Identifying and Reducing Workforce Fatigue in Rapid Renewal Projects (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

9C h a p t e r 2 Fatigue Factors in rapid renewal highway Construction The initial research focus was to identify and describe represen- tative rapid renewal construction scenarios that are expected to become more common in the future. The team reviewed current and past construction projects, solicited contacts within the state DOTs and construction communities, and reviewed current literature and DOT websites to help define these rapid renewal scenarios and illustrate each scenario with spe- cific projects. The team identified 13 examples of rapid renewal construc- tion projects across six states (California, Florida, Illinois, New York, Utah, and Washington) and developed draft descriptions of these scenarios that included a project description, contract type and elements, work-shift schedules, specific workforce fatigue factors, and specific workforce fatigue countermeasures employed. These draft scenarios were distributed to state DOT staff, who were part of a Project Working Group established for this effort, with a request for them to review and, as necessary, revise the draft scenarios. Key aspects of the scenarios (espe- cially the specific fatigue factors and fatigue countermeasures) were refined through field research. Appendix A provides the revised rapid renewal scenario descriptions. This scenario task succeeded in identifying certain aspects of the work process that would be implicated in fatigue. These characteristics include the use of night work and weekend clo- sures. Table 2.1 lists the specific work-shift schedules reported and areas for further research. The information provided by the working group was not sufficiently detailed to determine how much night work or how many weekend closures were used. Subsequent experi- ence in field work suggested that these elements of scheduling are fairly dynamic; use of any particular scheduling approach depends on the work to be done, the period in which it is required to be performed, weather factors, and so forth. For example, night work may be planned but not transpire because of delays in lane closure approval. For another example, a weather window anticipated to allow paving based on tem- peratures may be delayed due to low temperatures. The impact of contract type on scheduling of the workforce did not appear to be significant in the projects reviewed and observed in the field. One contractor the team met with was working on two adjacent projects on SR-50 in Florida, with one of the projects being Design-Bid-Build and the other being Design-Build. In both cases, they used the same work scheduling and construction approaches. The more general observation with respect to work scheduling is the tendency of those projects that use continuous day–night construction to use two 10-h shifts, such as 7:00 a.m. to 5:00 p.m. and 8:00 p.m. to 6:00 a.m. This allows for more work to be com- pleted following the set-up phase, and for activity to be cur- tailed during evening rush hour periods. Scenario data also included information concerning work- force fatigue factors that respondents considered noteworthy. Table 2.2 lists these, with the team’s interpretation of research issues and potential risk factors. Table 2.1 and Table 2.2 summarize commonly occurring sce- narios associated with rapid renewal highway work, although the extent of their usage in any particular project is unknown; determining it would require retrospective analysis of work schedule data, which was outside the scope of this task. While the schedule and fatigue factor descriptions provide some insight as to how the overall labor force is affected by rapid renewal scenarios, the impact on salaried workers is much less clear. Fatigue countermeasures were reported for a subset of the projects catalogued for Task 1. These countermeasures are presented in Table 2.3. The countermeasures described in Table 2.3 represent a range of approaches often employed in more comprehensive fatigue management programs, and include offsetting extended hours, stable assignments to specific shift times, limitations on extended work periods, and crew awareness and reporting. The Fatigue in Occupational Settings

10 Table 2.1. Types of Work Schedules Described for Rapid Renewal Projects Type of Schedule Reported Area for Research Normal day work Assumed to mean 8-h shift Night work Shift length not specified Weekend work Shift pattern/length not specified Extended day work Assumed to mean more than 8 h, typically 10 h Nonstop, multi-shift weekend work Number of separate shifts and crews not specified Extended weekend work—33 h Number of separate shifts and crews not specified 7 day per week operations Number of hours per day not specified Double shifts Assumed to mean same crew works 2 sequential 8-h shifts 2 or 3 shifts per day for one weekend Assumed to be a closure 24-h-per-day operations with 3 shifts Assumed to be 8-h shifts Table 2.2. Workforce Fatigue Factors Reported for Rapid Renewal Projects Workforce Fatigue Factors Reported Research Issues and Risks Shift schedules Rotation or fixed assignment not specified; start or end times not specified. Simultaneous operations in close proximity High workload can enhance fatigue. Deadline pressure to reopen lanes Same as above. High level of construction monitoring Same as above. Long-term night shifts Potential for cumulative fatigue due to less sleep obtained by personnel on night shift. Change shift to night work in middle of week No time for adaptation to new start time; first few nights can be very difficult. Overtime shifts follow 40-h week Extended work week. Table 2.3. Workforce Fatigue Countermeasures Used in Selected Projects Countermeasure Reported Implementation Considerations Salaried workers offset extended workdays with shorter workdays following the long hours Hourly workers not addressed by this approach Outside workers were given water coolers, ice water, and water bottles Addresses dehydration Stability in shift assignment: day workers kept on days, night workers on nights Reduces potential for disrupting established sleep schedules Limitation of crew to one 10-h shift per 55-h weekend closure Permits recovery day Crew instructed to alert supervisors to issues related to long work hours Crew awareness of fatigue effects incorporated into management policy

11 extent to which these measures are used is not clear, since they were voluntarily reported in an open-ended question about fatigue countermeasures. There does seem to be awareness of worker fatigue as a problem, especially in relation to extended work periods and night work, but the scope of this task did not permit a detailed assessment of the extent or formality of countermeasure usage. Review of Scientific and Technical Research Literature Introduction There exists an extensive scientific literature exploring the rela- tionship between fatigue and adverse outcomes in occupational and transportation settings. In occupational studies, aspects of work schedules are examined as preconditions for fatigue- related adverse outcomes such as error, injury, and death. In some cases, work performance is measured in the field, either as an indicator for fatigue or as a proxy for presumed risk of an accident or incident. Measurement of fatigue itself is problem- atic; it is not directly observable and must be inferred from other, measurable phenomena. Fatigue is the presumed mecha- nism linking long periods of wakefulness and performance- related adverse events (Williamson et al. 2011). Therefore, the effect of work schedule on safety and effective job performance is hypothesized to operate primarily through prior sleep– wake history and the time of day at which work takes place. This literature review is intended to provide the foundation for a model of rapid renewal workforce fatigue and its man- agement. First, this section will discuss the biological and environmental processes governing fatigue. Then, it reviews the effects of fatigue on job performance in operational set- tings. Next, this section examines evidence from selected occu- pational studies linking identified fatigue “precursors” with occupational injuries, followed by a review of findings from the limited set of studies of occupational injuries and deaths within the construction industry. Finally, it explores integrated models for fatigue management, including discrete counter- measures, some of which are appropriate for deployment on job sites. Biological and Environmental Processes Governing Fatigue Biological Processes There is a fundamental neurobiology underlying fatigue and how it affects neurobehavioral performance in the workplace. This neurobiology is tightly coupled with sleep regulation. The natural timing and duration of sleep is governed by two principal biological processes: a homeostatic process and a circadian process (Borbély 1982). The homeostatic process seeks to balance the amounts of wakefulness and sleep, such that extended wakefulness results in a buildup of pressure for sleep. If this pressure for sleep is not met, sleep deprivation occurs and fatigue results (Daan, Beersma et al. 1984). As such, fatigue is regulated by the homeostatic process as a function of sleep/wake history. Getting sleep is a necessity to reduce fatigue resulting from the homeostatic process. Restricting sleep leaves a person with leftover sleep pressure, which, when repeated over sev- eral days, causes an accumulation of fatigue (Van Dongen et al. 2003). Cumulative fatigue from chronic sleep restric- tion results in sensitization to further sleep loss and a need for multiple recovery sleep periods (days to weeks) (Belenky et al. 2003). The circadian process is a 24-h rhythm produced by the endogenous biological clock (located in the suprachiasmatic nucleus in the hypothalamus, shown in Figure 2.1). It pro- motes wakefulness during the day and sleepiness during the night, making staying awake through the night a challenge even if one manages to get enough sleep beforehand. Indeed, fatigue is regulated by the circadian process as a function of time of day. Moreover, the circadian process makes it difficult to get enough sleep during the day, especially during the “wake maintenance zone” in the early evening (Dijk and Czeisler 1994), thereby interfering with the homeostatic pro- cess by causing sleep restriction in people attempting to sleep during the day. During daily awake hours, circadian rhythms lead to pre- dictable changes in alertness, such as the tendency to feel sleepy at some point during the afternoon (often referred to as the Source: McCallum et al. 2003. Figure 2.1. Location of the brain’s biological pacemaker that controls circadian rhythms.

