Surprisingly, in the present study the explanatory effect of somatic and mental symptoms, mental well-being, and job strain, on this social gradient of sickness absence, was small and in some cases even non-existent. This was true for both women and men. However, physical work conditions (i.e. heavy lifting and awkward working postures) explained the entire association between socioeconomic position and sickness absence in women. Not only did the associations become statistically insignificant but also the point estimates were all close to one. Although physical work conditions had a strong explanatory effect even in men the association between the two manual levels and the Higher non-manual level in men remained unexplained. While the previous study on the present material emphasized the explanatory effect of self-reported physical work ability for the social gradient in sickness absence  it left us with the query of whether it was the individual resources (e.g. health) or the work environment that constituted the critical element in this effect. In the additional analyses of the present study, in men, we observed that although self-reported physical work ability had an explanatory effect alone, the unique effect of physical work conditions was clearly more important. Since physical work ability  and physical work conditions respectively explained the total associations in women, in the present sample, similar analyses was not conducted in this group. Hence, together with previous studies [1–3] the present study strengthens scientific evidence suggesting that individual resources and psychosocial work conditions may not have a large impact on the social gradient in focus, but are still important factors in sickness absence per se. That physical work conditions (i.e. ergonomic) had the strongest explanatory effect on the social gradient in sickness absence is in line with three recent studies [1–3] of which one was performed in a general working population in Denmark . Hereby, the present results support the idea that physical work conditions are the main explanatory factor for the social gradient in sickness absence. It is of particular importance that this result is observed in a random working population. The reason for this is that such a sample provides the full range of socioeconomic positions, which are not found in more specific workplace samples. One of the strengths of this study is also that the measure of sickness absence was not based on self-reports but on new cases of sick-listed employees (>14 days), reported by the employer.
The present methodology may bring about a concern of circulatory reasoning. That is, that socioeconomy and physical work conditions end up representing the same phenomena. However, this should have been a graver problem if the aim had been to explain the distribution in one cohort with a variable strongly correlated with how this distribution was accomplished in the first place. As we, in the present study, compared the distribution of socioeconomy in two different cohorts, finding that this difference was explained by physical work conditions, the issue of circulatory reasoning is not applicable in the same matter. Furthermore, although the proportional distribution (Table 3 above) showed an association between socioeconomy and physical work conditions neither the correlation matrix nor the Variance Inflation Factors indicated any worrying covariance (i.e. multicollinearity) between these two variables. Although one could expect a high correlation between socioeconomic position and physical work conditions many occupations labeled as manual are not very physically demanding as captured in the two items used in the present study, e.g. observational occupations in the industry. There are also non-manual occupations that actually comprise heavy lifting and particularly awkward working postures, e.g. nurses. Hence, the relation between socioeconomic position and physical work conditions is today more complex than previously.
Somatic and mental health symptoms had low explanatory effect in both women and men which might appear as somewhat odd since an individual’s health position is conceptually seen as one of the main constituents for being sick-listed. One could also assume that the previously well-supported social gradient in health  would play a major role even in the social gradient in sickness absence. Still, the present study, in line with some previous, observed only small or even inconclusive results for the explanatory effect of self-rated general health on the social gradient of sickness absence [2, 7, 10]. However, a health problem per se is not a sufficient reason for sick-leave in the Swedish and most other welfare state regulations. In Sweden, in order to fulfill the criteria for sick leave (i.e. >14 days) one’s work ability has to be reduced due to a clinical diagnosis. This may explain why physical and mental health symptoms did not attenuate the social gradient in sickness absence in this study. Yet, in order to state that socioeconomic health differences do not contribute to the social gradient in sickness absence future studies should involve more objective measurements of health. One example could be to investigate whether socioeconomic differences in the prevalence of specific clinical diagnoses could work as an explanation in this issue, particularly since the association between socioeconomy and sickness absence might be diagnose-specific .
