Cost-Utility Analysis of Deep Learning and Trained Human Graders for Diabetic Retinopathy Screening in a Nationwide...

Area of Science:

• Ophthalmology
• Health Economics
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) screening is crucial for preventing vision loss.
• Limited healthcare resources in low- and middle-income countries (LMICs) hinder widespread DR screening.
• Health economic evaluations are needed to guide the adoption of deep learning (DL) for DR screening in LMICs.

Purpose of the Study:

• To evaluate the cost-effectiveness of DL-based DR screening compared to trained human graders (HG) in a middle-income country (MIC) context.
• To provide an economic rationale for implementing DL-based DR screening programs in resource-limited settings.

Main Methods:

• A decision tree-Markov hybrid model was developed to simulate lifetime costs and outcomes.
• Thailand's national DR screening program was used as a model.
• Incremental Cost-Effectiveness Ratio (ICER) and sensitivity analyses were performed.

Main Results:

• DL screening was associated with cost savings of approximately US$2.70 and similar quality-adjusted life-years (QALYs) compared to HG.
• DL remained cost-effective across a range of price points and within Thailand's willingness-to-pay threshold.
• Improved treatment referral compliance with DL screening could yield additional benefits of US$20-US$50.

Conclusions:

• DL-based DR screening may offer societal cost-savings and comparable health outcomes to HG in MICs.
• DL presents an economically viable alternative to address shortages of skilled healthcare professionals for primary DR screening.
• Findings support the expansion of DL for DR screening in LMICs with high diabetes prevalence and low treatment referral compliance.