Long working hours and health in Denmark: a post hoc analysis of three cohort studies Authors: Harald Hannerz Published: 2024 ABSTRACT BackgroundTo protect the safety and health of workers, national legislations often stipulate an upper limit to weekly working time. The purpose of the present study was to provide information that may facilitate the obtainment of evidence-based limits to weekly working time.ObjectiveTo estimate incidence rate ratios (RRs) of (i) hospital treatment or death due to accidental injuries, (ii) redeemed prescriptions for psychotropic drugs and (iii) redeemed prescriptions for antihypertensive drugs, respectively, as continuous functions of weekly working hours among employees in the general population of Denmark.MethodsBaseline data on working time arrangements were obtained from the Danish Labor Force Surveys of 1999–2013. Follow-ups with a mean duration of 1.8, 4.0 and 6.6 years for accidental injuries, psychotropic drugs and antihypertensive drugs, respectively, were accomplished through national registers. Adjusted RRs were obtained through Poisson regression in the interval 32–80 weekly working hours (wwh). The reference (standard working hours) was set at 37 wwh. RR estimates were interpreted in accordance with a modified version of Monson’s guide to strength of associations, in which a RR in the open interval 0.83 to 1.2 is classified as ‘no association’ (too weak to be detected by observational studies).ResultsThe estimated RRs as well as the upper and lower limits of their 95% confidence intervals lied within the no-association zone throughout the intervals 32–76, 32–64 and 32–69 wwh, for accidental injury, psychotropic drug use and antihypertensive drug use, respectively. The estimated RRs for psychotropic drug use decreased from 1.08 (95% CI: 1.03–1.13) at 32 wwh to 0.96 (0.93–0.99) at 45 wwh. They thereafter increased to 1.00 (0.95–1.06) at 55 wwh and 1.31 (1.08–1.63) at 80 wwh. The estimated RRs for accidental injury and antihypertensive drug use increased slowly from 1.00 at 37 wwh to 1.04 (0.88–1.24) and 1.11 (0.94–1.31), respectively, at 80 wwh.ConclusionThese results suggest that, from a societal perspective, the effect of wwh on the rates of these types of morbidity is negligible throughout the interval 32–64 wwh. Introduction Long working hours are recognized as risk factors for death due to mental and cardiovascular disorders, by the governments of Japan, Taiwan and South Korea [1–3]. They are, moreover, recognized as occupational safety and health hazards (without further specification) by the EU Working Time Directive [4], which, accordingly, stipulates that member states shall take the measures necessary to ensure that a worker’s average working time for each 7-day period, including overtime, does not exceed 48 h. The main theoretical arguments for the assumption that risks of safety and health problems among employees depend on their weekly working hours are: Firstly, that a prolonged workweek, all else being equal, will lead to an increased income and thereby a decreased risk or intensity of financial strain, which is a very important stressor among contemporary employees [5]. Secondly, that very long working hours have been associated with short sleep [6–8] and fatigue due to insufficient recovery between work shifts [6, 8–11], which in turn are risk factors for psychiatric disorders [12–18], accidental injuries [19], hypertension [20], ischaemic heart disease [21] and stroke [22]. There are, in other words, both advantages and disadvantages associated with long working hours, which suggests a U-shaped relation, where the risk of ill health decreases with weekly working hours up to a point where they become too long to allow sufficient sleep and recovery. It could also be that long working time means less time for social relationships and exercise, for example, which may also lead to poor health [23, 24]. There are plenty of studies that have estimated rate ratios among workers with long (often categorized as > = 55) vs. standard (often categorized as 35–40) working hours per week. Consequently, a series of meta-analyses have been conducted to estimate rate ratios for safety and health outcomes among employees who work > = 55 vs. 35–40 h a week [25–33], and slightly increased risks among the employees with long hours have been reported for a variety outcomes. Some recent Meta-analyses, conducted by the World Health Organization (WHO) and International Labour Organization (ILO), found, for example, that the rates of stroke [33] and ischemic heart disease [32] were significantly higher among employees who worked > = 55 vs. 35–40 h a week. Such studies and findings are interesting. However, elevated rates among employees with 55–100 vs. 35–40 working hours a week do not necessarily mean that the rates are elevated throughout the interval 55–100 h. We can therefore not use such findings as evidence that a workweek of e.g. 60 h is associated with elevated rates of the concerned outcomes. We have previously published a series of cohort studies on health and safety in relation to weekly working hours among full-time employees in the general population of Denmark, which were based on a linkage between exposure data from the Danish Labor Force Survey samples of 1999–2013 and outcome data from national health registries. The studies aimed at estimating rate ratios for psychotropic drug use and psychiatric hospital treatment [34], antihypertensive drug use and ischemic heart disease [35] and accidental injuries [36] as a function of weekly working hours. The working hours were treated as a categorical variable, with 32–40 h (the reference) to represent normal weekly working hours, 41–48 h to represent overtime work that lies within the threshold of the EU working time directive [4], and 49–100 h to represent overtime work beyond the threshold of the directive. In spite of a sufficient statistical power, the studies did not find any statistically significant main effects of weekly working hours. Moreover, the studies did not find any statistically significant effects of interaction between weekly working hours and age, gender, socioeconomic