Assessing three process model predictions of mood and alertness levels using a sample of irregularly scheduled workers

The practice of working non-day shifts, and in particular the night shift, has been shown to both disrupt workers\u27 sleep/wake and circadian rhythms as well as increase their ratings of subjective sleepiness. A model has been developed which predicts subjective alertness in various shiftworker populations through the combination of the circadian and sleep/wake rhythms. This Three Process Model has clearly demonstrated its ability to generate accurate group predictions in samples of permanent shiftworkers and transmeridian flight crews. It is unclear, however, whether the constructed model variations are population specific or are transferable to other worker populations. These model variations can be used to make predictions of an individual\u27s sleepiness, but will these predictions be accurate? Lastly, given the static nature of the processes used to construct this model, it is possible that a different modeling approach based on schedule variability could produce more accurate predictions for a worker population on a highly variable work schedule. These were the issues explored by this dissertation. ^ Six variations of the Three Process Model were used in this study. These model variants differed with respect to the circadian acrophase and the sample population used for their development. Four measures of sleep/wake cycle variability, based on sleep start time and sleep duration, were also developed. Study participants were extraboard freight locomotive engineers on variable work schedules. These six prediction model variants and four variability measures were applied to these participants, and the ensuing predictions were compared to mood data collected from the engineers. The results suggest that neither the prediction model variants nor the variability measures predict mood for the entire group with any meaningful level of accuracy. There is evidence, however, that individual sleepiness reports are predictable, with moderate accuracy, with these model variants. The success of these predictions seemed to be based primarily upon the particular circadian acrophase employed.