Poor work ability increases sickness absence over 10 years

Abstract

Background

Little is known about the predictive value of single items from the work ability index (WAI) on the risk of sickness absence over several years, and whether such risk varies across age groups.

Aims

The aim of the study was to investigate whether poor self-perceived physical and mental work ability among employees in the public sector are associated with long-term sickness absence over a 10-year period.

Methods

The study was based on a prospective cohort of employees within the public sector in Sweden reporting ‘good health for working’. Baseline information was collected with questionnaires from 2000 to 2003. Poor physical and mental work ability in relation to work demands were assessed with two items from the WAI. The outcome was the number of years of long-term sickness absence between 2003 and 2012. Long-term sickness absence was defined as ≥28 days of sickness and this information was retrieved from Swedish National Registers. Crude and adjusted incidence rate ratios were calculated with analyses stratified by age.

Results

Compared with those reporting very good physical work ability, employees reporting any lower grade of physical work ability had a higher risk of long-term sickness absence across all age strata, with higher risk estimates in the highest age groups and 6-fold increased risk in the oldest age group. Similar results were found for mental work ability with an almost 4-fold increased risk in the highest age group.

Conclusions

Self-reported physical and mental poor work ability are associated with long-term sickness absence during the subsequent 10 years. The risk increases with age.

Introduction

Absenteeism due to ill-health is a costly individual and societal burden. It is more common in women, independent of occupation and the length of the absenteeism depends on several different factors [1]. Musculoskeletal disorders and common mental disorders account for the majority of sickness absences in Sweden and elsewhere [2–4].

Factors of importance for sickness absence have been extensively investigated. Reviews of systematic reviews conclude that both workplace factors, including physical and psychological work demands, as well as non-workplace factors, are associated with sickness absence and also with return to work [5,6]. Systematic reviews investigating the impact of work factors on sickness absence include studies with both generic and diagnosis-specific health measures, and with few exceptions the studies included are restricted to a maximum follow-up period of 5 years [5–8]. One exception is the Finnish FLAME study that concluded that poor self-perceived work ability is associated with permanent work disability at follow-up 28 years later [9].

Self-perceived work ability, commonly measured using the Work Ability Index (WAI) or subscales of the instrument [10], has been found to be associated with several health outcomes [11–13]. WAI consists of seven items including two sub-questions about perceived work ability in relation to physical and mental work demands. Ahlstrom et al. [14] used both the full WAI and a single item from the WAI (WAI-S) and found that both were associated with sickness absence over a 12-month period.

In a previous exploratory study of women in a cohort of public sector employees, we found that two single items from the WAI—poor self-perceived physical and mental work ability—were independently associated with at least one period of long-term (≥28 days) sickness absence during a follow-up period of 3 years [15]. In the present study, we conducted a 10-year follow-up on an extended cohort, and this was not restricted to women only.

The aim was to investigate whether poor self-perceived physical and mental work ability are associated with long-term sickness absence over a 10-year period in public sector employees.

Methods

Individuals working within the public sector in six municipalities and four county councils in Sweden were included in an original project; ‘Work and sustainable health in the public sector in Sweden’, from here on shortened to HAKuL [16]. The cohort consists of a variety of occupations, including cleaners, schoolteachers, social workers, assistant nurses, nurses, physiotherapists, psychologists and medical doctors.

Baseline questionnaires were sent out in 1999–2000 and follow-up questionnaires were distributed 3 years after baseline. Information on socio-demographic factors, health, sickness absence, sleep and recuperation, lifestyle and personal factors, and work-related factors were collected. In the present study, information from the 2003 follow-up questionnaires was used as baseline. If no information was available from 2003, the information from the 1999–2000 questionnaires was used. The outcome was the number of years including long-term sickness absence, and this was retrieved from national registers from 2003 to 2012. The Regional Ethic Review Board in Stockholm approved the study (Dnr:2014/111–31).

All individuals in the HAKuL database were considered eligible (n = 9588) for the study. To prevent a risk of reversed causality, we selected those who reported having rather good or good ‘health for working’ at baseline for the present study (n = 8005). An index for ‘health for working’ was created by combining the item ‘general health’ from the SF-36 health survey [17] with one question from the WAI about the individual’s self-rated likelihood of their ability to continue working in their current position for 2 or more years considering their health condition [10]. We excluded those who scored less than 65 on SF-36 general health item and who reported it ‘unlikely’ or ‘not certain’ that they could continue working during the two following years. Individuals with no follow-up information (retired, deceased or emigrated before follow-up) were excluded (n = 137). The study population consisted of 7868 individuals (79% had baseline information from year 2003).

