Beskæftigelsesudvalget 2022-23 (2. samling)
BEU Alm.del Bilag 67
Offentligt
2665204_0001.png
Downloaded from www.sjweh.fi on December 06, 2022
Scand J Work Environ Health
Online-first -article
Published online: 29 Nov 2022
doi:10.5271/sjweh.4074
Night and evening shifts and risk of calling in sick within the
next two days – a case-crossover study design based on
day-to-day payroll data
by
Larsen AD, Nielsen HB, Kirschheiner-Rasmussen J, Hansen J,
Hansen ÅM, Kolstad HA, Rugulies R, Garde AH
The study brings new insight into the acute effects of night and
evening shifts on risk of calling in sick, as it is the first study to
investigate this in a study population beyond pregnant women. The
results are useful when planning shift work with focus on reduction of
sickness absence for the employee and consequently, costs for the
employer.
Affiliation:
National Research Centre for the Working Environment,
Lersø Parkallé 105, DK-2100 Copenhagen, Denmark. [email protected]
Refers to the following texts of the Journal:
2008;34(6):483-486
2009;35(1):48-55 2013;39(2):134-143 2021;47(4):268-276
2022;48(7):549-559
Key terms:
case-crossover; evening shift; hospital staff; irregular
work hour; night shift; payroll data; register data; shift work; shift
worker; sickness absence
This article in PubMed:
www.ncbi.nlm.nih.gov/pubmed/36445985
This work is licensed under a
Creative Commons Attribution 4.0 International License.
Print ISSN: 0355-3140 Electronic ISSN: 1795-990X
 BEU, Alm.del - 2022-23 (2. samling) - Bilag 67: Orientering om nye forskningsresultater fra NFA
2665204_0002.png
O
riginal article
Scand J Work Environ Health – online first. doi:10.5271/sjweh.4074
This work is licensed under a Creative Commons Attribution
4.0 International License.
Night and evening shifts and risk of calling in sick within the next two days – a case-
crossover study design based on day-to-day payroll data
by Ann Dyreborg Larsen, PhD,
1
Helena Breth Nielsen, PhD,
1
Jonas Kirschheiner-Rasmussen, MSc,
1
Johnni Hansen, PhD,
2
Åse
Marie Hansen, PhD,
1, 3
Henrik Albert Kolstad, MD,
4
Reiner Rugulies, PhD,
1, 3
Anne Helene Garde, PhD
1, 3
Larsen AD, Nielsen HB, Kirschheiner-Rasmussen J, Hansen J, Hansen ÅM, Kolstad HA, Rugulies R, Garde AH. Night and
evening shifts and risk of calling in sick within the next two days – a case-crossover study design based on day-to-day payroll
data.
Scand J Work Environ Health
– online first. doi:10.5271/sjweh.4074
Objective
Night and evening work is associated with risk of sickness absence, but little is known about the acute
effects of these types of shifts on sickness absence. The aim of the current study is therefore to examine the risk
of calling in sick within two days after a night or an evening shift.
Methods
By use of a case-crossover design, odds of calling in sick within two days after a night or an evening
shift compared to day shifts were analyzed within the same person. Day-to-day information on shifts and sick-
ness absence were derived from the Danish Working Hour Database on 44 767 cases. Data were analyzed using
conditional logistic regression. The analyses were supplemented by extensive testing of methodological choices.
Results
Analyses showed higher odds of calling in sick after a night shift [odds ratio (OR) 1.22, 95% confidence
intervak (CI) 1.14–1.30] and lower odds after an evening shift (OR 0.89, 95% CI 0.84–0.93) compared to day
shifts within the same person. Testing of methodological choices suggested that in particular the duration of case
and control periods, time between these periods along with the number of control periods affected the results.
Conclusion
This large and unique within-person study among Danish hospital employees indicate that the risk
of calling in sick is affected by the types of shifts, independently of sex, age, and time-invariant confounding.
Extensive testing identified important methodological choices eg, length and number of included periods to
consider when choosing the case-crossover design.
Key terms
hospital staff; irregular work hour; register data; sickness absence.
A significant proportion of the workers in the EU work
nights (19%) and/or in shifts (22%) (1). In the Danish
hospital sector, it is estimated that approximately 24%
of workers have schedules including evening shifts
without nights, and 11% work night shifts permanently
or as part of 3-shift schedules (10.2%) (2). Previous
studies have suggested that night and shift work are
associated with an increased risk of certain cancers,
ischemic heart disease, and diabetes (3–5).
Sickness absence is a predictor of subsequent mor-
bidity, dissatisfaction, disability, and mortality (6–8) and
is often used as an indicator for work-related health (9).
In addition, sickness absence poses a daily organiza-
tional challenge to many workplaces, where substitutes
are needed to cover the staffing requirements.
1
2
3
4
A systematic review by Merkus et al (10) concluded
in 2012 that epidemiological evidence was inconclusive
regarding the impact of rotating shifts, night work, and
fixed night work on the risk of sickness absence. Fur-
ther, the authors called for more detailed and non-self-
reported exposure data. Since then, several studies have
included more detailed and register-based exposure data,
eg, payroll data when studying the association between
night or evening work and sickness absence (11–19).
Yet, the results of these studies are inconsistent. Some
studies found higher rates of sickness absence among
night shift workers (11–16, 18) compared to workers on
day shift, whereas other studies observed no relationship
(12, 17, 19).
There are several explanations for the inconsis-
The National Research Centre for the Working Environment, Copenhagen, Denmark.
