Nomogram-Based Chronic Kidney Disease Prediction Model for Type 1 Diabetes Mellitus Patients Using Routine...

Area of Science:

• Nephrology
• Endocrinology
• Data Science

Background:

• Type 1 diabetes mellitus (T1DM) poses a significant risk for developing chronic kidney disease (CKD).
• CKD often remains asymptomatic in T1DM patients, leading to delayed diagnosis as routine checkups may not include specific CKD tests.
• Early detection of CKD in T1DM is crucial for timely management and preventing disease progression.

Purpose of the Study:

• To develop and validate a predictive model for CKD in T1DM patients.
• To utilize readily available data from routine checkups for early CKD detection.
• To create a nomogram for simplified clinical application of the CKD prediction model.

Main Methods:

• Utilized longitudinal data from 1375 T1DM patients from the EDIC clinical trials (16 years, 28 sites).
• Applied feature ranking algorithms (XGB, RF, ERT) to 17 routinely available features.
• Developed a multivariate logistic regression model using top-ranked features and validated its performance.

Main Results:

• Identified hypertension, diabetes duration, drinking habit, triglycerides, ACE inhibitors, LDL cholesterol, age, and smoking habit as key predictors.
• The final multivariate logistic regression model achieved 90.04% accuracy on internal data and 88.59% on test data.
• A nomogram was generated for practical clinical use.

Conclusions:

• The developed prediction model demonstrates excellent performance for identifying CKD in T1DM patients.
• The model facilitates early CKD detection during routine patient checkups.
• This tool can significantly improve CKD management in the T1DM population.