status or night-time work on any of these outcomes. However, no difference between the average rates among employees who work 49–100 vs. 32–40 h a week does not necessarily mean that such a null finding holds good throughout the interval 49–100 h. We can therefore not know if the null findings in our previous studies hold good also for employees who work for e.g. 60 h a week. The present post hoc study makes use of the data from the three previously published cohort studies to estimate relative rates of accidental injuries, incident use of psychotropic drugs and incident use antihypertensive drugs, respectively, among full-time employees in the general population of Denmark as continuous functions of weekly working hours. The reason for not including ischemic heart disease and psychiatric hospital treatment in the study is that the numbers of cases were too low to allow analyses of these outcomes as continuous functions of working hours. The study deals with the topic of long working hours from a societal perspective. At the individual level, it is generally recognized that an employee’s usual working hours may constitute a health hazard if they are too long to allow sufficient time for sleep and recovery. It is moreover recognized that the limit for when the workweek becomes too long depends on who you are, what you work with and what you do in your leisure time [37–39]. For example, an elderly scientist might be able to conduct exciting research for 60 h a week without elevated health risks. Yet, if the same individual were employed as a workman in a moving company, moving furniture up and down stairs, then it is possible that a 10-hour workweek (2 h a day, 5 days a week) would be too long to allow sufficient recovery between work shifts. From a societal perspective, a workweek may be considered too long if the safety and health risks among employees with such workweeks are considerably higher than ditto among full-time employees with standard working hours, after adjustment for relevant covariates. The purpose of the present study is to shed light on the question of how long a workweek need to be before it becomes too long, from a societal health and safety perspective. The new angle here is that working hours are treated as continuous functions in primary analysis. Methods As mentioned above, the present paper deals with a series of post hoc analyses of data from three previously published cohort studies. A brief description of the material and methods of these post hoc analyses are given below. Further details can be found in the method sections of the previously published cohort studies [36, 40, 41]. Data sources Information on weekly working hours was based on questionnaire data obtained in a series of randomly selected samples of full-time employees in the general working population of Denmark. The follow-ups were accomplished through national registers. The linkage between the questionnaire data and the register data was based on the participants’ personal identification numbers. The questionnaire data were obtained from the Danish Labor Force Surveys (DLFS) of 1999–2013, which, in accordance with an EU-directive, are based on random samples of 15 to 74-year-old people in the Danish population. New samples were drawn each quarter of the calendar years and each of the drawn individuals were invited to participate in up to four interview rounds over a period of 1.5 years [42]. The response rates in the DLFS decreased from 70% in 2002 to 53% in 2013. Register data was available for the time period 1998–2014. The following national registers were used: The central person register (for information on dates of birth deaths and migrations) [43], the employment classification module (for information on occupation, industry and socio-economic status) [44], the national patient register (for information on dates and diagnoses of hospital contacts) [45], the national cause of death register (for information on death diagnoses) [46] and the national prescription register (for information on redeemed prescriptions) [47]. Ethics approval Participation in the Danish Labor Force Survey was voluntary. The project was approved by Statistics Denmark, account number 704,291, and complies with The Act on Processing of Personal Data, Denmark (Act No. 429 of May 31, 2000), which implements the European Union Directive 95/46/EC on the protection of individuals. Exposure The exposure variable “weekly working hours” was based on the sum of the self-reported usual working hours in the primary and secondary jobs. The information was based on answers to the question “How many hours a week do you usually work in this job?” which was asked, first in relation to the participant’s primary job and then (if relevant) in relation to each of his/her secondary jobs. Clinical endpoints The following endpoints were regarded: Redeemed prescription for psychotropic drugs (ATC-code: N05 psycholeptica or N06 psychoanaleptica).Redeemed prescription for anti-hypertensive drugs (ATC-code: C02 antihypertensives, C03 diuretics, C07 alpha- and beta-blockers, C08 calcium channel blockers or C09 angiotensin-converting enzyme inhibitors and angiotensin-II antagonists).Hospital treatment or death due to accidental injuries (“poisoning and certain other consequences of external causes” (ICD-10: S00-T98) as principal diagnosis, or death with ”external causes of morbidity and mortality” (ICD-10: V01-Y98) as underlying death cause). Redeemed prescription for psychotropic drugs (ATC-code: N05 psycholeptica or N06 psychoanaleptica). Redeemed prescription for psychotropic drugs (ATC-code: N05 psycholeptica or N06 psychoanaleptica). Redeemed prescription for anti-hypertensive drugs (ATC-code: C02 antihypertensives, C03 diuretics, C07 alpha- and beta-blockers, C08 calcium channel blockers or C09 angiotensin-converting enzyme inhibitors and angiotensin-II antagonists). Redeemed prescription for anti-hypertensive drugs (ATC-code: C02 antihypertensives, C03 diuretics, C07 alpha- and beta-blockers, C08 calcium channel blockers or C09 angiotensin-converting enzyme inhibitors and angiotensin-II antagonists). Hospital treatment or death