Modelling fatigue and the use of fatigue models in work settings

In recent years, theoretical models of the sleep and circadian system developed in laboratory settings have been adapted to predict fatigue and, by inference, performance. This is typically done using the timing of prior sleep and waking or working hours as the primary input and the time course of the predicted variables as the primary output. The aim of these models is to provide employers, unions and regulators with quantitative information on the likely average level of fatigue, or risk, associated with a given pattern of work and sleep with the goal of better managing the risk of fatigue-related errors and accidents/incidents. The first part of this review summarises the variables known to influence workplace fatigue and draws attention to the considerable variability attributable to individual and task variables not included in current models. The second part reviews the current fatigue models described in the scientific and technical literature and classifies them according to whether they predict fatigue directly by using the timing of prior sleep and wake (one-step models) or indirectly by using work schedules to infer an average sleep-wake pattern that is then used to predict fatigue (two-step models). The third part of the review looks at the current use of fatigue models in field settings by organizations and regulators. Given their limitations it is suggested that the current generation of models may be appropriate for use as one element in a fatigue risk management system. The final section of the review looks at the future of these models and recommends a standardised approach for their use as an element of the 'defenses-in-depth' approach to fatigue risk management.