SEARCH

SEARCH BY CITATION

Keywords:

  • air traffic control;
  • naps;
  • neurophysiological alertness;
  • night shift;
  • psychomotor performance;
  • within-subjects field study

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interests
  9. References

The aims of this study were to measure sleep during a planned nap on the night shift; and to use objective measures of performance and alertness to compare the effects of the nap opportunity versus staying awake. Twenty-eight air traffic controllers (mean age 36 years, nine women) completed four night shifts (two with early starts and two with late starts). Each type of night shift (early/late start) included a 40-min planned nap opportunity on one occasion and no nap on the other. Polysomnographic data were used to measure sleep and waking alertness [spectral power in the electroencephalogram (EEG) during the last hour of the night shift and the occurrence of slow rolling eye movements (SEMs) subsequent to the nap]. Psychomotor performance task [Psychomotor Vigilance Task (PVT)] was completed at the beginning and end of the shift, and after the nap (or an equivalent time if no nap was taken). Nap sleep latencies were relatively long (mean = 19 min) and total sleep time short (mean = 18 min), with minimal slow wave sleep (SWS, mean = 0%), and no rapid eye movement sleep. Nap sleep resulted in improved PVT performance (mean and slowest 10% of reaction time events), decreased spectral power in the EEG and reduced the likelihood of SEMs. The occurrence of SWS in the nap decreased spectral power in the EEG. This study indicates that although sleep taken at work is likely to be short and of poor quality it still results in an improvement in objective measures of alertness and performance.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interests
  9. References

Night workers commonly experience high levels of sleep loss and sleepiness (Åkerstedt, 1998; Åkerstedt et al., 2002), which can potentially lead to cumulative performance degradation (Belenkey et al., 2003; Dinges et al., 1997; Van Dongen et al., 2003) and increased accident risk (Åkerstedt et al., 2002; Dinges, 1995; Mitler et al., 1988). Short naps are frequently recommended as a strategy to minimise these effects (Åkerstedt, 2006; Bonneford et al., 2004; Schweitzer et al., 2006).

Napping preceding night work has been demonstrated to be an effective countermeasure, both in laboratory studies and field settings (Dinges et al., 1987; Harma et al., 1989; Schweitzer et al., 1992, 2006). However, depending on the shift start time, this strategy may be difficult to implement because of the evening wake maintenance zone and/or competing time demands. On the other hand, napping during a scheduled break during the night shift has the advantages that sleep is taken at a more propitious time in the circadian cycle, and that the duration of wakefulness is reduced prior to the circadian nadir and the end of the shift, when performance degradation is expected to be greatest.

Laboratory-based studies have demonstrated improvement on some performance measures following 1-h naps at night (Gillberg, 1984; Rogers et al., 1989), while 30–50 min naps appear to limit, rather than reverse, performance decline across the night (Sallinen et al., 1998). Sleep opportunities later in the night contain more total sleep (Gillberg, 1984) more non-rapid eye movement stage 2 (S2), more slow wave sleep (SWS), have shorter latencies, and fewer awakenings (Sallinen et al., 1998) and may be more effective in maintaining alertness and performance (Gillberg, 1984).

Less evidence is available from workplace napping studies. A 1-year study of industrial plant workers allowed a 1-h nap on the night shift noted that napping was feasible, accepted, improved worker’s satisfaction with night work, and resulted in higher self-reported vigilance after the nap and a general improvement in quality of life (Bonneford et al., 2001). Among aircraft engineers, a 20-min nap (relative to no nap) improved response times on a vigilance task at the end of a 12-h night shift, but only on the first of two night shifts (Purnell et al., 2002). Nap sleep was not recorded and the nap did not improve subjective ratings of fatigue or sleepiness while driving home from work. Flight crewmembers allowed a 40-min nap opportunity during night flights had faster reaction times and fewer micro-sleep events than crewmembers who did not nap (Rosekind et al., 1994). Similarly, among physicians and nurses in an Emergency Department, a 40-min nap opportunity on a 12-h night shift resulted in fewer lapses, faster intravenous catheter insertion, and improved subjective alertness and mood (Smith-Coggins et al., 2006). In the latter two studies, nap sleep was recorded polysomnographically. Average total sleep time (TST) was between 25 and 27 min, with short sleep latencies, no rapid eye movement (REM) sleep, and a limited amount of SWS.

As with flight crew and physicians, air traffic controllers (ATCOs) are required to work at night while also maintaining high standards of performance. The value of napping as a countermeasure to sleepiness in this work context has not previously been assessed. The aim of the present study was therefore to objectively evaluate the effectiveness of a planned 40-min nap opportunity on the night shift for maintaining the performance and alertness of ATCOs through to the end of their shift. We were also interested in determining how well ATCOs could nap in their work setting, and in exploring associations between nap sleep structure and subsequent performance and alertness.

