Self-rated health inequalities in the intersection of gender and social class in Spain: exploring contributions of material and psychosocial factors

Background Inequalities in health across social class, gender and regional context in Spain are well-known; however, there is a lack of research examining how these dimensions of inequality interact. This study explores self-rated health (SRH) inequalities across intersectional positions of gender, social class and region, and the contribution of material and psychosocial factors to these inequalities. Methods Participants were drawn from the cross-sectional 2015 National Living Conditions Survey of Spanish residents aged 19-88 years (N=27,215; 77% response rate). Eight intersectional positions were formed by combining dichotomous variables of gender, social class and regional development. Poisson regression was used to estimate intersectional inequalities in SRH as prevalence ratios, and the contributions of material and psychosocial factors. Results Women in manual social class from low development regions reported the worst SRH. Inequalities by the interaction of social class and regional development were best explained by the joint contributions of material and psychosocial factors, while gender inequalities among non-manual social class were better explained by material factors alone. Conclusions The results illustrate the complexity of interacting inequalities in health and their underpinnings in Spain. Local and national policies that take this complexity into account are needed to broadly improve equity in health in Spain.

The Commission on Social Determinants of Health of the World Health Organization 22 proposed a framework whereby intermediary, structural social and economic determinants unequally affect health and wellbeing. Structural social and economic determinants are built upon the socioeconomic and political context and upon the individual social position. Intermediary determinants comprise material circumstances, psychosocial processes and behaviours and biological factors 23 . Many studies have assessed the intermediary determinants of health in relation to social positions of gender or social class 24,25 , but less so their contribution to intersectional inequalities in health. A Spanish example demonstrates contribution of intermediary determinants, including material factors, on the intersection between gender and social class and its effects on SRH 26 . Nevertheless, very little is known about the contribution of material and psychosocial factors on SRH in the intersection between gender, social class and region in Spain.
The aims of this study are therefore to (1) explore how intersectional social positions of social class, gender and region are reflected in population patters of SRH among Spanish adults, and (2) examine the contribution of intermediary social processes material and psychosocial factors to these inequalities in SRH. To illustrate the added value of using an intersectionality-informed multiplicative approach, these questions will also be addressed from a conventional additive approach where gender, social class and regional inequalities in health are approached separately. Stratified random sampling was used to collect information from the non-institutionalised adult population aged 19-88 years. The sampling frame allowed the derivation of survey sampling weights, which were used in this analysis to aid generalization of the results to the target population. The average response rate was 77% 28 . After excluding records with incomplete data on the study variables a sample of 22,456 individuals was available for analysis.

Variables
The dependent variable self-rated health (SRH) was derived from the following question: Would you say that your overall health is either: very good, good, fair, poor or very poor? The answers were dichotomized into either good health (good or very good coded as 0) or poor health (fair, poor or very poor coded as 1).
SRH is a common measure of an individual's well-being and health status and has been shown to be a valid and reliable indicator of morbidity and early mortality 29 , and that displays social inequalities 30.

.
The three binary variables of social positions were social class (manual or non-manual), gender (man or woman) and development regions (high or low). Social class was coded according to the Spanish adaptation of the British Registrar General classification, based on the International Standard Classification of Occupation 2008 32, 33 , with manual class comprising the III-V groups and the nonmanual I-III groups of the British Registrar General classification. Gender was self-reported in the Living Conditions Survey with two options; woman or man. Regional development was derived from the Inequality-adjusted Human Development Index (IHDI) for each Autonomous Community and Autonomous city in Spain in 2010 34 . Those with the highest IHDI were considered high development regions and those with the lowest IHDI were considered low development regions. Variables potentially reflecting social processes underpinning intersectional inequalities in SRH were identified in the Living Conditions Survey. The three material factors material standards of living, employment conditions, and residential environment and the two psychosocial factors social support and social participation were selected.
For material standards of living the following nine binary items were selected and summed up into an index: having holidays at least one week a year away from home; a mobile phone; a television; a computer; access to internet; a washing machine; a car; a private shower, and spending discretionary money weekly on oneself. The index was dichotomised and when four or more items were lacking it was labelled Material scarcity 35 . Employment conditions was indicated by two items: employment status (wage worker full time, wage worker partial time, self-worker full time, self-worker partial time, student, retired, permanent incapable to work, household worker, other type of economic inactivity), and type of contract (employer, self-employed, permanent wage, temporary wage, and familiar help). Unstable employment index was defined when employment status was student, retired, permanent incapable to work, household worker or other type of economic inactivity and when type of contract was temporary wage or familiar help.
Residential environment was based on two yes/no questions: existence of delinquency problems and existence of vandalism in the respondent's residential area. Insecure residential area was defined among those with at least one 'yes' answer.
Social support was based on two yes/no questions: if the respondent had family or friends who they could ask for help and if the respondent had someone to talk to about personal issues. Poor social support was defined among those with at least one 'no' answer.
Social participation was derived from ten items referred to participation in activities the past year such as having: gone to the cinema; gone to the theatre; visited cultural places; gone to sport events; participated in voluntary activities, and participated in political activities; as well as frequency of meeting friends, contacting family members, contacting friends, and participating in social media.
Lack of social participation was defined as a negative response to seven or more items. A complete case analysis was conducted when missing data existed, such as for 4,580 subjects without classifiable social class as nothing was stated in their occupational status. Out of these, 47% were born 1990-1998 and were therefore students or unemployed young adults. All analyses were carried out with the Stata version 14 statistical package.

