Prediction model of postoperative pain exacerbation using an intravenous patient-controlled analgesia device and a wearable electrocardiogram sensor

There is a need to develop objective and realtime postoperative pain assessment methods in perioperative medicine. Few studies have evaluated the relationship between pain severity and temporal changes of physiological signals in actual postoperative patients. In this study, we developed a machine learning model which was trained from intravenous patient-controlled analgesia (IV-PCA) records and electrocardiogram (ECG) of postoperative patients to predict pain exacerbation. A self-attentive autoencoder (SA-AE) model achieved 54% of sensitivity and a 1.76 times/h of false positive rate.