Participants included 367 sworn police officers enrolled in the Buffalo Cardio-metabolic Occupational Police Stress (BCOPS) study. All officers were employed by the Buffalo, New York Police Department. Electronic day-by-day work history records and sick leave data, from 1994 to 2015, were obtained from the Buffalo, New York, Payroll department. The payroll data included the start time and duration of work periods and sick leave. Officers were paid for their overtime hours; the department does not use compensatory time off in lieu of paid overtime.

Biomathematical models of fatigue are methodological tools for predicting fatigue, sleepiness and/or performance impairment using sleep–wake data and/or work–rest schedules (Dawson et al., 2011). These tools incorporate our scientific understanding of sleep timing and sleep duration and their relationships with fatigue, sleepiness and performance. In operational settings, the model predictions are used to (1) construct work–rest schedules with adequate sleep opportunities (Dawson and McCulloch, 2005), (2) evaluate the likelihood of on-duty fatigue to determine the extent to which fatigue mitigation strategies are needed (Dawson et al., 2012), (3) develop safety cases (Lamp et al., 2018) and (4) investigate the likelihood of fatigue in post-accident investigations (Price and Coury, 2015). In operational research, biomathematical models of fatigue are useful tools for studying sleep–wake behaviors, on-duty fatigue and job performance during different work–rest schedules (Riedy et al., 2019; Hursh et al., 2006; Åkerstedt et al., 2008; Morris et al., 2018; Sagherian et al., 2018).

In the current operational research, we used two biomathematical models of fatigue – Fatigue Audit InterDyne (FAID) and Three-Process Model of Alertness (TPM) – incorporated into the analytical software FAID Quantum (Roach et al., 2004; Åkerstedt and Folkard, 1995; FAID Quantum User Guide v1.0, 2017). FAID predicted officers’ on-duty fatigue (i.e. FAID scores) directly from their work–rest schedules. FAID scores ranged from 0 to 100+, with higher FAID scores indicating greater likelihood of fatigue. The default threshold for a high likelihood of fatigue is a peak FAID score of 80, which is equivalent to fatigue-related impairment after approximately 21 h of total sleep deprivation (FAID Quantum User Guide v1.0, 2017; Dawson and Reid, 1997). For each shift, a peak FAID score was calculated. The peak FAID score represented the greatest fatigue (i.e. the highest FAID score) predicted during that shift.

FAID Quantum predicted officers’ sleep–wake behaviors directly from their work–rest schedules. TPM predicted on-duty sleepiness (i.e. Karolinska Sleepiness Scale (KSS) ratings) from officers’ predicted sleep–wake behaviors. KSS ratings ranged from 1 to 9, with higher KSS ratings indicating greater likelihood of sleepiness. The default threshold for a high likelihood of sleepiness is a peak KSS rating of 7 (FAID Quantum User Guide v1.0, 2017). Previous research has demonstrated that physiological signs of sleepiness (e.g. heavy eyelids, long eye blinks and slow eye movements) and behavioral signs of sleepiness (e.g. lane deviation and unintentional line crossings) begin to be seen at KSS ratings of 7 (see Åkerstedt et al., 2014 for a review). For each shift, a peak KSS rating was calculated. The peak KSS rating represented the greatest sleepiness (i.e. the highest KSS rating) predicted during that shift.

Data analyses focused on sworn police officers regularly working 10-h shifts. Detectives, lieutenants, sergeants, chief commissioners and other police department personnel were excluded from the data analyses. Sworn police officers regularly working 8-h shifts were also excluded from the data analyses because the 8-h shifts were usually day shifts and were rarely swing shifts or night shifts. Payroll data collected during a suspension or an injury were excluded from all data analyses. All data analyses were conducted using SAS 9.4 (Cary, North Carolina, USA).

