J. Human Ergol.

, 30: 53-58, 2001

THE EFFECTS OF DOUBLE-SHIFTS (15.5 HOURS) ON SLEEP,

FATIGUE AND HEALTH

Goran KECKLUND, Mirjam EKSTEDT, Torbjorn AKERSTEDT, Anna DAHLGREN and

Bjorn SAMUELSON
National Institute for Psychosocial Factors and Health, Karolinska Institutet, Stockholm,Sweden
The Central WorkEnvironment Council for the Construction Industry, Sweden

The aim of the present study was to investigate how "double-shifts" (15.5 hours)
affects sleep, fatigue and self-rated health. The study was carried out on male
construction workers of which 80% were long-distance commuters. The schedule
involved two work periods and each work period involved two double shifts in a row.
The subjects filled in a sleep/wake diary at 8 times across a year and a questionnaire at
3 times. They also wore an actigraph during one shift cycle. The results showed that
sleepiness, and to a certain extent, mental fatigue increased during double shifts and
accumulated across days. The short rest time (8.5 hours) between days caused
insufficient sleep and approximately 5.5 hours of sleep was obtained between double
shifts. Questionnaire data showed that complaints of insufficient sleep, exhaustion on
awakening and pain symptoms increased across the year. It was concluded that a shift
system involving double shifts has a negative effect on fatigue, recovery and health-
related well-being.

Introduction
Long work shifts have become more common during the last 20 years. Today, individuals have
increased possibilities to influence their schedules, and compressed work hour systems, based on
extended work shifts, are often selected. The main advantage with compressed work schedules is a
longer sequence of days off and less interference between work and social activities. One negative
effect of extended work shifts is increased fatigue and accident risk (ROSA, 1995; FOLKARD,
1997; HANECKE et al., 1998). The effects on sleep and health are less clear (SMITH et al., 1998)
although some studies have shown impaired health and insufficient sleep. Most of the previous
research has concerned 12-hour shifts, whereas longer shifts have seldom been studied. However,
longer shifts do exist, for example, in construction industry, among fire fighters, ambulance drivers
etc. The aim of the present study was to investigate how "double shifts" (15.5 hours) affects sleep,
fatigue and self-rated health. More specifically, we wanted to examine whether fatigue and sleep
problems would accumulate across days with double shifts and whether fatigue and health
complaints would increase across a year with a shift system involving double shifts.

Methods
The study involved 83 male construction workers employed at two of the largest construction
companies in Sweden. The subjects belonged to two work sites. One of the work sites made a
temporary change of their schedule during the measurement period and was therefore excluded from
the present analysis. Thus, this paper includes 48 subjects (mean age: 41 years, range: 22-62 years).
Their main work task was to drill for a road tunnel system in the Stockholm area. Approximately
80% of the subjects were long-distance commuters and lived in a caravan or a rented room close to
the work site during the work periods.
The shift cycle covered two weeks: ADDXXDDMXXXXXX (A=afternoon shift, 14:30-22:15h;
D=double shift, 06:45-22:15h; M=morning shift, 06:45-14:30h, x=days off). The shift cycle
included two work periods with two days off in-between. The afternoon shift started on Wednesday
and the weekends were free. During the double shifts, and if there were no disturbances during work,
54 G. KECKLUND et al.

