Developing and Validating Models to Predict Progression to Proliferative Diabetic Retinopathy.

Area of Science:

• Ophthalmology
• Medical Informatics
• Biostatistics

Background:

• Diabetic retinopathy (DR) is a leading cause of vision loss.
• Progression from nonproliferative diabetic retinopathy (NPDR) to proliferative diabetic retinopathy (PDR) necessitates timely intervention.
• Accurate predictive models are crucial for patient management and resource allocation.

Purpose of the Study:

• To develop and validate models predicting the progression of NPDR to PDR.
• To assess the impact of incorporating updated clinical information on model performance.

Main Methods:

• Retrospective cohort study utilizing electronic health record (EHR) data.
• Four survival models were developed: Cox proportional hazards, Cox with backward selection, Cox with LASSO, and Random Survival Forest.
• Models were evaluated using static and dynamic datasets, with internal (UCSF) and external (ZSFG) validation.

Main Results:

• The study included 1130 patients at UCSF and 687 at ZSFG, with progression rates of 8.1% and 4.4%, respectively.
• All models demonstrated similar internal validation performance (C-indices ~0.70).
• The Random Survival Forest model with 6-month updated data achieved the best external validation (C-index 0.76). Key predictors included insurance, age, NPDR severity, diabetic neuropathy, and HbA1c.

Conclusions:

• Developed NPDR to PDR progression models demonstrate acceptable predictive accuracy and external validity.
• Updating baseline variables with new clinical information did not consistently enhance predictive performance.
• The findings support the use of validated models for predicting diabetic retinopathy progression.