12 “post-lunch dip,” although the alertness drop has more to do with the point in the circadian cycle and less to do with whether you have eaten). Alertness in humans is typically lowest between midnight and 5:00 a.m., which corresponds to the period when melatonin levels are highest (Czeisler and Dijk 2001). Sleep inertia refers to a transient reduction in neurobehav- ioral performance capability immediately upon awakening (Milner and Cote 2009). Sleep inertia may be observed fol- lowing sleep periods of about 30 mins or more, especially in people who are sleep deprived or who are awakened early in the night (near the circadian nadir) and/or before the need for sleep is fulfilled (i.e., sleep is restricted) (Dinges 1990). Depend- ing on the magnitude of sleep inertia, it may adversely affect neurobehavioral performance for up to about 2 h after awak- ening (Achermann et al. 1995), but most of the impairment dissipates within the first 15 min. A person experiencing sleep inertia may not be suitable to get back to work until some time for recuperation has passed (time that can be usefully spent taking a shower or eating breakfast). Caffeine counter- acts sleep inertia (Van Dongen et al. 2001), and it has been suggested that taking caffeine prior to a nap might be a good strategy to avoid sleep inertia upon awakening (Reyner and Horne 1997). Environmental Processes The body’s circadian rhythms can be disrupted by external cues that are inconsistent, out of phase with the established rhythm, or both. Such disruptions can affect an individual’s ability to fall and stay asleep as well as the quality of the result- ing sleep. External cues, particularly light–dark patterns, can reset the circadian pacemaker. This enables individuals to adjust to schedule or time changes. The length of time required for the adjustment depends on how extreme the changes are and on individual variability (Suvanto et al. 1987). Jet lag, for example, occurs when an individual’s circadian rhythm is dif- ferent from the day–night and activity patterns of the local environment. Most people can adjust their sleep–wake cycle to a full 12-h time zone change within a few days, although recent research indicates that disruption of sleep stages may persist past the time an adjusted sleep time has been estab- lished. It is more difficult for people to adapt to work schedules that are 12 h out of phase with their circadian rhythm than to accommodate a 12-h time-zone change. This is because the pattern of light and dark, some surrounding activities, and the sleep–wake schedule continue to be in conflict—unlike the case with a time-zone shift where light–dark schedules, activities, and sleep–wake schedules all shift together. The cir- cadian pacemaker of individuals, such as shift workers, who switch temporarily from one activity–rest pattern to another, as on weekends, can become chronically disrupted and mis- aligned with external time (Monk 2000). Process Interactions The homeostatic process and the circadian process interact to determine the level of fatigue and neurobehavioral perfor- mance and the way in which they change over time (Van Dongen and Dinges 2005). As such, fatigue in the workplace is a function of both time awake (due to hours of service) and time of day (due to the work schedule) (Åkerstedt 2007) and thus of the nature of the work. Results from a study examin- ing work scheduling in U.S. construction workers have shown that overtime and irregular work schedules adversely affect worker safety (Dong 2005). Second jobs, family obligations, recreational activities, and other social factors similarly co- determine fatigue through time awake and time of day. A large body of empirical data shows that fatigue is reduced by sleeping longer or in a more consolidated manner (i.e., less frequently interrupted by brief awakenings) (Banks et al. 2010). In addition, prior sleep history (over days to weeks) determines vulnerability to fatigue (Belenky et al. 2003; Rupp et al. 2009). Inadequate sleep (excessive wakefulness) over successive sleep– wake cycles results in a “sleep debt” that has measurable neuro- physiological consequences (Spiegel et al. 1999; Sallinen et al. 2004; Van Dongen et al. 2003). There is substantial individual variation in the sleep homeostatic process; that is, there are long sleepers and short sleepers (Aeschbach et al. 2001; Van Dongen et al. 2005). Understanding the underlying neurobiology is critical to the issue of fatigue, because these homeostatic and circadian processes are active always in everyone. There are many other factors, both internal and external, that regulate fatigue, but by and large these other factors are determined by circumstance and their effects are primarily transient. However, the homeo- static and circadian regulation of fatigue is always present, comparatively powerful, and highly predictable (McCauley et al. 2009), making it both critical (Caruso et al. 2006; Ricci et al. 2007) as well as manageable (Smiley 1998; Dawson and McCulloch 2005) in operational settings. Effects of Fatigue Physiological effects of fatigue experienced by the person are different from performance effects. The physiological symp- toms or manifestations of fatigue, such as blinking, facial tone, and posture, might be better thought of as indicators of the state of fatigue. Indicators of the state of fatigue can have limited value if the primary issue is whether the person pre- sents a concern about degraded performance. Interpersonal variability in the extent to which indicators of the state of fatigue predict performance degradation (Oonk et al. 2008) might make it preferable to directly measure degraded perfor- mance. On the other hand, directly measuring performance degradation might not provide sufficiently early detection. Moreover, performance degradation can be caused by factors