due to accidental injuries (“poisoning and certain other consequences of external causes” (ICD-10: S00-T98) as principal diagnosis, or death with ”external causes of morbidity and mortality” (ICD-10: V01-Y98) as underlying death cause). Hospital treatment or death due to accidental injuries (“poisoning and certain other consequences of external causes” (ICD-10: S00-T98) as principal diagnosis, or death with ”external causes of morbidity and mortality” (ICD-10: V01-Y98) as underlying death cause). Follow-up and inclusion criteria The present post hoc analysis is based on data obtained in three separate studies with different sets of co-authors. To minimize the risk of post hoc bias, I have used exactly the same case definitions, covariates and follow-up strategies as in the original studies. The only difference is that the present analyses excluded employees who worked more than 80 h a week. The analysis of accidental injuries included employees who responded to DLFS sometime during the calendar years 1999–2013, while the analyses of redeemed prescriptions included responders from the calendar years 2000–2013. The analyses were based on the participants’ first interview in the concerned calendar period. Each of the included participants was followed from the beginning of the calendar year that succeeded that of his/her baseline interview. The participants were censored if they died or emigrated or the follow-up period ended before they reached the particular end-point of a study. Person years at risk was calculated for each of the included employees. The maximum follow-up was set at 2 years for accidental injuries, 5 years for incident use of psychotropic drugs and 14 years for incident use of anti-hypertensive drugs. The reason for the different maximum follow-up periods was the idea that the three outcomes have different latency periods. It may take a long time to develop and detect a hypertensive disorder. Hence, a longer follow-up period for this outcome. A possible effect of long working hours on the risk of injuries, on the other hand, would most probably end as soon as the employee changes from long to short or standard working hours. Hence, a shorter follow-up period for this outcome. The five-year follow-up period for psychotropic drugs was due to a decision to make the study compatible with a prior study of long working hours and psychotropic drug use [48]. To be eligible for inclusion, participants should be between 20 and 59 years old at the start of the follow-up period and employed with 32–80 weekly working hours at the time of the interview. Participants who had experienced the endpoint of the study during the calendar year preceding the start of the follow-up were excluded from the analysis. The analyses excluded, moreover, all participants who were registered in the Employment Classification Module as unemployed or otherwise not economically active (e.g. students, sickness absence beneficiaries and pensioners) during the main part of the calendar year preceding the start of the follow-up. Covariates Age, gender and socioeconomic status (SES), calendar time and night work (‘Yes, regularly’ or ‘Yes, occasionally’ vs. ‘No’) were included as covariates in all of the studies. A term for interaction between age and gender was included as a covariate in the study of accidental injuries. In the studies of accidental injuries and psychotropic drug use, the age variable was fixed at the start of the follow-up. In the study of antihypertensive drugs, it was treated as a dynamic (time-varying) variable. In the study of accidental injuries, it was categorized into 20–24, 25–29, 30–39, 40–49 and 50–59 years. In the other studies, it was categorized into 10-year classes. In the study on psychotropic drug use, SES was coded into “legislators, senior officials and managers; professionals; technicians and associate professionals; workers in occupations that require skills at a basic level; workers in elementary occupations; gainfully occupied people with an unknown occupation” in accordance with SOCIO (Statistics Denmark’s socioeconomic classification) [49]. In the other studies, it was coded into “low; intermediate; high; unknown” in accordance with ESeC (The European Socioeconomic Classification) [50]. Calendar time (2000–2004; 2005–2009; 2010–2014) and time passed since start of follow-up (0–4 years; 5–9 years; >=10 years) were included as dynamic variables in the study of antihypertensive drug usage. In the study of accidental injuries and the study on psychotropic drug use, we included calendar year of interview as a covariate. It was categorized as (1999–2000; 2001–2006; 2007–2013) in the study of accidental injuries and as (2000–2004; 2005–2009; 2010–2013) in the study on psychotropic drug use. Employment in the healthcare industry (Yes vs. No) was used as a covariate in the study on antihypertensive drug use. Industrial sector (agriculture and fishing; manufacturing; construction; wholesale and retail trade; transport and storage; accommodation and food service; human health and social work; other; unknown) was included as a covariate in the study on accidental injuries. Statistical analysis Poisson regression was used to estimate endpoint specific rate ratios (RR) as functions of weekly working hours (wwh). The analyses were adjusted for covariates as described above. In the primary analyses, weekly working hours (wwh) were treated as a continuous variable in accordance with the following equation:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:log\left(RR\right)=\alpha\:\left(log\left(\frac{wwh}{37}\right)\right)+\beta\:{\left(log\left(\frac{wwh}{37}\right)\right)}^{2}$$\end{document} The equation contains one linear and one quadratic term, which makes it possible to use it also for non-linear (e.g. J-shaped or U-shaped) associations. Goodness-of-fit is a measure of how close the observed numbers of cases are to the numbers that are expected according to the statistical model. Pearson’s chi-square statistic tests the null hypothesis that the model fits the data vs. the alternative that the model does not fit the data. The null hypothesis is rejected at the significance