Article Text

Download PDFPDF

Sick at work—a risk factor for long-term sickness absence at a later date?
  1. C D Hansen,
  2. J H Andersen
  1. Department of Occupational Medicine, Herning Hospital, Herning, Denmark
  1. Mr C D Hansen, Department of Occupational Medicine, Herning Hospital, Gl. Landevej 61, DK 7400 Herning, Denmark; clhan{at}


Background: Little is known about the long-term consequences of sickness presence (ie, going to work despite ill-health), although one study suggests an association with coronary heart disease. This study examined the effect of sickness presence on future long-term sickness absence.

Methods: Information from a random sample of 11 838 members of the Danish core workforce was collected from questionnaires, containing questions about work, family and attitudes towards sickness absence. Information on prospective sickness absence spells of at least 2 weeks was derived from an official register during a follow-up period of 1.5 years.

Results: Sickness presence is associated with long-term sickness absence of at least 2 weeks’ duration as well as with spells lasting at least 2 months. Participants who had gone to work ill more than six times in the year prior to baseline had a 74% higher risk of becoming sick-listed for more than 2 months, even when controlling for a wide range of potential confounders as well as baseline health status and previous long-term sickness absence. The association was consistent for most subgroups of employees reporting various symptoms, but either disappeared or became insignificant when analysing subgroups of employees with specific chronic diseases.

Conclusions: Going to work ill repeatedly is associated with long-term sickness absence at a later date. For this reason, researchers and policy-makers should consider this phenomenon more carefully when planning future studies of sickness absence or when laying out new policies.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Sickness presence (SP), ie, turning up at work despite ill-health, has been a subject of steadily increasing interest over the last decade, although the number of studies is still clearly inferior to that of its twin sister: sickness absence (SA).1 2 Most studies with an epidemiological or occupational health perspective have focused on finding those job and personal characteristics that predict going to work ill instead of staying home and taking time off.36 However, there are also a number of organisational studies primarily from the United States that have examined presenteeism.7 8

The idea of studying SP originates from three distinct sources: first, the hypothesis that SA serves as a way of coping with bad health for employees, enabling them to rest in order to recuperate more thoroughly than is possible when going to work.9 Those taking no short-term SA would thus comprise a group of employees particularly prone to becoming ill at a later date because of shortage of restitution. Second, a u-curved association between short-term absence and different indicators of ill-health, eg, mortality, has been observed in different studies indicating that a clear SA record is not always an indicator of good health.10 11 Finally, a number of studies have tried to measure the impact of SP on employees’ productivity, reaching the conclusion that being absent can sometimes be less expensive for companies in terms of productivity, because employees are underperforming, when they turn up at work ill.7 8

So far, only one study has explicitly focused on the possible detrimental effects of SP on health. In a paper from the Whitehall study, the authors found that, for “unhealthy” men with no recorded days of absenteeism, the risk of serious coronary disease was twice as high as it was for those “unhealthy” men who had a moderate amount of sick leave (1–14 days). They interpreted this as evidence of a detrimental effect of SP due to the lack of restitution in the group with no absences compared with those who had one or several days of absence.12

The aim of this paper is to replicate the findings of the Whitehall study utilising a more global measure of health. We examined whether self-reported SP was a risk factor for later long-term SA, even among the healthiest part of the workforce.


Study population

The data stem from an original random sample of 30 000 inhabitants in Denmark between the ages of 19 and 64 years obtained via the CPR (Central Person Register).13 Only healthy members of the core workforce of Danish origin were included in the study. This means that only employees who had been in employment for at least 80% of the time during the previous year or had been employed for 6 out of the 12 weeks preceding 1 July 2004 (which was the date at which this information was collected) were asked to fill out a questionnaire. Employees who had taken SA for more than 10 weeks in the 12 months preceding the baseline questionnaire as well as those sick-listed at the time of answering the questionnaire were excluded from the study population, as were students, people on parental leave and immigrants (because of possible language problems). DREAM, a register from the Danish Ministry of Employment which contains weekly information on all social welfare payments for the Danish population,14 was used to exclude those from the original sample who did not meet our inclusion criteria. This resulted in a total sample of 20 464 employees who received a questionnaire in September 2004.

