From desk to dashboard: A multi-faceted approach exploring the combined impact of prolonged sitting and sleep restriction on driving performance

Increasing driver safety is of critical importance and understanding a driver's recent sitting and sleep history map help to reduce risks on the road. Sleep restriction and prolonged sitting have been linked to reduced cognitive function, posing potential hazards to safe driving. Such hazards are of particular concern for office workers who have an increased prevalence of both prolonged sitting and sleep restriction. Regular bouts of light-intensity physical activity to break up prolonged sitting has shown promising results in improving cognitive performance. Despite these findings, it remains unclear whether breaking up sitting with light-intensity physical activity can effectively improve cognitive performance related to driving, especially when counteracting the performance decrements associated with sleep restriction. Another risk to driver safety is the misalignment of self-assessed driving ability and actual driving performance, leading to poor driver calibration and an increased risk of accidents and unsafe driving behaviours. Various factors, including sleep restriction and prolonged sitting can impact driver workload, which, in turn can impact driver calibration. Understanding the impact of sleep restriction and prolonged sitting and identifying new methodologies to classify problematic behaviours in at-risk population such as office-workers is crucial for identifying strategies to enhance driver safety. To accomplish this objective, the studies in this thesis utilised data from a 7-day laboratory study (1 Adaptation day, 5 Experimental days, 1 Recovery day). Three separate studies were conducted. 

The first study (Study One) examined the combined impact of breaking up sitting during the day and sleep restriction on driving performance during a simulated commute to and from work. 81 participants were assigned to one of four groups; a) breaking up sitting with a 9 h sleep opportunity, b) sitting with a 9 h sleep opportunity, c) breaking up sitting with a 5 h sleep opportunity and d) sitting with a 5 h sleep opportunity. Participants performed light-intensity physical activity (3 min of walking every 30 min) during a simulated workday (09:00 h-17:00 h). Participants with a 9 h sleep opportunity displayed better driving performance than participants with a 5 h sleep opportunity, regardless of physical activity levels during the day. Participants with 5 hours of sleep per night who broke up sitting during the day demonstrated higher lane variability and reported greater subjective sleepiness before each commute when compared to all other conditions. Additionally, survival analyses revealed that participants with a 5 h sleep opportunity had a higher likelihood of crashing earlier into the simulated commute than participants with 9 hours of sleep. There were no significant interactions observed between condition, day, or commute time for any of the outcome measures in Study One. The findings from this study suggest that the breaking up sitting protocol utilised in this study may not be an effective countermeasure for the cognitive performance detriments in driving performance associated with sleep restriction. 

The second study (Study Two) explored the combined impact of breaking up sitting during the day and sleep restriction on the accuracy of self–assessments of driving performance . Driver calibration in Study Two was measured with the Brier Score which provided a measurement of how accurate each participant was at assessing their driving ability in the areas of speed control, lane control and confidence. Findings from participants (n=84) revealed that participants with 9 hours of sleep per night exhibited better calibration for lane variability, lane position, and safe zone-lane parameters compared to participants with a 5 h sleep opportunity. Additionally, participants with 9 hours of sleep per night reported lower perceived workload scores compared to participants with 5 hours of sleep per night. The findings from this study suggest short bouts of physical activity during the day may not be adequate to increase driver calibration or reduce driver workload. 

The third study (Study Three) assessed the efficacy of a deep learning approach using movement data to classify prior physical activity levels and prior sleep history during a simulated driving commute. Two convolutional neural networks (Dixon-Net and ResNet-18) were trained to classify accelerometry data into four classes: sitting, breaking up sitting, 9 hour sleep history, and 5 hour sleep history. The model, ResNet-18, achieved higher accuracy scores for activity and sleep history classification compared to DixonNet. Class activation mapping revealed distinct patterns of movement and postural changes between the classes, demonstrating the suitability of Convolutional Neural Networks for identifying drivers at risk. 

Collectively, the three studies presented in this thesis provide valuable insights into the use of breaking up sitting during the day as a potential countermeasure to driver fatigue. The findings from these studies suggest that breaking up sitting with light-intensity physical activity may not counteract the cognitive deficits associated with fatigue during driving. Moreover, any potential benefits of breaking up sitting with physical activity may not be sufficient to overcome the detrimental impacts of sleep restriction on workload and subjective assessments of driving performance. The novel deep learning approach utilised in this thesis identified previously unknown patterns within human movement data which can be used to classify potentially at-risk driving behaviours. The overall thesis highlights the importance of understanding the interactions between prolonged sitting, physical activity and sleep restriction for increasing driver safety.