Risk prediction models for diabetic retinopathy: a systematic review

Area of Science:

• Ophthalmology
• Endocrinology
• Biostatistics

Background:

• Diabetic retinopathy is a significant complication of diabetes mellitus, posing a growing public health challenge.
• Predictive models are crucial for early identification of high-risk patients, enabling timely interventions.

Purpose of the Study:

• To systematically evaluate existing prediction models for diabetic retinopathy.
• To identify superior models for clinical application and guide future model development.

Main Methods:

• A comprehensive literature search was performed across major databases (PubMed, Web of Science, Embase, Cochrane Library) up to November 14, 2023.
• Risk of bias and applicability of included models were assessed using the PROBAST tool.

Main Results:

• Fifteen studies on diabetic retinopathy prediction models were included, with model performance (Area Under the Curve) ranging from 0.700 to 0.960.
• All 15 models exhibited a high risk of bias, though five demonstrated better applicability.
• Common predictors included diabetes duration, age, HbA1c, serum creatinine, and urine albumin-to-creatinine ratio.

Conclusions:

• Despite some predictive capability, the high risk of bias in current diabetic retinopathy models is a concern.
• Future studies need to prioritize transparency, large-scale machine learning, and multicenter external validation.
• This systematic review is registered with PROSPERO (CRD42023483749).