A total of 14 241 respondents returned the questionnaire resulting in a response rate of 70%. By comparing the distribution of certain key variables (ie, age, gender, region and occupational status) in our study with official statistics held by Statistics Denmark, we found a small under-representation of participants below the age of 30 years, a slight under-representation of male respondents and a similar small lack of respondents with lower occupational social status.

Sickness presence

SP was measured by a single-item question with a wording similar to that used in the few existing studies of SP.3 4 6 The respondents were asked: “How many times during the last 12 months have you gone to work even though it would have been reasonable to take sick leave?” (“none”, “once”, “2–3 times2, “4–5 times”, “6–10 times” and “more than 10 times”). Owing to the relatively small number of participants in the last two categories, we decided to recode the variable into three categories: <2 SP episodes, 2–5 SP episodes and ⩾6 SP episodes.

Long-term sickness absence

Long-term sickness absence was obtained from DREAM. All employees are entitled to compensation from the first day of SA. The first 2 weeks are paid for by the employer and, for this reason, the Ministry’s register only contains information for those spells of SA lasting 2 weeks or more.

We calculated the number of weeks between answering the questionnaire and the onset of the first spell of SA in a period of up to 1.5 years following baseline. This was done for all spells registered as well as for those spells lasting at least 2 months (ie, 8 weeks in the register) in order to isolate those spells that are most likely to be the result of more severe medical conditions.

Only those weeks where the participant was at risk of being sick-listed were counted (eg, if people were unemployed for some weeks, this would not count as time at risk). Those participants who emigrated, retired or died during the follow-up period were censored.

Known risk factors and other factors associated with SA

The following factors were included as covariates to control for potential confounding: demographic factors (age, gender, social class using the Eriksson–Goldthorpe–Portacarero (EGP) scheme);15 work-related factors (physical work environment: monotonous repetitive work (α = 0.83), heavy lifting (α = 0.89) using four items from the Dutch Musculoskeletal Questionnaire (DMQ);16 psychosocial work environment: single items measuring job demands, decision latitude, social support and time pressure); effort–reward imbalance using 10 items, effort (α = 0.75), reward (α = 0.79).17 Other work-related factors were measured using single items: job insecurity,18 degree of cooperation with colleagues, employment conditions (tenure vs fixed term), supervisor status and total work hours per week); family-related factors (a single item measuring satisfaction with family life, and work-to-family conflict measured with four items (α = 0.72), personal factors (lifestyle: body mass index (BMI), smoking, drinking and leisure-time physical activity (a single item measuring trait anxiety, overcommitment measured with six items (α = 0.78)17 and attitudes towards sickness absence measured using seven items (α = 0.69)). Information on number of children and cohabitation status was obtained from registers at Statistics Denmark.

Unless noted in table 1, these variables were recoded into dichotomous variables for use in the analysis. The cut-off points were chosen from theoretical considerations so that, in each case, they reflected a problematic situation vs a non-problematic situation. Details on the exact wording of most of the questions can be found in an earlier paper.6

Table 1 Sociodemographic characteristics, known risk factors for long-term sickness absence and additional factors associated with sickness absence by number of sickness presence (SP) episodes

Health status

Individual health status was assessed by asking the respondents: “In general, would you say your health is (excellent, very good, good, fair, poor)?”19 We also created a variable indicating whether the respondent was flagged in DREAM as having at least one spell of SA of 2 weeks’ duration in the 12 months preceding the questionnaire. Finally, to examine the consequences of SP for employees with specific chronic diseases, we created dichotomous variables for each of the following self-reported chronic diseases or symptoms: psychiatric disorder, high blood pressure, asthma, rhinitis, eczema, rheumatoid arthritis, severe musculoskeletal pain, mental health problems and, finally, one dichotomous variable indicating a “completely healthy” group (ie, those reporting no chronic diseases, low musculoskeletal pain, no mental health problems and having taken SA the year before only due to “influenza, cold or other passing conditions”).

Statistical analysis

Cox proportional hazards regression was used to analyse differences in the time until onset of the first spell of sick leave in the follow-up period. We ran the analyses both for spells of 2 weeks’ minimum duration and for spells lasting a minimum of 2 months—in both cases, the same procedure and models were tested. To examine whether the association between SP and SA changed in the event of the participants having reported health problems, the analyses were carried out for each of the subgroups mentioned above. This was done only for spells of a minimum of 2 weeks because the number of SA events of at least 2 months’ duration was too small to be analysed properly in the subgroups. All analyses were conducted using only respondents with no missing data for all the variables used, resulting in a total number of respondents of 11 838. The analyses were performed using SPSS 16.20


Table 1 presents the characteristics of the participants by number of SP episodes. Owing to the large sample, most of the factors are univariately associated with SP, although some of the differences are not substantially significant. Poor health, heavy work, work-to-family conflicts, social support, decision latitude and being obese are prevalent characteristics among those turning up ill at work most often.