The exposures were self-perceived physical work ability and mental work ability in relation to work demands and were measured with questions from the WAI. The psychometric properties of this instrument have been tested [18,19] and are considered stable, predictive and internally coherent. Physical work ability was measured with the question: ‘How do you rate your current work ability with respect to the physical demands of your work?’ response options were: ‘Very good’, ‘Rather good’, ‘Moderate’, ‘Rather poor’ and ‘Poor’. The latter two were combined for the analyses. Mental work ability was measured with the question: ‘How do you rate your current work ability with respect to the psychological and mental demands of your work?’ The response options for this were the same as for the physical work ability. The two questions used as exposures have an acceptable predictive ability in relation to sickness absence in Sweden [20].

The outcome was the number of years with long-term sickness absence. Long-term sickness absence was defined as ≥28 days of sickness absence compensated by the Swedish Social Insurance Agency. In the Swedish social security system, every period of inability to work compensated by the Swedish Social Insurance Agency is preceded by 2 weeks of sick pay remunerated by the employer. Thus, every year that a certain individual has received compensation from the Social Insurance Agency, at least 14 gross days, was counted as 1 year with long-term sick leave. Data were retrieved from ‘the longitudinal integration database for health insurance and labour market studies’ (or LISA) which holds the information from the Social Insurance Agency and includes all residents in Sweden [21].

Sex and age at baseline were a priori included as confounders for both models. Age was controlled by categorization according to the best fit to the model into age groups with a 5-year interval from 19–24 up to 60–65.

Other potential confounders for physical work ability and long-term sickness absence were marital status (married, single, divorced, widow/widower), education (basic education <9 years, basic education 9 years, upper secondary 2 years or less, upper secondary 3 years, university <3 years, university 3 years or more, doctoral education) and individual income (total income, taking into account wage earnings, complementary social allowances and tax deductions). Lifestyle factors tested were smoking (daily smoker versus not), physical activity during leisure time (mainly sedentary and active) and body mass index (BMI) tested both in continuous form and categorized into underweight, normal range, overweight and obese [22].

For mental work ability and long-term sickness absence, potential confounders included marital status, education, individual income, smoking and BMI.

Physical exertion at work was also tested as a potential confounder and was measured by Borg’s Rating of Perceived Exertion scale. The scale has previously been used and considered useful for measuring muscular and cardiovascular load during work [23]. The scale ranges from 6 to 20 and the answers were categorized into low (6–10), moderate (11–13) and high (14–20) physical exertion at work.

To assess whether the exposures affect age groups to a varying extent (effect measure modification), we categorized age in to three levels, 20–34, 35–49 and 50–65 years.

The associations between the exposures and outcome were investigated with negative binomial regression and the results presented with incidence rate ratios (IRR) and 95% confidence intervals (95% CI). One statistical model was built for each exposure. The potential confounding factors were tested by including the factors one at a time in to the basic model including only exposure, baseline age and sex. If the factor affected the basic model estimates by 10% or more, it was included in the final model. Potential effect measure modification was evaluated by stratified analyses.

Individuals considered at risk of the outcome every year were included in the LISA registers at the end of that specific year. Individuals were censored if deceased, emigrated or retired.

To test the stability of the exposures over time, sensitivity analyses were conducted by dividing the follow-up time into 2003–07 and 2008–12. A potential problem in dividing the analyses into separate time periods is the censoring, especially among participants belonging to the highest age group. To enable a proper comparison, we restricted the sensitivity analysis to those present during all 10 years. Another sensitivity analysis was performed with data based on baseline information from 2003 only. This was done to test for potential selection bias and exposure misclassification due to inclusion of 1999–2000 data in the cohort.

Results

The total number of years of long-term sickness absence for the whole cohort over the 10 years was 11 283. It was most common to have 1 year with long-term sickness absence (1391 individuals) followed by 2 years (718 individuals). A description of the study population at baseline is shown in Table 1. The majority were women (82%), and the mean age 46 years. During the follow-up, 49% (3168/6478) of the women and 30% (417/1390) of the men had at least 1 year of long-term sick absence.

None of the factors tested confounded the associations between the exposures and the outcome. Baseline age modified the effect in the associations between physical and mental work ability and long-term sickness absence (Table 2). Compared with those with very good physical work ability, individuals with any lower grade of physical work ability had a higher risk of long-term sickness absence in all age strata. Among 20–34 year olds, the IRR in the exposure category ‘rather poor/poor’ was 2.15 (95% CI 1.14–4.06), while 35–49 year olds in the same category had an IRR of 4.94 (95% CI 3.02–8.08), and individuals in the 50–65 age range had an IRR of 6.68 (95% CI 4.05–11.04). Reporting a moderate physical work capacity yielded an IRR of 1.69 (95% CI 1.24–2.29), 3.50 (95% CI 2.81–4.37) and 3.70 (95% CI 3.00–4.55) in age groups 20–34, 35–49 and 50–65, respectively.

The analyses of mental work ability showed that individuals in all age groups had a higher risk of long-term sickness absence if they reported any exposure level higher than the reference category ‘very good’. The strongest associations were found in those reporting ‘rather poor/poor’ mental work capacity with IRR of 2.00 (95% CI 1.26–3.16), 2.32 (95% CI 1.50–3.60) and 3.70 (95% CI 2.23–6.16) for age groups 20–34, 35–49 and 50–65, respectively.

