Development and Validation of Machine Learning–Based Models to Predict In-Hospital Mortality in Life-Threatening Ventricular Arrhythmias: Retrospective Cohort Study

• Published in: Journal of medical Internet research 2023-11, Vol.25 (11), p.e47664-e47664
• Main Authors: Li, Le, Ding, Ligang, Zhang, Zhuxin, Zhou, Likun, Zhang, Zhenhao, Xiong, Yulong, Hu, Zhao, Yao, Yan
• Format: Article
• Language: English
• Subjects: Algorithms; Blood; Calibration; Cardiac arrhythmia; Cardiomyopathy; Clinical medicine; Cohort analysis; Critical care; Decision making; Emergency medical services; Feature selection; High risk; Hospitals; Intensive care; Life threatening; Machine learning; Medical prognosis; Medical research; Medicine, Experimental; Missing data; Mortality; Myocardial infarction; Original Paper; Physiology; Prediction models; Review boards; Variables; Vital signs
• ISSN: 1438-8871; 1439-4456; 1438-8871
• DOI: 10.2196/47664

Life-threatening ventricular arrhythmias (LTVAs) are main causes of sudden cardiac arrest and are highly associated with an increased risk of mortality. A prediction model that enables early identification of the high-risk individuals is still lacking. We aimed to build machine learning (ML)–based models to predict in-hospital mortality in patients with LTVA. A total of 3140 patients with LTVA were randomly divided into training (n=2512, 80%) and internal validation (n=628, 20%) sets. Moreover, data of 2851 patients from another database were collected as the external validation set. The primary output was the probability of in-hospital mortality. The discriminatory ability was evaluated by the area under the receiver operating characteristic curve (AUC). The prediction performances of 5 ML algorithms were compared with 2 conventional scoring systems, namely, the simplified acute physiology score (SAPS-II) and the logistic organ dysfunction system (LODS). The prediction performance of the 5 ML algorithms significantly outperformed the traditional models in predicting in-hospital mortality. CatBoost showed the highest AUC of 90.5% (95% CI 87.5%-93.5%), followed by LightGBM with an AUC of 90.1% (95% CI 86.8%-93.4%). Conversely, the predictive values of SAPS-II and LODS were unsatisfactory, with AUCs of 78.0% (95% CI 71.7%-84.3%) and 74.9% (95% CI 67.2%-82.6%), respectively. The superiority of ML-based models was also shown in the external validation set. ML-based models could improve the predictive values of in-hospital mortality prediction for patients with LTVA compared with traditional scoring systems.