level alpha if the P-value of the test is less than alpha. The goodness-of-fit of the present statistical model was evaluated by use of Pearson’s chi-square test with wwh compartmented into the categories: 32–34, 35–39, 40–44, 45–49, 50–54, 55–59, 60–64, 65–69, 70–80. The goodness-of-fit tests yielded the p-values 0.61, 0.20 and 0.91 for accidental injuries, psychotropic drugs and antihypertensive drugs, respectively. Each primary analysis was supplemented with a sensitivity analysis in which wwh was treated as a categorical variable divided into 32–40 (reference group), 41–48, 49–59 and 60–70 h a week. Strength of associations To allow for possible effects of selection bias, misclassifications and uncontrolled confounding, rate ratios are interpreted in accordance with a slightly modified version of Monson’s guide to strength of association [51], in which a rate ratio in the open interval 0.83 to 1.2 is classified as ‘no association’ (too weak to be detected by observational studies), a rate ratio in the interval 0.67–0.83 or 1.2–1.5 is a weak association, a rate ratio in the interval 0.33–0.67 or 1.5–3.0 is a moderate association, a rate ratio in the interval 0.10–0.33 or 3.0–10.0 is a is a strong association and a rate ratio less than 0.1 or greater than 10.0 is a very strong association. The cut-off points for rate ratios above unity are exactly the same as the ones proposed by Monson. The cut-off points for rate ratios below unity has been modified (harmonized) to be equal to the reciprocals of the cut-off points for the rate ratios above unity. E.g. in Monson’s guide an RR is classified as a weak association if it lies in the interval 0.7–0.9 or 1.2–1.5. The corresponding intervals in the modified version are 1/1.5–1/1.2 and 1.2–1.5. In the present study, an estimated RR will be interpreted as “no association” if both the upper and lower limits of its 95% confidence interval lie within Monson’s no association zone. Here it should be noted that there is a difference between “not finding an association”, which could be due to an insufficient statistical power, and “finding that there is no association”, which can only happen in studies in which the statistical power to detect a weak association (i.e. a rate ratio of 1.2) is high (e.g. at least 95%). Results The respective inclusion criteria were fulfilled by 149 916 participants in the analysis of hospital treatment or death due to accidental injuries, 130 873 participants in the analysis of redeemed prescriptions for psychotropic drugs and 124 973 participants in the analysis of redeemed prescriptions for antihypertensive drugs. In total, 23 402 cases of accidental injury, 15 766 cases of psychotropic drug use and 20 570 cases of antihypertensive drug use were observed in 272 765, 520 296 and 830 021 person years at risk, respectively. The mean follow-up times were 1.8, 4.0 and 6.6 years for accidental injury, psychotropic drug use and antihypertensive drug use, respectively. Figures 1, 2 and 3 show the estimated rate ratios with 95% confidence intervals for accidental injury, psychotropic drug use and antihypertensive drug use, respectively, as a function of weekly working hours. The figures also contain horizontal lines at the rate ratio levels 0.83 and 1.20, which are there to denote the lower and upper limits of Monson’s no association zone. Here, we can see that the lower limit of the 95% confidence interval lies within Monson’s no association zone throughout the examined interval of 32–80 working hours a week, for each of the studied endpoints. We can also see that the upper limit of the 95% confidence interval lies within the no-association zone throughout the intervals 32–76, 32–64 and 32–69 weekly working hours, for accidental injury, psychotropic drug use and antihypertensive drug use, respectively. Which suggests that, from a societal perspective, the effect of weekly working hours on the rates of these end-point specific outcomes is negligible throughout these intervals. Fig. 1Estimated rate ratio with 95% confidence interval for hospital treatment or death due to accidental injury as a continuous function of hours worked per week, among employees in Denmark 1999–2013. The horizontal lines at 0.83 and 1.20 are the lower and upper limits of Monson’s no association zone Estimated rate ratio with 95% confidence interval for hospital treatment or death due to accidental injury as a continuous function of hours worked per week, among employees in Denmark 1999–2013. The horizontal lines at 0.83 and 1.20 are the lower and upper limits of Monson’s no association zone Fig. 2Estimated rate ratio with 95% confidence interval for incident use of psychotropic drugs as a continuous function of hours worked per week, among employees in Denmark 2000–2013. The horizontal lines at 0.83 and 1.20 are the lower and upper limits of Monson’s no association zone Estimated rate ratio with 95% confidence interval for incident use of psychotropic drugs as a continuous function of hours worked per week, among employees in Denmark 2000–2013. The horizontal lines at 0.83 and 1.20 are the lower and upper limits of Monson’s no association zone Fig. 3Estimated rate ratio with 95% confidence interval for incident use of antihypertensive drugs as a continuous function of hours worked per week, among employees in Denmark 2000–2013. The horizontal lines at 0.83 and 1.20 are the lower and upper limits of Monson’s no association zone Estimated rate ratio with 95% confidence interval for incident use of antihypertensive drugs as a continuous function of hours worked per week, among employees in Denmark 2000–2013. The horizontal lines at 0.83 and 1.20 are the lower and upper limits of Monson’s no association zone The results of the sensitivity analyses were quite similar to the ones obtained in the primary analyses (cf. Table 1). Table 1Endpoint specific rate ratios with 95% confidence interval (CI) as a function of weekly working hours among employees in the general population of DenmarkEndpointWeekly working hoursPersonsPerson yearsCasesRR95% CICases per 1000 Person yearsHospital treatment or death due to accidental