Aircrew: Repeated Sleep Loss and the Accumulation of Fatigue

Introduction: It is commonly assumed that two undisturbed nights of recovery reset the fatigue to normal. This is also the assumption in mathematical models designed to predict fatigue from work schedules which lacks long-term recovery processes. However, there is almost no research done on the effect of repeated sleep deprivation/sleep loss across longer periods of time. It is possible that fatigue/sleepiness accumulates across repeated bouts of sleep loss. Such findings should affect work hour regulations and mathematical models predicting fatigue from work schedules. The present study focused on aircrew, and the purpose was to study the association between repeated sleep loss and accumulation of fatigue across work rosters, and to derive metrics that can be used to improve the performance of mathematical models in predicting fatigue. As an indicator of sleep loss, we used the encroachment of work (roster based) on the window of circadian low – WOCL (timing between 0200–0600), since that was available directly from collected roster data.

Methods: The data was collected from long-haul operation conducted between March 2017 and December 2023. Flight deck crew assessed their sleepiness on the Karolinska Sleepiness Scale (KSS) at the top of descent for each flight. Each assessment was linked to crew id, flight number, and date of departure, and then merged with roster data from the day-of-operation system. In total 32,795 assessments could be matched with rostered activities. To represent WOCL encroachment, we calculated the amount of block time within the WOCL on three different time horizons: 60, 120 and 240 hours prior to each of the assessments. These three values were then added up to get one combined metric to use in the analysis with a greater weighting on more recent events.

Results: Fatigue accumulation was evident, with KSS scores increasing as WOCL encroachment increased: • Reported KSS increased from 5.7 ± 0.23 for 0–2 hours block time to 7.8 ± 0.23 at 30–34 hours (B = 0.25 ± 0.021, p < 0.00001). •Predicted KSS increased from 4.9 ± 0.35 at 0–2 hours block time to 5.5 ± 0.35 at 30–34 hours (B = 0.081 ± 0.033, p = 0.042). • The residual (reported KSS - predicted KSS) increased from 0.89 ± 0.18 at 0–2 hours block time to 2.3 ± 0.18 at 30–34 hours (B = 0.17 ± 0.016, p < 0.0001).

Conclusion: These findings indicate that fatigue accumulates as flight duty time repeatedly encroaches on the WOCL, contradicting the assumption that just two nights of recovery is sufficient. This effect should be accounted for in aircrew scheduling, as well as in other industries where shift work disrupts normal sleep patterns. Furthermore, fatigue prediction models may be improved by incorporating a longer- term process representing this fatigue accumulation.