A Machine Learning Tool to Predict Survival After First Surgery in Peripheral Artery Disease Patients

TL;DR

This study created a machine learning tool to predict survival rates in peripheral artery disease patients after their first surgery, using clinical and demographic data.

Contribution

The novel contribution is a validated machine learning model using gradient boosted decision trees for predicting long-term mortality in PAD patients.

Findings

The model achieved AUCs of 0.86, 0.84, and 0.80 for 1-, 3-, and 5-year mortality predictions.

Disease stage, age, and comorbidities were the most important predictors of survival.

Simple clinical parameters were sufficient for accurate mortality prediction.

Abstract

The aim of this study was to develop and validate a machine learning tool for predicting survival in PAD patients who received surgical treatment. We used the data from 1,615 patients who underwent PAD surgery from 2005 to 2020. Gradient boosted decision trees (GBDTs) were used to predict mortality at one, three and five years after the first surgery, while predictor importance was assessed using the SHAP values method. The area under the curve (AUC) of the receiver operating characteristic curve of the one-, three and five-year prediction models were 0.86, 0.84 and 0.80, respectively. Disease stage was the most important predictor, along with age, chronic kidney disease status, hospital length-of-stay and total number of comorbidities. Presence of dyslipidemia was slightly predictive of one- and three-year mortality. Simple clinical and demographic parameters can be used to train a GBDT model capable of predicting PAD follow-up mortality.

Figures