This study evaluated the cost-effectiveness of DR screening based on human assessor and AI grading in urban and rural areas in China. Apart from comparing the different grading accuracy of two screening strategy, our model also considered the difference in compliance and cost between the urban and rural areas. Our results suggested that the 2 screening strategies are both cost-effective, and the telemedicine-based screening with AI grading dominates the cost-effectiveness as compared to human assessor-based grading according to the threshold of 3 times per capita GDP of Chinese in 2022. As the result, the urban area shows less cost and higher QALY than the rural area, which could be most likely caused by more compliance in opportunistic case finding, screening, and treatment. With the progression of screening and treatment, it is assumed that the compliance would increase, and cause the difference in ICER. With the condition of same utility, the results could be most likely influenced by different screening cost and sensitivity and specificity of different screening strategies. The screening strategy and model design are basically depending on the standard of DR stages and the screening intervals. Promotion across the country can save the costs and resources, and reduce the occurrence of DR.
Assuming the model primarily reflecting the actual situation, the prevalence and transition probability were selected and calculated from the Lifeline Express Diabetic Retinopathy Screening Program in China. The screening initiative and the AI grading system were adhered to the grading criteria outlined by the English National Screening Programme[17]. Our findings could potentially apply to other screening program aligned with same grading standard.
Based on the different geographic and economic landscapes between the rural and urban areas, we demonstrated superior cost-effectiveness in telemedicine screening with AI-based grading. The ICER associated with the two telemedicine screening strategies in rural area is consistently below three times of the per capita GDP, while, in urban settings, it is below one time of the per capita GDP. The telemedicine-based screening is cost-effective because: First, the geographic, substantial population, and logistic challenges in remote areas of China are huge. With insufficient of medical resources, traditional screening methods rely on substantial workforces. Second, telemedicine-based screening can reduce the expenses in transportation and the wage loss due to time constraints so that the financial burden on individuals can be relieved. In urban area, the heightened awareness of general healthcare and willingness to receive regular ophthalmic examinations exists, but the wait times and demanding work schedules often impede regular hospital visits. The convenience and time-saving in telemedicine-based screening cater to the demands for comprehensive eye healthcare management.
In previous studies, fundus photography and telemedicine-based DR screening were compared in the family physician-based DR screening;[10, 31, 32] yet, both strategies graded the fundus photos highly depending on human assessors. With the development of digital devices, algorithms, and technologies, the cost of telemedicine-based screening with AI-based grading is low and with high diagnostic accuracy, which can assist real-time DR screening. In this study, we provided more evidences on the cost-effectiveness of telemedicine-based AI DR screening. Recently, several studies that AI-based DR screening was recommended in rural area, but not recommended in urban area while combining telemedicine technique.[33, 34]Previous studies showed that combining screening AMD and DR strategy and combining screening of multiple eye diseases are cost-effective in rural and urban China, [35, 36] suggesting that other modes of screening can help to improve the DR screening programs.
Charging a small out-of-pocket cost would reduce the willingness of potential DR patients to participate in screening, especially among low and middle-income populations. [37] Due to huge population and finite healthcare budget and resources in China, implementing free DR screening service would be unfeasible. Yet, the AI-based screening will considerably overcome the shortage of ophthalmologists across China. AI-based screening could potentially be promoted to other nations, with lacking of medical resources, human assessors, and routine DR screening programs.
Study limitations
First, the transition probabilities and the utility values were partly derived from the Asian studies other than China, which might be not same as that in China. Second, we simplified the national disease grading system according to the actual situation. Third, the benefits of the detection and treatment of DME were not considered in this study. Furthermore, we assumed that the direct costs were uniform across the urban and rural settings. Future investigations should consider these areas.