Danish Cancer Society Research Center, Copenhagen, Denmark.
Department of Public Health, University of Copenhagen, Copenhagen, Denmark.
Department of Occupational Medicine, Danish Ramazzini Centre, University of Aarhus, Aarhus, Denmark.
Correspondence to: Ann D Larsen, National Research Centre for the Working Environment, Lersø Parkallé 105, DK-2100 Copenhagen,
Denmark. [E-mail: [email protected]]
Scand J Work Environ Health – online first
1
 BEU, Alm.del - 2022-23 (2. samling) - Bilag 67: Orientering om nye forskningsresultater fra NFA
2665204_0003.png
Night and evening shifts and acute sickness absence
tencies: the studies used different lengths of sickness
absence eg, short-term (1–3 days) (15, 16, 18), 1–8 days
(12) or long-term (>3–30 consecutive days) (11–14, 19).
This makes it difficult to compare results as there may
be different mechanisms related to short- and long-term
sickness absence (20). Further, specifying the length of
sickness absence may compromise basic epidemiologi-
cal principles in regards to conditioning on a descendant
of the outcome (21). In addition, other methodological
challenges such as the high frequency of the outcome
make sickness absence a challenging outcome to study
with the traditional survival analysis with long follow-up
as most of the study population will experience sickness
absence at some point in time during observation. The
risk of sickness absence is also highly correlated to pre-
vious sickness absence (22) and inversely correlated to
working time, as one cannot register sickness absence
and be working at the same time. Sickness absence is
associated with many other factors, which can be dif-
ficult to obtain information on when using register data
eg, personality, genetic background and to some extent
work environment (23–26). Studies on sickness absence
are therefore vulnerable to between-subject differences
ie, that the risk of sickness absence is not the same in
the exposed group as in the reference group. These
challenges can be addressed by use of a case-crossover
design, which handles differences between employees
by self-matching and thereby excludes effects of time-
invariant covariates (27). This design has been used in
Finnish studies (15, 16, 18) to study the effects of work
schedules within the past 28 days on sickness absence,
but no studies have yet investigated acute effects of night
or evening shifts and the risk of sickness absence.
Against this background, we aim to examine if
night or evening shifts are associated with calling in
sick within the next two days, using a case-crossover
design in a large Danish cohort based on day-to-day
pay-roll data.
ables related to the specific employment conditions of
the employee. In addition, the database also includes
day-to-day information on sickness absence. Further
information on the DWHD is published elsewhere (2).
Case selection
We restricted the study population to adults below the
general retirement age (18–67 years old) in 2019 with
≥50% employment time were included. This allowed
us to include participants with part-time jobs, which is
quite common among hospital workers in Denmark, but
exclude participants with low employment degrees (mar-
ginal part-time work) eg, due to health issues, which
evidently could affect their sickness absence. We also
excluded those who were pregnant by excluding women
on parental leave within the 8 months after the sickness
absence as risk of sickness absence differs across the
pregnancy (28). Only employees with a change in expo-
sure (day, evening or night) between the case and the
control period(s) were included. We identified 44 767
cases according to the definition (see figure 1).
Exposure assessment – night and evening work
We defined night work as ≥3 hours of work between
23:00 and 06:00 similar to previous studies (29). Eve-
ning shifts were defined as 3 hours of work between
18:00 and 02:00. Day shifts were defined as shifts start-
ing after 06:00 and ending before 21:00. The definitions
Methods
Study design
We applied a case-crossover design with a unidirectional
approach and multiple control intervals. We analyzed
data from the Danish Working Hour Database (DWHD)
a Danish nationwide database including administrative
payroll data from 2007 onwards. The database holds
detailed information on actual working hours from >340
000 employees from the five Danish regions, including
all public hospitals. For all employees, we had precise
information on daily starting and ending time for each
shift worked as well as age and sex along with vari-
Figure 1.
Flowchart of the study population.
2
Scand J Work Environ Health – online first
 BEU, Alm.del - 2022-23 (2. samling) - Bilag 67: Orientering om nye forskningsresultater fra NFA
2665204_0004.png
Larsen et al
of shifts were not mutually exclusive: night work was
given the highest priority, then evening work and the
lowest priority was day work. The exposure assigned to
each case and control period were the most recent shift
to end or start within the case or control period. If the
employee was sick listed during work hours in the case
period, the most recent shift before was considered as
the exposure in the period.
Cases – sickness absence
Information on daily sickness absence was drawn from
DWHD. The cases were selected as each person’s first
day of sick-leave registration preceded with ≥90 days
without sickness absence after 1 January 2019. The
case period was defined as the two days leading up to
the day of the sick-leave registration. Control periods
were matched to the case period by weekday, to adjust
for differences across days of the week. Thus, for each
case period we selected all (up to five) two-day control
periods occurring day 28–56 before the case period. We
included only control periods that were followed by a
day with work, respectively 28, 35, 42, 49 and 56 days
before the day of sickness absence (see figure 2), as sick-
ness on days off are not registered in our payroll data.
Covariates
Sex (woman/man) and age groups (18–24, 25–34, 35–
44, 45–54, ≥55 years old) from DWHD were included
as effect modification and used in stratifications.
Statistical analyses
The case-crossover design is a type of fixed effects
models (30). Only employees with a change in exposure
between the case and control period (discordant pair)
contribute to the analysis (31, 32). Discordant
exposed
pairs
were calculated as the total number of pairs where
employees were exposed in the case period and unex-
posed during the control period. Discordant
unexposed
pairs
were calculated as the number of pairs where the
case period was unexposed and the control period was
exposed. Also, the number of employees with at least
one discordant pair was calculated.
