“Inclusive working life in Norway”: a registry-based five-year follow-up study

The dichotomous study outcome was having at least one spell of a long-term (>8 weeks), continuous sickness absence during the five-year period from 2001–2005. Eight weeks was chosen as the responsible doctor until 2012 was obliged to produce an “eight weeks sickness absence sertificate” including medical information and plans for treatment and rehabilitation. Therefore, sickness absence longer than eight weeks has been considered the starting point for long-term absence in Norway.

The cohort was divided into IW and non-IW groups. Serial numbers from the Norwegian Labour and Welfare Administration’s IW registry were linked to individual data from HUBRO and FD-Trygd. An “IW employee” was defined as a person employed in an IW enterprise after the agreement was enacted in 2001. Following this definition, 4,791 (2,733 women and 2,058 men) were IW employees (43.6%). The remaining 2,973 women and 3,231 men were classified as non-IW employees. Data on IW status was available from March 2003.

The SEP indicator was based on the education level classifications from Statistics Norway and categorised into five levels [15], as described in the footnotes of Tables 1, 2 and 3 and in more detail in our previous studies [5], [6].

We made the following industry group categorisation based on the Standard Industrial Classification (SIC2002) by Statistics Norway [16]: (1) secondary industry (industry, building / construction, etc., SIC2002 codes 10 – 45), (2) tertiary industry, heavy (retail, hotel / restaurant, transport / storage / communication, SIC2002 codes 50 – 64), (3) office work (service trades / insurance, civil service, SIC2002 codes 65 – 75), (4) the teaching sector (SIC2002 code 80), (5) the health sector (health and social care, social services, SIC2002 codes 85 – 93) and (6) other/unknown (SIC2002 codes 00 – 05, 95 – 99).

The following work-related variables were included in the present analyses: The work schedule variable was based on the question, “Do you do shift work, night work or have rotating hours of work?” The response categories were Yes or No. Physical job demands were measured by the question, “If you have paid work or do unpaid work, how would you describe your work? Responses were as follows: (one cross only) Mainly sedentary work?; Work involving a lot of walking?; Work involving a lot of walking and lifting?; or Heavy physical work” The two last categories were merged into one category since only 1% of the participants had “Heavy physical work”.

Stata/SE 11.2 was used for the analyses. Five-year risks of at least one spell of LSA in subgroups were computed with cross tabulations. Cox proportional hazards models were used to compute the LSA hazard ratios (HR) and corresponding 95% confidence intervals (CI). We adjusted for age group, educational level, industrial classification and the job exposure variables. All analyses were stratified by gender, and the main analyses were additionally stratified by IW status. We additionally performed a gender-stratified Cox regression to estimate the relative IW effect. Gender-stratified models were estimated in certain subgroups in which the results indicated heterogeneity between the IW and non-IW groups. The follow-up lasted until the end of 2005. Censoring occurred when individuals died (N=13), emigrated (N=130) or received a disability pension (N=30) or early retirement (N=603) during the follow-up period, i.e. they were no longer at risk for LSA. Of the total sample, 776 persons were censored.

We have found that the most important risk factors for an LSA were shift work/night work or rotating hours (strongest in men in the non-IW group) and heavy physical work or work involving a lot of walking and lifting (men only and strongest in the non-IW group). When estimating the IW effect in each gender separately, we found no association after adjustment. However, we found an IW effect in shift working men. Our results could suggest that IW companies that employ many men in shift work have implemented relevant efforts for reducing sickness absence.

Women exhibited a higher LSA risk than men, but the associations between a high physical workload and shift/night work and LSA risk were stronger for men than for women. Women working in the health sector had the highest LSA risk. The health sector is dominated by demanding work, such as hard physical work and shift and night work. Most enterprises in public industries, including the health sector, are IW enterprises. Some IW companies may have levels of high sickness absence and absenteeism because the workload is more demanding, which might have led to more interest in the IW agreement. We likely have elements of reverse causation in that high sickness absence levels in a company might have acted as an incentive to participate in the IW program. This is a plausible interpretation in cases in which the program was adopted late in the 2001–2005 period. The IW participation varies by industry. As an example, 62% of women in the health sector were in IW companies, compared with only 32% in the heavy tertiary industry (Table 1). Thus, industry may act as a confounder on the association between IW participation and LSA. IW and non-IW persons differ systematically by demographic characteristics and work environment. This is a challenge in interpreting the results. In addition, there are reasons to believe that not all enterprises with an IW agreement have achieved the IW intentions, and many non-IW businesses may have implemented sickness absence reduction efforts widely. Several measures that have arisen in the wake of the IW agreement are directed toward the entire working life, such as new sick leave rules, training of physicians and more dialogue between employers and sick-listed employees. Because no impact on the national sickness absence rates was observed by 2004, the government and social partners realised that there was a need to involve a patient’s physician because general practitioners issued approximately 80% of sickness absence certificates [1]. Since 2004, the physicians’ sickness absence certificates have been replaced by a “work ability certificate” to promote an early return to work. Sickness absence from work has gained more attention in the society as a whole after the IW agreement was implemented, which has affected the follow-up work for absentees in the entire labour market. These issues might additionally explain the lack of contrasts between the IW and non-IW groups. Many EU governments have introduced programmes aimed at encouraging long-term absentees back into work. There is still relatively limited documentation concerning the effect of workplace interventions on sickness absence [17]. The IW programme could possibly have more impact on shorter, self-certified sickness absence. However, we had no data available on this and could not explore this further.

The labour market in Norway is highly gender segregated. Studies have shown that occupational groups dominated by one gender have higher levels of absence [18], [19]. Others have found that the fact that women hold other occupations than men explain a considerable part of the differences in sickness absence [20]. Work tasks and working conditions may differ substantially between the genders or between female- and male-dominated occupations [21]. A low educational level was highly predictive of LSA, especially for men. Interventions for preventing LSA should be especially targeted to these groups.

The strengths of this study are a relatively large study sample, combining data obtained from a survey with register-based data and a longitudinal study design. Survey-based studies of work characteristics and health have been criticised on methodological grounds for an undue reliance on self-reported outcome measures and not addressing information bias [22]. The basic source of a dependent error [23] in such studies is usually a normal variation in certain personality traits, but the error may additionally be in more transitional moods in the study population or inadequate measurement tools. Registry-based rates of physician-certified sickness absence may be a more accurate measure of health status than self-reports, which are often biased [24]. The major precaution that should be taken to eliminate bias from dependent error is to break the bond between information on exposure and outcome by gathering data from two separate sources [25], which is what was done in this study. Additionally, the data on educational level were objective information from public registries. The low participation rate (46%) in HUBRO may have led to the self-selection of healthy subjects into the study. The predictors of participation and magnitude and direction of the non-response bias in prevalence estimates and association measures have been investigated based on information from all 40,888 invitees to the Oslo Health Study. The potential selection bias was studied by linking register-based data from Statistics Norway on demographics, lifestyle and social security benefits to the entire study population. Unhealthy persons participated to a lesser degree than healthy individuals, but the social inequality in health by different socio-demographic variables appeared unbiased [26]. In addition, the response rate to the supplementary questionnaires on working conditions was lower in subgroups with worse health [27]. Because poor health often is associated with a poor work environment, this finding may imply a lower response among subjects with poor work environments and thus conservative estimates of the effects of working conditions on LSA.