they had two 30-minute breaks when they could eat. However, they could often take some extra
pauses during blasting operations.
The design included 8 measurement periods when the subjects had to fill in a sleep wake-diary.
The measurement period covered a year and started in September 1999 and ended in September
2000. The first measurement was done approximately 5 months after the introduction of the shift
system. During the year they completed a questionnaire at three times (September 1999, February
2000, September 2000). The response rate was approximately 80% for the first seven measurements.
It should be pointed out that the number of participants decreased to 28 subjects for the last
measurement. The reason for the loss of subjects was that many workers were dismissed since the
work task was ended. During one shift cycle a smaller group (n=35) wore an actigraph (Cambridge
Neurotechnology Ltd., UK). They were instructed to push the event button at lights out and when
they rose in the morning. It was not possible to obtain a suitable control group since the 8-hour
workers at the site had very different work tasks.
The questionnaire included questions of sleep, work satisfaction and health complaints. The
instrument has been used in several other studies on shift workers (LOWDEN and AKERSTEDT,
1993). It also included a burnout instrument (MELAMED et al., 1992). We used several subscales
of this instrument and the result for the index "exhaustion" (including the items: "I feel physically
exhausted", "My batteries are empty", "I feel burned-out") will be presented in this paper. The
response scale varied between 1 "not at all" and 7 "almost all the time". The sleep and health
questions were grouped according to certain indices. An insomnia-index included the items "ease
falling asleep", "restless sleep", "premature awakening" and "repeated awakenings with difficulties
going back to sleep". Two pain indices were constructed; one was related to complaints of the upper
back and the shoulders, and the other index referred to other pain symptoms (legs, lower back etc.).
The response scale ranged from 1 "complaints almost all the time" to S "no complaints at all". A
question of general health was also used (1 very poor-S excellent).
The diary was filled in during S days for each measurement point. Measurement I, 3, 5, 7 and
8 involved the first half of the work period (ADDxx) and measurement 2, 4 and 6 the second half of
the work period (DDMxx). The diary included several sleep questions, but in the pzesent analysis
only the results for the sleep quality index ("ease falling asleep", "restless sleep", "sleep quality"
phrased as "How did you sleep", and "slept throughout") are presented. The diary also included
several ratings of fatigue. The subjects rated sleepiness (Karolinska Sleepiness Scale) and physical
exertion (CR-10) at six times during the day (07, 10, 14, 16, 19 and 22h). KSS (1 very alert - 9 very
sleepy) has been validated against physiological and behavioral indicators in several studies. Severe
sleepiness (=micro-sleeps) occurs when a value of 7 or higher is rated (AKERSTEDT and
GILLBERG, 1990). CR-10 (0 not at all - 10 extremely high) is also a well-established rating
instrument for physical fatigue that has been validated against physiological measures (BORG,
1990). The scale of mental fatigue (1 very slow, low energy, inactive - 9 very active, high energy)
has in a previous study showed increased fatigue in connection with insufficient sleep and early
morning shifts (SODERSTROM et al., 2001).
The statistical analyses were based mainly on repeated measurement ANOVA and t-tests.
When appropriate, the p-value was adjusted according to the correction presented by HUYNH and
FELDT (1976). Comparisons between separate shifts were done by t-tests. The alpha level was set
to 0.05, although statistical trends (p<0.10) will also be reported.

Results
Table 1 presents the results for the questionnaire data on self rated health symptoms and the
attitude to work hours. Since the number of subjects decreased for the last questionnaire, the
statistics are based on paired t-test with questionnaire 1 as a reference. The complaints of
insufficient sleep, exhausted at awakening and both pain indices increased across the year and the
difference between questionnaire 1 and 3 was significant. Poor sleep (insomnia) and exhaustion
showed a statistical trend to increase across the year. No significant differences appeared between
the first and the second questionnaire, except for some trends. The attitude to work hours was
 55

positive and did not change during the year. A question of self-rated health also showed high values
for all measurements and no significant change across the year. It should be pointed out that
questions of gastrointestinal symptoms, depressive/anxiety symptoms, other fatigue subscales and
sleepiness did not change across the year (means not presented in the table).

Table 1. Questions of health and attitude to work hours.

=1 very negative-5 very positive, 1=1 always (several times/week)-5 never, *=p<0.05; df for Q1 vs. Q2 varies between 31-
41, df for Q1 and Q3 varies between 19 and 23. Q=questionnaire, the questionnaires were distributed in Sept 1999 (Q1), Feb
2000 (Q2) and Sept 2000 (Q3).

