Study design and setting
The study included questionnaires and a 3-week observational field study with actigraphy, sleep diaries, Psychomotor Vigilance Tasks (PVT) (Dinges and Powell 1985), ratings of subjective sleepiness (Karolinska Sleepiness Scale, KSS) (Åkerstedt and Gillberg 1990), and EEG-based sleep recordings. Figure 1 presents the timing of the measurements in relation to shift types and days off. The study was carried out between May 2012 and February 2013.
The participants were shift workers from the maintenance, customer service, and catering units of a Finnish airline company. The volunteers were recruited on the basis of their responses to a questionnaire of shift-specific questions on insomnia and sleepiness (SS-Q, see ‘Shift work disorder’ section below) in our previous study (n = 17) (Viitasalo et al. 2015), in response to an announcement in the company (n = 70), or on the advice of occupational health care (n = 9).
All volunteers completed a pre-questionnaire on health, medication, working times, and shift-specific questions on insomnia and sleepiness (the SS-Q). They also reported symptoms or a diagnosis of obstructive sleep apnoea (OSA), restless legs syndrome (RLS), and depression. If they had symptoms, occupational health care diagnosed the condition in question. The volunteers were invited to the study if their shift schedule during the past 12 months had regularly included night shifts (at least 3 h of work between 23:00 and 06:00) and/or early morning shifts (starting by 06:00). Those with other shift schedules (n = 1), part-time work (n = 4), unsuccessful treatment of medically diagnosed OSA or untreated OSA (less than 5 h of continuous positive airway pressure treatment each night) (n = 3), RLS and/or depression (n = 2), or continual medication affecting sleep (sedatives, antidepressants, or soporific antihistamines) (n = 3) were excluded, as were volunteers who changed workplaces (n = 1). Volunteers who did not fulfil the pre-set criteria and could not be classified into either the SWD or the non-SWD group during the recruitment process were also excluded (too many holiday-related insomnia symptoms or sleepiness: n = 6; not sufficiently often shift-related insomnia symptom(s) and/or sleepiness: n = 22; see criteria for the groups in ‘Shift work disorder’ section below). The study included no pregnant or breastfeeding women.
Of the 54 eligible invited volunteers, ten discontinued for health, personal or unknown reasons. In all, 44 completed all the registrations, including a second round of the SS-Q as part of the research questionnaire. As the time interval between answering the SS-Q in the pre-questionnaire and in the research questionnaire was from weeks to months, there was a risk that insomnia symptoms and/or sleepiness may change. Thus, the study groups were formed on the basis of the second answers to the SS-Q. We excluded two participants due to too many holiday-related insomnia symptoms or sleepiness, and a further eleven because they did not have shift-related symptom(s) sufficiently often. Finally, we had 31 participants in our analyses of two groups: 22 in the SWD group (77% working in maintenance, 18% in customer service, and 5% in catering) and nine in the non-SWD group (89% working in maintenance and 11% in customer service). The selection criteria for the groups are presented in the “Shift work disorder”.
Shift work disorder
Shift work disorder was defined by the SS-Q, which was developed at the Finnish Institute of Occupational Health. The definition is based on the ICSD-2 and is consistent with the updated criteria in the ICSD-3. The SWD symptoms were assessed using six questions in the SS-Q presented in Supplementary table 1 (Online Resource).
Individuals belonged to the SWD group if they often had insomnia symptom(s) and/or sleepiness, which specifically manifested in relation to shifts, but not on holiday. The non-SWD group was composed of those who did not often have insomnia symptoms or sleepiness in relation to shifts or holiday. However, all participants were allowed to have one holiday-related symptom. The symptom that designated a participant as an SWD case could only appear in relation to morning, evening, and/or night shifts, but not in relation to holidays. Supplementary table 1 presents the detailed criteria for the two groups (Online Resource).
According to the ICSD, SWD symptoms should not be better explained by other conditions. To exclude primary conditions with continuing insomnia and/or sleepiness, we did not qualify holiday-related insomnia symptoms and/or sleepiness as indicators of SWD.
Compliance with ICSD-3
The ICSD-3 specifies that the symptoms of SWD should associate with a recurring work schedule that overlaps the usual time for sleep. The frequency of the symptoms was examined to determine whether the SWD cases’ SWD symptoms, defined by the SS-Q, recurred sufficiently frequently. The questions were part of the online questionnaire and were similar to those of the SS-Q (see Supplementary table 1, Online Resource), but the participants answered on a five-point scale, namely ‘never/less than once a month’, ‘less than once a week’, ‘1–2 times per week’, ‘3–5 times per week’, or ‘daily/almost daily’. All SWD cases reported at least one of the symptoms of insomnia (see Questions 1–4) at least once or twice a week (during the last 3 months) or sleepiness (see Questions 5–6), which occurred on holiday no more than ‘rather rarely’. In addition, as the ICSD-3 specifies TST reduction as a symptom of SWD, we examined the field data to determine whether the average daily TST of each SWD case decreased in comparison to that on days off. Each participant’s daily TST decreased by at least 1 h in relation to morning, evening, and/or night shifts, i.e., both SWD and non-SWD cases. Similarly, as the ICSD specifies that sleep diary and actigraphy monitoring should demonstrate a disturbed sleep and wake pattern, we observed that each SWD case’s timing of sleep changed in relation to the timing of shift type and/or day off.
