Development and validation of a predictive model for diabetic kidney disease risk in patients with T2DM: a hospital...

This study used clinical data from patients with type 2 diabetes mellitus (T2DM) and applied least absolute shrinkage and selection operator (LASSO) regression to identify risk factors for diabetic kidney disease (DKD). We then constructed a nomogram prediction model to support early clinical screening of high-risk populations.

Clinical data from patients with T2DM were collected from January 2020 to December 2025. Data from January 1, 2020, to December 31, 2023, were used as the training set for LASSO-based variable selection, model development, and internal validation. Data from January 1, 2024, to December 31, 2025, were used as a temporal internal validation set. Variables selected by LASSO regression were entered into a multivariable logistic regression model, and a nomogram was constructed from the regression results. Discrimination was assessed using the area under the receiver operating characteristic curve (AUC). Calibration was evaluated using calibration curves, and clinical net benefit was assessed using decision curve analysis (DCA). Internal validation was performed with 1,000 bootstrap replicates.

The retrospective study included 23,152 patients with T2DM, of whom 5,019 (21.68%) had DKD and 18,133 (78.32%) did not. LASSO regression identified nine candidate predictors: hypertension, diabetes duration, HbA1c, estimated glomerular filtration rate, urine protein, serum creatinine, uric acid, total cholesterol, and homocysteine. The nomogram showed moderate discrimination, with an AUC of 0.773 (95% CI: 0.764-0.782). In the temporal internal validation set, the AUC was 0.758 (95% CI: 0.743-0.774), indicating similar performance over time within the same hospital system. The calibration curves for both validation procedures were close to the diagonal, indicating agreement between predicted and observed probabilities.

Based on real-world clinical data, this study used LASSO regression to identify risk factors associated with DKD in patients with T2DM. We developed a nomogram that integrates multidimensional predictors to estimate individualized DKD risk. Internal validation and temporal internal validation showed acceptable discrimination, calibration, and clinical net benefit. The model may serve as an auxiliary decision-support tool for early screening of patients at high risk of DKD in clinical practice. Because the AUC values indicate moderate rather than strong discrimination, multicenter external validation is still needed before broad implementation. The model is intuitive, practical, and cost-effective, and it may help clinicians identify high-risk patients who warrant closer monitoring or intensive intervention.