Impact on blindness of organized diabetic retinopathy screening including artificial intelligence (AI) and optical coherence tomography (OCT) in urban China – a lifetime cost-effectiveness analysis (CEA)

Abstract

Purpose : Many middle-income countries including China have a scarcity of ophthalmologists and have not started organized diabetic retinopathy (DR) screening. Latest developments in DR screening such as AI-assisted grading, low-cost fundus cameras and OCT offer potential solutions in such settings. We investigated the long-term impact of such a next-generation screening programme in urban China on blindness prevention and cost-effectiveness.

Methods : Our CEA was from a government perspective. An individual Markov model simulated the natural history of 100,000 people with known diabetes (closed cohort design). Transitions were estimated from direct observations and published data sense-checked against Chinese data where available. We compared lifetime costs (2020 US dollars, 5% discount rate) and effectiveness (new DR blindness, quality-adjusted life years (QALY)) between: i) organized screening (technicians, AI photo grading, OCT, human arbitration) with 80% annual uptake; ii) current ophthalmologist-led opportunistic screening with 2% annual uptake. An incremental cost-effectiveness ratio (ICER) <1 GDP per capita (11,638 USD for China) was deemed highly cost-effective (WHO guidelines). Alternative treatment stategies for clinically significant macular edema (CSME) (laser only, antiVEGF+laser), and varying the timeline to 10 years were evaluated.

Results : Compared to current screening in China, with laser only for CSME, organized screening reduced blindness by 52.4% in the model across the lifetime and gained 0.23 QALYs with an incremental cost of 682 USD / individual. This generated an ICER of 2,942 USD/QALY gained, 0.25 times GDP per capita. With antiVEGF+laser for CSME, blindness reduced by 61.9%, gaining 0.31 QALYs with an incremental cost of 3,177 USD. The ICER was 10,259 USD/ QALY gained, 0.89 times GDP/capita. Results remained robust in sensitivity analyses.

Conclusions : Introducing systematic screening including AI and OCT in a middle-income setting with no prior organised screening appears to offer an ICER well within the WHO guidelines for effective interventions. These findings provide strong economic evidence to support the wide introduction of systematic screening in populations not currently undergoing organised screening.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

 

Table 1. Cost, effectiveness and ICERs

 

Table 2. Blindness avoided