13 other than fatigue and, as such, is not by itself a very accurate indication that fatigue is an issue. Cognitive indicators of fatigue can include degraded alert- ness and attention, problems with sustained concentration, tendency to be easily distracted, confusion, forgetfulness, memory problems, and performance worries. Psychomotor and cognitive speed, vigilant and executive attention, working memory, and higher cognitive abilities appear to be particu- larly affected by sleep loss (Lim and Dinges 2010). These cog- nitive decrements can accumulate to severe levels over periods of chronic sleep restriction without the full awareness of the affected individual (Van Dongen et al. 2003). Affect indica- tors can include demotivation (such as boredom, lack of desire and enthusiasm, or temporary feelings of depression) and coping, emotional, or interactional fatigue (such as anxi- ety, avoidance, comfort seeking, irritability, or feeling stressed) (Luna et al. 1997; Kamdar et al. 2004). These effects show con- siderable variability across individuals (Van Dongen et al. 2005). Microsleeps, sleep attacks, and lapses in cognition are considered to be an indication of state instability (i.e., short- duration transitions between sleep and wake states) (Doran et al. 2001). The following effects of fatigue are generally agreed upon in the scientific literature (Caldwell et al. 2008): • Accuracy and timing degrade; • Attentional resources are difficult to divide; • A tendency toward perseveration develops; • Social interactions decline; • The ability to logically reason is impaired; • Attention wanes; • Attitude and mood deteriorate; and • Involuntary lapses into sleep begin to occur. Sleep and Recovery: A Selective Survey of Recent Studies Sleep loss and disrupted sleep are the underlying causes of fatigue that are addressed by work-hour regulation and fatigue- management policies. Time on the job is clearly a factor, as shown by the risk profiles of time-into-shift by Folkard (1997), but it is increasingly recognized that sleep history and sleep that is obtained in the 24 h before the work period are critical elements of how rested and alert a worker will be (Dawson and McCulloch 2005). This section selectively reviews the state of knowledge regarding the impact of sleep on fatigue, alertness, and operational safety and recent work concerning sleep depri- vation and recovery opportunities. Sleep Need and Individual Differences Work-hour regulations have at their basis the assumption that rest and recovery time (hence the amount of sleep workers obtain) depends on how much time people are at work. There is some basis for this assumption, as shown by a time-use sur- vey of sleep in relation to waking activities (Basner et al. 2007). Analysis of extensive survey data obtained in the American Time Use Survey suggests that the only variable showing a reciprocal relationship with amount of time reported sleeping is the amount of time reported working. The correlation is not large (-0.36 for weekdays), but it is significant. More detailed analysis of the data indicates that average sleep time was low- est for the 45- to 54-year-old age group (7.8 h), and that sleep time decreases with increasing income. The time-use survey data raise the proverbial question of how much sleep is really needed. This issue was discussed extensively by Ferrara and De Gennaro (2001) in a review that addressed historical surveys of sleep habits; sleep depri- vation effects on sleep structure, performance and alertness; gradual and long-term sleep reduction; sleep extension; and individual differences and sleep typologies such as long and short sleepers. There has been a gradual curtailment of sleep with the increased pace and activity of modern society, from a modal number of nocturnal hours of 8 to 8.9 in 1959 to 7 to 7.9 in the mid-1980s. There are many methodological issues associated with epidemiological surveys of this sort, but the bulk of historical data support the conclusion that sleep time has decreased in the past 100 years. In terms of how much sleep is enough, the theories and data are quite varied, with notions of “core sleep” of 4.5 to 6 h (Horne 1988) being sufficient countered by sleep extension studies suggesting that people can sleep as much as 9 h when no restrictions are applied. “Long night” studies, in which subjects remained in bed for 14 h each night for a 4-week period, showed that subjects slept an average of 10.6 h per night, compared with a control group that slept 7.6 h when given 8 h in bed. However, by the 4th week of the study, the average total sleep time was about 8 h, suggesting that a pre- existing sleep debt had recovered in the first weeks, and that extra rest time available led to only an additional 30 to 60 min of sleep (Wehr et al. 1993). The general conclusion from this work is that sleep deprivation does affect performance and alertness, sleep extension appears to show no substantial ben- efits for average individuals, and large differences exist among individuals. This latest point is quite important in consider- ing fatigue management programs in industry, because it is not possible to establish a specific amount of time to satisfy the physiological sleep need, although it is clear that the num- ber of hours of prior wakefulness is positively correlated with subsequent sleep duration. That is, sleep duration depends on the amount of need for sleep (Åkerstedt 2007). Individual Differences in Response to Fatigue There are marked individual differences in the neurobehav- ioral response to fatigue. This is reflected in wide-ranging

14 levels of performance impairment between subjects that are stable within subjects over time (Van Dongen et al. 2004). Thus, some individuals are highly vulnerable to performance impairment due to sleep loss, while others are relatively resil- ient. Workers may not be aware of their own vulnerability to neurobehavioral performance impairment due to fatigue; there is little congruency between subjective reports of sleepi- ness and objectively measured performance following sleep loss (Leproult et al. 2003; Van Dongen et al. 2004). Individual differences can originate in multiple factors implicated in the regulation of fatigue, including sleep need (Aeschbach et al. 1996), circadian timing (commonly recognized as the trait of morningness/eveningness; Kerkhof and Van Dongen 1996), tolerance to shift work (Van Dongen and Belenky 2009), and sensitivity to caffeine (Retey et al. 2006). In addition, sleep disorders and other medical conditions may contribute to between-subjects variability in the effects of fatigue, in com- plex ways (Oonk et al. 2008). Research suggests that when workers are allowed to select their own occupations and shift schedules, they tend to select those that are least challenging, a situation that which has been an issue in many studies of shift work tolerance (Härmä 1995). However, there is evi- dence that self-selection does not eliminate individual vari- ability in vulnerability to fatigue (Van Dongen et al. 2006). Sleep Patterns, Fatigue, and Performance The findings reviewed indicate that sleep disruption and deprivation are associated with performance decrements. Most of the data come from laboratory settings where it is possible to control sleep time and to measure performance precisely. In controlled settings, it is clear that sleep depriva- tion results in dose-response relationships, with greater sleep deprivation leading to larger performance decrements, and that these effects are cumulative, that is, the longer a subject is sleep deprived, the greater the effect (Van Dongen et al. 2003). These effects are most pronounced for performance tests; subjective ratings show an initial increase in sleepiness that levels off after the 1st or 2nd day. The Van Dongen et al. (2003) results confirm and elaborate upon the basic notions of sleep debt and the cumulative impact of sleep deprivation proposed on the basis of early studies (Carskadon and Dement 1982). The basis for some performance decrements is the amount of time spent on the task, and ultimately how the work time affects the opportunity for sleep. Time-on-task effects are observed as a performance decrement on a single task the longer one is engaged in the task (Bills 1937). Time-on-task effects are par- ticularly evident in tasks requiring sustained attention (Basner et al. 2008). The effect appears to be exacerbated by monotony, particularly in tasks that are machine paced (Koslowsky and Babkoff 1992) so that compensatory slowing down is not an option. The homeostatic and circadian processes interact with time-on-task effects when the homeostatic drive for sleep is elevated, the circadian drive for wakefulness is reduced, or both (Van Dongen and Belenky 2008). Rest breaks, either with or without sleep, and task switching provide recuperation from the time-on-task effect (Van Dongen et al. 2010). Breaks are also useful to reduce the impact of fatigue on safety in that they tem- porarily remove a person from harm’s way. The relationship of laboratory findings to operational set- tings is suggestive but not complete. Philip and Åkerstedt (2006) review a substantial number of transport and indus- trial safety studies that point to disrupted sleep as an underly- ing factor in accidents. The principal finding from a number of on-the-road studies is that truckers and car drivers do not get enough sleep just before a long trip. Åkerstedt (2007) sug- gests that there is a need for addressing white-collar work in addition to transport and industrial settings. Evidence for the relationship of sleep deprivation to on-the-job safety in cogni- tive work comes from a study by Landrigan et al. (2004), which showed that reducing work hours for on-call physician interns reduced serious errors by 50%. This finding suggests that risky decision making is affected by sleep deprivation, a subject which was addressed by McKenna et al. (2007). This group evaluated performance on a choice task involving risk and how it was framed (potential gain or loss). The results suggest that sleep deprivation decreases risk taking when a potential loss is presented, but increases risk taking when the problem is framed as a potential gain. In terms of operational implica- tions, these results suggest that framing decisions in terms of conservative approaches for those working shifts with sleep- deprivation potential may be a reasonable strategy. A considerable amount of research concerning shift work has been performed over the years (Åkerstedt 1998), and con- tinues to be elaborated on and refined. The basic complaint of disturbed sleep in shift workers was investigated in a large study by Åkerstedt et al. (2008) to determine the specific com- plaints underlying “shift-work sleep disorder.” This study com- pared shift workers, day workers, and insomnia patients on responses to a 300-item health and work environment ques- tionnaire, which included items from the Karolinska Sleep Questionnaire. The principal finding for shift workers was that they reported more frequent “too little sleep” and “nodding off at work.” Insomniacs reported substantially more sleep prob- lems, and the authors conclude that shift-work disorder com- plaints are a distinct category. Introducing a single nap in the midst of a night shift improves mood and quality of work evaluations without disrupting the main sleep episode during the day (Bonnefond et al. 2001). Recovery from Sleep Deprivation Early studies of sleep deprivation suggested that following a single sleepless night, it takes more than a single full sleep