Socio-demographic characteristics of the samples
The sample analysed in this study comprised of 14,565 adults in the manual class and 7,891 in the non-manual class; 11,080 women and 11,376 men; and 11,461 people living in high development regions and 10,995 in low development regions ( Table 1).
Out of these, 40% of the respondents in high development regions were non-manual workers while in low development regions 30% were non-manual workers. Gender was homogenously distributed by region, but when it comes to social class men tended to belong to manual class more often that women did (68% v 61%).

Intersectional inequalities in material and psychosocial disadvantages
The distribution of material and psychosocial factors displayed distinctive inequalities between, but also within, the indicators of class, gender and regional development ( Table 1; Figure 1).
Firstly, considering the three indicators one by one, the largest inequalities were found for social class for which manual class consistently displayed more material and psychosocial disadvantages as compared to non-manual social class. There were eight times higher frequency of material scarcity and more than double the frequency of unstable employment, poor social support and lack of social participation, but with fairly similar prevalence of insecure residential area. Women reported unstable employment 80% more often than men, with the other indicators displaying smaller inequalities (<20% relative difference). The disadvantages for women were material scarcity and insecure residential area and the corresponding for men were social support and social participation.
Low development regions reported disadvantages more often (30-90%) than high development regions, except for insecure residential area which was slightly more common in privileged regions.
Secondly, distribution of material and psychosocial factors across intersectional social positions revealed more complex patterns of inequalities not discernible through single indicator inequalities. For example, although the triply disadvantaged group (women in manual class from low development regions) reported 10 times higher material scarcity than the 8 times difference between manual and non-manual social class (Table 1; Figure 1a); it also reported 12 times higher unstable employment than the triply advantaged group (men in non-manual class from high development regions) compared to the moderate 2-3 times higher within each social position (Table 1; Figure 1b).
This illustrates how the magnitude of the intersectional inequalities cannot be monotonously predicted from single inequalities, but depended on life conditions. The complexity become even more apparent when considering intersectional groups with mixed position of advantage and disadvantage; further illustrating the heterogeneity in life conditions not only between, but also within, the crude categories captured by the single indicators. For example, the intersectional position with the overall lowest material scarcity was women and not men, in non-manual occupations and high development regions. Moreover, whereas material scarcity, as noted above, was clearly patterned by social class, women in manual class from low development regions reported twofold material scarcity as men in manual class from high development regions (Table 1; Figure 1a). Additionally, the small relative advantage of women as a group when it comes to psychosocial resources was restricted only to non-manual class (Table 1; Figure 1d; Figure 1e).