injuries*71–80400719761.070.85–1.3410660–70200835923461.050.95–1.179649–59603710,9299691.000.94–1.078941–4814,85227,06222810.960.91–1.008432–40126,619230,46319,7301.00-86Redeemed prescription for psychotropic drugs**71–803501415411.020.75–1.382960–70178572412251.130.99–1.293149–59541022,3826521.060.98–1.152941–4813,59257,16415680.940.89–0.992732–40109,736432,09413,2801.00-31Redeemed prescription for antihypertensive drugs***71–803282164591.170.91–1.522760–70169512,1803111.070.95–1.192649–59515638,3479021.000.93–1.072441–4812,97899,09023390.990.95–1.042432–40104,816680,24016,9591.00-25* Adjusted for sex, age, night work, calendar time, socioeconomic status and industry** Adjusted for sex, age, night work, calendar time and socioeconomic status*** Adjusted for sex, age, night work, calendar time, socioeconomic status, time passed since start of follow-up and health-care work (Yes vs. No) Table 1Endpoint specific rate ratios with 95% confidence interval (CI) as a function of weekly working hours among employees in the general population of DenmarkEndpointWeekly working hoursPersonsPerson yearsCasesRR95% CICases per 1000 Person yearsHospital treatment or death due to accidental injuries*71–80400719761.070.85–1.3410660–70200835923461.050.95–1.179649–59603710,9299691.000.94–1.078941–4814,85227,06222810.960.91–1.008432–40126,619230,46319,7301.00-86Redeemed prescription for psychotropic drugs**71–803501415411.020.75–1.382960–70178572412251.130.99–1.293149–59541022,3826521.060.98–1.152941–4813,59257,16415680.940.89–0.992732–40109,736432,09413,2801.00-31Redeemed prescription for antihypertensive drugs***71–803282164591.170.91–1.522760–70169512,1803111.070.95–1.192649–59515638,3479021.000.93–1.072441–4812,97899,09023390.990.95–1.042432–40104,816680,24016,9591.00-25* Adjusted for sex, age, night work, calendar time, socioeconomic status and industry** Adjusted for sex, age, night work, calendar time and socioeconomic status*** Adjusted for sex, age, night work, calendar time, socioeconomic status, time passed since start of follow-up and health-care work (Yes vs. No) Endpoint specific rate ratios with 95% confidence interval (CI) as a function of weekly working hours among employees in the general population of Denmark * Adjusted for sex, age, night work, calendar time, socioeconomic status and industry ** Adjusted for sex, age, night work, calendar time and socioeconomic status *** Adjusted for sex, age, night work, calendar time, socioeconomic status, time passed since start of follow-up and health-care work (Yes vs. No) Table 2 (in combination with Eq. 1) contains all the information that is needed to find the numbers behind the curves in Figs. 1, 2 and 3. This information is given to facilitate the use of the regression-based rate ratio estimates in future meta-analyses. Table 2Estimates, variances and covariance of the endpoint specific regression coefficientsEndpointαβVar(α)Var(β)Cov(α,β)Hospital treatment or death due to accidental injuries-0.06320.15280.011420.05965-0.02202Redeemed prescription for psychotropic drugs-0.38920.97730.014610.07857-0.02775Redeemed prescription for antihypertensive drugs-0.04180.24100.011410.06403-0.02246 Table 2Estimates, variances and covariance of the endpoint specific regression coefficientsEndpointαβVar(α)Var(β)Cov(α,β)Hospital treatment or death due to accidental injuries-0.06320.15280.011420.05965-0.02202Redeemed prescription for psychotropic drugs-0.38920.97730.014610.07857-0.02775Redeemed prescription for antihypertensive drugs-0.04180.24100.011410.06403-0.02246 Estimates, variances and covariance of the endpoint specific regression coefficients The distributions of weekly working hours, stratified by age, gender and SES, are given in Tables 3, 4 and 5 for the study populations of accidental injury, psychotropic drug use and antihypertensive drug use, respectively. Here it should be noted that the percentages that are given in the tables concern the study populations only and should not be construed as prevalences in the total population of Denmark. Table 3Distributions of weekly working hours among the participants in the injury analysis, stratified by age, gender and socioeconomic status (SES)32–4041–4849–5960–7071–80 N % N % N % N % N %Total126,61984.514,8529.960374.020081.34000.3Age 20–24904887.18908.63213.11041.0300.3 25–2912,23684.5149510.35393.71681.2400.3 30–3933,88584.2414610.316074.04931.21050.3 40–4936,33283.6443210.218754.36751.61290.3 50–5935,11884.938899.416954.15681.4960.2Gender Men63,89880.2916411.546565.816182.03040.4 Women62,72189.256888.113812.03900.6960.1SES High35,38980.5534012.224155.57071.6900.2 Medium24,44987.723168.38533.12310.8340.1 Low54,06186.955478.917762.96311.01820.3 Unknown12,72080.0164910.49936.24392.8940.6 Table 3Distributions of weekly working hours among the participants in the injury analysis, stratified by age, gender and socioeconomic status (SES)32–4041–4849–5960–7071–80 N % N % N % N % N %Total126,61984.514,8529.960374.020081.34000.3Age 20–24904887.18908.63213.11041.0300.3 25–2912,23684.5149510.35393.71681.2400.3 30–3933,88584.2414610.316074.04931.21050.3 40–4936,33283.6443210.218754.36751.61290.3 50–5935,11884.938899.416954.15681.4960.2Gender Men63,89880.2916411.546565.816182.03040.4 Women62,72189.256888.113812.03900.6960.1SES High35,38980.5534012.224155.57071.6900.2 Medium24,44987.723168.38533.12310.8340.1 Low54,06186.955478.917762.96311.01820.3 Unknown12,72080.0164910.49936.24392.8940.6 Distributions of weekly working hours among the participants in the injury analysis, stratified by age, gender and socioeconomic status (SES) Table 4Distributions of weekly working hours among the participants in the analysis of antihypertensive drug use, stratified by age, gender and socioeconomic status (SES)32–4041–4849–5960–7071–80 N % N % N % N % N %Total104,81683.912,97810.451564.116951.43280.3Age 20–2917,32084.6212610.47303.62381.2580.3 30–3930,42383.8388310.714704.04511.2940.3 40–4931,06483.1394210.516614.45971.61120.3 50–5926,00984.430279.812954.24091.3640.2Gender Men52,89279.4809112.140046.013692.12530.4 Women51,92489.048878.411522.03260.6750.1SES High30,67180.3477512.520755.45901.5760.2 Medium20,42087.320378.77193.12000.9240.1 