Development and External Validation of a Clinically Applicable Model for Identifying Diabetic Kidney Disease in Population-Based and Clinical Cohorts.

Background: Diabetic kidney disease (DKD) is a leading cause of chronic kidney disease worldwide. Early identification of DKD remains challenging in routine clinical practice. Objective: We aimed to develop and externally validate a clinically accessible model integrating metabolic and renal biomarkers for identifying DKD among individuals with diabetes. Methods: Data from 3494 diabetic participants in the China Health and Retirement Longitudinal Study (CHARLS) were utilized for model development. Independent predictors were identified via multivariable logistic regression. To ensure robust generalizability, external validation was conducted in two independent cohorts: a national survey cohort from the National Health and Nutrition Examination Survey (NHANES) and a real-world clinical cohort from our hospital. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves and decision curve analysis (DCA). Results: The final model incorporated age, gender, the triglyceride-glucose (TyG) index, blood urea nitrogen and Cystatin C. In the CHARLS cohort, the model demonstrated robust discriminative power (AUC = 0.868) and excellent calibration. At the optimal cutoff (8.02%), the negative predictive value (NPV) reached 98.43%. In the NHANES validation, the model maintained high discriminative ability (AUC = 0.82). Crucially, in the hospital-based clinical cohort, the model demonstrated stable performance (AUC = 0.781, 95% CI: 0.648-0.888). DCA confirmed a consistent and significant net clinical benefit across a wide threshold range (0.2 to 0.8) in real world. Conclusion: The proposed model may serve as an adjunctive tool for identifying individuals with diabetes who have a high probability of prevalent DKD and who may therefore benefit from confirmatory kidney assessment.