Unlock the Recent Prediction Models using Clinical Variables for Diabetic Retinopathy

Abstract

Diabetic retinopathy (DR) poses a major challenge to clinician and public health personnel due to the potential complication of vision loss especially among young adults. Thus, many studies came out with artificial intelligence prediction models for diabetic retinopathy to enhance the management of potential complication. The aims of this systematic review are to identify published prediction models using clinical variables for diabetic retinopathy and to compare their accuracy and quality. A systematic search was conducted in PubMed, Scopus, ScienceDirect, ProQuest, Web of Science databases. Studies were included if the model was applicable in type I or type II diabetes mellitus and the outcome was diabetic retinopathy. The methods of Checklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS) and Prediction model risk Of Bias Assessment Tool (PROBAST) were used as a guide. Ten studies since 2019 were identified and included here. Results of each model were compared in terms of area under ROC curve (AUC), sensitivity, specificity and others. This review provides an insight about the existing DR prediction models and to foresee the future prospects. Future work includes combination of prediction models using clinical variables with fundus images of the patients to predict the area possible for development of diabetic retinopathy.