Low43,62386.247639.415393.05341.11470.3 Unknown10,10279.0140311.08236.43712.9810.6 Table 4Distributions of weekly working hours among the participants in the analysis of antihypertensive drug use, stratified by age, gender and socioeconomic status (SES)32–4041–4849–5960–7071–80 N % N % N % N % N %Total104,81683.912,97810.451564.116951.43280.3Age 20–2917,32084.6212610.47303.62381.2580.3 30–3930,42383.8388310.714704.04511.2940.3 40–4931,06483.1394210.516614.45971.61120.3 50–5926,00984.430279.812954.24091.3640.2Gender Men52,89279.4809112.140046.013692.12530.4 Women51,92489.048878.411522.03260.6750.1SES High30,67180.3477512.520755.45901.5760.2 Medium20,42087.320378.77193.12000.9240.1 Low43,62386.247639.415393.05341.11470.3 Unknown10,10279.0140311.08236.43712.9810.6 Distributions of weekly working hours among the participants in the analysis of antihypertensive drug use, stratified by age, gender and socioeconomic status (SES) Table 5Distributions of weekly working hours among the participants in the analysis of psychotropic drug use, stratified by age, gender and socioeconomic status (SES)32–4041–4849–5960–7071–80 N % N % N % N % N %Total109,73683.813,59210.454104.117851.43500.3Age 20–2918,72684.8224010.17963.62561.2650.3 30–3929,75583.7379410.714534.14431.2930.3 40–4931,48383.0405310.716794.46151.61140.3 50–5929,77284.335059.914824.24711.3780.2Gender Men56,25879.6849112.042045.914562.12690.4 Women53,47888.851018.512062.03290.5810.1SES Legislators and managers275059.694220.466714.52435.3110.2 Professionals20,79881.7314112.311654.62941.2440.2 Associate professionals19,89085.2234010.08463.62281.0450.2 Workers in occupations that require skills at a basic level45,41386.748909.314662.84710.91300.2 Unskillled workers944788.57437.03283.11261.2310.3 Unknown11,43879.3153610.69386.54232.9890.6 Table 5Distributions of weekly working hours among the participants in the analysis of psychotropic drug use, stratified by age, gender and socioeconomic status (SES)32–4041–4849–5960–7071–80 N % N % N % N % N %Total109,73683.813,59210.454104.117851.43500.3Age 20–2918,72684.8224010.17963.62561.2650.3 30–3929,75583.7379410.714534.14431.2930.3 40–4931,48383.0405310.716794.46151.61140.3 50–5929,77284.335059.914824.24711.3780.2Gender Men56,25879.6849112.042045.914562.12690.4 Women53,47888.851018.512062.03290.5810.1SES Legislators and managers275059.694220.466714.52435.3110.2 Professionals20,79881.7314112.311654.62941.2440.2 Associate professionals19,89085.2234010.08463.62281.0450.2 Workers in occupations that require skills at a basic level45,41386.748909.314662.84710.91300.2 Unskillled workers944788.57437.03283.11261.2310.3 Unknown11,43879.3153610.69386.54232.9890.6 Distributions of weekly working hours among the participants in the analysis of psychotropic drug use, stratified by age, gender and socioeconomic status (SES) Discussion The present study estimated rate ratios for psychotropic drug usage, antihypertensive drug usage and accidental injuries, as continuous functions of weekly working hours among full-time employees in the general population of Denmark. The purpose the study was to shed light on the question of how long a workweek need to be before it becomes too long, from a societal health and safety perspective. The estimated rate ratios as well as the upper and lower limits of their 95% confidence intervals lied within Monson’s no association zone throughout the interval 32–64 working hour a week, for all of the examined endpoints. The results of the present study thereby provide evidence against the notion of significantly elevated rates of morbidity among employees who choose to work up to 64 h a week compared with employees with standard full-time working hours, after adjustment for relevant covariates. The confidence intervals were, however, too wide to rule out the possibility of elevated morbidity rates among employees who choose to work for 65 or more hours a week. Since this is an observational study and not a randomized controlled trial, there are several possible explanations for the observed absence of an effect in the interval 32–64 working hour a week. It is, for example, possible that a workweek at 64 h is short enough to allow sufficient time for sleep and recovery among most of the employees who chose to work for that many hours. It is also possible that health selection effects “the broadest shoulders carry the heaviest load” have offset potentially detrimental effects of long working hours. Either way, the absence of an effect in the interval 32–64 working hours a week indicates that voluntary long workweeks for up to 64 h do not constitute a safety and health problem, from a societal perspective. Previous research The results of the present study are in line with the results obtained in a study by Alicandro et al. [52], in which the rate ratio for cardiovascular disease mortality in the time period 2012–2016, was estimated as a continuous function of weekly working hours, among employees in the general population of Italy. The analysis included employees who worked between 35 and 70 h a week at baseline and the reference was set at 40 weekly working hours. The estimated rate ratios were stratified by gender and adjusted for age, marital status, geographic area of residence, number of dependent children and type of occupation. Approximately 8.1 million men and 3.8 million women were included in the analysis and approximately 15,000 and 1900 cases of death due to cardiovascular disease were observed, among men and women, respectively, during the five-year follow-up period. The estimated rate ratios lied within Monson’s no association zone throughout the examined intervals while the upper limit of their 95% confidence intervals lied within no association zone throughout the weekly working hour interval 35–68 among the men and 35–60 among the women. I am not aware of any other prospective studies that have attempted to model rate ratios as a continuous function of weekly working hours. As mentioned in the introduction, several meta-analyses have been conducted to estimate rate ratios for safety and health outcomes among employees who work > = 55 vs. 35–40 h a week. The most comprehensive of these