Masters Thesis Defense - Modeling Performance Impairment due to Chronic Sleep Restriction
Date: August 23, 2005
Time: 10:00 AM
Location: LeBow Engineering Center, Room: 240
Supervisor: Hans P.A Van Dongen, Ph.D., University of Pennsylvania
Advisor: Donald McEachron, Ph.D.
Laboratory experiments have demonstrated that cognitive performance deteriorates due to sleep deprivation and sleep restriction (even if sleep is reduced only a few hours per day on a chronic basis). Many Biomathematical models have been developed to predict performance deficits resulting from sleep deprivation. One influential model is the “two-process model” of sleep regulation, which predicts sleep and performance on the basis of two interacting processes. The first process, referred to as the “sleep homeostat” or Process S, which seeks to balance time spent awake and time spent asleep. The second process, known as the “circadian rhythm” or Process C, is driven by the biological clock in the brain, which keeps track of the time of day.
The two-process model properly predicts the performance degradation associated with multiple days of total sleep deprivation, but does not accurately predict performance under conditions of chronic partial sleep restriction. The model predicts that chronic sleep restriction leads to relatively little cognitive impairment, whereas laboratory experiments have shown that performance deteriorates progressively across days of sleep restriction.
This thesis describes the development of an expansion of the two-process model to accurately predict the performance impairment resulting from chronic sleep loss, by integrating a novel Process U along with the original two processes S and C. The parameters of process U were estimated using statistical analysis. The parameter assessment was performed by maximum likelihood estimation using nonlinear mixed effects modeling (NONMEM) software.
The predictions of the expanded two-process model were compared to psychomotor vigilance task (PVT) performance data from a laboratory experiment involving 14 days of sleep restriction to 4 h, 6 h or 8 h time in bed (TIB) or 3 days of total sleep deprivation. Model predictions were fitted to experimental observations of PVT lapses (RT ≥ 500 ms), as measured every 2 h during scheduled wakefulness in our laboratory study. We used the observations from two baseline days (8 h TIB per day) and all experimental sleep loss days, for a total of n = 47 subjects; as well as data from one recovery day (8 h TIB) for the subset of 34 subjects exposed to chronic sleep restriction.
This model in future may become useful in operational environments faced with sleep loss, such as hospitals, emergency services, and transportation. The model would be used for scheduling work hours for people working in sleep-deprived environments. Thus the expanded model may help optimize safety and performance.
The LeBow Engineering Center is located at the corner of 31st and Market Streets.