15 episode to recover. This conclusion was based primarily on the sleep latency test and EEG parameters indicating that physi- ological changes persisted beyond the first recovery day. The issue of recovery is central to work-hour and scheduling poli- cies, so an understanding of the dynamics of rest periods in relation to alertness and performance is important (Belenky et al. 2003; McCauley et al. 2009; Banks et al. 2010). The main work in this area appears to consist of laboratory studies. Two recent examples provide illustrative findings. Lamond et al. (2007) induced moderate (1 sleepless night) and severe (2 sleepless nights) sleep deprivation in subjects, followed by five recovery nights that were either augmented (9 h) or restricted (6 h). After moderate sleep deprivation, a single night of 9 h sleep was sufficient to return alertness and performance to baseline levels; the physiological measure of sleep latency returned to baseline after two recovery nights. However, if the recovery opportunity was restricted to 6 h— in effect inducing additional sleep deprivation—none of the measures returned to baseline levels. Severe sleep depriva- tion was evaluated only with 9-h recovery periods. In this condition, alertness recovered after a single 9-h sleep, physio- logical measures (sleep latency) recovered after two 9-h sleeps, but performance remained below baseline for the entire recov- ery period. A similar study was reported by Axelsson et al. (2008), in which subjects were restricted to 4 h of sleep for 5 days, followed by 7 recovery days of 8-h duration. Sleepiness scores increased and performance scores deteriorated over the restricted sleep days. After 3 recovery days, sleepiness scores returned to base- line, but the performance scores did not. More detailed analy- sis of the data showed substantial individual differences in performance scores, but not in the sleepiness scores, suggest- ing that trait-like characteristics may lead to differential effects in performance across individuals that are not reflected in self ratings of sleepiness. The implication of this is that while two people may rate themselves as equally alert, one may perform substantially worse than the other. A report by Åkerstedt et al. (2000), integrating findings from a number of laboratory and operational studies, sug- gests that following extended and/or irregular work hours, 2 days and possibly more are necessary for full recovery. This must be balanced by workers preferring backward rotating shifts in order to obtain additional time off; these shifts have the adverse effect of progressively reducing sleep duration during the days on which work is assigned (Signal and Gander 2007). Age tends to influence the preference for vari- ous shift and recovery schedules, with younger people prefer- ring rapidly rotating systems that allow greater time off (Kecklund et al. 2008). The data concerning sleep deprivation and recovery peri- ods is particularly pertinent to the question of what consti- tutes an appropriate recovery period following work. Proposed rules for the nuclear industry, for example, mandate a 10-h period between successive shifts, which is comparable to other industries such as trucking (10 h) and aviation, which has a graded recovery period based on prior time spent flying. The key finding that bears on recovery period length is that when recovery sleep is limited, recovery of performance is insuffi- cient. Limited recovery might be analogous to lengthened operational shifts with 8 or fewer hours off between them. Fatigue and Job Performance in Operational Settings Published studies of task-specific performance decrements related to fatigue occur primarily in the domains of medicine and surface transportation, and to a lesser extent in aviation, rail, and military settings. Medicine In medicine, most studies focus on the effects of very long, overnight shifts of 24 h or more, which are typical of physician training programs; such schedules have been demonstrated to result in sleep loss among medical interns (Lockley et al. 2004; Barger et al. 2005). Studies of medical staff in the field and in the lab have repeatedly shown an association between sleep loss and impaired job performance, including attentional lapses and medical errors. Studies uniformly conclude that the 24+ h shifts typical of residency programs should be eliminated. A series of studies looked at medical error, injuries, and acci- dents for interns over various types of rotations (Barger et al. 2006; Barger et al. 2005; Ayas et al. 2006). These studies used case-crossover designs where subjects were their own controls, and the studies relied upon self-reporting to measure out- comes. In one study, they examined the risk of self-reported, fatigue-related medical error and associated adverse medical events during months when participating interns had no extended shifts (baseline), one to four extended shifts, and five or more extended shifts (Barger et al. 2006). An extended shift was defined as a period of at least 24 h continuously at work. Compared to baseline, odds ratios (ORs) for significant medi- cal errors were 3.5 [95% confidence interval (CI) 3.3 to 3.7] and 7.5 (95% CI 7.2 to 7.8), respectively. Odds ratios for related adverse events were 8.7 (95% CI 3.4 to 22) and 7.0 (95% CI 4.3 to 11), respectively, compared to baseline. A similar study used several methods, including a team of chart reviewers and physician observers, to record and rate errors made by interns during traditional (~80 h per week with regular extended shifts) versus intervention (no extended shifts) work schedules on an intensive care unit (Landrigan et al. 2004). Interns on the traditional schedule made 35.9% more errors overall, and 56.6% more non-intercepted errors

16 than those on the intervention schedule. Rates of serious error (22% higher), medication error (20.8% higher), and diagnos- tic error (OR 5.6) were all higher among interns on the tradi- tional schedule. An experiment using within-subjects comparison of 12 medical residents examined the effect of sleep loss and shift duration on “high-fidelity” simulated patient care over a 24-h period (Sharpe et al. 2010). Researchers found that errors during performance of routine tasks remained low over the period, but that residents committed increasing errors during complex tasks as wakefulness increased. How- ever, they also found that error rates and self-reported impact of wakefulness were not correlated. A study of generalized performance in a medical setting fol- lowed anesthesia specialists and trainees for 2 weeks, record- ing sleep loss (diary and actigraphy), work shift and duration, and Psychomotor Vigilance Tests (PVTs) at the beginning and end of each shift (Gander et al. 2008). Among trainees, who regularly worked night shifts, PVT was degraded at night compared with daytime shifts, and declined as the night shift progressed; acute sleep loss was associated with degraded per- formance on night shifts. Among specialists, PVT worsened over the course of 12 consecutive days on duty, and declines were associated with acute sleep loss. In another study of med- ical residents, attentional lapses, as measured by continuous electro-oculography, occurred twice as frequently during night shifts as during days, and 1.5 times as frequently on day shifts for those on the traditional (not intervention) work schedule (Lockley et al. 2004). Driving The relationship among self-report measures, EEG, and driv- ing performance was reported by Horne and Baulk (2004). In this study, participants were sleep restricted to 5 h the night before performing in a 2-h simulator drive. Sleepiness mea- sures were obtained every 200 s, and lane crossings were counted as safety incidents. The results showed very strong correlations between EEG and sleepiness scale scores (0.93) and between sleepiness scores and incidents (0.88). Horne and Baulk believe that the results demonstrate that drivers are aware of physiological sleepiness as it occurs. The implication of these results for actual driving or other operations is that people need to be educated regarding the meaning of their subjective sense of fatigue so that they can make responsible decisions about rest versus continued activity. For example, in the driving situation, Horne and Baulk (2004) discuss the likely actions drivers would be taking when they reach the lane-deviation state of fatigue; that is, sleepiness scale 8 or 9, which embodies “fighting sleep.” In this state, drivers tended to open the window and shift their postures in an attempt to stay awake. It is hoped that proper understanding of the meaning of this state by drivers would encourage them to take appropriate precautions by pulling off the road and resting. A series of similar studies to that of Horne and Baulk (2004) has also shown the close link between self-ratings of sleepi- ness and the occurrence of fatigue or actual incidents. Smith et al. (2005) conducted a diary study of young drivers that showed their self-rated alertness was below critical levels dur- ing 2.6% of their driving episodes. Ingre et al. (2006) showed that sleepiness ratings were strongly related to absolute risk of an accident, based on simulator performance. For example, with a Karolinska Sleepiness Scale (KSS) score of 9 (fighting sleep) the average subject’s risk increased by 185 times; how- ever, differences among individuals in accident-likelihood complicate the use of prediction at the individual level. Jones et al. (2006) studied sleep-deprived subjects in a per- formance and self-rating study with questions designed to assess their perceived ability to drive and the perceived ability of another who is similarly sleep deprived (a “hypothetical other”). Self-ratings declined substantially with increasing sleep deprivation and were correlated 0.7 to 0.76 with perfor- mance. Ratings for self-performance were somewhat higher than for the hypothetical other, suggesting that subjects had more confidence in themselves than someone else in the same condition. Dorrian et al. (2007) conducted a simulated train- driving study with actual locomotive drivers suggesting that as fatigue increases, certain error types increase as drivers become “disengaged” with the task. Aviation For more than 10 years, the National Transportation Safety Board (NTSB) has included fatigue on its “Most Wanted List” of advocacy priorities for aviation safety. In 2008 the NTSB recommended that the Federal Aviation Administration develop guidance, based on empirical studies and scientific research findings, to manage fatigue in aviation operations. This recommendation followed key fatigue-related mishaps, such as the crash of Corporate Airline flight 5966, the runway overrun by Delta connection flight 6488, the off-runway excursion of Pinnacle Airlines flight 4712, and the airport overshoot by go! flight 1002 (Caldwell 2009). While these mishaps implicated fatigue as a contributor, systematic data on the number of fatigue-related aviation accidents do not exist. A recent NTSB analysis of aviation accidents based on the Aviation Accident Database, which dates back to 1962, found that only 0.45% of them noted fatigue as a cause (Price 2009). Price concluded that analysis of this database shows that it drastically underestimates the role of fatigue because accident investigation methods did not include adequate methods for assessing whether fatigue was a potential contributor. Caldwell (2009) conducted a search of the scientific literature and other resources to explore the

