Predicting Physical Fatigue for Healthcare Workers During Mass Decedent Handling Using a Wearable Sensor-Based Machine Learning Approach

Work-related injuries, such as musculoskeletal disorders (MSDs) among healthcare workers lead to substantial physical, mental, and financial costs with increased turnover rates. During the COVID-19 pandemic, these issues further intensified, with healthcare workers frequently engaging in prolonged mass decedent handling tasks. Performing these tasks caused the workers to exceed their physical limits, leading them to physical fatigue. Physical fatigue is one of the major factors and significant precursors of workplace injuries, underscoring the need for effective monitoring. By classifying and predicting physical fatigue automatically in real-time, we can proactively identify at-risk workers before fatigue leads to potential accidents. To address this issue, the study conducted a controlled laboratory experiment simulating decedent handling tasks. Twelve participants performed standardized logrolling tasks on a mannequin. Physiological signals, including heart rate (HR), skin temperature (ST), and electrodermal activity (EDA), were collected using a wristband-type wearable sensor. The Borg’s RPE scale (6-20) was used to assess subjective fatigue, categorizing it into four distinct levels. Subsequently, three machine learning (ML) models, namely Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN), were developed to classify and predict these physical fatigue levels. The results showed that the ANN model achieved the highest accuracy rate at 80%, followed by the RF model at 78% and the SVM at 75%. Future research could prioritize the utilization of robust datasets to achieve higher accuracy in model performance. This study contributes to the body of knowledge on safer working conditions for healthcare workers in global health crises.