Compared with previous field research, this study has the particular strengths of employing a within-subjects design combined with polysomnographic measures of nap sleep and subsequent neurophysiological alertness. Alertness was recorded while controllers performed their normal duties, and performance was objectively assessed using the Psychomotor Vigilance Task (PVT) (Lisper and Kjellberg, 1972; Powell, 1999). The inclusion in the roster of night shifts starting and ending at two different times, allowed within-subjects evaluation of naps both early and late in the shift.

Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interests
  9. References

Procedure

Ethical approval for the study was received from the Wellington Regional Ethics Committee. Participation was voluntary and confidential, and written informed consent was obtained. Volunteers were sought from area ATCOs, who controlled upper level airspace, working in the primary air traffic control centre in New Zealand. All ATCOs complete an annual medical assessment for fitness for duty, have a legal responsibility to report fit for work, and none reported suffering a diagnosed sleep disorder. Of 35 eligible ATCOs approached, 28 agreed to participate in the study (19 men, age range 26–56 years, mean = 36 years; and nine women, age range 26–39 years, mean = 34 years). Participants had worked in air traffic control for a mean of 11.3 years (range 2–34 years).

Each ATCO was monitored across four night shifts: two ‘early’ night shifts (22:30–06:00) and two ‘late’ night shifts (23:30–06:30). On one night shift of each type, a 40-min nap opportunity began approximately 2 h into the shift (early shift, mean start time nap 00:23, SD = 13 min), or 3 h into the shift (late shift, mean nap start time nap 02:22, SD = 13 min) during a normal break from controlling. ATCOs were free to choose where to nap, and all selected a location where they could be completely supine (no ATCO chose to nap in a seat). Most ATCOs utilized a room that could be completely darkened and was located in a quiet area of the building away from the radar facility. All naps occurred during the planned opportunities only. On the remaining two night shifts, ATCOs were asked to remain awake during their break.

The four study conditions were thus: early night shift with a nap (EN); early night shift without a nap (EW); late night shift with a nap (LN); and a late night shift without a nap (LW). It was not possible to reorganise the roster to achieve counter-balancing of the four study conditions or identical amounts of time between each study condition. All data were collected over a period of 8 months, with ATCOs completing the four study conditions over a mean of 12 weeks (SD = 6.5 weeks). There was a mean of 27 days (SD = 27 days) between the start of one data collection period and the start of another.

The night shifts were the last shift in a rapid backward rotating sequence of four shifts (evening, day, morning and night). Sleep across 2 days off, the sequence of 4 shifts, and the subsequent 2 days off was monitored using actigraphy (Mini-Mitter/Respironics; Bend, OR, USA) and sleep diaries (reported in Signal and Gander, 2007). The morning shift prior to the night shift was rostered to start at 06:30 and finish between 11:00–12:00, with an afternoon break of 10.5–11.5 h before the night shift.

Immediately prior to each night shift, ATCOs had electrodes connected for sleep and alertness monitoring. They completed the PVT three times across each night shift: immediately prior to the shift, after the nap opportunity and just before returning to work (or an equivalent time on the non-napping shifts), and at the end of the shift. To measure neurophysiological alertness, electro-oculographic (EOG) recordings across the second half of the night shift were viewed for the presence of slow rolling eye movements (SEMs) and electroencephalographic (EEG) recordings across the last hour of the night shift were evaluated using spectral analysis. Fig. 1 summarises the timing of the performance tests, the duration of the post-nap EOG data analysed for SEMs, and the timing of the 1-h sections of EEG data subject to spectral analysis.

image

Figure 1.  Timing of events across the night shift. inline image, psychomotor vigilance task (PVT) completed [mean test time (SD) 1E = 22:34 (0:28), 2E = 02:00 (0:27), 3E = 04:31 (0:25), 1L = 23:50 (0:24), 2L = 03:55 (0:24), 3L = 06:42 (0:22)]; solid thick lines with single arrows, nap periods (if provided); solid lines with double arrows, work periods; dashed lines, sections of recording assessed for slow rolling eye movements [mean duration (SD) SEM-E = 2.1 h (0.2), SEM-L = 2.5 (0.2)]; dotted lines, sections of electroencephalogram recording subject to spectral analysis [mean start time (SD) EEG-E = 03:16 (0:14), EEG-L = 05:29 (0:06)].

Download figure to PowerPoint

Measures

Sleep and waking alertness were monitored using an A10 recorder (Embla, Broomfield, CO, USA) with the following electrode montage: EEG from standard 10–20 sites, C4, P4, O2, Oz, and A1; EOG from the left and right outer canthi (referenced to A1); and bipolar electromyogram from the mentalis/submentalis muscles. Participants were able to move around freely, with the recorder either carried in a small pouch on the waist or over the shoulder. All input signals were sampled simultaneously at 2000 Hz and digitised to 16 bit resolution by a Sigma-Delta AD converter. Data was then down sampled to 200 Hz, pass-band filtered at 0.5–90 Hz, and stored on a removable hard disk in the A10 recorder.

