Understanding fatigue in a naval submarine: Applying biomathematical models and workload measurement in an intensive longitudinal design

Fatigue is a critically important aspect of crew endurance in submarine operations, with continuously high fatigue being associated with increased risk of human error and long-term negative health ramifications. Submarines pose several unique challenges to fatigue mitigation, including requirements for continuous manning for long durations, a lack of access to critical environmental zeitgebers (stimuli pertinent to circadian physiology; e.g., natural sunlight), and work, rest and sleep occurring within an encapsulated environment. In this paper, we examine the factors that underlie fatigue in such a context with the aim of evaluating the predictive utility of a biomathematical model (BMM) of fatigue. Three experience sampling studies were conducted with submarine crews using a participant-led measurement protocol that included assessments of subjective sleepiness, workload (NASA-Task Load Index [TLX] and a bespoke underload-overload scale), and sleep. As expected, results indicated that predicting KSS with a BMM approach outperformed more conventional linear modelling approaches (e.g., time-of-day, sleep duration, time awake). Both the homeostatic and circadian components of the BMM were significantly associated with KSS and used as controls in the workload models. We found increased NASA-TLX workload was significantly associated with increased average KSS ratings at the between-person level. However, counter to expectations, the two workload measures were not found to have significant linear or quadratic relationship with fatigue at the within-person level. An important outcome of the research is that applied fatigue researchers should be extremely cautious applying conventional linear predictors when predicting fatigue. Practical implications for the submarine and related extreme work context are discussed. Important avenues for continued research are outlined, including directly estimating BMM parameters.