Sensitivity analyses indicated a washout effect, with somewhat lower estimates in the second period for the physical work ability, but the effect was still significant in all age groups at all exposure levels (data not shown). The second sensitivity analysis including individuals with baseline information from 2003 only indicated a somewhat stronger effect in the two highest categories but with wider confidence intervals (Table 3).

Discussion

We found a greater risk of long-term sickness absence during the subsequent 10 years in employees who rated their physical or mental work ability in relation to work demands as low at baseline. The risk tended to be higher in the older age groups.

This study has several strengths; it is based on a large study population of public sector employees from several geographical areas in Sweden. It has representation from a variety of occupations and levels of education. Furthermore, we used register data to create the outcome variable, which ensures follow-up data of all study participants until censored. During the study period, there were few changes in the regulations for sickness benefits and processing routines. The most notable changes appeared after 2008, when diagnosis-based guidelines on when and how long patients should be on sick leave, was introduced, and in 2009, when the number of days on sick benefits was maximized to 365, with exceptions for very severe diseases. Lastly, a new regulation was introduced demanding a structural investigation is performed by the Social Insurance Office, when a person exceeded 180 days of sickness absence. Return to work was the focus, and measures were taken to gradually re-introduce the employee to the labour market (same or new job) [24,25]. If these changes, except potentially reducing long-term sickness absences, were associated with the exposures in our study, the results may have been over- or underestimated. Since we performed separate analyses of the first and the second 5-year follow-up period, and did not find any major differences between the two periods, we felt it reasonable to conclude that these changes did not have a major impact on the results.

Common mental disorders and musculoskeletal disorders account for the majority of all sickness episodes in Sweden [26]. One can assume that lower mental work capacity would more often result in a diagnosis relating to common mental disorders, and lower physical work capacity more often lead to musculoskeletal disorders. It would, therefore, have been preferable to include information on diagnoses related to the sickness absences in this study; however, such information was not available, and we may have underestimated the associations. Still we found a rather strong increased risk, especially in older ages.

Another limitation is the unconventional definition of long-term sickness absence. This is due to the restricted register information containing only the total number of days of sickness absence per year, but not specifying whether they were consecutive or not. Thus, according to our definition, there may be several periods of sickness absence in 1 year of long-term sickness absence. Furthermore, according to our definition sickness absence including a year-end may be counted as 2 years, which could risk distorting the findings in the study. However, the majority of individuals with years of absence had only one or two registered, which makes this risk low. Thirdly, 88% of the study population were women, reflecting the gender distribution in the public labour sector in Sweden. We adjusted for sex and age in all multivariable models, but it is possible that sex instead is an effect modifier.

The questions regarding work ability are derived from the frequently used WAI questionnaire which has been tested with acceptable psychometric results [18]. A Dutch study of construction workers, followed for 23 months, found that the complete WAI, as well as all separate scales in the WAI, were predictive of receiving disability pension. The question about work ability in relation to work demands had the highest predictive validity [27]. The magnitude of the effect in the Dutch study is difficult to compare with our results, since a previous version of WAI was used and the analysis approach was different. In contrast, a recent validation study found that individual items of WAI were less predictable of sickness absence compared with the full WAI [20]. However, we found strong associations and if any misclassification was present, it would have diluted our results.

Work ability is related to work tasks and personal resources and the balance between the two. These may change with time (age, morbidity, change of work demands etc.), making self-perceived work ability time dependent [28]. The result of the sensitivity analysis in our study did not reveal any strong support for this hypothesis. Notably, the vast majority of our study population included employees from smaller cities or towns in Sweden where labour market mobility is limited and employees in the public sector seem to stay within their area of work.

Both physical and mental work ability were modified by age in their association with long-term sickness absence. It seems reasonable to assume that older age entails a higher sensitivity, especially to physical overexertion. Individuals in professions with a high physical work load often have a low level of education, too, and therefore less opportunity to change to less physically demanding work. This group of individuals is very important to target for improving work ability as they are also a group with a high sickness rate and morbidity overall [5]. To counteract the risk of reversed causality, we excluded individuals who rated their health as poor at the study start and doubted whether they could continue working over the two following years, and still we found strong associations with future sickness absence.

To the best of our knowledge, this is the first study on this topic with a 10-year follow-up. The fact that two WAI questions seem to predict sickness absences over such a long period indicates that there is a lot to gain, both in terms of reducing human suffering and national costs, by working towards preventing individuals from having inadequate physical and mental resources in relation to their job demands. This should be done at the organizational as well as individual level. Future studies on this topic should be large enough to allow testing for the effect modification of both gender and age.

In conclusion, self-reported physical and mental poor work ability in relation to work demands are independently associated with long-term sickness absence during the subsequent 10 years. The risk tends to increase in older age.

Funding

This work was supported by AFA Insurance, Sweden [grant no. 130316].

Competing interests

None declared.

References