Participants were asked to push the event marker button when they started and stopped trying to sleep during the nap opportunity. Where this information was unavailable, actigraphy and sleep diary data were used, together with examination of the polysomnographic data, to determine the start and end of sleep. Two experienced sleep scorers (LS and MM) independently scored the nap sleep according to the criteria of Rechtschaffen and Kales (1968) with EEG arousals scored according to the criteria of the American Sleep Disorders Association (1992). For 11/54 available naps, the inter-rater reliability was just below 70%. These recordings were re-scored on a consensus basis. Mean inter-rater reliability on the remaining sessions was 92.9% (SD = 4.69%).

Air traffic controllers completed a questionnaire at the end of each night shift which included questions on the nap opportunity. All ratings were made on a 10-cm visual analogue scale and included ratings of the degree of difficulty falling asleep during the nap (0 = not at all difficult, 10 = very difficult), the quality of the nap sleep (0 = very poor, 10 = very good), the extent to which they woke refreshed (0 = not at all, 10 = completely), and whether the nap was helpful in managing fatigue across the remainder of the night shift (0 = very unhelpful, 10 = very helpful).

Electroencephalogram alertness during the last hour of the night shift was evaluated with the help of a purpose-built LabVIEW programme. EEG data from Oz– O2 were visually screened in 5.12 s epochs to remove artefact, then subject to Fast Fourier Transform (50% overlapping epochs with a Hanning window applied). Two individuals (LS and JG) undertook the artefact screening, with a resulting kappa of 0.60 for the 6 h of data that were double-scored (0.4–0.75 is regarded as ‘fair to good agreement’) (Fleiss, 1981). Spectral power in artefact-free data was summed into four frequency bands: delta (1.56–3.91 Hz); theta (4.10–7.81 Hz); alpha (8.01–11.91 Hz); and beta (12.11–15.82 Hz). These data were log transformed for subsequent statistical analyses. A square root function was also applied to beta spectral power to reduce the impact of the skewed distribution.

Electro-oculographic recordings during the second half of the shift (after the nap and the second performance test session) were viewed for the presence of SEMs with a minimum amplitude of 10 μV per cm, a frequency between 0.2 and 0.67 Hz, and a relatively smooth movement of the trace (Santamaria and Chiappa, 1987). Sharper, faster eye movements were categorised as SEMs if there was clear alpha activity in the EEG traces and/or alpha could be seen in the EOG traces. A single trained sleep scorer (LS) viewed all recordings for SEMs and a second trained sleep scorer (NM) independently scored 20% of the recordings. In 47% of recordings, there was complete agreement, while in the remaining recordings NM marked 1–2 more SEMs. The more conservative scores (LS) were used in analyses. The total time during which SEMs were scored was presented as a percentage of the recording duration. The distribution of this variable was highly skewed so it was dichotomised (no SEMS versus SEMS) for subsequent statistical analyses.

In addition to scoring post-nap recordings for SEMs, a single trained sleep scorer (LS) scored each 30-s epoch for sleep, according to the criteria of Rechtschaffen and Kales (1968). A second trained sleep scorer (NM) independently scored 20% of these recordings. Inter-rater agreement was 100%.

Performance was assessed with the PVT, using standard test parameters (test length = 10 min and inter-stimulus interval ranging between 2 and 10 s) (Powell, 1999). All tests were completed in a room without distractions. Mean reaction time, slowest 10% of responses and fastest 10% of responses were used as dependent variables (for analyses all were reciprocally transformed) and lapses (any response longer than 500 ms) were dichotomised.

Statistical analyses

Data available for analysis included all four study nights from 25 participants, two study nights (LN and LW) from one participant, and one study night from one participant (excluded from analyses). For one further participant, on one occasion, because of an anomalous combination of operational and staffing demands, the nap on the late night shift was taken at a similar time to the nap on the early night shift. Data from this shift were excluded from analyses. The final numbers of participants for each condition are: EN = 26, EW = 26, LN = 26 and LW = 27. One alertness recording was unusable. The amount of missing PVT data was negligible (4%) and spread across three of the four study conditions.

For continuous variables, mixed linear analyses of covariance (ancova) were carried out (SAS system for Windows, version 8.02, Cary, NC, USA), while mixed model logistic regression was used for dichotomised variables. Correlation matrices were used to determine the most suitable covariance structures for repeated measurements (Littell et al., 1998) and checked for fit using the Schwarz Bayesian Criteria (generally a compound symmetric or autoregressive structure gave the best fit). Where appropriate, the repeated effect was coupled with an inter-individual random effect. When main and interaction effects were statistically significant, post hoc t tests were used to investigate comparisons of interest.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interests
  9. References

Nap sleep

During the night shift, on 5/52 nap opportunities five different ATCOs obtained no sleep (three on the early night shift and two on the late night shift). Three of these ATCOs obtained <1 min sleep on the other napping occasion, while the other two obtained 6 and 17 min of sleep, respectively. Table 1 details the characteristics of nap sleep on each night shift. Excluding those who did not sleep, mean sleep duration on the early shift = 19 min, and on the late shift = 20 min.