17 impact of fatigue on aviation safety; he found no other stud- ies that estimated the percent of aviation accidents that involved fatigue. Although pilots and crew have been the primary focus of concern in the aviation sector, fatigue among aviation main- tenance personnel may be a concern as well. However, data to assess the extent to which fatigue poses a hazard to aviation maintenance do not exist. At best, one study estimated that approximately 12% of major aircraft accidents and 50% of engine-related flight delays and cancellations worldwide are due to maintenance deficiencies (Marx and Graeber 1994). That study only points to the fact that aviation maintenance may have an operational impact, but it does not necessarily implicate fatigue as being responsible for maintenance issues. However, aviation maintenance personnel have reported fatigue to be one of the most prevalent causes of accidents and deficiencies (Hobbs and Williamson 2000). Rail The rail sector has long recognized fatigue to be a concern. In fact, the first “hours of service” (HOS) regulations in the United States were established in 1907 for the rail industry. The 2008 head-on collision between a Union Pacific freight train and a Metrolink commuter train in the Chatsworth district of Los Angeles, California, heightened this concern due to the likelihood that fatigue was a major contributor to the accident. HOS have been further restricted and recently the U.S. Congress mandated fatigue management be adopted in the railroad industry. The Chatsworth accident and the mandate from Congress spurred several studies. A recent study indicated that between 1996 and 2002 accidents involving fatigue or alertness occurred on average three times per year (Oman et al. 2009). Gerson et al. (2009) screened a sample of railroad workers using the Epworth Sleepiness Scale. More than 40% of the respondents reported that they experienced Excessive Day- time Sleepiness, which is significantly higher than the esti- mates (ranging from 2 to 8%) for the general population. Rail workers experience irregular schedules and sometimes backward shift rotations. Research has demonstrated that both these practices aggravate fatigue by hindering an indi- vidual’s ability to biologically adjust to disruptions of their circadian rhythms. The Federal Railroad Administration (FRA) validated and calibrated a biomathematical fatigue model to assess fatigue in the rail industry. Examining the 30-day work histories of locomotive crews prior to 400 “human-factors” accidents and 1,000 “nonhuman-factors” accidents, this assessment found a strong relationship between crew fatigue scores based on the biomathematical model analysis and the probability of a “human-factors” accident (Hursh et al. 2006). Alerter systems have existed in locomotives for approxi- mately 15 years but these systems have proven to be inade- quate. The alerter system requires the locomotive engineer to hit a button every few minutes (the frequency increasing or decreasing depending on the extent of the potential safety hazards given the time and place) or an alert is sounded. The recent study by Oman et al. (2009) found that approximately 70% of fatigue-related accidents involved alerter equipped locomotives. This finding suggests that fatigued persons are able to hit a button every few minutes and that this action is insufficient to keep them alert. Locomotive engineers have reported that this motion becomes so habitual that they sometimes move their arms to push the button when they are sleeping at home in their beds (Oman et al. 2009). New loco- motives will have a positive train control system installed. The rail industry is investigating whether other technologies might also be beneficial in new locomotives or for scheduling purposes (Oman et al. 2009). Military and Space Fatigue has been a long-term concern for the military and the space sectors in the United States and other countries. Each of the U.S. military branches has a division that conducts, spon- sors, and reviews research on fatigue and tests technologies and methods to manage fatigue. These efforts are typically part of broader human factors initiatives. Military research and assessment efforts have focused particularly on fatigue of the war-fighter during sustained operations and, in collabo- ration with the non-military sector, on fatigue in transpor- tation modes (land, air, and sea). The military has focused particularly on methods for predicting fatigue and its impacts, countermeasures including pharmaceuticals, and on incor- porating fatigue in operations planning and management (Caldwell et al. 2009; King 2005; Storm 2008; Kronauer and Stone 2004). The military has also emphasized examination of the effects of fatigue on team performance (see for exam- ple, Darlington et al. 2006) and on performance in complex, multisystem environments (Lawton et al. 2005). In related work, the U.S. military has invested heavily in the development of human performance modeling. As part of this effort, the U.S. Department of Defense funded a project by the Sandia National Laboratories to develop a model of soldier fatigue and its potential impacts on a “system of systems” that better reflects the operational context than previous efforts that focused only on a single system (for example, cockpit operations) (Lawton et al. 2005). The military branches have monitored or collaborated in many of the efforts to develop predictive models that better represent the complex opera- tional environment and individual variability, including the Sleep, Activity, Fatigue, and Task Effectiveness (SAFTE) and the Fatigue Avoidance Scheduling Tool (FAST) models.

