Study population, sampling and procedures
The study population (
N = 27,215) comprised respondents of the 2015 annual national cross-sectional Living Conditions Survey ‘Encuesta de Condiciones de Vida’ [
27], carried out by the National Institute of Statistics of Spain, Instituto Nacional de Estadística (INE). The survey is the source of the Spanish data that is part of the European Union Statistics on Income and Living Conditions (EU-SILC), one of the statistical operations that have been harmonised for EU countries and that is governed by European Union Statistical Office (Eurostat) of the European Commission. The Living Conditions Survey has been ongoing since 2004 and comprises four independent sub-samples, each of which is a four-year panel, and with the sample rotated in one of the panels each year. The 2015 survey corresponds to rotational group “3”. The survey procedures are described in detail in the technical reports by INE [
27,
28] and in the European Commission Regulations describing definitions [
29], fieldwork and imputation procedures [
30], sampling and monitoring [
31], target variables [
32] and quality reports [
33].
The target population for the 2015 survey comprised all non-institutionalised people resident in Spain aged 16 years old or older as of 31st December 2014 [
27].The municipal population register was used to identify a stratified random sample, with an independent sample drawn within each Autonomous Community (self-governing region of Spain) as the INE is required to produce reliable data at this level of disaggregation as well as nationally. First, census sections, each consisting of about 400 addresses, were selected stratified for municipality size, with a selection probability proportional to the number of households within each census. Second, addresses were randomly sampled within each municipality, and personal interviews were held with all eligible household members. The sampling frame enables derivation of survey sampling weights, which were used in this analysis to aid generalization of the results to the target population.
Data collection was carried out through computer-assisted personal interviews with all participants in March – July 2015, with an average interview duration of 27 min. The content of the questionnaire is defined by the European Commission Regulations [
32]. Training of interviewers was done at the provincial area level according to training manuals. If a personal interview was impracticable because the subject was temporarily absent or was unable to respond, a telephone interview or interview with another household member was conducted, and later the information was corroborated with the participant in question.
The INE ensures confidentiality during the data collection process and provides information on the use and confidentiality of the data to the respondents, and all participants give informed consent to the use of the data for research purposes. Quality assurance by the INE is based on the European Statistics Code of Practice made by Eurostat, and Eurostat also carries out independent review of the survey data before the results are published [
27]. One of the weak points of the survey pointed out by Eurostat is the relatively high fraction of proxy interviews, the extension of which, however, has decreased since 2010 [
27].
Complete public domain microdata files from the survey are disseminated free of charge by the INE in anonymized form and was for the present project retrieved from the official microdata download page of INE [
34]. The overall individual response rate of the 2015 survey was 80.4% [
35]. After excluding records with incomplete data on the study variables, a sample of 22,456 individuals was available for analysis in the present study. The demographic, material and psychosocial characteristics of the effective sample are described in Table
1.
Table 1
Demographic, health and material and psychosocial characteristics by intersectional social positions in Spanish adults
Total (N) | 7891 | 14,565 | 11,376 | 11,080 | 11,461 | 10,995 | 2069 | 2503 | 1539 | 1780 | 6655 | 3234 | 41,113 | 3563 |
Age (mean, range) | 51.97 (20–88) | 55.49 (19–88) | 54.53 (19–88) | 53.97 (20–88) | 54.62 (19–88) | 53.88 (19–88) | 54.12 (21–88) | 50.69 (22–88) | 53.98 (22–88) | 49.54 (20–88) | 55.03 (19–88) | 57.50 (20–88) | 54.51 (19–88) | 55.28 (20–88) |
Poor SRH (%) | 17.96 | 34.99 | 27.94 | 30.10 | 26.76 | 31.34 | 17.88 | 16.26 | 20.60 | 18.15 | 30.26 | 36.61 | 33.67 | 39.88 |
Material factors (%) | Material scarcity | 1.80 | 14.42 | 9.27 | 10.72 | 7.72 | 12.35 | 1.79 | 1.60 | 1.88 | 2.02 | 9.90 | 13.79 | 15.24 | 18.69 |
Unstable employ- ment | 6.59 | 16.87 | 9.56 | 17.05 | 9.05 | 17.64 | 2.32 | 7.39 | 4.35 | 12.36 | 8.81 | 14.90 | 15.83 | 28.12 |
Insecure residen- tial area | 8.97 | 10.11 | 9.11 | 10.33 | 10.94 | 8.43 | 9.18 | 11.15 | 6.82 | 7.53 | 10.73 | 12.15 | 8.49 | 9.51 |
Psycho-social factors (%) | Poor social support | 2.43 | 5.60 | 4.80 | 4.16 | 3.96 | 5.03 | 2.61 | 1.96 | 2.86 | 2.53 | 5.20 | 4.98 | 6.27 | 5.78 |
Lack of social partici- pation | 17.27 | 42.52 | 34.89 | 32.37 | 29.67 | 37.79 | 18.12 | 14.62 | 21.12 | 16.69 | 38.00 | 39.30 | 45.71 | 46.39 |
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 [
36], and that displays social inequalities [
37,
38].
The three binary variables of
social positions were
social class (manual or non-manual),
gender (man or woman) and
regional development (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 [
39,
40], with manual class comprising the III-V groups and the non-manual 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 [
41]. Those with the highest IHDI were considered High development regions and those with the lowest IHDI were considered Low development regions.
The three social position variables were combined to form eight intersectional social positions: Men Non-manual social class High development regions (MNH); Women Non-manual social class High development regions (WNH); Men Non-manual social class in Low development regions (MNL); Women Non-manual social class in Low development regions (WNL); Men Manual social class in High development regions (MMH); Women Manual social class in High development regions (WMH); Men Manual social class in Low development regions (MML); and Women Manual social class in Low development regions (WML).
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 1 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 [
42].
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.
Statistical analysis
Descriptive statistics comprised frequencies of SRH, age, and explanatory variables across social positions and eight intersectional social positions.
Two different sets of analysis were undertaken to address the aims, all using of multivariate Poisson regressions to estimate prevalence ratios (PR) [
43] with SRH as the outcome and social positions as main exposures. The first set of analysis included the three indicators of social positions as mutually adjusted independent variables (corresponding to an additive approach, which does not illustrate the health risks in the intersecting social positions). The second instead utilized the indicator of eight intersectional social positions according to an intersectionality-informed multiplicative approach (which discloses the health risks of multiple intersecting social positions) – with the best-off group (men in non-manual social class from high development regions) as the reference category. For each of the two approaches, four models were created. Model A was adjusted only by age, Model B was adjusted for age and all psychosocial factors, Model C was adjusted for age and material factors, and Model D was adjusted for all factors together. The explained fraction (EF) of each social position and intersectional social position was calculated after every adjusted model versus the crude model given the following equation: ((PR
A-PR
B)*100/(PR
A-1)).
A complete case analysis was conducted when missing data existed, such as for 4580 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.