Machine learning algorithm for early detection of end-stage renal disease.

Area of Science:

• Nephrology
• Artificial Intelligence
• Health Informatics

Background:

• End-stage renal disease (ESRD) is the most severe stage of chronic kidney disease (CKD).
• Delayed recognition and treatment of CKD etiologies contribute to disease progression.
• Accurate prediction of ESRD is crucial for timely medical intervention.

Purpose of the Study:

• To develop a machine learning model for predicting progression to ESRD.
• To utilize a large-scale, multidimensional patient database for model development.
• To identify key predictors of CKD progression to ESRD.

Main Methods:

• Analysis of 10 million medical insurance claims from 550,000 patients with CKD Stages 1-4.
• Feature engineering included demographics, chronic conditions, diagnoses, procedures, medications, costs, and episode counts.
• Word2Vec for temporal feature extraction and XGBoost for predictive modeling.

Main Results:

• The XGBoost model achieved a C-statistic of 0.93, with high sensitivity (0.715) and specificity (0.958).
• The model demonstrated strong predictive power, with a Negative Predictive Value (NPV) of 0.981.
• Key predictors included comorbidities like chronic heart disease, patient age, and hypertensive crisis events.

Conclusions:

• Machine learning models can effectively predict ESRD progression.
• Early identification of high-risk patients can trigger timely nephrology referrals.
• Proactive management initiated through electronic alerts can improve patient care and outcomes.