AUTHOR=Zhao Lina , Wang Yunying , Ge Zengzheng , Zhu Huadong , Li Yi 

TITLE=Mechanical Learning for Prediction of Sepsis-Associated Encephalopathy

JOURNAL=Frontiers in Computational Neuroscience

VOLUME=Volume 15 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/computational-neuroscience/articles/10.3389/fncom.2021.739265

DOI=10.3389/fncom.2021.739265

ISSN=1662-5188

ABSTRACT=Objective: The study aims to develop a mechanical learning model as a predictive model for 19 predicting the appearance of sepsis associated encephalopathy (SAE). 20 Materials and Methods: The prediction model was developed in a primary cohort of 2028 sepsis 21 patients from June 2001 to October 2012 who were retrieved from the MIMIC III Database. Lasso 22 regression model was used for data dimension reduction, and feature selection. The model was 23 developed using multivariable logistic regression analysis. The performance of the nomogram has 24 been evaluated in terms of calibration, discrimination and clinical utility. 25 Results: There were nine particular features in septic patients that were significantly associated with 26 SAE. Predictors of individualized prediction nomograms included age, rapid sequential evaluation of 27 organ failure (qSOFA), drugs including carbapenem antibiotics, quinolone antibiotics, steroids, 28 midazolam, H2 antagonist, diphenhydraine hydrochloride, heparin sodium injection. The area under 29 the curve (AUC) was 0.743, indicating good discrimination. The prediction model showed  sepsis-associated encephalopathy  2 This is a provisional file, not the final typeset article 30 calibration curves with minor deviations from the ideal predictions. Decision curve analysis (DCA) 31 suggested that the nomogram is clinically useful. 32 Conclusion: We propose a nomogram for the individualized prediction of sepsis associated 33 encephalopathy with satisfactory performance and clinical utility, which may aid the clinician in the 34 early detection and management of SAE.