Table 2 exhibits an association between SA of at least 2 weeks as well as of at least 2 months’ duration and the number of SP episodes in the year prior to baseline. This relationship persists even after controlling for other known risk factors for SA and SP, as well as for baseline health status and prior long-term SA. When adjusting for these factors, the strength of the association diminishes but remains statistically significant.

Table 2 Cox proportional hazard ratios of the associations between sickness presence (SP) with incidence of long-term sickness absence (SA) of at least 2 weeks as well as of at least 2 months over 1.5 years of follow-up

Table 3 shows the association between self-reported SP and long-term SA for subgroups with different (degrees of) health problems. Taking into account the relatively small number of events analysed, it appears that the association between SP and long-term SA is almost consistent across groups with different symptoms, whereas the association disappears or becomes insignificant for those indicating specific chronic diseases. Overall, SP seems to be more strongly associated with future SA among those not reporting any health problems than among those reporting some degree of health problems.

Table 3 Cox proportional hazard ratios of the associations between sickness presence (SP) with incidence of long-term sickness absence (SA) of at least 2 weeks over 1.5 years of follow-up for different subgroups of employees with different types of chronic diseases and other symptoms


This study indicates a linear association between going to work ill and long-term SA during the follow-up period. Participants with six or more episodes of SP were 53% more likely to become sick-listed for at least 2 weeks in the follow-up period compared with those reporting no or one such episode. The risk of becoming sick-listed for more than 2 months was even higher, at 74%, adjusted for a wide range of potential confounders.

In the analyses of subgroups with specific health problems, the association differed slightly between the groups. It seems that SP is not associated with future SA for those indicating that they have a specific chronic disease. Some caution should be taken with respect to this result because of the inclusion criteria in the study: chronic disease patients with the most severe health problems and some degree of work participation will most probably have been excluded from the study. This could result in underestimating the true association between SP and SA for these groups. Because of these circumstances and because we do not know what kind of symptoms or diseases people brought to work in their self-reported episodes of SP, there are several ways to interpret these findings. It might be the case that those having a specific health condition are more cautious and take fewer risks in the event of bad health, making the events of SP among those with no or few health problems more severe. This would mean that people with well-regulated chronic conditions such as high blood pressure would not be doing any harm to their health by going to work ill. The differences in hazard ratios between the healthy and the unhealthy could, however, also be interpreted as an indication that SP is a proxy for some unmeasured disease/illness present at baseline that breaks out at a later stage (ie, in the follow-up period). However, having adjusted the results for general health as well as prior SA reduces the plausibility of this explanation. Finally, the association between SP and SA could have been produced by a behavioural mechanism: perhaps those turning up at work ill repeatedly are more reluctant to return to work after a spell of SA, thereby making their spells of SA longer on average. If this group on average had longer spells of SA, they would more frequently have a spell of at least 2 weeks compared with other groups with similar health status, resulting in an association between SP and SA such as the one observed in our study. This hypothesis could have been examined if we had access to spells of SA lasting less than 2 weeks but, unfortunately, we did not have that. The possibility that those taking SP repeatedly on average would be more reluctant to return to work after a spell of SA could be plausible if the work environment had influenced their sick-listing and there were no prospects of the work environment improving. Regardless of these possible explanations, it seems safe to conclude that SP is associated with long-term SA only under certain circumstances.