The results for the ratings of fatigue and sleep quality, as well as the sleep parameters based on
actigraphy are listed in table 2. For each day in the shift cycle a mean was computed and the
statistical analyses are based on a repeated measurements ANOVA. The means for the fatigue
ratings during workdays do not include ratings during non-work times. Sleep length was clearly
reduced in connection with the short rest periods and reached approximately 5.5 hours prior to the
double shifts and the morning shift. However, sleep efficiency showed no significant variation
across days. Subjective sleep quality showed a significant variation across days. The sleep prior to
the last shift (morning-shift) was rated as poorer than the rest of the sleep episodes. However, sleep
quality prior to the double shifts was only marginally affected compared to the days off and sleep
prior to the afternoon shift.
Sleepiness (KSS), mental fatigue and CR-10 showed significant variation across days. The
highest sleepiness level occurred during the last double shift and the morning shift. Pair-wise t-tests
between the double shifts demonstrated that the fourth double shift included significantly higher
mean sleepiness than the other double shifts. For the sleepiness ratings we also calculated the
relative frequency of ratings of severe sleepiness (7 or higher). This analysis of severe sleepiness
showed that severe sleepiness was rare but had a minor peak for the second double shift. Severe
sleepiness decreased during the days off and was low for the third double shift and increased for the
morning-shift. When days off were excluded from the ANOVA, the F-ratio increased to 3.6 and the
p-value decreased to p<0.01. Ratings of physical exertion (CR-10) showed significantly lower
exertion on days off. When days off were excluded from the ANOVA, a significant trend was found
(F=2.6, p=0.06). Pair-wise t-tests indicated that the morning shift and the second double shift had
significantly lower exertion than the third and the fourth double shift. Mental fatigue, on the other
hand, was higher on days off compared to workdays. When days off were excluded from the
analysis, a significant trend appeared (F=2.3, p=0.075). Pair-wise t-tests showed that the morning
shift, the first double shift and the fourth double shift differed from the afternoon shift with respect
to mental fatigue.
The time of day pattern for sleepiness is presented in figure 1. In this figure a mean was
calculated for the double shifts in work period 1 (D1 and D2) and work period 2 (D3 and D4),
respectively. The afternoon shift was plotted together with the morning shift. A two-way ANOVA
for the 8-h shifts demonstrated a higher sleepiness level of the morning shift (p<0.001). The highest
sleepiness ratings occurred in the evening after the morning shift. There was no significant effect of
period for the double shifts. The interaction between period and time of day was, however,
 56 G. KECKLUND et al.

Table 2. Means and standard error of the mean for diary ratings of fatigue, sleep quality and sleep parameters based on
actigraphy. F-value represents a repeated measurements ANOVA.

Sleep length=hours; Sleep efficiency=% (sleep time/time in bed x 100); SQI=sleep quality index (1 very poor-5 very good);
KSS=Karolinska Sleepiness Scale, 1 very alert-9 very sleepy), CR-10=physical exertion (1 not at all-10 extremely high),
Mental fatigue=1 high-9 low. *=p<0.05, ***=p<0.001, df at least 7/315 (for sleep length/efficiency; df=7/189).

significant (p<0.01). A paired t-test showed higher sleepiness for the double shifts in period 1(D1
and D2) at 07h (p<0.05). The ANOVA for the first days off (X3 represents the values for the first
day off after period 2) showed a significant main effect for period (p<0.01). The first day off after
period 1(Saturday) showed significantly higher sleepiness, in particular in the morning (significant
t-test at 10h, p<0.05). The interaction between period and time of day was also significant (p<0.01).

Fig. 1. Mean•}standard error for ratings of sleepiness (KSS). Black squares represents work period 1 and white circles
represents work period 2. X3 represents the first day off (Thursday) after work period 2. The results for the double shifts are
based on means for both shifts within the same period.

In order to test whether fatigue would accumulate across the year, a repeated measurement
ANOVA was computed with time as the only factor. Means (during work time) were calculated for
each measurement occasion (8 values) for KSS, CR-10 and mental fatigue. No significant variation
was found and the difference between the highest and the lowest mean was approximately 0.3 scale
units.
 57

