Development and Validation of a Machine Learning Algorithm for Predicting Diabetes Retinopathy in Patients With Type 2 Diabetes: Algorithm Development Study

• DOI: 10.2196/58107
• Publication Date: 2025-02-10T02:20:36Z
• Authors: Sunyoung Kim; Jaeyu Park; Yejun Son; Hojae Lee; Selin Woo; Myeongcheol Lee; Hayeon Lee; Hyunji Sang; Dong Keon Yon; Sang Youl Rhee
• Keywords: Dyslipidemia; Medical record

Abstract Background Diabetic retinopathy (DR) is the leading cause of preventable blindness worldwide. Machine learning (ML) systems can enhance DR in community-based screening. However, predictive power models for usability and performance are still being determined. Objective This study used data from 3 university hospitals South Korea to conduct a simple accurate assessment ML-based risk prediction development that be universally applied adults with type 2 diabetes mellitus (T2DM). Methods was predicted using independent electronic medical records: discovery cohort (one hospital, n=14,694) validation (2 hospitals, n=1856). The primary outcome presence at years. Different were selected through hyperparameter tuning cohort, area under receiver operating characteristic (ROC) curve analyzed both cohorts. Results Among 14,694 patients screened inclusion, 348 (2.37%) diagnosed DR. For DR, extreme gradient boosting (XGBoost) system had an accuracy 75.13% (95% CI 74.10‐76.17), sensitivity 71.00% 66.83‐75.17), specificity 75.23% 74.16‐76.31) original dataset. datasets, XGBoost 65.14%, 64.96%, 65.15%. most common feature model dyslipidemia, followed by cancer, hypertension, chronic kidney disease, neuropathy, cardiovascular disease. Conclusions approach shows potential patient outcomes enabling timely interventions T2DM, improving our understanding contributing factors, reducing DR-related complications. proposed expected competitive cost-effective, particularly care settings Korea.