Unfortunately, we did not have access to any information about occupation. Still, we must recognize that the labor market in Sweden is highly segregated between women and men where about 90% of all health care personnel are women . Consequently, it would not be surprising if the explanatory factors for the social gradient in sickness absence would differ between women and men. For example, one reason for physical work conditions having such a strong explanatory effect on the social gradient specifically in women may be the so-called horizontal segregation. Many women in lower socioeconomic groups work within the health care sector where ergonomic exposures like lifting and awkward working postures are common . It is also plausible that the historical development of ergonomic assistance has been more successfully implemented in male-dominated occupational groups like industrial or construction workers compared to care organizations that generally include more women. A previous study observed that assistant nurses have a six fold higher risk of over-exertion back injuries compared to other employed women in Sweden . Consequently, it is possible that by solely focusing the measurement of physical exposure on heavy lifting and awkward working postures that may be essential predictors of the social gradient in sickness absence among women, exposures more prevalent in occupations dominated by men were left out. Eng and colleagues (2011) recently observed that male workers are two to four times more likely to report exposure to dust, chemicals, load noise, irregular hours, and vibrating tools than female workers. Even when comparing women and men with the same occupations clear gender differences in exposures are observed .
Furthermore one must recognize that in the present study, like previous studies (e.g. ), the differences between socioeconomic groups regarding sickness absence were less pronounced among women. One reason for the smaller gradient in women may be that work-related mental disorders due to stress, where the social gradient is reversed, are more common in women. This reversed social gradient is also steeper in women . Correspondingly, Lahelma and colleagues (2005) observed a social gradient in global and physical health but not in mental health . Another reason for the less pronounced social gradient in women could be that measures of socioeconomic position may have less precision in women since they fail to capture significant elements of gendered structures including the distribution not only over different types of occupations discussed above but also the distribution of management positions and responsibilities of the unpaid work [27–29].
One surprising finding was that although having socioeconomic position in the same models ‘physical and mental symptoms’, ‘low mental well-being’, ‘jobs strain’, and ‘physical work exposures’ were associated with sickness absence, in both women and men. This observation further implies the complexity between the social gradient of sickness absence and its potential explanations.
Acknowledging the recognized difficulties in persistently changing individual health-related behavior and psychosocial work conditions, the present result does in fact leave us with a relatively attainable goal of reducing the socioeconomic differences in physical work conditions. Yet, in order to provide better knowledge for interventional design future studies should be more specific and discriminate between the importance of different dimensions of the physical work environment in relation to socioeconomy and sickness absence.
It is important to recognize that the overall response rate in the present study was rather low (i.e. 52%). The analysis on non-respondents (figures not shown) showed that the proportions of men, younger individuals, individuals with the lowest income level, and individuals born outside the Nordic countries were lower than in the total population. Although the overrepresentation of these groups in the dropouts is a limitation that this study shares with most population-based studies it must be assumed that these groups may be highly important in studies investigating health in relation to socioeconomic circumstances. However, the specific patterns according to age, income and gender of the non-respondents were quite similar across the both samples that make comparisons possible. That we have a lower response rate in individuals with the lowest income level and individuals born outside of the Nordic countries could have resulted in somewhat lower Odds ratios. Yet, the relational patterns would probably be very similar. The drop-out from younger individuals may not have had a major impact on the results since long-term sickness absence is rare in this group. However, since the data collection was based on postal questionnaires no information on potential differences in health was available. Finally, although constituting a type of case–control design, the data used in the present study are collected at one point in time with the limitations that this design brings.
Not being a traditional longitudinal study it could be stated that sick-listed individuals with for example lower back pain may report differently on items capturing physical work environment. However, since back pain is rather common even in the general population potential reporting bias should be quite similar in the two samples although individuals in the sick-listed sample may have had more acute problems. The present analyses were also adjusted for both physical and mental symptoms.
The results of the present study emphasize that even in a so-called post-industrial and high-income country like Sweden physical work conditions like heavy lifting and awkward working postures still seem to play a very special role in the relation between socioeconomy and sickness absence. Beyond the use of more sophisticated data materials and analyses (e.g. ) future studies should dig deeper into what predictors may affect this association in men but also into the importance of socioeconomic differences regarding the prevalence of specific diagnoses. Finally, it should also be investigated whether the predictors of sickness absence may act in divergent patterns in the different socioeconomic groups.