Discussion
The results showed that the double shifts were associated with increased sleepiness compared
with days off and an 8-h afternoon shift. However, the increase was relatively modest and severe
sleepiness was rare even on the extended shifts. In fact, sleepiness was also increased for the last 8h
morning shift in the work period. Severe sleepiness increased across the first 3 workdays, decreased
during the two days off, and then increased again for the next work period. This pattern suggested
accumulation across days. Ratings of mental fatigue showed a different pattern. The highest mental
fatigue occurred during the days off between the work periods. However, within each work period,
mental fatigue seemed to increase slightly across days and the highest value occurred for the last
work shift (8 hour morning shift). The increase in mental fatigue during days off in connection with
extended work shifts corresponds with previous findings from a study on construction workers
having an 84-hour workweek. This study showed that a mental fatigue index peaked during the first
two days off after the long workweek (ORBAEK et al., 2000). There can be two reasons why
mental fatigue increased across days off. Firstly, mental fatigue may be masked by activity and
stress during work, which suggests that the workers underestimated fatigue. Secondly, mental
fatigue symptoms during days off can be an effect of the effort and strain associated with the
compressed work hour system.
Physical exertion, as a measure of physical fatigue, showed no signs of accumulation across
work days. On the contrary, physical exertion decreased slightly across workdays and the lowest
value was found for the last shift of each work period. This could be interpreted as a decreased work
capacity across shifts, possibly, due to accumulated fatigue or insufficient recovery. However, an
alternative explanation could be that the subjects decided to work more slowly on the last work shift
in order to avoid fatigue and stress. The subjects had relatively good possibilities to influence their
work situation and could decrease work pace (for example, by taking some extra micro-breaks) or
perhaps avoid some heavy work tasks. Thus, the second suggestion, that the workers decided to
decrease their workload, is probably the most likely explanation.
Sleep length was highly affected by the short rest periods in connection with the double-shifts
and the morning-shift. Thus, the subjects obtained approximately 5.5 hours of sleep (or even
shorter) during the 8.5 hour rest time. It is likely that the sleep deficit between workdays is a cause
of the accumulation of sleepiness that we observed across work shifts. Still, the amount of sleep was
relatively high compared to what we have found in previous studies on short rest times. For example,
studies on quick changeovers (with 8 hour rest between shifts) often showed less then 5 hours of
sleep between shifts (LOWDEN et al., 1998). The fact that most subjects had almost no travel time
to work is probably one reason why the sleep debt was not larger. Another reason may be the
commuters' lack of family responsibilities, which could otherwise have decreased the time for sleep.
It should also be pointed out that the actigraphy recorded sleeps showed high sleep efficiency,
which indicates that, the amount of wake time after sleep onset was low. Subjective sleep quality
was also good, although it decreased somewhat towards the end of the work period. This decrease
could be due to the built-up sleep deficit across the workdays.
The questions of health showed a moderate increase in some symptoms during the year with
double-shifts. Thus, pain symptoms, feelings of not getting enough sleep, feeling exhausted at
wake-up, and possibly also exhaustion (based on burnout symptoms) increased (although the
exhaustion-index only showed a statistical trend). There was almost no increase between the first
and the second questionnaires whereas most of the increase occurred between the second and third
questionnaire (the time period between these questionnaires was 6 months). A methodological
problem with the questionnaire results is the lack of control group. Thus, we don't know whether
the increase in symptoms could depend on other factors than the work schedule. Still, many of the
health complaints, like gastrointestinal and depressive/anxiety complaints, did not change across the
year. The fact that it was pain, sleep and exhaustion that responded is also in line with some
previous findings (WAERSTED and WESTGAARD, 1991; SMITH et al., 1998). However, the
question of general self-rated health did not change which suggests that the increase in symptoms
 58 G. KECKLUND et al.

did not have any impact on their general health perception. Thus, it is still an open question whether
long work shifts have any effect on health. This question definitely needs further research.
Another limitation with the present study was that we did not measure performance. Due to the
work task it was not possible to administer performance tests during work. Thus, we cannot say
whether the increase in severe sleepiness and mental fatigue affected performance. However,
according to the statistics from the company there has not been any serious accidents or near misses
among these workers.
Another question is whether the obtained results are representative of other groups and
occupations. Clearly, the present group consisted of a selected sample that may have better health
and sleep than the average construction worker. The workers were also positive to the schedule,
which was initiated on the request of the workers. Thus, the results are probably not representative
of groups that are being forced to work double shifts. Furthermore, groups who have a high work
pace and high work demands (stress) would probably have more difficulties tolerating this extreme
schedule. Preliminary findings related to a comparison between long-distance and short-distance
commuters suggest that long-distance commuters have a lower level of sleepiness and obtain more
sleep between the double shifts (EKSTEDT et al., 2001). Thus, those subjects that have to combine
family life and double shifts may also have more problems tolerating extended work shifts.
In conclusion, the double-shifts were associated with increased sleepiness and mental fatigue,
as well as insufficient sleep. The increase in sleepiness and fatigue was, however, rather moderate.
Severe sleepiness and mental fatigue did accumulate across shifts and the last shift (8h morning
shift) also showed elevated values. Symptoms of pain and insufficient sleep also increased across
the year with double-shifts, although the lack of a control group, suggests some caution with the
interpretation of the latter findings.

