Postoperative Delirium Prediction Based on Preoperative Electrocardiogram and Electroencephalogram

Preventing postoperative delirium is of great importance because its onset causes significant social and health problems. We aimed to develop a model to predict the postoperative delirium using data measured preoperatively with electrocardiogram (ECG) sensors and simplified electroencephalogram (EEG) devices in a less invasive situation. The data were collected from patients undergoing oral surgery, specifically from 59 patients (including 8 who developed delirium) from 9 AM until 1 PM and 56 patients (including 8 who developed delirium) from 4 PM until bedtime. We developed binary classification models using support vector machine (SVM) and random forest (RF). We tested three patterns of input variables: EEG indicators, heart rate variability (HRV) indicators, and both. Finally, the SVM model with both indicators was able to predict with a sensitivity of 0.63 and a specificity of 0.69.