Abstract
Background
The incidence rate, mortality rate and readmission rate of acute heart failure (AHF) are high, and the in-hospital mortality of AHF patients in ICU is higher. However, there is no method to accurately predict the mortality of AHF patients at present.
Methods
The Medical Information Mart for Intensive Care Ⅳ (MIMIC-Ⅳ) database was used to perform a retrospective study. Patients meeting the inclusion criteria were identified from the MIMIC-Ⅳ database and randomly divided into training set (n = 3580, 70%) and validation set (n = 1534, 30%). The variates we collected include demographic data, vital signs, comorbidities, laboratory test results and treatment information within 24 hours of ICU admission. By using the Least Absolute Shrinkage and Selection Operator (LASSO) regression model in the training set, we screened variates that affect the in-hospital mortality of AHF patients. Subsequently, in the training set, five common machine learning (ML) algorithms were applied to construct models using variates selected by LASSO to predict the in-hospital mortality of AHF patients. We evaluated the predictive ability of the models by sensitivity, specificity, accuracy, the area under the curve (AUC) of receiver operating characteristics (ROC), and clinical net benefit in the validation set. In order to obtain a model with the best predictive ability, we compared the predictive ability of common scoring systems with the best ML model.
Results
Among the 5114 patients, in-hospital mortality was 12.5%. By comparing AUC, the XGBoost model had the best predictive ability among all ML models, and the XGBoost model was chosen as our final model for its higher net benefit. Meanwhile, its predictive ability is superior to common scoring systems.
Conclusions
The XGBoost model can effectively predict the in-hospital mortality of AHF patients admitted to the ICU, which may assist clinicians in precise management and early intervention of patients with AHF to reduce mortality.