Presentation
Investigating the Utility of Artificial Intelligence Models to Distinguish Between Physical and Mental Stress Among ICU Nurses
DescriptionNurses experience significant stress levels (physical or mental) due to the demanding nature of their work. Both stress types affect heart activities. However, the literature has often failed to distinguish between the heart-related variations associated with mental or physical stress. To overcome this, wrist-based inter-beat-interval and accelerometer data of ICU nurses were collected in a naturalistic environment during their shifts. First, stress levels (derived from Baevsky’s stress index) and activity levels (derived from hand accelerometer data) were determined for each time window. Then, a labeling rule was proposed to label specific windows (mental or mental/physical labels). Next, several classification techniques were utilized to train the data. The best model (random forest with AUC = 86.05) was then used to predict labels for unlabeled windows. SHAP values showed that HF power, RMSSD, and LF/HF ratio were the most heart associated variables, which can help us to distinguish mental and physical stress.
Contributors
Associate Professor
Assistant Professor
Event Type
Lecture
TimeThursday, September 12th3pm - 3:20pm MST
LocationFLW Salon I
Student Forum