The case-crossover matched-pair interval approach
was used to compare each employee’s exposures in
the case period with exposures in the control periods.
Using conditional logistic regression with the employee
used as strata, we calculated odds ratios (OR) with 95%
confidence intervals (CI). All analyses were conducted
in SAS 9.4 (SAS Institute, Cary NC, USA) with Proc
Logistic in accordance with previously described tech-
niques (33).
Main analyses.
We compared the exposure in the refer-
ence periods (control periods) with the exposure before
calling in sick (case period). Exposure was measured as
having a night or an evening shift within two days of
the reference. Day shifts in the case period were used
as reference.
The analyses were carried out in the total study
population and stratified populations in regards to sex
and age groups. Further we compared risk of calling in
sick after a second, third or fourth consecutive night
shift (exposure) to the first night shift (reference).
Sensitivity analyses for check of methodological choices.
All
methodological choices were made
a priori.
Therefore,
to gain insight into the consequences of the epidemio-
logical and methodological decisions of the design, we
performed sensitivity analyses of night shift to test: (i)
effect of the duration of the case and control periods,
the main analysis (night versus day) was repeated with
a case and control periods of 24 hours and 72 hours;
(ii) the effect of number of control periods, the main
analysis was repeated with 1, 3, 6 and 9 control periods;
(iii) the effect of days between control periods, the main
analysis was repeated with 5-day intervals and 5 days
between case and control in one analysis, and with 10
days intervals and 14 days between case and control
periods in another analysis; (iv) the effect of excluding
participants with sickness absence 90 days before case
periods, the main analysis was repeated while including
all participants regardless of preceding sickness absence,
control periods with sickness absence were treated as
cases; (v) if degree of full-time/part-time work affected
the results, the main analysis was repeated with >75%
and 100% employment, respectively.
As previous studies have been conducted on nursing
personnel only, looking into short-term sickness absence
(1–3 days) (15, 16, 18) the main analyses were repeated
restricted to nursing personnel and sickness absence up
to three days.
To test robustness of the main analyses, we repeated
the setup from the main analyses but used non-night
shift (day and evening combined) as reference.
Scand J Work Environ Health – online first
Figure 2.
Illustration of the case-crossover study design. The case period
(with the circle) covers 48 hours preceding a sick leave registration (the
X). Five control periods of 48 hours (prior square boxes) were matched by
weekday 28–56 days prior to the sick leave registration.
3
 BEU, Alm.del - 2022-23 (2. samling) - Bilag 67: Orientering om nye forskningsresultater fra NFA
2665204_0005.png
Night and evening shifts and acute sickness absence
Results
The studied population included primarily women
(79.7%) (table 1) with few subjects <24 years (2.7%).
In the remaining of the age groups, the distribution
was similar (22.0–26.4%). Most participants worked
as nursing personnel including nurses, nurse assistants,
midwives etc (51.7%), but also as physicians (13.1%)
and other professions including physical therapists,
occupational therapists etc (35.2%).
Results from the main and supplementary analyses
Table 2 shows the assessment of risk of sickness absence
after a night or evening shift (the exposures) compared
to a day shifts (the references). Results from the analy-
ses on all employees showed higher odds of sickness
absence after a night shift compared to a day shift (OR
1.22, 95% CI 1.14–1.30) Evening shifts were associated
with lower odds of sickness absence (OR 0.89, 95% CI
0.84–0.93) compared to day shifts.
Table 2 shows the odds of sickness absence after
consecutive night shifts compared to the odds of sick-
Table 1.
Characteristics of employees in the main analysis.
Cases (N=44 767)
N
Age (years)
18–24
25–34
35–44
45–54
55–67
Sex
Women
Men
Occupation
Physician
Nurse
Others
Missing
1194
9840
11 440
11 832
10 461
35 658
9109
5843
23 124
13 501
2299
%
2.7
22.0
25.6
26.4
23.4
79.7
20.4
13.1
51.7
30.7
5.1
ness absence after the first night shifts. Results show
that the highest odds was after the third night shift in a
row compared to the first night shift (OR 1.44, 95% CI
1.20–1.73).
When stratifying by sex, men’s risk estimates were
higher than women’s. However, both sex showed similar
results as in the first analyses: higher odds of sickness
absence after night shifts compared to day (OR
men
1.38,
Table 2.
