and yielded 17.96 QALYs. GP dominated no intervention. GP+AI increased costs by CNY2,838 and QALYs by 0.071 versus GP, resulting in an ICER of CNY40,216/QALY. Lifetime first MI risks were 0.076, 0.071, and 0.068 for no intervention, GP, and GP+AI, respectively; corresponding first IS risks were 0.240, 0.226, and 0.217. ASCVD-free life-years were 28.039, 28.502, and 28.784. One-way sensitivity analyses kept incremental net monetary benefits positive. Probabilistic analysis showed GP+AI was the most cost-effective strategy in 92.68% of simulations.
CONCLUSIONS: Under current evidence and plausible assumptions, GP+AI appeared likely to be cost-effective for community hyperlipidemia management in Shanghai. Further real-world validation is needed.
METHODS: A sex-stratified Markov cohort model simulated adults aged 20-59 years without ASCVD from a Chinese healthcare system perspective. Strategies were no intervention, defined as usual care outside the GP management pathway and relying on routine healthcare use and self-management; GP management; and AI-assisted GP management (GP+AI). The model used annual cycles until age 85 across hyperlipidemia states, first and recurrent myocardial infarction (MI), ischemic stroke (IS), and death. In primary prevention, management effects were modeled as improvements in statin use among eligible individuals and lifestyle modification. MI, IS, and mortality rates were adjusted using the resulting low-density lipoprotein cholesterol (LDL-C) reductions and established LDL-C-cardiovascular risk relationships. Annual per-person AI and GP costs were CNY200 and CNY50. Costs and QALYs were discounted at 3%; the willingness-to-pay threshold was CNY99,665/QALY. Sensitivity and scenario analyses assessed uncertainty.
RESULTS: No intervention cost CNY21,388 and yielded 17.78 QALYs; GP cost CNY21,006 and yielded 17.89 QALYs; GP+AI cost CNY23,844 and yielded 17.96 QALYs. GP dominated no intervention. GP+AI increased costs by CNY2,838 and QALYs by 0.071 versus GP, resulting in an ICER of CNY40,216/QALY. Lifetime first MI risks were 0.076, 0.071, and 0.068 for no intervention, GP, and GP+AI, respectively; corresponding first IS risks were 0.240, 0.226, and 0.217. ASCVD-free life-years were 28.039, 28.502, and 28.784. One-way sensitivity analyses kept incremental net monetary benefits positive. Probabilistic analysis showed GP+AI was the most cost-effective strategy in 92.68% of simulations.
CONCLUSIONS: Under current evidence and plausible assumptions, GP+AI appeared likely to be cost-effective for community hyperlipidemia management in Shanghai. Further real-world validation is needed.
Conference/Value in Health Info
2026-09, ISPOR Asia Pacific 2026, Bangkok, Thailand
Value in Health, Volume 55, Issue S1
Code
EE1
Topic
Economic Evaluation
Disease
SDC: Cardiovascular Disorders (including MI, Stroke, Circulatory), SDC: Diabetes/Endocrine/Metabolic Disorders (including obesity)