18 Shift Schedule and Observed Risk of Injury Information on fatigue risk factors associated with injuries and “accidents” or incidents at work from actual occupa- tional settings offers good face validity. However, because these are not controlled experiments, there is always the risk of unknown biases and confounding factors. To the degree that these studies can avoid such pitfalls, they offer observed relative risks which take into account real-life reactions and strategies for managing various shift and extended work- hour situations. Thus, for example, whether workers choose sleep or socializing during recovery periods is accounted for and averaged out over each workforce. Because it is unlikely that these “recovery” factors will be controlled under real highway construction activities, these studies, if well con- ducted, offer a good estimate of how shift and extended work impact safety. Among all publications reviewed, a series from Folkard and colleagues at the Université Paris Descartes, Paris, France, stand out as providing convincing risk factors across a range of industries (Folkard and Tucker 2003; Folkard and Åkerstedt 2004; Folkard and Lombardi 2006; Folkard et al. 2006). These researchers carefully screened published industrial studies from the 1960s through 2001 for sufficient information and potential confounding factors, including changes in work by shift, numbers of workers per shift, availability of first aid, and factors impacting reporting by shift. The researchers con- ducted meta-analyses, but also demonstrated consistency between the individual studies. These studies contain a number of important findings, summarized here. The risk of injury increases in a linear fashion from morning (baseline) to afternoon (18.3%) to night (30.4%). Risk also increases by number of consecutive days on shift, and the risk increase is substantially greater for night versus day shifts. Thus, increased risks (compared to the first day on shift) for the second, third and fourth consecutive days are as follows: for day shift, 2%, 7%, and 17% higher; for night shift, 6%, 17%, and 36% higher. Increased risks associated with increased hours worked per week vary by how these hours are organized. Examples are given for 48-h and 60-h weeks (compared to five 8-h shifts, i.e., a standard 40-h week). In a 48-h work week, increased risk associated with six 8-h shifts is 3% [relative risk (RR) 1.03], and for four 12-h shifts, 25% (RR 1.25). In a 60-h work week, increased risks are even greater: 54% for six 10-hr shifts (RR 1.54); and 62% for five 12-h shifts (RR 1.62). Finally, risk of injury increases linearly with time since last break, by 110% (RR 2.1) at 90 to 120 min compared with risk at 0 to 30 min. Two other studies in the series conducted at Brigham and Women’s Hospital examined fatigue-related injuries and acci- dents among medical interns. In one study, they examined the risk of self-reported percutaneous injuries (PI) among partici- pating interns doing extended work (day shift after night shift) versus non-extended work (day shift with no previous night shift); and night work versus day shift (Ayas et al. 2006). Results were reported per 1,000 opportunities. During extended versus non-extended shift, interns reported 1.310 versus 0.757 PI/1,000 (OR 1.61, 95% CI 1.46 to 1.78), and dur- ing night versus day shift, interns reported 1.48 versus 0.70 PI/1,000 (OR 2.04, 95% CI 1.98 to 2.11). The results from both studies are much less specific than those reported by Folkard and colleagues. However, they provide evidence for the signifi- cant impact of both extended work shifts and type of shift on performance and safety. Another study examined the risk of motor vehicle accident (MVA) as a function of time on shift (Barger et al. 2005). They found that extended shifts of 30 h were common and that interns reported significant sleep restriction during extended shifts. Interns were 2.3 times more likely to report an MVA while driving home from extended shifts compared with shorter shifts, and 5.9 times more likely to report a near miss under the same conditions. Moreover, each additional extended shift per month was associated with a 9.1% increase in the monthly risk of MVAs overall and a 16.2% increase in the monthly risk of an MVA during the commute home after an extended shift. Find- ings from another case-crossover design study in Italy sug- gested an association between long hours at work or without sleep in the prior 24 h and elevated risk for MVAs (Valent et al. 2010). The Construction Industry Injury and Fatality Rates The U.S. construction industry is among the most hazardous for workers. In 2008, construction laborer was one of seven occupations where the rate of injury requiring time away from work was greater than three times the all-worker rate [Bureau of Labor Statistics (BLS), 2009a]. This rate was esti- mated at 383.1 per 10,000 full-time workers for construction laborers, similar to the rate for truck drivers (362.0), and substantially higher than the rate for retail salespersons (90.1), though lower than the rates for occupational groups containing nursing and moving freight by hand (449.0 and 440.3, respectively). The leading cause of these injuries among construction laborers was contact with object or equipment (40%). Of private-sector industries, construction accounted for 11% of all illnesses and injuries requiring time away from work (BLS 2009a). The construction industry experienced the greatest number of fatal occupational inju- ries in 2008 (969), and the fourth highest rate, 9.6 per 100,000 full-time workers, nearly three times the all-worker rate of 3.6 (BLS 2009b).

19 Hispanic Workers Hispanic construction workers have elevated risks for occupa- tional injuries and fatalities. In 2005, Hispanics made up 23% of construction employees overall and 27% of employees in con- struction production; also, 75% of the 2.6 million Hispanics employed in construction overall were foreign born (Center for Construction Research and Training 2008). An analysis of national survey data from 1996 to 2002 found that Hispanics employed in construction were significantly more likely to work in production occupations or to be temporary workers than were non-Hispanic whites; they were also significantly more likely to have had a work-related injury, even when the effects of occupation and education were taken into account (Dong et al. 2010). In 2000, Hispanic construction workers were nearly twice as likely to die from occupational injuries than non-His- panic workers (Dong and Platner 2004), and between 2003 and 2006 the construction industry led occupational deaths among Hispanics at 34% (Centers for Disease Control and Prevention 2008). In addition, occupational deaths among Hispanic con- struction workers may be undercounted due to these workers’ overrepresentation in the temporary or informal sector work- force, whose members are not typically eligible for workers compensation or other benefits (Dong and Platner 2004). Finally, communication barriers between mostly non-Hispanic managers and a growing Hispanic workforce, many of whom report not speaking English comfortably, are a significant chal- lenge to efforts to improve workplace safety (Center for Con- struction Research and Training 2008; Harrington et al. 2009; Centers for Disease Control and Prevention 2008). Road Construction Workers at road construction sites are at risk from both auto- mobile traffic passing by and from construction equipment and car and truck traffic on the site. From 1995 to 2002, according to the BLS Census of Fatal Occupational Injuries, 844 workers were killed while working at road construction sites (Pegula 2004). Of these, 509, or about 60%, were struck by a vehicle or mobile equipment. Of those 509 fatalities, only 28% involved automobiles; the majority of workers were struck by trucks (54%), and others were struck by construc- tion machinery (12%). Almost half of workers killed by vehi- cle or mobile equipment were flagging or directing traffic (18%) or walking in or near the roadway (28%) when they were struck. Another source indicates that of all traffic work zone fatalities, 71% were the result of the worker being struck by a vehicle working in the zone or entering the work zone (Center for Construction Research and Training 2008). An analysis of New York State Department of Transportation construction project accidents from 1990 through 2001 found that 22 accidents involving worker fatalities and 356 accidents involving hospitalizations of workers occurred in the con- struction work area, as compared with 14 and 91, respectively, in traffic (Mohan and Zech 2005). The leading source of fatal and hospital-level accidents in the construction area was large equipment or worker vehicle (45.5% and 38.3%, respectively). A study of accident data in the state of Illinois concluded that highway construction workers have approximately five times more risk of an occupational accident during nighttime versus daytime work (Arditi et al. 2007). Fatigue and Construction Injuries While there exists a volume of research demonstrating links between extended work hours, overtime, and shift work with occupational injuries and with motor vehicle accidents while commuting home, research on the relationship between fatigue and injuries or fatalities specific to the construction industry is limited. Two different studies using the same national survey dataset (National Longitudinal Survey of Youth) found links between workplace injuries and work schedule among con- struction workers. In the first, construction workers working more than 8 h per day (OR 1.02), more than 50 h per week (OR 1.98), or starting work before 7:00 a.m. (OR 1.28) were all found to be more likely than others to report a work-related injury after adjustment for other factors (Dong 2005). Signifi- cant findings from the second study also include a positive association between overtime work and likelihood of injury relative to not working overtime (OR 1.48), as well as increased risk of injury related to working the evening shift compared with working the day shift (OR 2.86) (Dembe et al. 2008). An analysis of workers compensation claims and payroll data for 2,843 construction contracts associated payroll overtime of >20% with higher rates of non-lost-time injuries among workers (Lowery et al. 1998). Two related studies linked sleep disorders (determined from self-reported factors) with ele- vated risk for work-related injuries requiring substantial time off work or hospitalization and for injuries from moving objects, which the authors speculated could be due to impaired vigilance (Chau et al. 2004; Chau et al. 2004). None of the studies of construction workers, and only a few occupational studies, were able to establish timing of injuries in relation to work or sleep schedule, or to directly measure neuro- biological factors underlying fatigue. However, these studies do provide some contextual evidence that supports the hypoth- esized relationships among work scheduling, fatigue, and adverse safety outcomes. Work schedules may affect the neu ro- behavioral performance of construction workers through cir- cadian influence, affecting safety. Also, extended duration at work and poor sleep quality may each affect the safe perfor- mance of tasks because of decreased opportunity for sleep. Construction laborers may be at increased risk for occupational injury and death for reasons related to both a riskier-than-usual

