Machine learning model based on routine blood and biochemical parameters for early diagnosis of diabetic kidney disease - PubMed
8 days ago
- #diabetic kidney disease
- #machine learning
- #early diagnosis
- Machine learning model developed for early diagnosis of diabetic kidney disease (DKD) using routine blood and biochemical parameters.
- Study included 3,114 diabetic patients for development and 1,496 from NHANES for external validation.
- Early DKD defined as UACR 30-300 mg/g with eGFR ≥60 ml/min/1.73m².
- Logistic regression model performed best with AUC = 0.689, sensitivity=40.5%, specificity=81.3%.
- Top predictors identified: triglyceride-glucose index (TyG), gender, creatinine, globulin, and age.
- External validation confirmed key associations for HbA1c, globulin, TyG, and neutrophil-to-albumin ratio.
- Model demonstrates potential as a cost-effective screening tool for early DKD diagnosis.