Dimensions of social inequality in the health of women in England: occupational, material and behavioural pathways

Abstract

This paper examines the role of behavioural and psychosocial risk and protective factors in explaining social inequalities in the general self-assessed health of women. Using path analysis, data from the Health Survey for England (1993) are used to demonstrate how different dimensions of social position (working conditions, general social advantage and material deprivation) have distinct pathways to ill-health. Smoking, diet, alcohol consumption, exercise, social support and job strain were all related to poorer health, but not always in the predicted direction. The effects of social position on health were not fully mediated through these risk and protective factors. Each dimension of social position had unique pathways to ill-health via other unidentified mechanisms. Furthermore, the salience of the three dimensions of social position differed according to the level of labour market attachment. Different path models are required to fit the data for women at home or in full-time or part-time work.

Introduction

Many studies, using a variety of measures of social position and circumstances, have demonstrated social variation in the health of women (Arber and Lahelma, 1993, Gregorio et al., 1997, Chandola, 1998). It is now being argued that in order to advance our understanding of the social determinants of population health, more work needs to be carried out to investigate the underlying processes, to which correlations between class measures and health measures only give aetiological hints (Macintyre, 1997, Marmot et al., 1997). The need for greater understanding of these mechanisms is highlighted in research on health inequality in women, where attempts to transfer theoretical assumptions underlying research on men have produced conflicting and contradictory results (Arber, 1991). One reason for these contradictory findings has been that insufficient attention has been paid to the fact that health measures are inconsistently related to indicators of different dimensions of social inequality such as class, employment status and education (Arber, 1997).

In this paper we address these concerns by treating social inequality as multi-dimensional and building separate models of the pathways from each dimension of social inequality through risk and protective behaviours to health for women with different degrees of attachment to the labour market. Each measure of social position is intended to capture a different component of inequality so that the contribution from each can be identified (Krieger, Williams & Moss, 1997). The advantage of such a conceptual framework when investigating social variations in women health is that it acknowledges that different dimensions of inequality may be more or less important in those who have full or part time paid work, or who work only in the home.

Recently, researchers have moved towards the use of theoretically based and validated measures of social position. In the UK, the two most well known of these are the Erikson–Goldthorpe schema (Erikson & Goldthorpe, 1992) and the Cambridge scale (Prandy, 1990). The Erikson–Goldthorpe (E–G) schema is the basis for the new socio-economic classification to be implemented in the 2001 census (Rose & O’Reilly, 1997). A similarly based measure has been used in the work of the European Union’s working group on Socio-economic Inequality in Health (Kunst, 1997, Mackenbach and Kunst, 1997). The Erikson–Goldthorpe schema distinguishes between those who are employers or employees, perform manual or non-manual work, and have a ‘service’ versus ‘labour’ form of employment contract. A ‘service’ contract is characterised as one of trust, with higher levels of autonomy and job security, with the possibility of advancement through a clear career structure. A ‘labour’ contract is one where supervision is tighter, and motivation to work is gained through the exchange of wages for set amounts of work. Levels of job security are lower, and there is no career structure.

The Cambridge scale (CS) is described as an indicator of general social advantage and lifestyle. It is based on the analysis of friendship choices, judged to be the most accurate indication of perceived and experienced social distance between members of different occupations (Prandy, 1990). Originally, respondents to surveys were asked the occupations of five friends, and the number of times each pair of occupations cited each other was treated as a measure of social distance. These values were entered into a multi-dimensional scaling analysis, yielding one major factor: the score of each occupation on this factor is the Cambridge score. The advantages of the method, in the eyes of its originators, is that it does not force individuals into discrete groups (“classes”), it is based on concrete social experience, and it forms a clear hierarchy. Men’s and women’s occupations are rated separately, recognising that men and women in the same job may occupy different social positions.

The use of these theoretically based measures in the analysis of social variations in health in women overcomes some of the problems and contradictions of previous research. This is because it acknowledges that different dimensions of inequality may be more strongly related to health not only in men and women, but in women in different situations. In the past, researchers have had to use narrowly based occupational schemas such as the Registrar General’s social class or the Socio-economic Groups used in the British official statistics (Leete and Fox, 1977, Boston, 1984). The only alternatives were to use the woman’s own occupation, including any past occupation if she was not in paid work, or that of a husband or economically active partner, a problematic strategy (Arber, 1991, Arber, 1997). Theoretically based measures of social position force us to recognise that health inequality is a complex and multi-dimensional phenomenon: it may be more or less marked in different groups according to the measure of social position used. Most obviously, occupationally based measures of social class give lower estimates of health inequality in women who are not engaged in paid work than in men (Arber, 1996). However, this may not mean that health inequality is greater in men.

A further advantage in regarding social inequality as multi-dimensional and using theoretically grounded measures is that this approach helps to clarify what we are doing when we build statistical models to explain health inequality. Lynch, Kaplan & Salonen (1997) have pointed out that the reasons why behavioural, psychological and social risk and protective factors are unequally socially distributed are not well understood. Many studies examine social differences in some disease by arbitrarily choosing a single measure of social position, without explicit theoretical justification. Lists of risk factors are then entered into models as confounders. Whereas the relevant risk factors are chosen on the basis of evidence linking risk factors to disease, the reasons for links between risk factors and social position are left to common sense implicit assumptions (Bartley, Sacker, Firth & Fitzpatrick, 1999b). This results in further inconsistencies, even between studies of men only.

