Biomathematical modeling for the prediction of sleep behavior and comparison against cognitive performance in firefighters

Abstract

Biomathematical fatigue models, like the Sleep, Activity, Fatigue, and Task Effectiveness Fatigue Avoidance Scheduling Tool (SAFTE-FAST) could be adapted to serve as a fatigue mitigation tool for firefighting operations. The first aim of this study was to compare model predictions of effectiveness (Predicted Effectiveness) to actual Psychomotor Vigilance Task (PVT) speed (Actual Effectiveness) in a firefighter population. SAFTE-FAST also has a sleep prediction algorithm called AutoSleep that can be calibrated against actigraphy data in order to determine the best fit parameters for sleep estimation. The second aim of this study was to explore the ability of AutoSleep to predict sleep timing and duration in firefighters. N = 238 firefighters provided sleep and work data, with a subset of N = 151 firefighters who also provided PVT data, for up to two weeks. Firefighters slept between 6 and 8 h per 24-hour period whether on or off-shift. Linear regression revealed significant correlations between Actigraphy Predicted Effectiveness and PVT Actual Effectiveness scores (R2 = 0.68, p = 0.012). Best-fit AutoSleep parameters matched sleep duration in 99 % of predictions and sleep timing for 75 % of predicted sleep events compared to actigraphy, but there was no significant correlation between AutoSleep Predicted Effectiveness and PVT Actual Effectiveness scores (R2 = 0.34, p = 0.13). These findings highlight areas where the model must be refined in order to reliably predict firefighter performance. This study is the first step in adapting SAFTE-FAST for work shifts lasting 24 h or longer and an important step towards fatigue risk management for first responders.