Abstract
Continuous kidney replacement therapy (CKRT) is an essential treatment for uncontrolled severe metabolic acidosis. However, CKRT can increase workload and lead to complications, thus necessitating its selective application to patients who stand to benefit significantly. This study aimed to investigate the therapeutic effect of CKRT in patients with severe acidosis by utilizing a deep learning-based causal inference model to assess the potential impact of CKRT on the risk of in-hospital mortality. The Medical Information Mart for Intensive Care III (MIMIC-III) database was utilized to select patients with available data collected within the first 48 h after intensive care unit (ICU) admission. Patients who experienced severe acidosis with a pH < 7.2 within the initial 48 h were selected. Treatment was defined as the application of CKRT within 48 h of ICU admission, and the outcome was defined as in-hospital mortality. The dataset was randomly divided at an 85:15 ratio for the training and test datasets. The Generative Adversarial Nets for Inference of Individualized Treatment Effects model was used to train the model, and the model performance was evaluated on the test dataset. In the training dataset, the model generated values of the accuracy and area under the receiver operating characteristic curve as 0.883 and 0.887 (0.880-0.893), respectively, while in the test dataset, it showed 0.841 and 0.824 (0.804-0.843), respectively. The average probability change in in-hospital mortality due to CKRT treatment was predicted to increase by 15% and 14% in the training and test datasets, respectively. However, in patients who received CKRT, the application of CKRT resulted in an average reduction in the mortality risk of 13% in both the training and test datasets. Developing a model that strategically represents the therapeutic effects of CKRT for individual patients could aid decision-making in patients with severe acidosis.