References
BORG,G. (1990)Psychophysical
scalingwithapplications
in physicalworkandtheperception
of exertion.Scandinavian
Journalof WorkEnvironment
andHealth,16(suppl1):55-58.
EKSTEDT, M., KECKLUND, G., DAHLGREN, A., AKERSTEDT, T. and SAMUELSON, B. (2001) Kan dubbelskift vara
en acceptabel arbetstidslosning? - effekter pa aterhamtning, stress och halsa (In english: Can double shifts be an
acceptable work hours solution? - effects on recovery, stress and health). Stress Research Reports (Karolinska
Institutet, Sweden), 299, 1-75.
FOLKARD, S. (1997) Black times: Temporal determinants'of transport safety. AccidentAnalysis & Prevention, 29: 417-430.
HUYNH, H. and FELDT, L. S. (1976) Estimation at the box correction for degrees of freedom for sample data in randomised
block and split-plot designs. Journal of Educational Statistics, 1: 69-82.
HANECKE, K.,TIEDEMANN, S., NACHREINER, F. and GRZECH-SUKALO, H. (1998) Accident risk as a function of
hour at work and time of day as determined from accident data and exposure models for the German working
population. Scandinavian Journal of WorkEnvironment and Health, 24 (suppl 3): 43-48.
LOWDEN, A., KECKLUND, G., AXELSSON, J. and AKERSTEDT, T. (1998) Change from an 8-hour shift to a 12-hour
shift, attitudes, sleep, sleepiness and performance. Scandinavian Journal of Work Environment and Health, 24

(suppl 3): 69-75.

LOWDEN, A. and AKERSTEDT, T. (1993) Jet lag in air crew - a questionnaire study. Stress Research Reports (Karolinska

Institutet, Sweden), 237: 1-20.

MELAMED, S., KUSHNIR, T. and SHIROM, A. (1992) Burnout and risk factors for cardiovascular health. Behavioral
Medicine, 18: 53-60.
ROSA, R. (1995) Extended workshifts and excessive fatigue. Journal of Sleep Research, 4 (suppi 2): 51-56.

SMITH, L., FOLKARD, S., TUCKER, P. and MACDONALD, I. (1998) Work shift duration: a review comparing eight hour
and 12 hour shift systems. Occupational and Environmental Medicine, 55: 217-229.
SODERSTROM, M., INGRE, M., KECKLUND, G., AKERSTEDT, T. and KECKLUND, L. (2001) Lokforarens
arbetsbelastning: paverkan pa somn, trotthet och prestation (In english: The train drivers' workload: effects on sleep,

fatigue and performance). Report from the National Institute for Psychosocial Factors and Health, Sweden.
WAERSTED, M. and WESTGAARD, R.H. (1991) Working hours as a risk factor in the development of musculoskeletal
complaints. Ergonomics, 34: 265-276.
AKERSTEDT, T. and GILLBERG, M. (1990) Subjective and objective sleepiness in the active individual. Int. J.
Neuroscience, 52: 29-37.

ƒÕ RBAEK, P., KECKLUND, G., SEGER, L. and AKERSTEDT, T. (2000) Arbetsvecka pa 84 timmar. Trotthet,
somnmonster och stressreaktioner. (In English: 84 hour working week. Fatigue, sleep and stress reactions). Report
from the Dept. of Occupational and Environmental Medicine, University Hospital in Lund, Sweden.