Electrodermal activity as a physiological marker of stress and experience-related differences in workers performing work at heights: A real-world functional data analysis study of certified and novice firefighters

This is the authors' abstract. We don't add key points for this paper.

Falls from heights remain a leading cause of fatal workplace accidents, with psychological stress contributing to human error. Electrodermal activity (EDA) and heart rate (HR) are promising stress markers whose real-world, physically demanding applicability remains underexplored. This study examined whether EDA and HR discriminate stress responses between certified and basically trained firefighters, and how these responses evolve across task phases. Seventeen male firefighters from the Moravian-Silesian Region, Czech Republic, completed an ascent, tower-top stay, and rappel from a 50-meter industrial cooling tower. EDA was recorded at the wrist and lumbar region using Sensetio wearable sensors, and HR via a Polar H10+ chest strap. After visual inspection, SCL was extracted from EDA; SCL and HR were both resampled to 600 samples (200 per phase) and min–max normalized. Functional data analysis using point-wise Mann–Whitney U tests, complemented by a cluster-based permutation test for multiple comparisons, compared SCL and HR between groups. SCL increased in both groups, but novices showed an earlier, steeper rise during the ascent and tower-top phases, forming two significant clusters after permutation correction ( t ≈ 59 –213, p c l u s t e r = 0.026 ; t ≈ 219 –421, p c l u s t e r = 0.017 ) with large effect sizes (median r = 0.71 ), whereas certified firefighters showed marked increases mainly during rappelling, where the difference narrowed and lost significance (median r = 0.46 ). No significant between-group HR differences emerged. Normalized SCL trajectories thus captured experience-related timing differences that HR did not, suggesting EDA warrants further investigation as a complement to safety training and monitoring, rather than as a validated readiness or error-risk-prediction tool at this stage.