Case-crossover design
a
. Conditional logistic regression analyses of risk of sickness absence after a night or evening shift (exposure) com-
pared to the risk after a day or non-night shift (reference). [D=discordant; OR=odds ratio; CI=confidence interval]
Work shift characteristics
All employees
Night vs. day (ref)
Evening vs. day (ref)
2
nd
vs. 1
st
night shift
3
rd
vs. 1
st
night shift
4
th
vs. 1
st
night shift
Sex
Men
Night vs. day (ref)
Evening vs. day (ref)
Women
Night vs. day (ref)
Evening vs. day (ref)
Age (years)
18–24
Night vs. day (ref)
Evening vs. day (ref)
25–34
Night vs. day (ref)
Evening vs. day (ref)
35-44
Night vs. day (ref)
Evening vs. day (ref)
45–54
Night vs. day (ref)
Evening vs. day (ref)
55–67
Night vs. day (ref)
Evening vs. day (ref)
a
b
Exposed in
case period
3352
4412
1765
1306
444
769
816
2583
1596
115
170
1090
1097
834
886
748
1066
565
1193
Unexposed in
case period
37 003
37 003
4024
4024
4024
7524
7524
29 479
29 479
909
909
7653
7653
9720
9720
10 018
10 018
8703
8703
D-exposed
pairs
b
3294
4847
563
346
150
841
910
2453
3937
134
206
1287
1408
905
1033
638
1150
330
1050
D-unexposed
pairs
c
2702
5433
514
237
111
604
1052
2098
4381
81
231
1056
1526
682
1078
556
1384
327
1214
D-cases
d
Conditional logistic regression analysis
OR
95% CI
P-value
1.22
0.89
1.13
1.44
1.31
1.38
0.86
1.18
0.89
1.60
0.90
1.22
0.91
1.27
0.95
1.19
0.82
1.06
0.86
1.14–1.30
0.84–0.93
0.99–1.30
1.20–1.73
0.99–1.73
1.20–1.58
0.77–0.97
1.09–1.27
0.84–0.94
1.14–2.27
0.70–1.15
1.10–1.35
0.83–1.01
1.12–1.44
0.85–1.07
1.02–1.38
0.74–0.91
0.87–1.29
0.77–0.96
<0.001
<0.001
0.070
0.001
0.06
<0.001
0.020
<0.001
<0.001
0.009
0.400
<0.001
0.07
<0.001
0.38
0.02
<0.001
0.60
0.96
1844
2720
418
208
97
421
395
1423
2225
76
118
727
774
487
572
362
642
192
614
Case-crossover design with up to 5 control periods per case period.
Discordant exposed pairs: number of pairs where case is exposed and control is unexposed.
c
Discordant unexposed pairs: number of pairs where case is unexposed and control is exposed.
d
Discordant cases: number of employees with at least one discordant pair.
4
Scand J Work Environ Health – online first
 BEU, Alm.del - 2022-23 (2. samling) - Bilag 67: Orientering om nye forskningsresultater fra NFA
2665204_0006.png
Larsen et al
95% CI 1.20–1.58, OR
women
1.18, 95% CI 1.09–1.27) and
lower odds of sickness absence after an evening shift
(OR
men
0.86, 95% CI 0.77–0.97, OR
women
0.89 95% CI
0.84–0.94) compared to after a day shift.
When stratifying on age groups the results were
similar to previous analyses; higher odds of sickness
absence after a night shift in the age groups 18–54 years
of age. Further, the analyses showed lower odds of sick-
ness absence after an evening shift compared to a day
shift when ≥45 years of age.
Results from the sensitivity analyses
Table 3 includes all the sensitivity analyses. Results from
the main analysis on the total population comparing night
to day shifts are included at the top.
When testing the
duration
of periods, analyses
showed that including periods of only one day the OR
decreased (1.06, 95% CI 0.97–1.15), whereas the inclu-
sion of three days were close to the main result (OR
1.17, 95% CI 1.11–1.24). When testing the
number
of
control periods, results showed that when including only
one control period the risk estimate was lower than in
the main analysis. However, risk estimates maintained
the size regardless of including three, six or nine con-
trol periods. There was a slight change in risk estimates
when changing the length between case and control peri-
ods, where seven days (or the same weekday) showed
the highest risk estimates.
It made no difference to the results, if participants
with sickness absence 90 days prior to the case period
were excluded or not (OR 1.21, 95% CI 1.13–1.29).
Similarly the results were not affected when restrict-
ing to participants working 75% of full-time (OR 1.22,
95% CI 1.14–1.30) or only full-time (OR 1.25, 95% CI
1.15–1.35).
When restricting to nursing personnel and short-term
sickness absence (1–3 days), the results showed the
same tendency as the main results but with lower risk
estimates. Similarly, when including day and evening
shifts as reference (= non-night shifts), the risk estimate
was lower, but still showed higher odds of calling in sick
after a night shift compared to non-night shifts (OR 1.19,
95% CI 1.12–1.27).
Discussion
The main analyses showed that the risk of calling in
sick was 22% higher after a night compared to day
shift within the same person in a population of >44 000
included individuals. The odds of calling in sick were,
however, 11% lower after an evening compared to day
shift. Further, supplementary analyses showed higher
odds of sickness absence after the third night shift in a
row compared to the first. When stratifying on sex and
age groups, the results from the main analyses were
corroborated.
Sensitivity analyses checking the methodological
choices showed that changing the duration of the case
and control period (1–3 days) and the time between
Table 3.
Case-crossover design a. Sensitivity analyses. Conditional logistic regression analyses of risk (OR) of sickness absence with 95% CI. First
line presents the results from the main analysis (in italics). [D=discordant; OR=odds ratio; CI=confidence interval]
Analysis
Total population
Nursing personnel and
short-term sickness absence
Duration of periods (days)
1
3
Number of control periods
1
3
6
9
Different lengths between case period and
control period (days)
5
10
First sickness absence in 2019 as case
Degree of full-time work
≥0.75
≥1
Night vs non- night (reference)
a
b
Exposed in Unexposed in D-exposed D-unexposed
case period case period
pairs
b
pairs
c
3352
2332
2142
4125
1765
2984
3446
3557
3402
3407
3529
3245
1770
3352
37 003
16 959
31 908
49 635
25 335
31 443
37 382
37 876
38 036
37 466
39 788
36 028
25 797
41 415
3294
2818
1726
3722
664
2006
3939
5832
3174
2953
3475
3270
2338
3914
2702
1795
1620
3239
576
1644
3249
4817
2971
2781
2862
2674
1853
3472
D-cases
d
1844
1196
1009
2289
664
1441
1991
2273
1966
1929
1950
1826
1212
2149
Conditional logistic regression analysis
OR
1.22
1.08
1.06
1.17
1.15
1.21
1.22
1.22
1.07
1.09
1.21
1.22
1.25
1.19
95% CI
1.14–1.30
1.00–1.17
0.97–1.15
1.11–1.24
1.10–1.29
1.12–1.30
1.14–1.30
1.15–1.30
1.01–1.34
1.02–1.16
1.13–1.29
1.14–1.31
1.15–1.35
1.12 -1.27
P-value
<0.001
0.06
0.22
<0.001
0.01
<0.001
<0.001
<0.001
0.03
0.008
<0.001
<0.001
<0.001
<0.001
Case-crossover design with up to 5 control periods per case period.
