Real-time Stress Monitoring for Intensive Care Unit (ICU) Nurses

Abstract

The aim of this study is to explore real-time stress monitoring models (based on physiological features) for intensive care unit (ICU) nurses. The quantification of stress in ICU nurses has been limited to subjective ratings, with a general gap in continuous measurement; real-time stress monitoring based on continuous physiological measurement is needed to assess the negative outcome of stress. Electrodermal activity, eye tracking, accelerometer data, and skin temperatures were recorded continuously through 12-hour shifts for ICU nurses (23 participants). A machine learning algorithm was applied to identify stress over time based on physiological features. eXtreme Gradient Boosting (XGBoost) was performed with an accuracy of 0.88. Skin temperature contributed the most to real-time stress identification for monitoring. Future work should investigate the efficacy of using skin temperature for stress identification in real-time for ICU nurses.