Accepted for/Published in: JMIR Medical Informatics
Date Submitted: Mar 1, 2025
Open Peer Review Period: Feb 28, 2025 - Apr 25, 2025
Date Accepted: Sep 26, 2025
(closed for review but you can still tweet)
Early Prediction of Delirium in Post-Cardiac Surgery Patients: Machine Learning Model Development and External Validation
ABSTRACT
Background:
Delirium is one of the complications with higher incidence after cardiac surgery, accounting for about 10% -40%.There are studies predicting severe mortality after cardiac surgery, and few studies reporting machine learning predicting the occurrence of cardiac surgery.
Objective:
The objective of this investigation is to develop and verify a variety of machine learning algorithms (ML) that are capable of predicting delirium in patients who have undergone cardiac surgery in the intensive care unit (ICU).
Methods:
We collected patient data from the Medical Information Mart for Intensive Care Database-IV (MIMIC-V 2.0) and the EICU Collaborative Research Database (EICU-CRD). The MIMIC-IV dataset was divided into a training set. Furthermore, the EICU-CRD dataset was specifically chosen as the external verification set. The predictive criteria encompassed the demographic characteristics upon admission to the hospital, the existence of conjunctive dysfunction, vital signs, laboratory analysis, scoring systems and specifics of treatment. The study outcome was established as delirium. Delirium was set as the study outcome. We assessed our models using stratified repeated cross-validation with four algorithms: logistic regression (LR), eXtreme Gradient Boosting (XGB), Support Vector Machine (SVR), and Random Forest Classifier (RFC). The model's performance is evaluated using metrics such as the area under the receiver operating characteristic curve (AUC), accuracy (ACC), sensitivity, precision, and F1 score
Results:
The XGB model achieved the greatest Area Under the Curve (AUC) score of 0.801. Additionally, using the XGB model, we identified age, ICU length of stay (ICU los), weight, heart rate, respiration rate, and urine output as the most significant predictors.
Conclusions:
The XGB model has significant potential for developing early warning systems for patients after cardiac surgery, as it outperforms the other three machine learning models in predicting delirium after cardiac surgery.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.