Developing the ‘fatigue impairment and performance simulator’: a computational implementation and application in a naval context.

Abstract

Background: Effective fatigue management is critical to maintaining operational effectiveness and mission endurance. To mitigate fatigue-related risks, there is growing interest in applying computational biomathematical models of fatigue (BMMs) for forecasting fatigue under various operational scenarios and understanding the situational factors that underlie fatigue. BMM’s are a class of biological phenomenological models which are used to predict the neuro-behavioural outcomes of fatigue (e.g., alertness, cognitive performance) using sleep-wake history, and are frequently applied in defence settings to support system safety as part of broader fatigue management strategies. Despite their utility, BMMs have a number of limitations. For instance, most current models are designed to predict fatigue outcomes over short time frames and have been parameterised based on short-term sleep deprivation laboratory studies (36-72 h). Consequently, they underestimate the cumulative and chronic effects of fatigue that unfold over missions spanning weeks to months. While newer models have been developed that can account for these longer-term effects, they are rarely applied in practice. A substantial limitation is that most BMMs are either ‘black-box’ propriety software, prohibiting independent evaluation and extension for decision-making support systems, or require substantial mathematical expertise to implement from scratch. Method and Results: This presentation will introduce a new open-source tool currently under development that addresses these challenges: the “Fatigue Impairment and Performance Simulator” (FIPS). FIPS is the first open-source BMM framework enabling defence researchers to inspect, validate, and ideally extend the underlying code. FIPS is implemented in the R programming language and provides a comprehensive set of functions for estimating, applying, and simulating from several classes of BMM under a Bayesian probabilistic framework. FIPS also includes an application enabling simulations without prior programming expertise. The initial motivation for FIPS’s development was to understand and forecast submariner fatigue. Submariners have no access to sunlight and are constrained to artificial shift patterns for weeks at a time, violating many additional assumptions underlying traditional BMMs. The current presentation reports on an application of FIPS to an intensive longitudinal dataset obtained from submariners during at-sea operations. Results of the study indicated that there was considerable variability in estimated model parameters (e.g., circadian rhythm) across individuals in our trials, highlighting the need for individualised models of fatigue (rather than a one-size fits all approach as is typically conducted). Model comparisons also revealed that a ‘Unified Model’ incorporating cumulative sleep debt provided the best account of the observed fatigue ratings. Thus, we validate assumptions that submariner fatigue is likely to be driven by cumulative sleep debt processes; and that submariners may take some time to recover from sleep debt. Conclusion: In summary, BMM tools as they are commonly applied may have serious limitations in unique defence contexts because they fail to account for relevant environmental constraints and individual differences. Given the common application of BMMs for defence decision-making, the present research has inherent ramifications for defence practitioners and researchers. We seek to foster collaborations with other fatigue researchers and defence decision-makers who might consider implementing their own models and methods under the FIPS framework.