This project is the first to evaluate the ability of wearable accelerometer-based sensors to predict subjective self-reported pain levels in critically ill patients. This study shows that although pain episodes can be distinguished from those without pain, the models fail to classify different pain levels based on wearable information in this patient population.
Quantification of pain in patients admitted to intensive care units (ICUs) is challenging due to the increased prevalence of communication barriers among this patient population. Previous research has posited a positive correlation between pain and physical activity in critically ill patients. In this study, we advance this hypothesis by constructing machine learning models to examine the ability of accelerometer data collected from daily wearables to predict the level of pain experienced by ICU patients. We evaluated four methods, three of them deep learning. The best model was a Convolutional Neural Network (CNN) encoder with Long Short Term Memory (LSTM) projector that achieved 68.37% Area Under the Curve (AUC) and 72.19% Precision when classifying pain versus no pain.
Data privacy requirements of the Institutional Review Board prevent us from sharing data or code related to data input to the models. Therefore, the data and part of the code supporting this study's findings are restricted.
Don't hesitate to contact us with any questions about our research work.
Mrs. Jessica Sena (jessicasena@ufmg.br)