Machine learning-based prediction for outcomes of cardiac arrest in intensive care units: Model Development and Validation Study (Preprint)

• DOI: 10.2196/preprints.45741
• Publication Date: 2023-02-05T01:04:15Z
• Authors: Peifeng Ni; Sheng Zhang; Yan Xiao; Weidong Zhang; Huan Yu; Chenghao Wu; Hongwei Zhang; Ying Zhu; Mengyuan Diao; Wei Hu
• Keywords: Preprint; Feature Engineering; Feature (linguistics)

<sec> <title>BACKGROUND</title> Cardiac arrest (CA) is a global public health challenge. Accurate prediction of outcomes critical aspect the management patients after CA. This study will develop and validate an applicable machine learning (ML) model to predict in-hospital mortality CA in intensive care units. ICU. </sec> <title>OBJECTIVE</title> The aims models <title>METHODS</title> Patients with were extracted from Medical Information Mart for Intensive Care (MIMIC)-IV database, further divided into training set (80%) validation (20%). primary outcome was mortality. best selected 11 ML algorithms 3 time periods 24 hours, 48 72 hours. SHapley Additive exPlanations (SHAP) applied visualize importance features, while recursive feature elimination (RFE) performed figure out key features. optimal compact developed based on its performance proven set. In addition, Web-based tool designed this clinical practice. <title>RESULTS</title> 721 included study, dividing (80%, n=576) (20%, n=145). 72-hour CatBoost obtained highest area under receiver operating characteristic (AUROC) 0.839. Thirteen variables ultimately as each visualized by SHAP. achieved greatest AUROC 0.862 validation, which better than other SOFA (AUROC: 0.650). convenient clinicians use our model. <title>CONCLUSIONS</title> had great patients.