Explainable machine learning algorithm to predict cardiovascular event in patients undergoing peritoneal dialysis

Area of Science:

• Nephrology and Cardiovascular Medicine
• Biostatistics and Machine Learning

Background:

• Cardiovascular events (CVE) are a major cause of morbidity and mortality in patients undergoing peritoneal dialysis (PD).
• Accurate prediction of CVE risk is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To compare the predictive performance of machine learning algorithms (extreme gradient boosting and random survival forest) against Cox proportional hazard regression for cardiovascular events in PD patients.
• To identify key predictors of CVE in this patient population.

Main Methods:

• A cohort of 318 patients undergoing PD catheterization was retrospectively analyzed.
• Patients were randomly assigned to training (70%) and validation (30%) sets.
• Cox regression, XGBoost, and RSF models were developed and validated using time-dependent area under the curve (AUC) and concordance index (C-index).

Main Results:

• The random survival forest (RSF) model demonstrated superior predictive performance with a C-index of 0.725 and favorable time-dependent AUC values (1-year: 0.812, 3-year: 0.836, 5-year: 0.706) in the validation set.
• Key predictors identified included platelet count, age, 4-hour dialysate to creatinine ratio (4hD/Pcr), left atrium diameter, and left ventricular diameter.
• A significant difference in cumulative CVE-free survival was observed between high-risk and low-risk groups stratified by the RSF model.

Conclusions:

• The random survival forest (RSF) model offers a robust and potentially superior method for assessing cardiovascular event risk in patients on peritoneal dialysis.
• This machine learning approach can aid clinicians in stratifying patients and implementing personalized preventive strategies.