There are several strengths to this study. First, the study is longitudinal and makes use of official registers to obtain information on long-term SA. Second, the study uses a large representative sample of the Danish core workforce, which strengthens the external validity. Third, the inclusion of a broad range of potential confounders makes the findings robust in the face of a very complex and multifactorial outcome such as SA. However, there are also a number of weaknesses that should be taken into account when assessing the validity of the study. First, some of the measures used were not validated prior to the study or consisted of single items. Second, the predictors were only measured at baseline, making it impossible to assess the nature of the relationship between these. This is particularly problematic with respect to the association between SP and the health measures. One potent criticism would be the claim that the measure of SP used in this study is just another way of measuring chronic disease and that there is nothing strange about a proxy for disease predicting future long-term SA. Finally, the fact that those who had been sick-listed for a long period in the year prior to baseline were excluded makes the participants in this study on average healthier than the workforce at large. Depending on perspective, this can be seen as either a strength or a weakness: a strength because the results of the study indicate that SP is followed by higher rates of SA among the healthiest part of the workforce; and a weakness because the study will probably underestimate the association between SP and long-term SA.

What is already known on this subject

  • Earlier results indicate that sickness presence may be detrimental to your health in the long run. In the Whitehall II study, an elevated risk of serious coronary heart disease was found for those men with bad self-rated health who did not take any sick days at all compared with those unhealthy men who had a moderate number of sick days (ie, 1–14 days).

  • Apart from this study, no evidence about the possible effects of sickness presence on health is known, and nothing is known about its possible effects on diseases other than coronary heart disease.

What this study adds

  • This study establishes an association between the self-reported number of sickness presence episodes and future long-term sickness absence of both at least 2 weeks' duration as well as at least 2 months’ duration.

  • This association persists even when controlling for other known risk factors for long-term sickness absence as well as baseline health and prior long-term sickness absence.

  • Sickness presence is associated with future poor general health (when using long-term sickness absence as a global measure of health).

If we compare this study with the only other study of the possible health effects of SP, there are again both strengths and weaknesses. This study benefits from having a distinct measure of SP instead of employing an indirect one such as that used in the Whitehall II study. In addition, the fact that this study consists of a random sample of the core workforce instead of a sample of a distinct subgroup of employees serves to strengthen the validity of its findings. Finally, one can question the utility of using long-term SA as the outcome measure. On the one hand, this is an improvement over the Whitehall II study because we use a more global measure of health instead of focusing exclusively on cardiovascular health. We can imagine several other types of disease where going to work despite experiencing symptoms may be detrimental (eg, slipped disc, asthma or some ongoing infections).

On the other hand, the use of a behaviour-based outcome measure such as long-term SA makes the evidence of a “biological” association between SP and health less convincing than a more “hardcore” biological outcome measure such as the one used in Whitehall II (ie, serious coronary disease). In other words, we do not know whether the association between self-reported SP and SA is the result of an unobserved behavioural mechanism that is associated with both outcome and determinants. Finally, differences in welfare state regimes will probably produce differences in the prevalence of SP across countries, possibly lowering it in countries where employees have the right to full wages in cases of short-term SA.6 21 This could have affected the threshold for reporting SP in our study, leading to a situation where our measure of SP registers more minor ailments than the measure used in the Whitehall II study.

Despite these reservations, we believe SP should be given more attention in public debate as well as in SA research where SA and efforts to reduce this receive the highest attention.2 Policies aiming to reduce SA without taking the possible effects of SP into account could encourage people to go to work ill. If employees are in fact substituting presenteeism for absenteeism, the effect of changes in SA policies might reduce SA without improving the employee’s health.5 For these reasons, SA researchers should try to incorporate a measure of SP into their studies. If short-term SA indeed serves as a coping mechanism and a way for employees to rest and recuperate from high job demands, this perspective should be discussed more openly than it is today and possibly implemented into future SA policies. This could be done by urging employers to monitor their employee’s health instead of only their levels of SA.2 9

More studies on the association between SA and SP are needed, and our knowledge base could be expanded with the use of more detailed reports of the specific symptoms and disorders that lead to both SA and SP. This is also warranted because there is evidence that long-term SA leads to marginalisation from the labour market.22 In addition, some studies of common physical and mental health problems point to an overall beneficial effect of work on health.23 In sum, these contradictory findings can only be reconciled using better measures of SP.


The authors would like to thank the Danish Health and Safety Research Fund for financing the study. The authors also wish to thank the two anonymous reviewers for their many helpful suggestions.



  • Funding: The Danish Health and Safety Research Fund.

  • Competing interests: None.

  • Ethics approval: In Denmark, when doing surveys (without medical examination, etc.), this is not required by law. However, we did inform the participants and obtained their consent to retrieve information from the official registers containing information about sickness absence. This retrieval was also approved by the Danish Data Protection Agency.