The prediction power of machine learning on estimating the sepsis mortality at intensive care unit

• Published in: Journal of critical care Vol. 81; p. 154720
• Main Authors: Selcuk, Mehtap, Koc, Oguz, Kestel, Sevtap
• Format: Journal Article
• Language: English
• Published: Philadelphia Elsevier Inc 01.06.2024; Elsevier Limited
• Subjects: Accuracy; APACHE II; Intensive care; Machine learning; Mortality; SAPS II; Sepsis; Sepsis mortality; SOFA; Sepsis mortality; SOFA; SAPS II; APACHE II; Machine learning
• ISSN: 0883-9441; 1557-8615
• DOI: 10.1016/j.jcrc.2024.154720

The prediction of sepsis mortality of intensive care unit (ICU) observations using machine learning (ML) methods are hypothesized to yield better or as good as performance compared to the prognostic scores. This paper aims to show that the accuracy of ML in mortality estimation may be superior and supportive knowledge to SAPS II, APACHE II and SOFA (traditional) scores. The retrospective collection of data from the patients (n = 200) admitted to ICU of Acibadem Hospital, Istanbul-Turkey, between 2015 and 2020 is utilized to detect the sepsis mortality risk using eight ML methods along with the traditional prognostic scores. The mortality as decisive indicator is evaluated according to the explanatory variables included to quantify the traditional scores. In the calibration of the data, five different predetermined split of the data is used for training and the validation of the ML methods. The efficiency of the prediction results of ML methods and the ICU scoring methods are investigated by AUC-ROC curves and other ML accuracy indicators. Consecutive processes of Box-Cox and Min-Max transformations on data, and parameter optimization (Grid Search) are performed to increase the efficiency of ML methods. The accuracy in the mortality prediction is achieved the best by Multi-Layer Perceptron method compared to SAPS II and APACHE II methods, and is as good the one with what SOFA predicts. Prediction accuracies (AUC) in ML methods ranges between 75% -86%, 63%–82% and 73%–81% compared to the ones obtained for these three scores, respectively. Additionally, the outcomes show that non-tree based algorithms implemented first time in this paper can be utilized as ML tool in predicting mortality even for small sample. The outcomes of this study have clinical merits in evaluating on the potential use of ML methods in predicting ICU mortality superior to traditional scores APACHE II, SAPS II and as good as SOFA. Additionally, it highlights which of the variables contributing to the risk of sepsis mortality should be taken as apriori information in treating the patients.