If we define different dimensions of inequality and combine them in the same models, these problems may begin to be overcome. We can then allow for the pathways from social position to health to vary according to what aspect of social inequality we are concentrating on, and compare their relative importance. We are also forced to address issues left implicit in most of the literature, such as ‘why it is that people with lower incomes are more likely to smoke?’. This cannot literally be a consequence of income although other inequalities such as diet and housing quality may be. Similarly, why should people with more routine work, less work autonomy and security (the conceptual basis for modern class measures such as those of Erikson and Goldthorpe) tend to have poorer diets or take less exercise? In many papers, there is a clear, if often implicit line of argument appealing to social status or prestige and its cultural and psycho-social effects (Stronks et al., 1997, Bosma et al., 1999). Low income and disadvantaged work conditions co-vary, this argument claims, with differences in culture and coping abilities, and it is these which determine the social distribution of behavioural mediators to social inequality in disease.

In order to move forward, there is a need for explicit and plausible hypothetical pathways to be spelt out and tested. Here we propose that all three dimensions: general social advantage and lifestyle, social class based on employment relations, and material living standards will be related to health independently, but through different pathways. The pathway from the Cambridge scale will more prominently involve health behaviours because this measure taps aspects of lifestyle which have different levels of acceptability in social groups of different status. The pathway from E–G class will involve work strain; and that from material living standards will have the strongest independent component, unmediated by any of the risk and protective factors.

The dimensions of social position have been conceptualised in the following way. Social class based on employment relations (E–G class) is based on the present occupation of each woman; for those outside the labour force the last held occupation has been used. The hypothesised pathways relating employment to health involve aspects such as hazardous conditions and work strain. The slight risks that the employment relations and conditions of a spouse may have for their partner’s health are not considered here. The dimension of general social advantage is regarded as affecting health because of the relationship of prestige to health behaviour. In this case, it is likely that behaviour will be influenced by that which is acceptable in the most advantaged group to which either partner has access. In modern industrial societies, groups with higher levels of social advantage and prestige have increasingly adopted ‘healthy’ behaviours as part of their accepted lifestyle. For example, it is likely that a secretary married to a senior manager will have fewer social opportunities to smoke and more opportunities to take leisure time exercise than one who is married to a construction worker. Therefore the Cambridge score is allocated on the basis of the higher scoring occupation of either spouse (whether or not they were in paid employment).

The outcome health variable used here, general self-assessed health, is a well-known predictor of mortality (Mossey & Shapiro, 1982; Kaplan & Camacho, 1993; Wannamethec & Shaper, 1991; Schoenfeld et al., 1994; Appels et al., 1996; Moller, Kristensen & Hollnagel, 1996; Idler & Benyamini, 1997). The model examines the effects of six well-known behavioural and psychosocial risk and protective factors for morbidity and mortality. The relationship between general self-assessed health and smoking, alcohol consumption, recreational exercise, diet, social support and job strain has been confirmed in many studies (Segovia et al., 1991, Joung et al., 1995, Bobak et al., 1998, Litwin, 1998, Manderbacka et al., 1998, Power et al., 1998, Schrijvers et al., 1998, Kawachi et al., 1999, Poikolainen and Vartiainen, 1999). Thus, the model predicts that job strain, a poor diet, lack of exercise for pleasure, low social support, high alcohol consumption and cigarette smoking are all associated with poorer general self-assessed health. A simplified path diagram for the proposed theoretical model from social position to ill-health is shown in Fig. 1.

Occupational conditions, operationalised by the Erikson–Goldthorpe class schema, are hypothesised to affect health indirectly via job strain. Those with a ‘labour’ form of contract are predicted to have more strain than those in ‘service’ contracts which in turn will increase the risk for poor health. In addition, occupational conditions which are not made explicit in the model are also hypothesised to have effects on health. These might include the physical conditions in the workplace, job insecurity or psychosocial support by colleagues, all of which have been reported to be precursors of ill-health (Stansfeld et al., 1997, Amick 3rd et al., 1998, Ferrie et al., 1998, Schrijvers et al., 1998).

General social advantage (the Cambridge scale) is predicted to affect health via its influence on the behavioural risk and protective factors. The relationship between heart disease and the Cambridge scale has previously been shown to be more strongly attenuated by behavioural variables than that between heart disease and Erikson–Goldthorpe class (Chandola, 1998). Indirect evidence also comes from studies which find education more strongly related to behavioural risk factors than income or occupation because of its influence on membership in peer groups that promote healthy behaviours (Winkleby et al., 1992, Whitlock et al., 1997). Thus, the Cambridge scale, based on friendship patterns, is predicted to be associated with a poorer diet, less participation in sport, more alcohol consumption and smoking, and empirically with lower social support (Bartley et al., 1999b). The analyses also aim to test whether there is empirical support for other pathways from general social advantage to poor general self-assessed health which are not mediated by the risk and protective factors identified in the model.

Material deprivation is hypothesised to have indirect effects on health via diet and sport due to financial limitations and also direct effects due to adverse living conditions and other factors associated with financial disadvantage and material conditions. Because of the empirical evidence for the association of material deprivation with low social support and increased prevalence of smoking, which in turn affect health (Goodwin et al., 1991, Bernard and Smith, 1998, Law and Morris, 1998, McKee et al., 1998), these pathways are also included in the model.

Path analysis techniques are used to examine the mechanisms whereby social position affects health by simultaneously modelling the relationships between these dimensions of social position and the risk and protective factors, as well as the relationships between the factors and health. The paper addresses the following key questions: (1) Which aspects of social position are most important for understanding the health gradient in women? (2) Do behavioural and psychosocial risk and protective factors fully mediate the relationship between social position and the ill-health of women? (3) Do the same measures of social position best capture the variability in behavioural and health status of women with different labour force attachments? and (4) How do the pathways from social position to ill-health differ for women in and out of the labour market?