Table 1.   Details of nap sleep duration and structure, and factors influencing nap sleep
 Early (n = 26)Late (n = 26)ANCOVA results
Mean (SD)RangeMean (SD)Range
  1. TST, total sleep time; SOL, sleep onset latency; WASO, wake after sleep onset; SE, sleep efficiency; SWS, slow wave sleep.

  2. Variables not normally distributed, median presented not mean.

  3. *Early n = 22, and Late n = 23, one individual on each shift had 30 s TST and therefore variables not calculated.

  4. Variables transformed for ANCOVA after inspection of residuals: WASO log transformed; S1 sleep square root transformed; S2 sleep square root transformed; SWS dichotomised and logistic regression used for analysis; awakening index log transformed.

Start of nap (hours)00:23 (13 min)00:03–00:5302:22 (13 min)02:03–02:56NA
Duration of nap opportunity (min)44 (3)38–5142 (5)28–50NA
TST (min)17 (12)0–3719 (12)0–47Prior wake F(1,42) = 6.94, P = 0.012
WASO (min)* 20.5–14 20–20No significant effects
SOL (min)*21 (11)6–4717 (10)2–37Prior wake F(1,36) = 10.21, P = .003
SE (%)43 (25)1–8148 (26)1–95Prior wake F(1,37) = 8.64, P = .006
S1 (% TST)†‡275–100324–100No significant effects
S2 (% TST)†‡700–96620–96No significant effects
SWS (% TST)†‡ 00–59 00–51Prior wake F(1,37) = 6.68, P = .014
Awakening index†‡ 72–60 60–67No significant effects
Arousal index*243–96247–88No significant effects

No REM sleep was observed during the naps. On the early shift, three participants obtained non-REM stage 1 (S1) sleep only, while nine participants entered SWS. On the late shift, one participant had S1 sleep only, and 11 entered SWS. The majority of ATCOs were woken from S2 at the end of the nap opportunity (early 52%, late 63%), fewer from SWS (early 26%, late 12%) and only a small percentage from S1 (early 9%, late 8%). The remainder were already awake (early 13%, late 17%).

Air traffic controllers reported that they found it moderately difficult to fall asleep (mean = 5.97, SD = 2.90), and that the quality of the nap sleep was relatively poor (mean = 3.80, SD = 2.84), but that they woke feeling somewhat refreshed (mean = 4.48, SD = 2.13) and that the nap was reasonably helpful in helping to manage fatigue across the remainder of the night shift (mean = 5.63, SD = 2.24).

On the night prior the night shift, ATCOs averaged 6.0 h of sleep (SD = 0.82, range 4.3–8.1), and 90% napped between the end of the morning shift and the start of the night shift (mean nap duration = 2.2 h, SD = 0.90 h, range 0.52–4.85 h) (Signal and Gander, 2007). The mean sleep obtained in the 24 h prior to early shift was 7.1 h (SD = 1.2 h, range 4.5–9.2 h), while for the late shift the mean was 5.4 h (SD = 1.1 h, range 3.5–8.0 h). The median duration of continuous wakefulness prior to the naps on early shift was 7.2 h (range 2.6–19.5 h) while for naps on the late shifts the median was 6.3 h (range 3.0–20.9 h).

Mixed model ancovas were conducted to determine whether the shift worked (early/late), duration of prior wakefulness, or the amount of sleep in the 24 h preceding the night shift influenced the duration and/or structure of the nap sleep (Table 1). Longer wakefulness prior to the night shift increased the amount of sleep obtained during the nap, the efficiency of sleep taken during the nap, the likelihood of SWS occurring in the nap, and decreased sleep onset latency (SOL).

Sleep outside the nap opportunity

On 2/102 night shifts (one LN shift, one LW shift), two participants fell asleep outside the nap opportunity. In one case, two 30-s epochs of S1 were scored 5 min apart at approximately 05:00 h (sleep on the night prior = 5.5 h, afternoon nap prior to the night shift = 3.0 h). On the second occasion, a participant entered S1 for 60 s at approximately 05:15 h (sleep on the night prior = 5.3 h, afternoon nap prior to the night shift = 1.1 h, nap sleep during the night shift = 24 min).

Effects of nap sleep

Table 2 summarises analyses that examine the relationship between the opportunity to nap (regardless of the amount of sleep obtained) and subsequent PVT performance and alertness. Table 3 presents a similar series of analyses that are focussed on the amount of sleep obtained during the nap and subsequent performance and alertness. All models included interactions between the independent variables, but findings related to the interactions are only presented where these were statistically significant. The provision of a nap opportunity (Table 2) and greater amounts of nap sleep (Table 3) were associated with faster mean reaction times and faster performance in the slowest 10% domain. There was no effect of shift type (early/late) on reaction time performance, but across the shift mean reaction time and the slowest 10% of responses slowed, and the likelihood of lapsing increased. Post-hoc tests from the ancovas that modelled the amount of sleep obtained during the nap indicated that for mean reaction time: test 3 > test 1, t = 2.81, P = 0.007; for slowest 10%, test 2 > test 1, t = 2.15, P = 0.039, and test 3 > test 1, t = 3.27, P = 0.003; for number of lapses, test 3 > test 1 t = −3.46, P < 0.001; test 3 > test 2, t = −3.55, P < 0.001.