meta-analyses [31] included rate ratio estimates for 46 physical and mental diseases and 4 mortality endpoints, among employees with > = 55 vs. 35–40 weekly working hours, in eleven cohorts from four European countries: Finland, Sweden, Denmark, and the United Kingdom. Elevated rates were reported for cardiovascular death, hospital-treated infections, diabetes, injuries, and musculoskeletal disorders. The found associations were, however, quite weak, and the authors concluded that “In this European context, long working hours seem not to be a major health risk factor at the population level”. This conclusion aligns well with the results of the present study. Strengths, weaknesses and limitations The main strengths of the present study were that the registers used to determine the clinical endpoints of the study covered all residents of the target population, the analyses were censored for deaths and emigrations, and prevalent cases were excluded in the analyses of psychotropic and antihypertensive drug usage. The study was large enough to afford statistical evidence against the notion of significantly elevated rates of safety and health problems among employees in a general working population who choose to work up to 64 h a week compared with standard full-time working hours, after adjustment for relevant covariates. The study was, however, too small to afford evidence for or against the notion of elevated rates of mental ill health among employees who chose to work between 65 and 80 h a week. It was, moreover, too small to allow analyses of working hours as a continuous function stratified by gender, age, occupational based SES or night-time work, respectively. Our previous cohort studies, which treated working hours as a categorical variable, did, however, not find any statistically significant effects of interaction between weekly working hours and age, gender, SES or night-time work on any of the outcomes that were analyzed in the present post-hoc analyses [34–36]. The major weaknesses of the study are that the analyses were conducted post hoc and that the exposures to long working hours were self-reported. If workers perceive their work as more arduous when they are in a poor health condition than they do when they are in good physical or mental health then it is possible that they also perceive (recall) their working hours as longer when they are in a poor health condition than they do when they are in good health, and if this is the case it would bias the results towards the hypothesis of an increased risk among workers with long hours. Another limitation is that the data material did not differentiate between time spent “on-call” and time spent on active provision of services. Yet another drawback of DLFS is that it does not request information about the duration of the exposure. Hence, we can only know what the participants’ usual working hours were at the time of the interview. It has, however, been shown that weekly working hours tend to be quite stable over time [53]. A panel study, that included a random sample of Danish inhabitants who were 30 years or older and economically active in 2000 and 2005, has estimated that 68% of the workers with long working hours (> 40 h a week) and 83% with standard working hours (32–40 h a week) in 2005 belonged to the same category five years earlier [53]. Next, it should be noted that rate ratios of medical treatment were used as proxy measures for morbidity rate ratios. Since such rate ratios may depend not only on differences in the need of treatment but also on differences in the propensity to seek medical treatment when needed, we cannot rule out the possibility that the estimated rate ratios are biased by differences in the propensity to seek medical treatment. Finally, it should be noted that the response rates at the baseline interviews were quite low (53–70%) and that it is possible that the association between weekly working hours and rates of ill health among the responders differed from ditto among the non-responders. Generalizability It should be noted that the present study as well as the cohort studies conducted by Alicandro et al. [52] and Ervasti et al. [31], were conducted in nations where working time arrangements are/were regulated by the EU working time directive [4]. This directive prohibits employers from ordering employees to work more than an average of 48 h a week against their will. The above-mentioned cohort studies thereby concerned comparisons of health and safety among employees who had chosen to work long hours with ditto among employees with standard working hours. It is possible that the findings of these studies do not hold good in nations without such working time restrictions. It should also be noted that the absence of a general effect does not exclude the possibility of detrimental effects among subgroups of the workforce. Implications for practice In keeping with the need to protect the safety and health of workers the EU Working Time Directive [4] stipulates that Member States shall take the measures necessary to ensure that: Article 3: “every worker is entitled to a minimum daily rest period of 11 consecutive hours per 24-hour period.”Article 4: “where the working day is longer than six hours, every worker is entitled to a rest break.”Article 5: “per each seven-day period, every worker is entitled to a minimum uninterrupted rest period of 24 hours plus the 11 h’ daily rest referred to in Article 3.“Article 6: “(a) the period of weekly working time is limited by means of laws, regulations or administrative provisions or by collective agreements or agreements between the two sides of industry; (b) the average working time for each seven-day period, including overtime, does not exceed 48 hours.”Article 7: “every worker is entitled to paid annual leave of at least four weeks in accordance with the conditions for entitlement to, and granting of, such leave laid down by national legislation and/or practice.” Article 3: “every worker is entitled to a minimum daily rest period of 11 consecutive hours per 24-hour period.” Article 3: “every worker is entitled to a minimum daily rest period of 11 consecutive hours per 24-hour period.” Article 4: “where the working day is longer than six hours, every worker is entitled to a rest break.” Article 4: “where the working day is longer than six hours, every worker is entitled to a rest break.” Article 5: “per each seven-day period, every worker is entitled to a minimum uninterrupted rest period of 24 hours plus the 11 h’ daily rest referred to in Article 3.