JMIR Publications

ABSTRACT

Background:

The occurrence of delirium is a prevalent phenomenon among patients admitted to the geriatric intensive care unit (ICU), with the potential to adversely impact prognosis and augment the risk of complications.

Objective:

This study aimed to construct and validate a predictive model for postoperative delirium in elderly patients in ICUs, providing timely and effective early identification of high-risk individuals and assisting clinicians in decision-making.

Methods:

The data from patients admitted to the ICU for over 24 hours were extracted from the Medical Information Marketplace for Intensive Care IV (MIMIC-IV) database and the eICU Collaborative Research Database (eICU-CRD). The MIMIC-IV data were split into a training set and an internal validation set (7:3 ratio), while the eICU-CRD data served as an external validation set. A delirium prediction was conducted for the subsequent prediction windows (12h, 24h, 48h, and whole stay time) utilising data from the first 24 hours post-admission. The corresponding feature variables were subjected to Boruta feature selection, and the prediction models were constructed using logistic regression, support vector classifier, random forest classifier, and extreme gradient boosting (XGB). Subsequently, the model performance was evaluated using receiver operating characteristic curves, calibration curves, decision curve analysis, and external validation.

Results:

The MIMIC-IV and eICU-CRD datasets comprised 5897 and 618 patients, respectively, who were included in the analysis. A total of 57 features were selected for the construction of the predictive model. In the context of internal validation, the XGB model demonstrated the most effective prediction of delirium across different prediction windows. The Area under the curve values for the four prediction windows (12h, 24h, 48h, and whole stay time) were 0.860(95% CI: 0.839-0.880), 0.871(95% CI: 0.850-0.889), 0.851(95% CI: 0.829-0.871), and 0.846(95% CI: 0.827-0.867), respectively. The Area under the curve values for the external validation set were 0.828(95% CI: 0.768-0.880), 0.811(95% CI: 0.762-0.855), 0.756(95% CI: 0.705-0.803), and 0.750(95% CI: 0.701-0.795). Furthermore, the XGB model demonstrated the most accurate calibration across all prediction windows, with values of 0.115, 0.119, 0.136, and 0.144, respectively. Additionally, the decision curve analysis revealed that the XGB model outperformed the other models in terms of net gain for the majority of threshold probability values. The five most significant predictive features identified were the first day's delirium assessment results, invasive ventilation, Sequential Organ Failure Assessment score, minimum Glasgow Coma Scale score, and type of first care unit.

Conclusions:

The high-performance XGB model for predicting postoperative delirium in elderly ICU patients has been successfully developed and validated. The model predicts the incidence of delirium at 12h, 24h, 48h, and whole stay time after the first day of hospitalisation within ICU. This enables physicians to identify high-risk patients early, thus facilitating the optimisation of personalised management strategies and care plans.