Table 2.   Relationships between study condition, and PVT performance and alertness (ANCOVAS)
 Shift (early and late)Test session (1st, 2nd and 3rd)Nap (nap and no nap) Shift × nap
  1. NS, not significant; NA, not applicable to these analyses; SEMs, slow rolling eye movements.

  2. Logistic multiple regression.

Mean reaction time (reciprocal)NSF(2,44.1) = 3.96, P = 0.026F(1,65.8) = 5.85, P = 0.018NS
Fastest 10% responses (reciprocal)NSNSNSNS
Slowest 10% responses (reciprocal)NSF(2,33) = 5.30, P = 0.010F(1,46.5) = 5.79, P = 0.020NS
Lapses (dichotomised)NSF(2,78) = 8.02, < 0.001NSNS
Delta power (log)F(1,501) = 44.94, < 0.001NAF(1,497) = 5.52, P = 0.019F(1,494) = 9.93, P = 0.002
Theta power (log)F(1,453) = 51.76, < 0.001NAF(1,450) = 6.69, P = 0.010F(1,449) = 5.34, P = 0.021
Alpha power (log)F(1,486) = 38.99, < 0.001NAF(1,482) = 7.65, P = 0.006F(1,481) = 9.47, P = 0.002
Beta power (log and square root)F(1,404) = 38.04, < 0.001NANSF(1,400) = 12.64, < 0.001
SEMs (dichotomised)NSNAF(1,74) = 5.88, P = 0.018NS
Table 3.   Relationships between nap sleep duration, and PVT performance and alertness (ANCOVAS)
 Shift (early and late)Test session (1st, 2nd and 3rd)TST (min)
  1. PVT, psychomotor vigilance task; NS, not significant; NA, not applicable to these analyses; TST, total sleep time; SEMs, slow rolling eye movements.

  2. Logistic multiple regression.

Mean reaction time (reciprocal)NSF(2,46.2) = 4.66, P = 0.014F(1,70.4) = 4.88, P = 0.030
Fastest 10% responses (reciprocal)NSNSNS
Slowest 10% responses (reciprocal)NSF(2,35.4) = 5.79, P = 0.007F(1,51.8) = 6.40, P = 0.015
Lapses (dichotomised)NSF(2,78) = 6.07, P = 0.004NS
Delta power (log)F(1,511) = 28.48, < 0.001NAF(1,479) = 14.30, < 0.001
Theta power (log)F(1,464) = 27.28, < 0.001NAF(1,435) = 15.59, < 0.001
Alpha power (log)F(1,498) = 26.18, < 0.001NAF(1,467) = 24.21, < 0.001
Beta power (log and square root)F(1,412) = 26.79, < 0.001NAF(1,387) = 24.61, < 0.001
SEMs (dichotomised)NSNAF(1,74) = 6.02, P = 0.017

The provision of a nap opportunity (Table 2) and greater amounts of nap sleep (Table 3) were also associated with increased alertness (decreased spectral power in all EEG frequency bands and a reduced probability of SEMs occurring), with the exception that a nap opportunity did not significantly alter spectral power in the beta band.

In both sets of analyses (Tables 2 and 3), results indicated that alertness was higher (spectral power in all frequency bands was lower) during the last hour of an early night shift compared with the last hour of a late night shift. In the models focussed on the effect of the amount of sleep obtained during the nap opportunity, there was no significant interaction between the amount of sleep and the timing of the night shift (early or late starting) on spectral power in the EEG (Table 3). However, for models that included the opportunity to nap, there was a significant interaction between the shift worked (early/late) and the provision of a nap or not (Table 2). Post-hoc tests for all frequency bands indicate that there is no difference in alertness on an early shift if a nap is taken or not (results for delta power t = 0.60, P = 0.552) but the nap does make a difference on the late night shift (delta power t = −3.80, P < 0.001) (see Fig. 2 also). Alertness is also lower (spectral power in all frequency bands higher) on the late night shift when no nap is taken, compared to the early shift with a nap (delta power t = −6.40, P < 0.001), and the early shift without a nap (delta power t = −6.98, P < 0.001). Alertness is also lower (spectral power in all frequency bands higher) on the late shift when a nap is taken compared to the early shift with no nap (delta power t = 3.14, P = 0.002).

image

Figure 2.  Change in performance and alertness for each of the four study conditions as represented by mean speed on the PVT and mean spectral power in the theta and alpha bands of the electroencephalogram (EEG). PVT performance was measured at the beginning of the night shift (time 1), after the nap or at an equivalent time if no nap was taken (time 2) and the end of the night shift (time 3). Theta and alpha power were measured across the last 60 min of the night shift.

