Detecting Cognitive Fatigue in 12-Hour Shifts via Multimodal Wearables

Cognitive fatigue during 12-hr shifts can reduce attention, slow responses, and increase error risk in safety-critical healthcare and manufacturing settings. This study evaluated a multimodal wearable system for detecting fatigue and delivering timely micro-break prompts. Fifty shift workers wore sensors measuring photoplethysmography, electrodermal activity, and inertial movement during normal work. These signals were combined with brief micro-task measures and analyzed using a gradient boosting fusion model. Ground truth indicators included Psychomotor Vigilance Task lapses and self-reported fatigue. The model achieved an AUC of approximately 0.82, indicating useful field-based detection performance. Workers receiving fatigue-triggered micro-break prompts showed approximately 21% fewer late-shift vigilance lapses than controls, without measurable productivity loss. Individual calibration improved model reliability and prompt timing across workers. Participants also emphasized privacy, transparency, and data governance as essential for adoption. Overall, the findings support wearable, individualized, lowburden fatigue management to improve vigilance during extended shifts in diverse real-world occupational environments.