Discordant exposed pairs: number of pairs where case is exposed and control is unexposed.
c
Discordant unexposed pairs: number of pairs where case is unexposed and control is exposed.
d
Discordant cases: number of employees with at least one discordant pair.
Scand J Work Environ Health – online first
5
 BEU, Alm.del - 2022-23 (2. samling) - Bilag 67: Orientering om nye forskningsresultater fra NFA
2665204_0007.png
Night and evening shifts and acute sickness absence
case and control period (5–10 days), had the largest
effect on the results in terms of lower risk estimates. In
contrast, there was little effect of changes in number of
control periods (>5) and degree of full-time employment
or when including individuals with previous sickness
absence.
New insight
The study brings new insight to the acute effect of night
shift work on risk of calling in sick, as this is the first
study to investigate this in a study population beyond
pregnant women (28). It is also, to the authors’ knowl-
edge, the first study to observe acute effects of evening
shifts on sickness absence. Further, the current study
includes a large number of individuals and is the largest
study to date to investigate sickness absence due to night
shift work using the case-crossover design. The results
from the extensive testing of methodological choices
indicate that although the case-crossover design meet
many of the previous challenges, one should be aware
of the consequences of these choices.
Previous literature
No previous studies have addressed the acute risk of
calling in sick after a night shift except a single study on
pregnant women which found higher risk of calling in
sick after a night shift in the 1
st
(OR 1.28, 95% CI 1.19–
1.37) and 2
nd
trimester (OR 1.27, 95% CI 1.17–1.39)
(28). The results support, to some extent, the findings in
the current study. However, as sickness absence during
pregnancy may have other reasons or mechanism, these
studies are not fully comparable.
An association between night shift work and
increased risk of sickness absence in general (not acute)
has been observed in other recent studies (11, 13–16,
18). Studies from Finland using the case-crossover
design, reported that night shift work during the last
28 days (case exposure window) was associated with
increased odds of short-term sickness absence (1–3
days) compared to working time 28 days earlier (case
control window) (15, 16, 18). Several of the previ-
ous case-crossover studies on night work and sickness
absence have been conducted in a study population of
nurses only and with short-term sickness absence of 1–3
days as outcome. We therefore repeated the main analy-
sis in a sub-group of nurses limiting sickness absence
to 1–3 days. When comparing the results from our main
analysis, the OR for the analysis on nurses in this study
was lower than for all employees, but still indicating
higher odds of calling in sick after a night shift com-
pared to a day shift (OR 1.08, 95% CI 1.00–1.17). The
OR for nurses further corroborated previous findings
(16) and extended them to another national context. The
differences in risk estimates between countries, might be
caused by slightly different study designs, and a reflec-
tion of contextual differences. In some countries, pay-
ment of sickness absence benefits require medical cer-
tification – in Finland after three days (16), in Norway
after eight days (12) and in Denmark after four weeks
(34) – which may affect the incentive for return to work.
We observed lower odds of sickness absence among
those working evenings compared to day shifts, which
confirms previous findings in a study using the same
method (16). It also indicates that night and evening
shifts have different mechanisms in regards to sickness
absence. Being awake at night and the resulting circa-
dian rhythm disturbances, lack of restitution, and fatigue
are the strongest explanatory parameters in relation to
night work and the risk of sick leave. However, the same
parameters cannot explain the reduced risk of sickness
absence when working evening shifts. Rather, we expect
evening and day shifts to be comparable in regards to the
biological mechanisms and, therefore, in risk of sickness
absence. One could speculate that lower demands for
amount of work on evening shifts compared to day time
work (31) could contribute to reduced sickness absence
(and more sickness presenteeism on evening shifts).
However we were unable to study this in the current
data. In the current study, we included the acute effect of
sickness absence, however previous studies have found
higher risk of long-term sickness absence when working
evening shifts (13, 35).
Ropponen et al (16) found a higher risk of sickness
absence after ≥2 consecutive night shifts (OR 1.24,
95% CI 1.12–1.38) and especially after ≥4 consecutive
night shifts (OR 1.54, 95% CI 1.10–2.15) indicating a
dose–response relationship. This could not be fully cor-
roborated in the current study as results showed only
statistically significant results for the third consecutive
night shift compared to the first and not for the second
or the fourth consecutive shift compared to the first.
Stratification on sex showed that both women and
men had statistically higher odds of calling in sick after
a night shift compared to day shift. Highest odds were
found among men in contrast to a previous case-cross-
over study (18). The other two case-crossover studies
were restricted to women only (15, 16).
Stratification on age groups showed higher odds of
sickness absence after a night shift and lower odds after
an evening shift compared to day shifts for most age
groups. The similar results across age groups is in line
with previous case-crossover studies (15, 18).