Intersectional inequalities in SRH
Descriptive patterns indicating complex inequalities between intersectional positions were also found when it comes to SRH (Figure 2). For example, whereas all manual workers' intersectional positions displayed higher frequencies of poor SRH than all non-manual positions, there was considerable heterogeneity especially within manual groups, with prevalence ranging between 30% frequency for men in high development regions to 40% for women in low development regions. The most advantageous position was women in non-manual social class from high development regions, while the most disadvantageous position was women in manual social class from low development regions.
Differences between gender and regional development were larger among those in manual classes than non-manual classes.
Role of material and psychosocial factors in gender, social class and regional inequalities Poisson regression analyses were carried out to estimate social inequalities in SRH by social positions of gender, social class and regional development, The additive approach revealed that social class inequalities was the most remarkable inequality indicator, amounting to 61% higher prevalence of poor SRH among manual compared to non-manual social class. Minor but significant inequalities were found for gender and development regions. As indicated by the explained fraction (EF), psychosocial and material factors partially, but not completely, explained these inequalities. Psychosocial factors (Model B) explained about a fourth of the large class inequalities (EF=26%) and the smaller regional inequalities (EF=26%) in SRH but not substantially gender inequalities (EF=-5%). Material factors (Model C), had a greater relative importance for gender (EF=36%) than social class (EF=17%) or regional (EF=19%) inequalities. As a result, all factors together (Model D) explained a larger portion of social class (EF=34%) and regional inequalities (EF=36%) but less of gender inequalities (EF=23%). All inequality estimates remained statistically significant (p<0.001) even after full adjustment (see Table 2).
Analysing the inequalities according to a contrasting multiplicative approach the reference category was the best-off intersectional position after adjusting by age (the triply advantaged group men in non-manual social class in high development regions) (see Table 3).
The multiplicative approach showed cumulative effects of disadvantages. Whereas the additive approach estimated a 61% higher prevalence of poor SRH among social class groups (Model A) (see Table 2), intersectional social positions revealed up to 111% higher prevalence of poor SRH for manual social class women from low development regions compared to the reference group.
Moreover, there was heterogeneity in prevalence ratios; ranging from 1 (reference) to 1.25 within the intersectional non-manual social class groups, and between 1.61 and 2.11 among intersectional manual social class groups (Model A) (see Table 3).
Discrepancies between the additive and multiplicative approach were also evident when it comes to the effect on the estimates when taking the indicators of social processes into account.
Overall, psychosocial factors (Model B) explained inequalities mostly involving those intersectional social positions which had a higher relative frequency in both psychosocial factors (see Table 3); EF WML =24%; EF MML =28%; EF MMH =24%. Specifically the increased gender inequality when taking psychosocial factors into account in the additive approach (EF=-5%) (see Table 2) was only evident for women in non-manual social class from high development regions (EF WNH =-25%) (see Table 3).
Moreover the sizeable explanation by psychosocial indicator of social class (EF=26%) and regional (EF=26%) inequalities (see Table 2) were in the multiplicative approach comparable only for the specific intersectional position of men in manual social class from low development regions (EF MML =28%) (see Table 3).
Whereas material factors explained a large portion of the overall gender inequalities (EF=36%) (see Table 2) the multiplicative approach showed their importance differed markedly for intersectional social positions within the same gender; from 12.5% to 21.4% for women and from 1.5% to 16.6% for men. A similar variation in explanatory power reflecting the intersectional inequalities was seen when adjusting for material factors: prevalence ratios for non-manual class groups ranged from 1.03 to 1.22 while prevalence ratios for manual class groups ranged from 1.54 to 1.86 (see Table 3).
The full model (Model D) involved the greatest explanatory fractions for all intersectional positions except for women of non-manual class which were better explained by material factors only (EF WNH = 41%; EF WNL =12%) (Model C) and men from non-manual class and low development regions which were better explained by psychological factors only (EF MNL = 17%) (Model B) (see Table 3).
Among all intersectional inequalities, the prevalence ratios of manual class groups decreased the most when adjusting by all factors (Model D). However, the considerable explanatory fractions for social class (EF=34%) and regional (EF=36%) inequalities (see Table 2) were only seen for manual social class from low development regions (EF MML =36%; EF WML =36%), (see Table 3).

Discussion
This cross-sectional study on Spanish adults is the first of its kind to investigate how intersectional social positions of gender, social class and regional development are reflected in population patterns of SRH, and to examine the contributions of the intermediary social processes material and psychosocial factors to inequalities in SRH.
Our results suggest that women in non-manual social class from high development regions had the most advantageous position when it comes to SRH, while women in manual social class from low development regions had the most disadvantageous position. Multiplicative analysis revealed cumulative albeit not monotonous health effects of multiple disadvantages, as has been found by others 11 . For example, the largest difference in prevalence ratios of poor SRH was displayed among the intersectional social position women in manual social class from low development regions compared to the reference category (see Table 3).
Malmusi 26 suggested that individual income contributes importantly to gender inequalities in health.
Similarly, we found that unstable employment contributed substantially to gender inequalities in SRH, especially when considering intersectional social positions, which were underpinned by the ubiquitous material disadvantage in the triply disadvantaged group. For example, although material scarcity was clearly patterned by social class, women in manual class in low development regions reported twice as much material scarcity as men in manual class in high development regions did. This emphasizes how inequity in access to material resources plays a widespread role for social inequalities in Spain.

Strengths and limitations
One of the study's strengths is the large and rich random population-based sample with rather high participation rate that allows creating intersectional categories with enough statistical power. However, the cross-sectional design limits causal inference. Selection bias might be present since

Ethics approval and consent to participate
The study is based on secondary data that is routinely collected and processed by the National

Consent for publication
Not applicable.

Availability of data and materials
Datasets are available in a publicly accessible repository. The datasets analysed for this study can be found in the Instituto Nacional de Estadística de España: https://www.ine.es/dyngs/INEbase/es/operacion.htm? c=Estadistica_C&cid=1254736176807&menu=resultados&secc=1254736195153&idp=12547359766 08

Competing interests
The authors declared that they have no competing interests.

Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Authors' contributions
NP conceived and designed the study with the help of EE. NP acquired the data, and carried out and interpreted the final analysis with guidance from PEG. NP prepared the drafts which were revised critically by EE and PEG. All authors have read and approved the final manuscript and are accountable for its content. 1 The first letter of the intersectional social position is gender (M=Men, W=Women), the second letter is the social class (M=Manual, N=Non-manual) and the third letter is the regional development (H=High, L=Low).  *p-value < 0.05; **p-value < 0.01; *** p-value < 0.001.
1 The first letter of the intersectional social position is gender (M=Men, W=Women), the second letter is the social class (M=Manual, N=Non-manual) and the third letter is the regional development (H=High, L=Low).  Spanish adults. 1The mean of self-rated health was based of the dichotomised answer 0=good health and 1=poor health. 2The first letter of the intersectional identity is gender (M=Men, W=Women), the second letter is the social class (M=Manual, N=Non-manual) and the third letter is the regional development (H=High, L=Low).