AUTHOR=Wang Kai , Gu Longyuan , Liu Wencai , Xu Chan , Yin Chengliang , Liu Haiyan , Rong Liangqun , Li Wenle , Wei Xiu'e 

TITLE=The predictors of death within 1 year in acute ischemic stroke patients based on machine learning

JOURNAL=Frontiers in Neurology

VOLUME=Volume 14 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2023.1092534

DOI=10.3389/fneur.2023.1092534

ISSN=1664-2295

ABSTRACT=Objective: To explore the predictors of death in acute ischemic stroke (AIS) patients within one year based on machine learning (ML) algorithms.
Methods: This study retrospectively analyzed the clinical data of patients hospitalized and diagnosed as AIS in the Second Affiliated Hospital of Xuzhou Medical University between August 2017 and July 2019. Patients were randomly divided into training and validation sets at ratio of 7:3, and clinical characteristic variables of patients were screened by using univariate and multivariate logistics regression. Furthermore, six ML algorithms were applied to develop models to predict the death in AIS patients within one year, including Logistic Regression (LR), Gradient Boosting Machine (GBM), Extreme Gradient Boosting (XGB), Random Forest (RF), Decision Tree (DT) and Naive Bayes Classifier (NBC). During training, a 10-fold cross-validation approach was used to validate the training set internally and the models were interpreted using important ranking and the SHAP principle. In addition, the validation set was used to externally validate the models. Ultimately, the highest performing model was selected to build the web page calculator.
Results: Multivariate logistic regression analysis showed C-reactive protein (CRP), Homocysteine (HCY), Stroke severity (SS) and Number of stroke lesions (NOS) were independent risk factors for death within one year in AIS patients. In the prediction of death within one year, the XGB model had the highest AUC value of 0.846 among the six ML algorithms. Thus, the XGB model had the strongest performance in predicting the death. Further, a ML web calculator (https://mlmedicine-de-stroke-de-stroke-m5pijk.streamlitapp.com/) was constructed to predict the death within one year in AIS patients.
Conclutions: In this study the ML model found that serum inflammatory markers (CRP, HCY), SS, and NOS were good predictors of death within one year in AIS patients. The web calculator based on the XGB model can help clinicians make more personalised and rational clinical decisions.