WHAT IS THE ROLE OF HUMANS IN A WORLD OF ARTIFICIAL INTELLIGENCE: AN ECONOMIC EVALUATION OF HUMAN-AI COLLABORATION
Author(s)
Yueye Wang, PhD1, Xiaotong Han, PhD2, Mingguang He, PhD3, Lei Zhang4.
1The Hong Kong Polytechnic University, Hong Kong, China, 2Zhongshan Ophthalmic Center, Guangzhou, China, 3The Hong Kong Polytechnic University, Hong, China, 4Professor, Monash University, Carlton, Australia.
1The Hong Kong Polytechnic University, Hong Kong, China, 2Zhongshan Ophthalmic Center, Guangzhou, China, 3The Hong Kong Polytechnic University, Hong, China, 4Professor, Monash University, Carlton, Australia.
OBJECTIVES: This study aims to evaluate human-AI collaboration strategies in diabetic retinopathy (DR) screening and identify the most cost-effective strategy in clinical practice.
METHODS: From a healthcare provider’s perspective, we developed a hybrid human-AI decision tree/Markov model to simulate DR screening pathways and disease progression in China, using a hypothetical cohort of 100,000 individuals aged 18-79 years, followed over a lifespan. Compared with manual screening, we assessed the costs and effectiveness of eight potential human-AI collaborative strategies. All nine screening strategies were assessed across five age groups and six DR screening intervals, resulting in 270 screening scenarios. Main outcome measures included the incremental cost-effectiveness ratio (ICER), evaluating quality-adjusted life-years (QALYs), and cost for each blindness year averted. We used both 1- and 3-time GDP/capita (US$ 12,684 and US$ 38,052 in 2023) as the willingness-to-pay thresholds in China.
RESULTS: Annual ‘Copilot human-AI’ screening in the 20-79 age group, where AI and humans performed independent grading with disagreement reviewed by a second human grader for a final decision, was the most cost-effective among all 270 scenarios. Compared with manual screening, this strategy would save US$1.32 million while adding 49.33 blindness-free years and 14.73 QALYs in the simulated population, making it cost-saving. The health benefits of this strategy were equivalent to a net monetary benefit of US$1.88 million per 100,000 individuals. The ‘Copilot human-AI’ screening strategy remained the most cost-effective strategy across a wide range of model parameters, except in two scenarios: if the specificity of the secondary grader decreased to below 88.7%, manual screening would become the most cost-effective strategy; and if the specificity of AI exceeded 96.9%, AI-only screening would become cost-effective.
CONCLUSIONS: The ‘copilot human-AI’ screening is the most cost-effective human-AI collaborative strategy for DR screening in China. Human involvement remains essential and cost-effective in a setting where AI is highly mature and efficient.
METHODS: From a healthcare provider’s perspective, we developed a hybrid human-AI decision tree/Markov model to simulate DR screening pathways and disease progression in China, using a hypothetical cohort of 100,000 individuals aged 18-79 years, followed over a lifespan. Compared with manual screening, we assessed the costs and effectiveness of eight potential human-AI collaborative strategies. All nine screening strategies were assessed across five age groups and six DR screening intervals, resulting in 270 screening scenarios. Main outcome measures included the incremental cost-effectiveness ratio (ICER), evaluating quality-adjusted life-years (QALYs), and cost for each blindness year averted. We used both 1- and 3-time GDP/capita (US$ 12,684 and US$ 38,052 in 2023) as the willingness-to-pay thresholds in China.
RESULTS: Annual ‘Copilot human-AI’ screening in the 20-79 age group, where AI and humans performed independent grading with disagreement reviewed by a second human grader for a final decision, was the most cost-effective among all 270 scenarios. Compared with manual screening, this strategy would save US$1.32 million while adding 49.33 blindness-free years and 14.73 QALYs in the simulated population, making it cost-saving. The health benefits of this strategy were equivalent to a net monetary benefit of US$1.88 million per 100,000 individuals. The ‘Copilot human-AI’ screening strategy remained the most cost-effective strategy across a wide range of model parameters, except in two scenarios: if the specificity of the secondary grader decreased to below 88.7%, manual screening would become the most cost-effective strategy; and if the specificity of AI exceeded 96.9%, AI-only screening would become cost-effective.
CONCLUSIONS: The ‘copilot human-AI’ screening is the most cost-effective human-AI collaborative strategy for DR screening in China. Human involvement remains essential and cost-effective in a setting where AI is highly mature and efficient.
Conference/Value in Health Info
2026-09, ISPOR Asia Pacific 2026, Bangkok, Thailand
Value in Health, Volume 55, Issue S1
Code
HTA38
Topic
Health Technology Assessment
Topic Subcategory
Systems & Structure
Disease
SDC: Sensory System Disorders (Ear, Eye, Dental, Skin)