Download figure to PowerPoint

Table 4 summarises the analyses examining the relationships between nap sleep quality and subsequent PVT performance and alertness. Only data from the post-nap performance test were considered, on the expectation that this test was most likely to be affected by nap sleep quality. Factors considered for inclusion in these models included the shift worked, TST during the nap, wake after sleep onset (WASO), S1%, SWS (dichotomised), sleep stage at the end of the nap, and the number of American Sleep Disorders Association arousals per hour. However, due to co-linearity between many variables, only the shift worked, S1%, and SWS (dichotomised) were included in the final models.

Table 4.   Relationships between nap sleep quality measures, and PVT performance and alertness (ANCOVAS)
 Shift (early/late)S1 (% TST)SWS (dichotomised)
  1. PVT, psychomotor vigilance task; TST, totla sleep time; SWS, slow wave sleep; NS, not significant; SEMs, slow rolling eye movements.

  2. Logistic multiple regression.

Mean reaction time (reciprocal)NSNSNS
Fastest 10% responses (reciprocal)NSNSNS
Slowest 10% responses (reciprocal)NSNSNS
Lapses (dichotomised)NSNSNS
Delta power (log)F(1,231) = 9.44, P = 0.002NSNS
Theta power (log)F(1,218) = 23.68, < 0.001F(1,172) = 4.31, P = 0.039F(1,222) = 13.17, < 0.001
Alpha power (log)F(1,207) = 12.43, < 0.001NSF(1,205) = 7.60, P = 0.006
Beta power (log and square root)F(1,1772) = 6.59, P = 0.011NSF(1,174) = 6.43, P = 0.012
SEMs (dichotomised)NSNSNS

Performance immediately after the nap was not related to the shift worked, percentage of S1 sleep, or the occurrence of SWS during the nap. Alertness during the last hour of the night shift was greater (spectral power in the theta, alpha and beta bands lower) after naps containing SWS. Spectral power in the theta band during the last hour of the night shift was lower after naps with a higher proportion of S1 sleep. As noted previously, alertness was higher (spectral power in all frequency bands was lower) at the end of an early night shift than at the end of late night shift.

Sleep after the night shift

The analyses summarised in Table 5 indicate that obtaining sleep during the night shift did not affect the timing, duration or efficiency of sleep at home immediately after the night shift. Analyses were repeated with the nap opportunity variable (nap/no nap) in place of TST and findings did not change. Following an early night shift, sleep at home began earlier and was longer than after a late night shift.

Table 5.   Relationships between nap sleep duration, shift timing, and sleep after the night shift
 Early mean (range)Late mean (range)ANCOVA results
No nap sleepNap sleepNo nap sleepNap sleep
  1. Median reported as variable not normally distributed.

  2. Bedtime log transformed; sleep efficiency square root transformed.

Bedtime (h:min)†‡5:24 (4:34–7:13)5:18 (4:31–7:11)7:21 (6:31–10:06)7:43 (7:07–11:01)Shift F(1,52.4) = 248.81, < 0.001
Sleep duration (h)4.3 (1.6–6.0)4.7 (3.1–7.2)4.2 (1.8–6.7)3.6 (1.2–5.5)Shift F(1,43) = 6.70, P = 0.013
Sleep efficiency (%)84 (70–99)89 (72–98)88 (62–96)88 (72–95)No significant effects

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interests
  9. References

Using a within-subjects design across four night shifts, and with objective measures, this study confirms the benefits of a short nap on the night shift for subsequent performance and alertness of ATCOs. Although the nap improved mean and slowest 10% PVT reaction times compared with no nap, this improvement was not sufficient to fully overcome the expected performance decline by the end of the night shift, and napping did not reduce the occurrence of lapses. Napping also reduced objective signs of sleepiness during the last hour of the night shift, (i.e. it decreased spectral power in the delta, theta, alpha and beta bands, and reduced the likelihood of SEMs). Nevertheless, on one occasion an individual who had napped, subsequently fell asleep when back on duty. This suggests that, although the nap affords some performance protection, it is not a fail-safe countermeasure to eliminate sleepiness on the night shift. The study also confirmed that napping during the night shift did not affect the amount or quality of sleep obtained subsequent to the night shift.

By comparison with two other field studies in which nap sleep was recorded polysomnographically (Rosekind et al., 1994; Smith-Coggins et al., 2006), ATCOs took a long time to fall asleep. This is somewhat surprising, given the circadian phase at which the nap was initiated, and for most individuals the relatively high homeostatic drive for sleep, because of the relatively small amount of sleep obtained in the prior 24-h. ATCOs were asked to initiate their nap as soon as possible after commencing their break, to maximize the amount of time between the end of the nap and returning to work and thus reduce any likelihood of impairment at work because of sleep inertia. This protocol may not have allowed them sufficient time to relax and unwind from the high level of vigilance and complete engagement demanded by the controlling task. These considerations point to the need for the nap opportunity to be embedded in a considerably longer period away from duty.