Absenteeism was analyzed as a dichotomous dependent variable (two levels: at work, absent from work). To be included as an outcome in the analyses, the shift or absence had to occur on a regularly scheduled workday and had to be preceded by at least one shift in the prior 24 h. The following continuous variables were tested as the main predictors of absenteeism in separate mixed-effects logistic regression models with a logit link function and a random intercept over subjects (PROC GLIMMIX, SAS 9.4): peak FAID scores during the prior shift, sleep in the 24 h prior to the shift and peak KSS ratings during the prior shift. Officers’ sex (categorical: male, female) was included as a main effect in each model. Two-way interactions between the main continuous predictor variables and shift type (categorical: day, swing, night) and the continuous variables and officers’ sex were included. The two-way interactions examined whether the effects of prior fatigue, sleep and/or sleepiness on absenteeism varied as a function of shift type and/or officers’ sex. Continuous predictors were group-mean centered by shift type. As such, the odds ratios are interpreted as deviations from the group (i.e. day, swing or night) averages for prior sleep, on-duty fatigue and on-duty sleepiness.

Absenteeism places an enormous economic burden on employers. The nature of absenteeism in the workplace is likely multifactorial and complex. However, the World Economic Forum has previously reported that the costs of sleep loss-related absenteeism, presenteeism and other negative occupational outcomes total over US$150bn each year in the USA (Azor Hui and Grandner, 2016; World Economic Forum, 2010). In US policing, specifically, absenteeism is, in part, an administrative problem. Absenteeism requires police departments to reassign duties (Violanti et al., 2015) or requires other officers to work overtime to fill in for their absent colleagues. As such, absenteeism aggravates the officer shortage problem at understaffed police departments, and it strains departmental budgets.

The conversation of unplanned absences in policing is thus often focused around the costs to police department budgets and the need to control the abuse of sick leave. However, absenteeism is not strictly an administrative problem. The current research suggests that short-term unplanned absences may represent a response to reduced sleep opportunity. That is, the results suggest that police officers may use absenteeism to cope with and reduce the amount of sleep loss and fatigue they experience. This could be a self-management strategy for fatigue, or it could be unintentional compensatory response to reduced sleep opportunity. Sleep loss and police fatigue resulting from long and irregular work hours may be administratively controllable culprits of unplanned absences in policing. It remains to be determined whether regulating and/or restricting use of sick leave increases police fatigue and fatigue-related risks.

Cross-sectional research conducted in policing and other operational settings has previously suggested that shift work, long work hours, sleep disturbances and sleep disorders increase workplace absenteeism. The longitudinal research presented here suggests that these factors may be related to absenteeism indirectly through sleep loss, fatigue and sleepiness. Officers predicted to obtain less sleep and have greater on-duty fatigue and on-duty sleepiness were more likely to call in sick from work during their next shift. This relationship was particularly evident for the swing shift workers and night shift workers who were predicted to obtain less sleep prior to a shift and have greater on-duty fatigue and sleepiness than day shift workers. On average, officers working night shifts had high levels of sleepiness – sleepiness levels that have previously been associated with increased lane deviation, unintentional lane crossing, slow rolling eye movements and other behavioral and physiological signs of sleepiness (Åkerstedt et al., 2014).

Furthermore, female officers had a greater likelihood of absenteeism than male officers. This is a common finding in previous research, and it is often explained by childcare and family responsibilities (Patton and Johns, 2007). In the policing literature specifically, however, there have been inconsistent findings. In 2009, a systematic review was conducted on all scientific studies on absenteeism among male and female police officers (Körlin et al., 2009). It was concluded that research on sex differences in absenteeism rates among police officers was scarce, and there was an obvious need for additional studies (Körlin et al., 2009). Since this systematic review was published, additional studies have found that female officers tend to have a greater likelihood of absenteeism than male officers (e.g. Fekedulegn et al., 2013b; Svedberg and Alexanderson, 2012). The results presented here add to this growing literature. These results could not be explained by a differential response to prior sleep loss, on-duty fatigue or on-duty sleepiness. The fatigue and sleepiness predictions did not significantly differ for female officers versus male officers. The sleep predictions did significantly differ for female officers versus male officers; however, the difference was negligible, and the significance was likely driven by the low variability in sleep predictions. Violanti et al. (2018) reported similar findings using BCOPS study data and actual self-reported fatigue ratings, i.e. fatigue ratings did not differ by officers’ sex after controlling for shift type.

