Résumé
Objective:To develop a risk prediction model for early cardiac arrest in emergency sepsis utilizing a machine learning algorithm to enhance the quality and efficiency of patient treatment.Methods:This study focused on patients with sepsis who received treatment at the emergency room of the First Medical Center of Chinese PLA General Hospital from January 1, 2020 to June 1, 2023. The basic clinical characteristics such as vital signs and laboratory results were collected. Patients who fulfilled the specified inclusion criteria were allocated randomly into a training group and a testing group with a ratio of 8:2. A CatBoost model was constructed using Python software, and the prediction efficiency of the model was assessed by calculating the area under the receiver operating characteristic curve (AUC). Furthermore, the performance of the model was compared to that of other widely employed clinical scores.Results:This study included a cohort of 2 131 patients diagnosed with sepsis, among whom 449 experienced cardiac arrest. The CatBoost model demonstrated an AUC of 0.760, surpassing other scores. Notably, the top 10 predictors in the model were identified as age, lactate, interleukin -6, oxygen saturation, albumin, N-terminal pro-B-type natriuretic peptide, potassium, sodium, creatinine, and platelets.Conclusions:The utilization of this machine learning algorithm-based prediction model offers a more precise basis for predicting cardiac arrest in emergency sepsis patients, thereby potentially improving the treatment efficacy for this disease.