The occurrence of SWS is thought to increase the likelihood of sleep inertia (Dinges et al., 1981, 1985). In the present study, minimal SWS was recorded, even by comparison with other field studies examining nap opportunities of similar length (Rosekind et al., 1994; Smith-Coggins et al., 2006). One contributing factor may have been the longer average sleep latency and slightly shorter average sleep duration in the present study, which would have reduced the likelihood of SWS occurring. On the other hand, alertness (but not performance) was improved after naps containing SWS, compared to those without SWS. Thus, depending on the work context, the short-term increased likelihood of sleep inertia may need to be weighed against the greater long-term improvements associated with a nap containing SWS.

Longer prior wake increased the amount of SWS obtained in the nap, as well as improving other measures of sleep quality. However, using extended wakefulness as a strategy to improve nap sleep would not be recommended, because of the risk of being unable to fall asleep during the nap opportunity. The proportion of ATCOs (90%) able to fall asleep during the nap opportunity was comparable to other field studies, as were the findings that sleep was light, fragmented, and that no REM sleep was recorded (Rosekind et al., 1994; Smith-Coggins et al., 2006). Importantly, there was considerable individual variability, with some individuals consistently experiencing much greater difficulty initiating and maintaining sleep.

In laboratory studies of naps at a similar time and of similar duration (Gillberg, 1984; Rogers et al., 1989; Sallinen et al., 1998), sleep efficiency (SE) is comparatively high (on average >70%), and significantly more SWS is typically obtained (>30% TST). These differences may be related to the physical sleeping environment and the particular context of sleeping at work. They indicate a need for caution about the extent to which the positive (or negative) effects of nap sleep in the laboratory can be generalised to occupational settings.

No information was collected from the 20% of ATCOs who declined to participate in the study. It is possible that some of these individuals chose not to participate because they considered napping at work difficult. Given the high participation rate, the results are likely to be generalizable to similar populations but this potential ‘selection’ effect must also be taken into account.

In this study, some variability in the timing of the naps was unavoidable because operational demands necessarily took precedence. The EEG and EOG recordings obtained while ATCOs were on position, during the last hour of their shifts, contained a large amount of artifact, the removal of which may have reduced the power of the study to detect differences in spectral power associated with nap sleep. On the other hand, the study has the strength that the intervention was tested under realistic ‘noisy’ conditions.

In summary, the present within-subjects field study confirms the efficacy of a nap on the night shift for maintaining the performance and alertness of individuals required to perform safety-critical tasks. If workplace napping is to occur, we would recommend that it be governed by appropriate procedures and that suitable facilities are provided. Further research on sleep inertia is needed so that more specific recommendations can be made on the duration of time required between waking and return to work.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interests
  9. References

This study was possible only through the cooperation of the remarkably patient and accommodating ATCOs in Christchurch Area Control. The authors are thankful to Airways New Zealand for welcoming and partly funding this research and also Mr Gordon Purdie for his statistical advice and Dr Michelle Millar, Dr Nathanial Marshall and Dr Jesse Gale for their assistance in screening data. This research was also funded by the Health Research Council of New Zealand through a limited budget grant (no. 99/595) and Training Fellowship awarded to Dr Signal (grant no. 99/252).