Actigraphy and sleep diary
We measured sleep–wake rhythm using a wrist-worn actigraph (Actiwatch AW7, Cambridge Neurotechnology Ltd, Cambs, UK) and a sleep diary. One-minute epochs were analysed using Actiwatch Activity and Sleep Analysis 7 software (Cambridge Neurotechnology Ltd, Cambs, UK). Actigraphy variables included TST, sleep latency, sleep efficiency, and a fragmentation index of sleep.
Participants filled in a sleep diary twice a day, at bedtime and awakening times, and kept a record of bed, wake-up, shift start, and shift end times. The sleep latency of main sleep, bedtime stress scale from 1 (very calm and relaxed) to 9 (extremely stressed and tense), quality of sleep from 1 (good) to 5 (poor), number of awakenings, and the greatest KSS were evaluated daily (Ingre et al. 2004) (Fig. 1).
The TST of main sleep and the preceding naps, based on actigraphy and sleep diaries, were combined as daily TST. Sleep debt was calculated by subtracting daily TST, based on the sleep diary, from sleep need (see ‘Questionnaires’ section below). We averaged the variables based on actigraphy and the sleep diary from the following periods: (1) days off: from bedtime to bedtime prior to and following all days off that were preceded by a day off (the days on which a night shift ended were not considered days off), (2) morning shifts: from bedtime to bedtime prior to and following all morning shifts, (3) evening shifts: from wake-up time to wake-up time prior to and following all evening shifts that were not followed by a morning shift, and (4) night shifts: from wake-up time to wake-up time prior to and following all night shifts.
Participants performed 10-min PVTs that measured vigilant attention on an HP ipaq 514 mobile phone (Hewlett-Packard Company, Palo Alto, CA, USA) (Karhula et al. 2013), and were instructed to practise these three times (Dinges et al. 1997). The PVTs were performed on six pre-selected days: on 2 days off, during two morning shifts starting by 06:00, and during two night shifts. The participants were instructed to complete the PVT in a quiet place with minimal disturbance between 10:00 and 13:00 on days off, and 1–2 h after the start and 1–2 h before the end of the pre-selected shifts (Fig. 1). The PVT variables included mean reaction time and the number of lapses, that is, reaction times of > 500 ms.
Subjective sleepiness was assessed using the nine-point KSS: Participants evaluated their sleepiness (in the last 5 min) prior to each PVT on a mobile phone and their greatest sleepiness each day in their sleep diary (Fig. 1).
EEG-based sleep recordings
Sleep EEG was recorded from the forehead using a Zeo Sleep Manager (Zeo, Inc., Newton, MA, USA), a consumer-friendly validated wireless single-channel system (Griessenberger et al. 2013). Participants were asked to sleep one rehearsal night wearing the device before the first recording and to consume no alcohol the day before each recording. Recordings were completed at home at the same pre-selected times as the PVT (Fig. 1): after days off, before morning shifts, and after night shifts. Sleep stages were reported every 30 s. Outcome measures included the TST of main sleep, stage R sleep, light sleep (stages 1–2), and slow-wave sleep (SWS).
Participants completed an online questionnaire including items on demographics, chronotype (1 = absolutely morning type, to 4 = absolutely evening type), flexibility in sleeping habits from the Circadian Type Inventory in the Standard Shift Work Index (scale 8–40, high scores indicate a tendency towards flexibility) (Barton et al. 1995), shift work experience, physical exercise during leisure time (1 = not so much, to 4 = several times a week/competitive type), daily consumption of caffeinated drinks (one dose equals 1 dl of coffee, 2 dl of tea, 5 dl of cola, or 3.3 dl of energy drink), alcohol consumption (1 = never, to 5 = at least 4 times a week), smoking (yes or no), the amount of sleep needed to feel rested the next day (sleep need), number of < 11 h returns to work per month, the SS-Q, and the frequency of SS-Q symptoms.
Work shift characteristics, calculated from roster and sleep diary data, were investigated using the Mann–Whitney U test to compare the SWD and non-SWD groups. Questionnaire data were investigated using the independent samples T test in cases of normally distributed scaled and continuous variables, the Mann–Whitney U test in cases of non-normally distributed continuous variables, and Fisher’s exact test in cases of categorical variables to compare the SWD and non-SWD groups.
Each participant had a unique amount of morning, evening, and night shifts and days off in their field data. To consider this, we conducted linear mixed model analysis (LMM) to compare the groups in cases of normally distributed continuous variables. Due to the difference between the ages of the groups, we used Age, in addition to Group as the main effect in the LMMs. Similarly, we performed LMMs using Chronotype, in addition to Group as the main effect. However, the results for the latter are not shown, because they were similar to the results of the LMM adjusted for age. Prior to model fitting, variables were transformed to meet the model assumptions, where necessary (√x: EEG-based TST of main sleep on days off, stage R sleep after night shifts, SWS on days off, SWS after night shifts, and fragmentation index of sleep after evening shifts; √(xmax + 1 − x): light sleep before morning shifts, sleep efficiency on days off, sleep efficiency after evening shifts, and sleep efficiency after night shifts; log10x: SWS before morning shifts; log10(xmax + 1 − x): sleep efficiency before morning shifts; and x0.7: fragmentation index of sleep before morning shifts). We applied the Mann–Whitney U test when transformations of repeated scaled and continuous variables were unsuccessful. We calculated the means, standard deviations (SD), medians, interquartile ranges (IQR), and the U test values of repeated measures from each participant’s averaged values. We used IBM SPSS Statistics 20.0 for the analyses.