Methodological considerations
The use of the case-crossover design enabled us to
handle several previous challenges in studies of night
shifts and sickness absence. In this design, the partici-
6
Scand J Work Environ Health – online first
 BEU, Alm.del - 2022-23 (2. samling) - Bilag 67: Orientering om nye forskningsresultater fra NFA
2665204_0008.png
Larsen et al
pant was compared with herself/himself, minimizing the
need for adjustment for time-invariant covariates. With
all methodological choices made
a priori,
it is critical
to choose the right control periods in order to minimize
risk of bias, both in regards to duration and number (30).
We therefore conducted a number of sensitivity analyses
exploring this. By selecting control periods close in time
to the case period and matching by week day (in case
eg, Mondays had a higher level of sickness absence),
we reduced some time-variant confounding similar to
previous studies (32). To challenge our choice of 7
days between control periods, we tested 5 or 10 days
between control periods. Results showed higher odds of
sickness absence in all analyses. The risk estimate was
highest when including 7 days between control periods,
indicating effect of weekday. In the main analyses, we
a priori
selected case periods and control periods of 2
days as night shift work often is followed by a day off,
where sickness absence is not recorded in the payroll
data. To challenge this, we conducted analyses with case
and control periods of 1 or 3 days. When restricting to 1
day, the results became statistically insignificant, when
including 3 days, results were close to those of 2 days.
Hence, there might be a risk of missing associations if
choosing a case period that does not cover a period with
full information.
When challenging the number of control periods,
results showed lower risk estimates when choosing 1
period compared to the 5 periods chosen
a priori.
How-
ever, choosing more periods eg, 6 or 9, did not affect the
results. This is supported by the literature, where previ-
ous studies have found stable estimates when including
≥5 control periods (36). We excluded participants with
sickness absence 90 days prior to the case period in
order to avoid bias due to previous sickness absence.
However, sensitivity analyses showed no differences in
risk estimates when not doing so.
Strengths and limitations
The case-crossover design meets many of the limita-
tions related to previously used study designs eg, vul-
nerability to between-subject differences. Further, by
using DWHD, we were able to include a large number
of participants directly from the database, yielding a
100% participation rate during follow-up. The database
includes the accurate daily information on working time
and sickness absence, both of which have been validated
(2). We thereby excluded recall bias in relation to both
working time and sickness absence.
Some limitations to the study need to be addressed.
As in all observational studies addressing night and
shift work, there is a risk of selection bias in terms of
a survivor population, ie, a population of night workers
who have the highest tolerance for night and shift work,
and thereby a risk of affecting comparability between
groups of night and non-night shift workers. There is
therefore also a risk of selection bias. However, the
advantages using the case-crossover design, where a
person is her or his own control along with objective
exposure assessment gives us the possibility to handle
bias due to misclassification of exposure and therefore
increase the exchangeability.
The study includes a very high percentage of women,
which usually decreases the possibility of generaliz-
ing the results in regards to sex. However, due to the
large population size, there is still almost 1200 cases
among men, and the results from the stratified analyses
showed higher odds of sickness absence for both men
and women. Accordingly, we have no reasons to believe
results would be very different among men.
The case-crossover design allowed us to handle
time-invariant confounding such as sex and occupation
along with, to some extent, personality, genetic differ-
ences etc. as we assume that these factors did not change
during the 56 days of the control periods. Still, it may be
the case that there are time-varying differences between
the case and control periods that are not accounted for
by the case-crossover design, eg, task and workload.
This would add to residual confounding in regards to
the biological mechanisms, where task and workload
could be considered confounders. However, differences
in tasks and workload could also be considered as part
of the mechanism and then, consequently, should not
be regarded as confounding. Further, the use of hospital
employees only may limit generalization to other pro-
fessions as differences in workload in day versus night
might not be the same issue in other sectors.
Concluding remarks
This large and unique study among Danish hospital
employees indicates that the risk of calling in sick is
affected by the type of shift worked prior to the sick list-
ing, independently of sex, age and time-invariant con-
founders. Extensive testing showed that methodological
choices were important to consider when choosing the
case-crossover design.
Ethical approval
Approval from the Danish National Committee on
Biomedical Research Ethics is not required for Danish
questionnaire- and register-based studies (33).
Funding
The National Research Centre for the Working Environ-
ment, Denmark funded this study under the Governmen-
tal grant for research on sickness absence. Further, the
Scand J Work Environ Health – online first
7
 BEU, Alm.del - 2022-23 (2. samling) - Bilag 67: Orientering om nye forskningsresultater fra NFA
2665204_0009.png
Night and evening shifts and acute sickness absence
study was based on data reprocessed in prior projects
funded by NordForsk, Nordic Program on Health and
Welfare [grant number 74809] and the Danish Working
Environment Research Fund [grant number 23-2012-
09/20120220951]. No one but the authors had a role in
planning, executing and interpretation of the study or in
the decision of publishing.
The authors declare no conflicts of interest.
5.
Li W, Chen Z, Ruan W, Yi G, Wang D, Lu Z. A meta-analysis
of cohort studies including dose-response relationship
between shift work and the risk of diabetes mellitus.
Eur J Epidemiol 2019 Nov;34(11):1013–24.
https://doi.
org/10.1007/s10654-019-00561-y.
Olsson D, Alexanderson K, Bottai M. Sickness absence and
the time-varying excess risk of premature death: a Swedish
population-based prospective cohort study. J Epidemiol
Community Health 2015 Nov;69(11):1052–7.
https://doi.
org/10.1136/jech-2015-205886.
