Machine learning identifies clinical and sociodemographic factors contributing to post-COVID hospitalization and cardiovascular risk in patients with peripheral artery disease.

This study aims to assess the impact of preexisting peripheral arterial disease (PAD) on adverse outcomes-including hospitalization and cardiovascular events-in COVID-19 patients, and to identify risk factors associated with these outcomes using both multivariate logistic regression (MLR) and machine learning (ML) models. The data were extracted from the TriNetX Research Network on February 14 2023, including 2,354 matched COVID-19 patients diagnosed from January 1, 2020 to July 7, 2021, evenly divided between those with and without PAD. The primary outcomes measured included mortality, ventilation, and hospitalization, while secondary outcomes consisted of stroke, dysrhythmia, myocardial infarction, thrombosis, pneumonia, acute pulmonary edema, and pulmonary embolism. Statistical analyses were conducted using MLR, and the performance of ML models, specifically XGBoost, was compared using AUROC values. Shapley values were also calculated to identify key predictors for hospitalization. COVID-19 patients with PAD had a significantly higher risk of hospitalization (OR 1.2170, p = 0.0448) and cardiovascular events, including stroke (OR 1.5250, p = 0.0067) and dysrhythmia (OR 1.5650, p = 0.0494), compared to those without PAD. In addition to PAD, Type II diabetes mellitus and non-White race also emerged as significant independent contributors to hospitalization risk based on machine learning analyses. Both MLR and XGBoost models provided comparable predictive performance, with AUROC values of 0.6651 for XGBoost and 0.6526 for MLR in predicting hospitalization. Shapley value analysis revealed PAD and a history of smoking amongst the most important predictors for hospitalization. Patients with preexisting PAD are at a significantly higher risk for hospitalization and adverse cardiovascular outcomes within 30 days of COVID-19 diagnosis. These findings underscore the need for early intervention and risk stratification in PAD patients to mitigate the effects of COVID-19. The study also highlights the utility of ML models in identifying key risk factors, aiding in precise clinical decision-making.