Health inequalities by wage income in Sweden: The role of work environment

Abstract

The main aim of this study was to explore the mediating role made by work environment to health inequalities by wage income in Sweden. Gender differences were also analysed. Data from the Swedish Survey of Living Conditions for the years 1998 and 1999 were analysed. Employed 20–64-year olds with a registered wage were included (nearly 6000 respondents). Sex-specific logistic regressions in relation to global self-rated health were applied.

Those in the lowest income quintile had 2.4 times (men) and 4.3 times (women) higher probability of less than good health than did those in the highest quintile (adjusted for age, family status, country of birth, education level, smoking and full-time work). The mediating contribution of work environment factors to the health gradient by income was 25 per cent (men) and 29 per cent (women), respectively. This contribution was observed mainly from ergonomic and physical exposure, decision authority and skill discretion. Psychological demands did not contribute to such inequalities because mentally demanding work tasks are more common in high income as compared with low income jobs. Using sex-specific income quintiles, instead of income quintiles for the entire sample, gave very similar results.

In conclusion, work environment factors can be seen as important mediators for the association between wage income and ill health in Sweden. A larger residual effect of income on health for women as compared with men suggests that one's own income from work is a more important determinant of women's than men's ill health in Sweden.

Introduction

Socioeconomic health inequalities have been reported in Sweden to a similar extent as in comparable neighbouring countries (Mackenbach et al., 1997). A large number of mechanisms for understanding these health inequalities have been found. These include childhood conditions, selection factors, current living conditions (work factors, material deprivation), behaviours (e.g., smoking) and psychosocial factors such as perceived control (Bosma, Schrijvers, & Mackenbach, 1999; Davey Smith, Hart, Blane, & Hole, 1998; Lundberg (1991), Lundberg (1993); Mäkelä, Valkonen, & Martelin, 1997; Townsend & Davidson, 1992). The explanatory mechanisms are highly complex and far from fully understood, as noted by Lahelma (2001) and in a review by Davey Smith, Blane, and Bartley (1994).

A majority of the explanatory work mentioned is based on research from a class context. Recently, however, Geyer and Peter (2000) observe an increasing interest in income and material deprivation as central in understanding population health inequalities. Income is hypothesized to be related to a large number of outcomes in life such as behaviour, consumption and basic needs. Moreover, income is likely to be redistributed by egalitarian social policies (Duncan, 1996). Studies from a large number of Western countries generally report large health inequalities by income, whether using relative or absolute income variables, or household or individual income (Backlund, Sorlie, & Johnson, 1996; Benzeval & Judge, 2001; Der, Macintyre, Ford, Hunt, & West, 1999; Ecob & Davey Smith, 1999; Fritzell & Lundberg, 1994; Fritzell, Nermo, & Lundberg, 2004; Geyer & Peter, 2000; Lynch, Kaplan, Cohen, Tuomilehto, & Salonen, 1996; MacDonough, Duncan, Williams, & House, 1997). In sum, most of these studies report on the form of the association between income and poor health or mortality risks. Findings across income measures tend to show that those classified as having high income have better health than do those with a low income, often with a gradient for intermediate income groups.

Only few studies try to disentangle the contributing factors to health inequalities by income, although those of the theoretically preceding indicators of social stratification—education and class—or a systematic mobility of sick people to the lower end of the income distribution are central (Benzeval, Judge, & Shouls, 1996; Rahkonen, Arber, Lahelma, Martikainen, & Silventoinen, 2000; Stronks, van den Mheen, van den Bos, & Mackenbach, 1997; Rahkonen, Lahelma, Martikainen, & Silventoinen, 2002). Although income is generally seen as an outcome of education level and occupational class (e.g., Lahelma, 2001), it is important to observe that they are not too highly correlated and measure social stratification at various levels of abstraction (individual, occupation or household level). An association between income and health tend to persist also after education and class has been adjusted for (Fritzell et al., 2004; Geyer & Peter, 2000). This is possibly due to the dynamic nature of income since it (wage income in particular) can vary to a high extent within occupational titles (such as across regions and work places), and often men receive a higher wage than women in the same job, or that women's work often is devalued as compared with men's work (Thoursie, 1998).

Thus, it seems relevant to analyse in some detail potentially mediating factors to health inequalities by wage income. I see the job situation as most central here, in particular the potential association between wages, work conditions and health. A major part of the income dynamics over the life course is also determined by time spent in paid work. There are some, but few, studies suggesting that work factors and income tend to combine into health inequalities (Lynch, Krause, Kaplan, Tuomilehto, & Salonen, 1997). My general aim here is to estimate the mediating role made by work factors to health inequalities by wage income in Sweden. I will also analyse eventual gender differences in this.

There is an extensive and multifaceted literature on work and health. Occupational epidemiology has focused either on environmental risks (chemical risk factors) or individual risk factors (Kristensen, 1995). In the more recent period, psychosocial job factors, in particular psychological demands, job control, and aspects of efforts and rewards have received great attention (Karasek, 1979; Karasek & Theorell, 1990; Siegrist, 1996). When combining various psychosocial job dimensions, situations of job stress has been hypothesized as particularly unhealthy, such as high psychological demands and low job control in the demand-control (D/C) model (Karasek, 1979; Karasek & Theorell, 1990), and high effort and low reward situations in the effort reward imbalance (ERI) model (Siegrist, 1996). In the ERI-model temporary employment is also considered as an unrewarding aspect of work.

Does work stress contribute to socioeconomic health inequalities? It is suggested that work stress is more common in lower compared to higher occupational classes, and hence could contribute to such inequalities (Siegrist, 2002). But Kristensen, Borg, and Hannerz (2002) observe that unskilled manual work is not typically characterized by job strain but rather by low demands and low control (passive jobs in the D/C-model). As can be expected from this, it is concluded in a review that there has been clearly more support for the control aspect as compared with the demand aspect of the D/C model (Schnall, Landsbergis, & Baker, 1994). Not surprisingly, job control has been found to contribute considerably to occupational class inequalities in a non-manual sample (Bosma et al., 1997; Marmot, Bosma, Hemmingway, Brunner, & Stansfeld, 1997). But this also holds for ergonomic and physical job demands (Borg & Kristensen, 2000; Lundberg, 1991).

Previous studies have found contributions of working conditions to socioeconomic health inequalities (measured by occupational class), clearly for ergonomic and physical demands and for job control. It seems more uncertain whether such contributions are similar for income measures as for occupational class, or whether different aspects of job control (skill discretion and decision authority) in the D/C-model are more important than the other here. Some but few studies separate the two in the analysis (e.g., de Jonge, Reuvers, Houtman, & Kompier, 2000; Stansfeld, Fuhrer, Head, Ferrie, & Shipley, 1997). Job stress does not seem to be correlated with occupational class position in a straightforward way (Kristensen et al., 2002), and did not seem to contribute to class inequalities in an earlier Swedish study (Lundberg, 1991).