Officers’ age and years of experience were intentionally not controlled for in the analyses for two reasons. The primary reason is that officers’ age and years of experience significantly differ by shift type. Younger officers tend to work more night shifts, while the older officers tend to work more day shifts. The secondary reason is that biomathematical models’ predictions are derived from workers’ work–rest schedules and/or sleep–wake behaviors. The model estimates do not differ by age or years of experience.

Fatigue in shift work is inevitable; however, mitigating and managing fatigue in policing is critical. Police fatigue has been associated with degraded operational performance, increased risk of injuries and accidents, changes in police–community relationships, long-term health consequences and now workplace absenteeism. Approaches for mitigating and managing fatigue in the operation should be considered. Such an approach should also provide police departments with the operational flexibility needed to schedule officers around-the-clock and meet demands for service.

In policing, a one-size-fits-all approach such as federal hours-of-service regulations that specify duty time limitations and minimum rest break requirements are likely impractical. There is considerable variability from one police department to the next on various environmental, social, economic and labor factors. Many police departments are also chronically understaffed and need scheduling flexibility to meet demands for service. Furthermore, the efficacy of hours-of-service regulations to manage fatigue and improve safety in an operation is continuously being challenged (McDonald, 1981; Jones et al., 2005; Gander et al., 2015; Honn et al., 2019). Operations such as aviation and rail have developed fatigue risk management systems as an alternative, more comprehensive risk management approach for addressing fatigue in the operation while also increasing scheduling flexibility. In fatigue risk management systems, managing fatigue is viewed as a shared responsibility between employees and employers because fatigue is driven by both work demands and other waking activities (Gander et al., 2011; Zaslona et al., 2018).

In 2019, the Working Time Society published a series of consensus statements and best practices with the aim of providing stakeholders with the best information available when developing working time policies and regulations in different operational settings (Popkin and Fischer, 2019). Within the consensus statements, it was stated that the purpose of fatigue risk management systems is, in part, to develop procedures and processes for mitigating and managing fatigue predictively, proactively and reactively (Honn et al., 2019). Examples of such procedures and processes include identifying fatigue hazards in an operation; use of predictive fatigue modeling (i.e. biomathematical models) in work scheduling; implementation of proactive fatigue monitoring; and development of education and training programs, non-punitive fatigue reporting systems and reactive procedures for investigating possible fatigue-related incidents or accidents (Balkin et al., 2011; Satterfield and Van Dongen, 2013; Dawson et al., 2016; Honn et al., 2019).

In policing, there will be situations where the consequences of not doing the job outweigh the consequences of sleep loss and fatigue. In these situations, police and first responders will be at the forefront, and overtime and long and irregular work hours will be inevitable. In everyday policing, however, fatigue should be mitigated and managed. The research presented here suggests that absenteeism may being used as a self-management strategy for fatigue and/or may represent a compensatory response to sleep loss and fatigue.

Limitations of this study include: (1) shifts and absences had to be preceded by a shift in the prior 24 h to be included in the analyses. As a result, the first shift in a work week is not included, and the absenteeism rate is lower than previously reported. Approximately 0.9 percent of the outcomes were absences. The Bureau of Labor Statistics (2020) reported that the absence rate of police and other protective service personnel is closer to 2.5; however, this absence rate includes more than sick leave. (2) Commute times were not accounted for. The FAID Quantum User Guide suggests adjusting shift start and end times for commutes longer than 45 min. Commute data were unavailable between 1994 and 2004. (3) Sleep, fatigue and sleepiness were predicted by a biomathematical model of fatigue. As such, some research questions could not be addressed, such as determining whether there were age effects independent of shift type. (4) Sleep timing, sleep duration, fatigue and sleepiness are predicted at the group level (i.e. for the average individual). Biomathematical models of fatigue do not account for untreated sleep disorders; workplace stress; family demands; and other physiological, non-physiological and psychosocial factors that impact sleep–wake behaviors and on-duty fatigue and sleepiness.