“ Article 5: “per each seven-day period, every worker is entitled to a minimum uninterrupted rest period of 24 hours plus the 11 h’ daily rest referred to in Article 3.“ Article 6: “(a) the period of weekly working time is limited by means of laws, regulations or administrative provisions or by collective agreements or agreements between the two sides of industry; (b) the average working time for each seven-day period, including overtime, does not exceed 48 hours.” Article 6: “(a) the period of weekly working time is limited by means of laws, regulations or administrative provisions or by collective agreements or agreements between the two sides of industry; (b) the average working time for each seven-day period, including overtime, does not exceed 48 hours.” Article 7: “every worker is entitled to paid annual leave of at least four weeks in accordance with the conditions for entitlement to, and granting of, such leave laid down by national legislation and/or practice.” Article 7: “every worker is entitled to paid annual leave of at least four weeks in accordance with the conditions for entitlement to, and granting of, such leave laid down by national legislation and/or practice.” In accordance with Article 18 of the directive, Articles three to five may be derogated by collective agreements on the condition that equivalent compensating rest periods are granted to the workers concerned. In accordance with Article 22 of the directive, Member States may opt out of Article 6, under the condition that the general principles of the protection of the safety and health of workers are respected. They may, however, not opt out of Article 3, 4, 5 and 7. An opt out of Article 6 allows Member States to permit non-application of the 48-hour limit to average weekly working time, subject to agreement of the worker concerned. In other words, the opt-out does not allow an employer to order employees to work more than an average of 48 h a week against their will. Moreover, it does not allow an employer to discriminate against employees who are unwilling to give their agreement to perform such work. Member States can decide to opt out of Article 6 without specifying an upper limit to weekly working time for the employees who consent to work for more than 48 h per week. Member States can also decide to opt-out with the condition that the average working time for each seven-day period, including overtime, does not exceed e.g. 60 h. Alternatively, they can decide to not opt out. Denmark did not apply the opt out clause during the period covered by the present study, which means that employees ordinarily needed more than one job and more than one employer in order to exceed the 48-hour limit to weekly working time. From July 2024, the Danish law permits the most representative social partners of the two sides of industry to introduce collective agreements that allow employees who are covered by collective agreement provisions on on-call shifts, and who perform socially critical functions within a series of pre-specified areas, to opt out of the 48-hour limit [54]. It, moreover, authorizes the Minister of Employment to, upon joint request by the most representative social partners of the two sides of industry, lay down further rules on the possibility of deviations from the 48-hour limit in other areas. The present study sheds light on the question of how long a workweek needs to be before it becomes too long from a societal safety and health perspective. The information provided by the study is therefore of special interest to anyone who aims at obtaining evidence-based decisions about the application or non-application of the opt-out clause of the European Working Time Directive. Recommendations for further research The information given in Figs. 1, 2 and 3 provides evidence against the notion of significantly elevated rates of safety and health problems among employees who choose to work up to 64 h a week compared with employees with standard full-time working hours in a general working population, after adjustment for relevant covariates. I propose that this type of information is valuable to anyone who aims at obtaining evidence-based working time regulations. The study was, however, conducted post hoc and it was too small to afford evidences for or against the notion of elevated rates of morbidity among employees who chose to work between 65 and 80 h a week. I therefore recommend that future research should aim at strengthening the evidence of the present study through replication of the analyses in a larger data set. I moreover propose that future research should aim at constructing graphs similar to the ones in the present paper also for other safety and health outcomes, e.g. long-term sickness absence. Finally, I recommend that future surveys and registers of working time arrangements should be designed in a way that makes it possible to differentiate between passive and active working time, i.e. between on-call time at the workplace and time spent on active provision of services [55]. Conclusion The results of the present study constitute statistical evidence against the notion of significantly elevated rates of accidental injuries, mental ill health and hypertension related disorders among employees who choose to work up to 64 h a week compared with employees with standard full-time working hours in the general population of Denmark, after adjustment for relevant covariates. This evidence implies that an opt-out of the 48-hour limit of the EU Working Time Directive combined with a condition that the average working time does not exceed 64 h a week would afford sufficient protection against excess rates of these types of morbidity among the employees who choose to work for more than 48 h a week.