Conflict of Interests

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interests
  9. References

This study was funded by the Health Research Council of New Zealand (limited budget grant no. 99/595 and a Training Fellowship awarded to Dr Signal, grant no. 99/252) and Airways New Zealand. Dr Signal and Prof. Gander are independent from Airways New Zealand. Mr Anderson and Ms Brash were employed by Airways New Zealand during the data collection period.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interests
  9. References
  • Åkerstedt, T. Shift work and disturbed sleep/wakefulness. Sleep Med. Rev., 1998, 2: 117128.
  • Åkerstedt, T. Searching for the countermeasure to night-shift sleepiness (editorial). Sleep, 2006, 29: 1920.
  • Åkerstedt, T., Knutsson, A., Westerholm, P., Theorell, T., Alfredsson, L. and Kecklund, G. Work organisation and unintentional sleep: results from the WOLF study. Occup. Environ. Med., 2002, 59: 595600.
  • American Sleep Disorders Association. EEG arousals: Scoring rules and examples. Sleep, 1992, 15: 174184.
  • Belenkey, G., Wesenten, N. J., Thorne, D., Thomas, M. L., Sing, H. C., Redmond, D. P., Russo, M. B. and Balkin, T. J. Patterns of performance degradation and restoration during sleep restriction and subsequent recovery: a sleep dose-response study. J. Sleep Res., 2003, 12: 112.
  • Bonneford, A., Muzet, A., Winter-Dill, A.-S., Bailloeuil, C., Bitouze, F. and Bonneau, A. Innovative working schedule: introducing one short nap during the night shift. Ergonomics, 2001, 44: 937945.
  • Bonneford, A., Tassi, P., Roge, J. and Muzet, A. A critical review of techniques aimed at enhancing and sustaining worker’s alertness during the night shift. Ind. Health, 2004, 42: 114.
  • Dinges, D. F. An overview of sleepiness and accidents. J. Sleep Res., 1995, 4: 414.
  • Dinges, D. F., Orne, E. C., Evans, F. J. and Orne, M. T. Performance after naps in sleep-conducive and alerting environments. In: L.Johnson, D.Tepas, W.Colquhoun and M. J.Colligan (Eds) Biological Rhythms, Sleep and Shiftwork: Advances in Sleep Research. Spectrum, New York, 1981: 539552.
  • Dinges, D. F., Orne, M. T. and Orne, E. C. Assessing performance upon abrupt awakening from naps during quasi-continuous performance. Behav. Res. Methods, 1985, 17: 3745.
  • Dinges, D. F., Orne, M. T., Whitehouse, W. G. and Orne, E. C. Temporal placement of a nap for alertness: contributions of circadian phase and prior wakefulness. Sleep, 1987, 10: 313329.
  • Dinges, D. F., Pack, F., Williams, K., Gillen, K. A., Powell, J. W., Ott, G. E., Aptowicz, C. and Pack, A. I. Cumulative sleepiness, mood disturbance, and psychomotor vigilance performance decrements during a week of sleep restricted to 4–5 hours per night. Sleep, 1997, 20: 267277.
  • Fleiss, J. L. The measurement of interrater agreement. In: J. L.Fleiss (Ed.) Statistical Methods for Rates and Proportions, 2nd edn. John Wiley & Sons, New York, 1981: 378382.
  • Gillberg, M. The effects of two alternative timings of a one-hour nap on early morning performance. Biol. Psychol., 1984, 19: 4554.
  • Harma, M., Knauth, P. and Ilmarinen, J. Daytime napping and its effects on alertness and short-term memory performance in shiftworkers. Int. Arch. Occup. Environ. Health, 1989, 61: 341345.
  • Lisper, H.-O. and Kjellberg, A. Effects of 24-hours of sleep deprivation on rate of decrement in a 10-minute auditory reaction time task. J. Exp. Psychol., 1972, 96: 287290.
  • Littell, R. C., Henry, P. R. and Ammerman, C. B. Statistical analysis of repeated measures data using SAS procedures. J. Anim. Sci., 1998, 76: 12161231.
  • Mitler, M. M., Carskadon, M. A., Czeisler, C. A., Dement, W. C., Dinges, D. F. and Graeber, R. C. Catastrophes, sleep, and public policy: Consensus report. Sleep, 1988, 11: 100109.
  • Powell, J. W. PVT-192 and Analysis Software Reference Manual. Unit for Experimental Psychiatry, University of Pennsylvania, Philadelphia, 1999.
  • Purnell, M. T., Feyer, A.-M. and Herbison, G. P. The impact of a nap opportunity during the night shift on the performance and alertness of 12-h shift workers. J. Sleep Res., 2002, 11: 219227.
  • Rechtschaffen, A. and Kales, A. A Manual of Standardised Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects. Brain Information Service/Brain Research Institute, UCLA, Los Angeles, 1968.
  • Rogers, A. S., Spencer, M. B., Stone, B. M. and Nicholson, A. N. The influence of a 1 h nap on performance overnight. Ergonomics, 1989, 32: 11931205.
  • Rosekind, M. R., Dinges, D. F., Connell, L. J., Rountree, M. S., Spinweber, C. L. and Gillen, K. A. Crew Factors in Flight Operations IX: Effects of Planned Cockpit Rest on Crew Performance and Alertness in Long-Haul Operations. NASA, Moffett Field, California, 1994.
  • Sallinen, M., Harma, M., Akerstedt, T., Rosa, R. and Lillqvist, O. Promoting alertness with a short nap during a night shift. J. Sleep Res., 1998, 7: 240247.
  • Santamaria, J. and Chiappa, K. H. The EEG of drowsiness in normal adults. J. Clin. Neurophysiol., 1987, 4: 327382.
  • Schweitzer, P. K., Muehlback, M. J. and Walsh, J. K. Countermeasures for night work performance deficits: the effect of napping or caffeine on continuous performance at night. Work Stress, 1992, 6: 355365.
  • Schweitzer, P., Randazzo, A., Stone, K., Erman, M. and Walsh, J. Laboratory and field studies of naps and caffeine as practical countermeasures for sleep-wake problems associated with night work. Sleep, 2006, 29: 3950.
  • Signal, T. L. and Gander, P. H. Rapid counterclockwise shift rotation in air traffic control: effects on sleep and night work. Aviat. Space Environ. Med., 2007, 78: 878885.
  • Smith-Coggins, R., Howard, S. K., Mac, D. T., Wang, C., Kwan, S., Rosekind, M. R., Sowb, Y., Balise, R., Levis, J. and Gaba, D. M. Improving alertness and performance in emergency department physicians and nurses: the use of planned naps. Ann. Emerg. Med., 2006, 48: 596604.
  • Van Dongen, H., Maislin, G., Mullington, J. and Dinges, D. F. The cumulative cost of additional wakefulness: dose-response effects on neurobehavioral functions and sleep physiology from chronic sleep restriction and total sleep deprivation. Sleep, 2003, 26: 117126.