20 work environment, which may be especially hazardous for fatigued workers through impaired vigilance, and scheduling practices and social norms, which can contribute to experi- ence of fatigue at work. An ethnographic study of construc- tion workers indicates that both management and personal pressures to work extended hours are substantial, which work- ers recognize can lead to elevated risk of adverse events at work (Goldenhar et al. 2003). Integrated Fatigue Risk Management Approaches The data reviewed above, as well as the general trend in fatigue and sleep research, suggest that the best way to address the problem of fatigue in operational settings is with an integrated approach. That is, no single measure or intervention is likely to be particularly effective, but requires a combination of training and education, schedule risk assessment, healthy sleep, and fatigue countermeasures (Rosekind et al. 2006; Caldwell et al. 2008). This section reviews the elements of inte- grated approaches based on recent reports of implementa- tions in operational settings. Background Integrated approaches to fatigue management trace their ori- gin to the National Aeronautics and Space Administration (NASA) Ames Fatigue/Jet Lag program, later renamed the Fatigue Countermeasures Program. This nearly 20-year research and outreach program performed groundbreaking research on the causes and consequences of fatigue in aviation operations, and it established best research practices that are carried out today. Additionally, an education and training module was developed and presented to more than 2,500 indi- viduals, and the material in the program served as a basis for individual airlines to develop their own alertness management systems. Many of the aspects of integrated approaches, such as education and training, countermeasures, and healthy sleep habits were originally developed and articulated in this pro- gram. Currently this program is not supported within NASA, but the website, at human-factors.arc.nasa.gov, is an excellent resource for the various research studies and interventions evaluated, and the program was enormously influential in helping to establish similar activities across the U.S. DOT. Components of Integrated Fatigue Management Caldwell et al. (2008) delineate the principal strategies for managing alertness in operational contexts. The first element, and arguably the most important, is that management and staff understand the nature of fatigue. There has been a ten- dency over time to think of fatigue simply as a “state of mind” that can be overcome with “professionalism” or “endurance.” However, these philosophies lead to undesirable results from persons working to the point of safety vulnerability and beyond (e.g., falling asleep on the job). Thus, establishing this understanding concerning the physiological basis of fatigue, how it is manifest in work situations, and what can be done about it are the key components of a training and education program. A variety of material exists for an organization to create its own educational programs, such as the Fatigue Management Reference developed by the DOT Human Fac- tors Coordinating Committee (McCallum et al. 2003). Addi- tional elements described by Caldwell et al. (2008) include • Recognition of individual differences in fatigue vulnerability; • Strategies for improving work and rest scheduling; • Techniques for optimizing sleep (including good sleep hab- its and use of sleep inducing medication when necessary); • Techniques for optimizing circadian adjustments; and • Techniques for temporarily mitigating fatigue (“counter- measures” including rest breaks, napping, and judicious use of caffeine). Caldwell et al. (2008) delineate the major factors in opera- tional environments that are associated with fatigue (Table 2.4); these factors operate across a range of work environments. The challenge for researchers and practitioners in specific operational domains is to discover how these factors manifest themselves and impact the workforce. Implementation of integrated alertness or fatigue manage- ment programs has been shown to have measurable, beneficial impacts. Rosekind et al. (2006) report on the development and evaluation of an alertness management program for a major commercial airline. The program included education involv- ing the basics of sleep and fatigue, alertness strategies, assess- ment of schedule risks, and testing of alternative schedules. Table 2.4. Factors Leading to Fatigue Factors Leading to Fatigue Long hours in a given shift Working long shifts several days in a row Work and/or sleep schedules irregular or unpredictable Critical tasks performed during circadian low points (night, mid-afternoon) Insufficient sleep immediately prior to work shift Insufficient sleep for several days prior to work shift Work requires sustained attention Work environment is dimly lit and quiet Physical or mental stress present (e.g., high-speed traffic, noise, and vibration from heavy equipment)

21 Pilots were measured on a variety of variables before and after the alertness program intervention (knowledge, sleep dura- tion, PVT performance), and the results indicated that all measures improved significantly following the implementa- tion of the alertness management program. In particular, fol- lowing the program intervention, pilots slept an average of 1 h and 9 min longer while on flight trips. Dawson and McCulloch (2005) discuss managing the oppor- tunities for sleep as a key component of managing fatigue in the operational setting. Their concept is based on a hierarchi- cal model of hazards, error trajectories, and control mecha- nisms. At the base of the model is the opportunity for sleep and the average sleep obtained; the key control mechanisms for this level are work-hour rule (“hours of service” or HOS in their model) and aggregate prior sleep/wake modeling (PSWM) to assess the likelihood of fatigue. They focus par- ticularly on prior sleep in the past 24 and 48 h, as this would be relatively simple to measure from employees in an opera- tional setting. The opportunities for intervention, indicated by the breadth of the inverted pyramid in the figure, are more numerous at the base, and become increasingly more difficult to implement at higher levels on the risk trajectory. At increas- ing levels of the risk trajectory, more refined data and control mechanisms are used, and the opportunities for intervention are less available. Based on various modeling and empirical studies, Dawson and McCulloch suggest that sleep reduction of less than 5 h in the previous 24, or 12 h in the prior 48 is likely to lead to impairment. An adaptation of Dawson and McCulloch’s model includes focused countermeasures spe- cific to the operational environment of road construction that address different levels in the error trajectory (Figure 2.2). Fatigue Countermeasures Hours of Service (HOS) Most fatigue-oriented regulations in the United States consist of HOS limits and rest break requirements. For example, the current HOS regulations for the U.S. DOT specify both driv- ing time limits as well as an hours-off component; that is, after Adapted from Dawson and McCulloch 2005. Risk Factor Error Trajectory Focused Countermeasure Sleep obtained Individual choice of time use Training and strategies to optimize available break time for recovery Fatigue-related errors Task performance lapses, injuries, accidents Work zone safety practices, oversight procedures Sleep opportunity and Circadian factors On-the-job fatigue Insufficient break length Behavioral symptoms, inattention, poor decision- making Revised scheduling Symptom checklists, supervisor and peer observation, rest/nap breaks, appropriate use of caffeine Figure 2.2. Fatigue risk trajectory linking risk factors, fatigue impacts, and potential classes of countermeasure.