Kivimäki M, Forma P, Wikström J, Halmeenmäki T, Pentti
J, Elovainio M et al. Sickness absence as a risk marker of
future disability pension: the 10-town study. J Epidemiol
Community Health 2004 Aug;58(8):710–1.
https://doi.
org/10.1136/jech.2003.015842.
Schuring M, Robroek SJ, Otten FW, Arts CH, Burdorf A.
The effect of ill health and socioeconomic status on labor
force exit and re-employment: a prospective study with ten
years follow-up in the Netherlands. Scand J Work Environ
Health 2013 Mar;39(2):134–43.
https://doi.org/10.5271/
sjweh.3321.
Kreis JB. Indicators for work-related health monitoring in
Europe. Federal Association of Company Health Insurance
Funds, Essen, Germany. 2004.
6.
Acknowledgments
The DWHD was established in collaboration between
National Research Centre for the Working Environ-
ment, Denmark; The Danish Cancer Society Research
Centre, Denmark; Danish Ramazzini Centre, Depart-
ment of Occupational Medicine, Aarhus University
Hospital, Denmark; and Department of Public Health,
University of Copenhagen, Denmark. The establish-
ment of the DWHD was financed by research grants
from The Danish Working Environment Research Fund,
Denmark, (23-2012-09), NordForsk, Nordic Program on
Health and Welfare, Norway, (74809), and the National
Research Centre for the Working Environment, Den-
mark. The Danish regions have partially financed the
transfer of data to the cohort. We would like to acknowl-
edge the Danish regions for their participation and
willingness to provide data to the DWHD. Moreover,
we thank Jens Worm Begtrup, Anders Ørberg and Lis-
beth Nielsen from The National Research Centre for
the Working Environment for valuable work with data
management in establishing the DWHD
7.
8.
9.
10. Merkus SL, van Drongelen A, Holte KA, Labriola M, Lund
T, van Mechelen W et al. The association between shift
work and sick leave: a systematic review. Occup Environ
Med 2012 Oct;69(10):701–12.
https://doi.org/10.1136/
oemed-2011-100488.
11. Fekedulegn D, Burchfiel CM, Hartley TA, Andrew
ME, Charles LE, Tinney-Zara CA et al. Shiftwork and
sickness absence among police officers: the BCOPS study.
Chronobiol Int 2013 Aug;30(7):930–41.
https://doi.org/10.
3109/07420528.2013.790043.
12. Bernstrøm VH, Houkes I. Shift work and sickness absence
at a Norwegian hospital: a longitudinal multilevel study.
Occup Environ Med 2020 Aug;77(8):555–63.
https://doi.
org/10.1136/oemed-2019-106240.
13. Larsen AD, Ropponen A, Hansen J, Hansen ÅM, Kolstad
HA, Koskinen A et al. Working time characteristics and long-
term sickness absence among Danish and Finnish nurses: A
register-based study. Int J Nurs Stud 2020 Dec;112:103639.
https://doi.org/10.1016/j.ijnurstu.2020.103639.
14. Dall’Ora C, Ball J, Redfern OC, Griffiths P. Night work for
hospital nurses and sickness absence: a retrospective study
using electronic rostering systems. Chronobiol Int 2020
Sep-Oct;37(9-10):1357–64.
https://doi.org/10.1080/07420
528.2020.1806290.
15. Ropponen A, Koskinen A, Puttonen S, Härmä M. A case-
crossover study of age group differences in objective
working-hour characteristics and short sickness absence.
J Nurs Manag 2020 May;28(4):787–96.
https://doi.
org/10.1111/jonm.12992.
16. Ropponen A, Koskinen A, Puttonen S, Härmä M. Exposure
to working-hour characteristics and short sickness absence
References
1.
Eurofound. Sixth European Working Conditions Survey
2015. Available from:
https://wwweurofoundeuropaeu/
surveys/data-visualisation/sixth-european-working-
conditions-survey-2015
[Internet]. 2015.
Garde AH, Hansen J, Kolstad HA, Larsen AD, Pedersen
J, Petersen JD et al. Payroll data based description of
working hours in the Danish regions. Chronobiol Int 2018
Jun;35(6):795–800.
https://doi.org/10.1080/07420528.2018
.1466797.
Monographs Vol IA; IARC Monographs Vol 124 group.
Carcinogenicity of night shift work. Lancet Oncol
2019 Aug;20(8):1058–9.
https://doi.org/10.1016/S1470-
2045(19)30455-3.
Cheng M, He H, Wang D, Xu L, Wang B, Ho KM et al.
Shift work and ischaemic heart disease: meta-analysis
and dose-response relationship. Occup Med (Lond) 2019
May;69(3):182–8.
https://doi.org/10.1093/occmed/kqz020.
Scand J Work Environ Health – online first
2.
3.
4.
8
 BEU, Alm.del - 2022-23 (2. samling) - Bilag 67: Orientering om nye forskningsresultater fra NFA
2665204_0010.png
Larsen et al
in hospital workers: A case-crossover study using objective
data. Int J Nurs Stud 2019 Mar;91:14–21.
https://doi.
org/10.1016/j.ijnurstu.2018.11.002.
17. Peutere L, Rosenström T, Koskinen A, Härmä M, Kivimäki
M, Virtanen M et al. Length of exposure to long working
hours and night work and risk of sickness absence: a
register-based cohort study. BMC Health Serv Res 2021
Nov;21(1):1199.
https://doi.org/10.1186/s12913-021-
07231-4.
