Dynamic models of occupational fatigue: a systematic review for risk management advancement

Occupational fatigue is a major safety and health risk across work sectors. Because fatigue accumulates and fluctuates over time, effective fatigue management requires dynamic approaches that can anticipate risk and inform preventive controls rather than rely on point-in-time assessment. Although dynamic fatigue modelling has expanded rapidly, evidence remains fragmented across modelling paradigms and sectors, limiting clarity on which approaches are most suitable for implementation. This systematic review synthesises recent evidence on dynamic models of occupational fatigue and evaluates their relevance for fatigue management and risk control.

Records published from 2019 to 2025 were identified from Scopus using DOI-indexed retrieval and screened with Watase Uake Tools to support article mining and Preferred Reporting Items for Systematic Reviews. Searches were complemented in PubMed, ProQuest, Wiley Online Library, and ScienceDirect. References were managed in Zotero, and screening, selection, and data extraction followed a structured process with manual searching and citation tracking.

From 180 records, 12 peer-reviewed studies met the inclusion criteria. Four modelling approaches were identified: physiology-informed dynamic simulation, machine learning and deep learning for near real-time monitoring, probabilistic dynamic models such as Dynamic Bayesian Networks, and symptom-network or Bayesian graph approaches. Models were most useful for fatigue management when they integrated multimodal inputs and produced actionable outputs such as fatigue risk levels and intervention triggers, but implementation was limited by small samples and insufficient field validation.

Future research should prioritize longitudinal workplace validation, harmonized benchmarking, and deployable hybrid models integrating physiological, psychosocial, and contextual determinants.