22 34 consecutive hours of off-duty time drivers can begin a new 7-day period during which they can drive or be on duty for a cumulative total of 70 h. (On-duty time can include non- driving time, such as rest breaks and time loading or unload- ing, time spent at weigh stations, etc.) This means the 7-day clock restarts after a 34-h off-duty period. In some sectors, these regulations have not been updated for decades. These early regulations did not have a strong scientific basis (Jackson et al. 2009; Hursh 2009). Regulators, fatigue scientists, and technol- ogy developers are attempting to address this situation. Work Schedule Assessment and Management Until recently, biomathematical models of the sleep–wake cycle and fatigue have primarily been tools for basic research, seek- ing to integrate the empirical data obtained from studies by researchers from a wide range of disciplines. However, applied biomathematical fatigue technologies are now being used to predict the prevalence and extent of fatigue for work groups and to evaluate alternative work and non-work schedules. These biomathematical tools do not directly measure individuals’ bio- logical processes; instead, they use projected or past work hours and work schedules to predict workforce fatigue levels based on how these factors, on average, are expected to impact the sleep and wake cycle. Torgovitsky et al. (2009) are building a model specifically designed to assess an individual’s response (rather than a group average response) to work schedules. Others have proposed adaptive modeling techniques for individualized fatigue predictions (Van Dongen et al. 2007). Researchers have historically used biomathematical fatigue models to investigate how endogenous sleep and wake pro- cesses correlate with indications of fatigue and/or perfor- mance impairment. A key goal was to improve the validity of these models with respect to predicting, detecting, and estimating fatigue. Tools now exists that use biomathemati- cal fatigue models to evaluate alternative work schedules and optimize schedule solutions by balancing fatigue-informed practices and business needs. These biomathematical sched- uling technologies provide feedback regarding “time-at- risk” for particular schedule solutions and predict the likelihood of fatigue for the work group. Individual variability on mul- tiple parameters has thus far thwarted efforts to develop models capable of predicting fatigue for a particular indi- vidual. Advanced scheduling tools incorporate additional factors besides work schedules, such as workload, nature of the task (especially monotony), environmental conditions, and naps. Napping Naps are among the most efficacious countermeasures for fatigue, as they contribute approximately hour-for-hour to total amount of sleep during the day (Mollicone et al. 2008). Naps thus contribute to reducing the homeostatic pressure for sleep in proportion to nap duration (Sallinen et al. 1998; Schweitzer et al. 2006). Napping is particularly effective when scheduled defensively, that is, before sleep loss is incurred (Dinges et al. 1986; Cote and Milner 2009). Very brief naps (10 to 20 min, sometimes called “power naps”) may also be effective to counteract fatigue (Brooks and Lack 2006), but the duration of the alerting effect of such brief naps has not been determined and is probably relatively short. A side effect of napping is the potential for sleep inertia. Caffeine The most widely used operational countermeasure for fatigue is caffeine (Roehrs and Roth 2008), consumed through a variety of beverages and foods (of widely varying dosing) and even available in the form of a precisely dosed, fast-acting gum (Kamimori et al. 2002). Higher doses increase the alerting effect of caffeine, but the marginal benefit becomes smaller the higher the dose (Kaplan et al. 1997). For greatest effectiveness, caffeine should be used judiciously at the times when it is needed most, as repeated dosing reduces its efficacy (Roehrs and Roth 2008). Furthermore, with repeated use, tolerance builds up, reducing caffeine’s effects in habitual users and leading to withdrawal effects when one’s daily dose of caffeine is not available (Nehlig 1999). Carefully timed, strategic combinations of caffeine with napping may constitute a countermeasure approach that is superior to either caffeine or napping alone (Schweitzer et al. 2006). However, if caffeine is taken relatively close to bedtime, it may interfere with sleep quality, duration, or both, thereby reducing the homeostatic recuperation of sleep (Drake et al. 2006) and thus becoming counterproductive. Motivation Incentives and other sources of motivation can help with the effort to stay awake and maintain performance in the face of sleepiness (Horne and Pettitt 1985), although there is a limit to how long this compensatory effort can be maintained until the endogenous neurobiological drive for sleep takes over (Doran et al. 2001). Difficult to control because of numerous external influences potentially involved, motivation in most work envi- ronments is not a reliable countermeasure for fatigue. Light Light exposure affects the timing of the circadian process in a phase-response (Minors et al. 1991) and dose-response (Boivin et al. 1996) manner. Brighter light is more effective at shifting the circadian rhythm, but the timing of light expo- sure is also critical in determining the degree of shifting and

23 even the direction of shifting (Duffy and Czeisler 2009). Exposure to light in the late evening delays circadian rhythms, whereas light in the morning advances circadian timing. Daily exposure to (morning) light in individuals on a normal daytime schedule serves to keep the circadian process syn- chronized (entrained) to the 24-h clock (Beersma et al. 1999), but in night and shift workers, light exposure tends to disrupt circadian rhythmicity (Santhi et al. 2008). Yet, planned light exposure patterns can be helpful in managing the circadian process and mitigating fatigue (Lee et al. 2006). Laboratory studies of simulated night shifts have yielded evidence that exposure to intermittent bright light pulses during the shift may be somewhat successful in delaying circadian rhythms and improving reaction times (Smith et al. 2008). Light expo- sure also has an acute (but highly transient) alerting effect (Cajochen 2007), which is particularly strong for short wave- lengths (i.e., from blue and full-spectrum light sources) (Lockley et al. 2006). Summary and Conclusion Biological and environmental factors influence an individual’s need for sleep. Fatigue is a function of both time awake (the homeostatic process) and time of day (the circadian process). Work-related fatigue is therefore a function of shift duration and shift timing: Long shifts may not provide adequate oppor- tunity for rest; work at odd hours may require sleep during the day, resulting in poor sleep quality and less sleep obtained. Sleep deprivation affects both sensations of fatigue and cogni- tion. Cognitive effects can include degraded alertness, diffi- culty concentrating, forgetfulness, and confusion as well as degraded performance on specific tasks. Accumulated sleep deprivation can result in severe performance decrements. Individuals vary in their need for sleep, and therefore, in their responses to sleep loss. Sleep loss can only be countered by obtaining additional sleep; severe sleep loss can require sub- stantial recovery time. Following sleep deprivation, subjective ratings of fatigue return to baseline more quickly than does cognitive performance. Studies of the effects of restricted sleep in operational set- tings demonstrate that task performance is degraded with sleep loss, resulting in, for example, higher rates of medical error, unsafe driving behaviors, and motor vehicle accidents. Fatigue has been implicated as well in aviation and rail acci- dents. Shift workers are particularly prone to fatigue-related injuries if they work night shift, extended shifts (10 or more hours), or work weeks longer than 40 h. Construction work- ers have elevated risks for occupational injury and fatality relative to most other occupational groups. In construction, the likelihood of occupational injury is higher for those work- ing extended shifts, night and evening shifts, and weekly over- time. A work culture that expects workers to be available for overtime work likely contributes to the relatively high risk of adverse safety outcomes. An integrated approach to fatigue risk management is essen- tial and includes instilling in workers and management an understanding of the nature of fatigue. Components of such an approach include improving work/rest scheduling (strategic), developing countermeasures for temporarily mitigating fatigue (immediate), such as caffeine and napping, as well as appro- priate training and monitoring for symptoms of fatigue and reduced performance. Rapid renewal highway construction projects are likely to become more common in the future as planners strive to work around dense and growing populations with increasing traffic volumes. Understanding how the scheduling constraints of rapid renewal projects affect worker fatigue, and therefore pro- ductivity and safety, is key to developing a fatigue management plan that provides effective organizational- and individual- level countermeasures for adverse safety outcomes.