Predictive Biomathematical Modeling Compared to Objective Sleep During COVID-19 Humanitarian Flights.

Biomathematical modeling software like the Sleep, Activity, Fatigue, and Task Effectiveness (SAFTE) model and Fatigue Avoidance Scheduling Tool (FAST) help carriers predict fatigue risk for planned rosters. The ability of a biomathematical model to accurately estimate fatigue risk during unprecedented operations, such as COVID-19 humanitarian ultra-long-range flights, is unknown. Azul Cargo Express organized and conducted five separate humanitarian missions to China between May and July 2020. Prior to conducting the missions, a sleep-prediction algorithm (AutoSleep) within SAFTE-FAST was used to predict in-flight sleep duration and pilot effectiveness. Here we compare AutoSleep predictions against pilots' sleep diary and a sleep-tracking actigraphy device (Zulu watch, Institutes for Behavior Resources) from Azul's COVID-19 humanitarian missions. Pilots wore Zulu watches throughout the mission period and reported sleep duration for their in-flight rest periods using a sleep diary. Agreement between AutoSleep, diary, and Zulu watch measures was compared using intraclass correlation coefficients (ICC). Goodness-of-fit between predicted effectiveness distribution between scenarios was evaluated using the R² statistic. A total of 20 ( = 20) pilots flying across 5 humanitarian missions provided sleep diary and actigraphy data. ICC and R² values were >0.90, indicating excellent agreement between sleep measures and predicted effectiveness distribution, respectively. Biomathematical predictions of in-flight sleep during unprecedented humanitarian missions were in agreement with actual sleep patterns during flights. These findings indicate that biomathematical models may retain accuracy even under extreme circumstances. Pilots may overestimate the amount of sleep that they receive during extreme flight-duty periods, which could constitute a fatigue risk.