Machine learning algorithm model for predicting heart failure and all-cause death in peritoneal dialysis patients (Preprint)

UNSTRUCTURED Objective The study presented here aimed to establish a predictive model for heart failure and all-cause mortality in PD patients with machine learning algorithm. Methods The baseline clinical variables of 606 patients on PD were collected. We implemented machine learning algorithm to construct the models to predictive the risk of the heart failure and all-cause mortality. In addition, the prediction performance of machine learning methods and cox regression was compared. Results Over a median follow-up of 49 months.298 patients developed heart failure required hospitalization. According to randomforest model (AUC=0.82), the risk factors were cci2, SBP, BMI, etc. 79 patients developed heart failure during the first year follow-up. According to randomforest (AUC=0.82) model, the risk factors were BMI, age, SBP, etc. 246 patients developed heart failure during 5-year follow-up. According to xgboost (AUC=0.81) model, the risk factors were cci score, BMI, SBP，etc. The machine learning model (AUC =0.885) demonstrated better discrimination than cox regression (C-Index=0.77) for heart failure. Regarding all-cause mortality, 199 patients died during the follow-up. According to randomforest model (AUC=0.86), the risk factors predicting death were age, cci score, creatinine, etc. 64 patients died during the first year follow-up. According to xgboost model (AUC=0.84), the predictive risk factors for 1-year all-cause mortality were age, HDL-C, Total cholesterol, etc. A total of 161 patients died during 5-year follow-up. According to randomforest model (AUC=0.84), the predictive risk factors were age, cci score, eGFR, etc. The machine learning model (AUC=0.83) demonstrated better discrimination than cox regression (C-Index=0.79) for all-cause mortality. Conclusion We developed a novel method to predict the risks of heart failure and all-cause mortality on PD patients that integrates readily available clinical, laboratory variables.