Abstract Visual impairment has been recognized as a potential risk factor for depressive symptoms (DS) in diabetes patients, yet the role of visual function in predicting DS remains unexplored. This study aims to develop and validate a predictive model for DS risk in type 2 diabetes mellitus (T2DM) patients in community health settings, incorporating a visual function index (VF14). We conducted a cross‐sectional study involving 542 T2DM patients from four community health centers in Guiyang. Univariate and multivariate logistic regressions identified significant predictors, while 10 machine learning algorithms were employed to construct the predictive model. Model performance was assessed using such metrics as receiver operating characteristic curves, accuracy, sensitivity, specificity, F1 score, Brier score, C‐index, calibration curves, and decision curve analysis. A restricted cubic spline (RCS) analysis evaluated the score‐dependent risk profiles between the VF14 and DS. Key predictors included body mass index (BMI), self‐reported glycemic status, age‐related macular degeneration, glycated hemoglobin (HbA1c), and VF14. Among the models, the gradient boosting machine exhibited the robust predictive performance, with an area under the curve of 0.73 and sensitivity of 0.72. The Shapley additive explanations analysis identified VF14, BMI, and HbA1c as the top risk factors. RCS analysis revealed a score‐dependent risk profile between VF14 and DS risk. This study introduces a clinically interpretable tool for early DS risk stratification in T2DM patients, offering potential for improved risk assessment and timely intervention in community health settings.
Sun et al. (2026) studied this question.