Linking Physiological Stress Detection Using Machine Learning with Evidence-Based Hospital Architecture

This study presents an integrated framework for predicting nurses' stress by combining physiological data from wearable sensors with hospital architectural and environmental features. Over ۱۱ million samples from ۱۵ nurses during real hospital shifts were analyzed, including electrodermal activity, heart rate, and body temperature, using machine learning models such as Random Forest, XGBoost, LightGBM, CatBoost, and a Voting Ensemble. Shorter time windows (۳۰ seconds) and class-weight adjustment improved prediction accuracy, with the Voting Ensemble achieving ۸۲% accuracy and an F۱-score of ۰.۸۰. Simulated mapping of stress levels onto hospital environments demonstrated that natural light, short circulation paths, greenery, and quiet rest areas significantly reduce physiological stress. This multidimensional approach provides actionable insights for evidence-based hospital design, promoting staff well-being, reducing fatigue, and enhancing operational efficiency.

Keywords:

Occupational Stress Prediction , Wearable Physiological Sensors , Machine Learning Algorithms , Evidence-Based Hospital Design , Environmental and Architectural Factors , Healthcare Staff Well-being