18. Shiri R, Hakola T, Härmä M, Ropponen
A.
The associations
of working hour characteristics with short sickness absence
among part- and full-time retail workers. on labor force
exit and re-employment: a prospective study with ten years
follow-up in the Netherlands. Scand J Work Environ Health
2021;47(4):268–76.
https://doi.org/10.5271/sjweh.3952.
19. van Drongelen A, Boot CR, Hlobil H, van der Beek AJ,
Smid T. Cumulative exposure to shift work and sickness
absence: associations in a five-year historic cohort. BMC
Public Health 2017 Jan;17(1):67.
https://doi.org/10.1186/
s12889-016-3906-z.
20. Marmot M, Feeney A, Shipley M, North F, Syme SL.
Sickness absence as a measure of health status and
functioning: from the UK Whitehall II study. J Epidemiol
Community Health 1995 Apr;49(2):124–30.
https://doi.
org/10.1136/jech.49.2.124.
21. Rothman JK, Greenland S, Lash TL. Modern Epidemiology.
3rd ed. ed: Lippincott Williams & Wilkins; 2008.
22. Laaksonen M, He L, Pitkäniemi J. The durations of past
sickness absences predict future absence episodes. J
Occup Environ Med 2013 Jan;55(1):87–92.
https://doi.
org/10.1097/JOM.0b013e318270d724.
23. Raynik YI, König HH, Hajek A. Personality Factors and Sick
Leave Days. Evidence from a Nationally Representative
Longitudinal Study in Germany. Int J Environ Res Public
Health 2020 Feb;17(3):E1089.
https://doi.org/10.3390/
ijerph17031089.
24. Svedberg P, Ropponen A, Alexanderson K, Lichtenstein
P, Narusyte J. Genetic susceptibility to sickness absence is
similar among women and men: findings from a Swedish
twin cohort. Twin Res Hum Genet 2012 Oct;15(5):642–8.
https://doi.org/10.1017/thg.2012.47.
25. Andersen LL, Vinstrup J, Thorsen SV, Pedersen J, Sundstrup
E, Rugulies R. Combined psychosocial work factors and
risk of long-term sickness absence in the general working
population: prospective cohort with register follow-up
among 69 371 workers. Scand J Work Environ Health 2022
Sep;48(7):549–59.
https://doi.org/10.5271/sjweh.4035.
26. Leineweber C, Marklund S, Gustafsson K, Helgesson M.
Work environment risk factors for the duration of all cause
and diagnose-specific sickness absence among healthcare
workers in Sweden: a prospective study. Occup Environ
Med 2020 Nov;77(11):782–9.
https://doi.org/10.1136/
oemed-2020-106510.
27. Mittleman MA, Mostofsky E. Exchangeability in the case-
crossover design. Int J Epidemiol 2014 Oct;43(5):1645–55.
https://doi.org/10.1093/ije/dyu081.
28. Hammer PE, Garde AH, Begtrup LM, Flachs EM, Hansen
J, Hansen AM et al. Night work and sick leave during
pregnancy: a national register-based within-worker cohort
study. Occup Environ Med 2019 Mar;76(3):163–8.
https://
doi.org/10.1136/oemed-2018-105331.
29. Garde AH, Hansen J, Kolstad HA, Larsen AD, Hansen
ÅM. How do different definitions of night shift affect
the exposure assessment of night work? Chronobiol Int
2016;33(6):595–8.
https://doi.org/10.3109/07420528.2016
.1167729.
30. Allison PD, Christakis NA. Fixed-Effects Methods
for the Analysis of Nonrepeated Events. University of
Pennsylvania. 2006;36(1):155–72.
31. Nabe-Nielsen K, Tüchsen F, Christensen KB, Garde AH,
Diderichsen F. Differences between day and nonday workers
in exposure to physical and psychosocial work factors in the
Danish eldercare sector. Scand J Work Environ Health 2009
Jan;35(1):48–55.
https://doi.org/10.5271/sjweh.1307.
32. Nielsen HB, Dyreborg J, Hansen AM, Kolstad HA,
Larsen AD, Nabe-Nielsen K, et al. Shift work and risk of
occupational, transport and leisuret-time injury. A register-
based case-crossover study of Danish hospital workers.
Saf Sci 2019;120:728–34.
https://doi.org/10.1016/j.
ssci.2019.07.006.
33. Wang SV, Coull BA, Schwartz J, Mittleman MA, Wellenius
GA. Potential for bias in case-crossover studies with shared
exposures analyzed using SAS. Am J Epidemiol 2011
Jul;174(1):118–24.
https://doi.org/10.1093/aje/kwr038.
34. Nordic Council of Ministers. Danish sickness benefit.
Available from:
https://www.norden.org/en/info-norden/
danish-sickness-benefit.
35. Tüchsen F, Christensen KB, Nabe-Nielsen K, Lund T.
Does evening work predict sickness absence among female
carers of the elderly? Scand J Work Environ Health 2008
Dec;34(6):483–6.
https://doi.org/10.5271/sjweh.1287.
36. Janes H, Sheppard L, Lumley T. Case-crossover
analyses of air pollution exposure data: referent selection
strategies and their implications for bias. Epidemiology
2005 Nov;16(6):717–26.
https://doi.org/10.1097/01.
ede.0000181315.18836.9d.
Received for publication: 29 September 2022
Scand J Work Environ Health – online first
9