To develop and validate a machine learning model using routine clinical parameters to predict progression in type 2 diabetic kidney disease (T2DKD), addressing the need for accessible predictive tools in primary care.
Key Findings:
Core predictors identified: metabolic dysfunction-associated steatotic liver disease (MASLD), aspartate aminotransferase (AST), diabetic peripheral neuropathy (DPN), and age, which are critical for clinical assessment.
The neural network model achieved an AUC of 0.742, indicating good discrimination, satisfactory calibration (Hosmer–Lemeshow P = 0.2020), and the lowest Brier score (0.2105), suggesting reliable predictions.
SHAP analysis confirmed stable feature importance rankings and revealed synergistic interactions among MASLD, AST, and DPN, highlighting their combined effect on T2DKD progression.
Interpretation:
The NN model effectively predicts T2DKD progression risk using routine clinical indicators, potentially guiding early interventions in clinical settings.
Limitations:
Single-center study may limit generalizability to broader populations.
Retrospective design may introduce bias, affecting the reliability of the findings.
Conclusion:
The NN model is a cost-effective tool suitable for clinical practice and community health services, offering a scalable solution for early intervention and necessitating further validation in diverse populations.
