Introduction

The protective health effect of living with a partner is well-known,^1-3 but we know less about health differences among partnered individuals. Although the link between health and individual socio-economic information from each partner member has already been studied,^{4 5} such research tested the effect of the partner's socio-economic characteristics in addition to the ego's characteristics rather than combining the socio-economic characteristics of both partners. The latter is the aim of the present study.

The health advantage of people co-residing with a partner has been extensively explained by a set of benefits related to living with a partner: first, being subjected to social control, particularly from partners. Especially partnered men may be more often encouraged to abandon risky behaviours than single individuals.³ Second, the combination of both partners' social nets creates a wider source of possible resources to reduce the impact of events that may negatively impact health, such as becoming unemployed or widowed.⁶ And third, the optimization of common resources through economies of scale and specialization of tasks within the household enhances the socioeconomic status of both partners.⁶ The latter two health benefits also have a socio-economic gradient.⁷

When measuring socioeconomic status of individuals, education is a frequently used indicator of adult socioeconomic status in studies on health differences because of its stability over time (unlike income or employment status) given that it is typically completed early in adulthood.⁸ Using education also avoids problems of reverse causation when linking it with health outcomes in older ages.⁹ It also has a well-established negative association with long-term risky health behaviours like smoking or drinking¹⁰ and facilitates the effective use of health services.¹¹ According to the social causation hypothesis, an individual's educational level is also assumed to affect material and psychosocial factors that in turn have effects on health. In this study, we extent this assertion by purporting that partner's education has an additional effect on health.

Indeed, previous studies have found positive associations between partner's educational attainment and individual's health, including self-perceived health,⁵ risky behaviour¹² as well as mortality.⁴ All studies stated that the inclusion of partner's education adds meaningful information to better understand health inequalities, even after controlling for own educational level. A common explanation was that couples usually pool their resources (material and non-material).

In line with the latter, previous studies on health inequalities have indicated household capacity to make ends meet to be an accurate indicator of poverty as it summarises the economic context of a household.¹³ This is of special interest to our study as relatively few elderly Spanish women participated in the labour market, meaning that their socioeconomic status is dependent on their partner's labour trajectory. Moreover, household income is known to be associated with individual mental and self-perceived health.^{14 15}

There are also other partner's features that can modify individual health and mortality risks. The spousal age gap has been associated with survival probabilities, although differently by gender: having a younger spouse is beneficial for men but detrimental for women, whereas having an older spouse is detrimental for both.¹⁶ Positive effects include psychological improvements of having a younger partner as well as selection effects (healthier individuals have higher chances of attracting younger partners).¹⁶ Conversely, being the caregiver of one's partner is a known risk factor for physical and mental health,¹⁷ a task that women fulfill more often than men due to their higher life expectancy and the traditional gender role norms roles.¹⁸ Partner's health status is another characteristic of the partner affecting individual health. Reasons that have been linked to this include shared living conditions,¹⁹ similar health-related behaviour²⁰ and the already mentioned health risks when caring for an ill partner.²¹

Why study Spain? While studies on the relationship between joint (partner and own) education profiles and health status have focused on the United States or northern, western and central European countries, no country from southern Europe has yet been analysed. Moreover, Spain is an especially interesting case study because during the last half-a-century it experienced a process of rapid political and social change, transitioning from an ultra-conservative dictatorship (1939-1975) to a democracy-based market economy. The educational expansion and the massive insertion of women into the labour market started later in Spain than in other Western European countries.²² Accordingly, the productive and reproductive roles are still very much gendered for the older cohorts studied here, whereby the female socio-economic status is almost totally determined by their partners' status.²³

Within these socioeconomic, gender and cohort specificities of Spain, the study's aim is to determine whether, besides own educational attainment also characteristics of the partner (educational attainment and health status) and union (household capacity to make ends meet and the spousal age gap) affect individual health status among co-residing individuals aged 65 years and older. We contextualize our results within the high level of gender inequality typical of older Spanish cohorts. Our research questions are: 1) do partner and couple features complement what we currently know about health status variability of the Spanish older population? and 2) how does this vary by sex?

Method

Data come from the Spanish cross-sectional sample of the European Union Statistics on Income and Living Conditions (EU-SILC) survey (2015 wave). Our target population are partnered Spaniards where both members of the couple are aged 65-81 (N=6569). The lower age boundary was fixed at 65 because this was the legal age at retirement at the time of the interview; 81 is the age of the oldest respondents. EU-SILC has information on socioeconomic status and health from all members living in a private household, although 32 couples which accomplished the age criteria were discarded because of incomplete data. Only native Spaniards who live with a partner were analysed (56% of the age-specific sample), irrespective of their marital status (due to the low (<1%) prevalence of marital statuses other than married). Given our interest in possible sex differences in the effect of partner characteristics on individual health, same sex couples were also removed from the analysis. This left us with 1787 couples with complete information on all variables.

Our dependent variable is individual self-reported health status, measured by the survey question ‘How is your health in general?'. The possible answers were dichotomised into good health (“good” and “very good” health) and less than good health (“fair”, “bad” and “very bad” health).

The co-variables are:

The possible effects of the variables of interest on individuals' health status were assessed by separate logistic regression models for men and women to assess possible gender differences. We present predicted probabilities of poor health with 95% confidence intervals that include all variables. Predicted probabilities allow the direct comparison of results from different models²⁴ (odds ratios are reported in Table II in online ). We additionally conducted a dominance analysis to determine the relative contribution from each variable in the total explained variance.²⁵ Besides education and household ability to make ends meet, we also tested the effect of overall household income and the individual relative contribution to the overall household income. However, they were excluded from the final model because they did not produce a significant contribution (see Table III in online ^Appendix). To check the robustness of our results, we replicated our models for two other health outcomes available in EU-SILC (suffer from any chronic [long-standing] illness or condition; limitation in activities because of health problems). Both outcomes were operationalized as a dichotomous variable (1=yes; 0=no) and results are provided in Tables IV and V in online ^Appendix, and are briefly commented on in the discussion.

Results

Discussion

Our study contributes to the growing body of research on health among partnered individuals. Moreover, we did this by also including characteristics from both partners combined rather than that from each partner separately. Results showed for Spanish elderly aged 65-81 that women's health is more sensitive to the characteristics of the partner and household as more of the variability in female health status was explained by the tested variables. Co-residing with a partner not in good health is associated with almost double the probability of not having good health. One possible reason for this association is selective mating as two individuals with similar behaviour are more likely to match up, irrespective of socioeconomic status.² Partners also influence each other's lifestyle during their relationship (for instance, it is more difficult to quit smoking if one's partner also smokes) and partner resemblance in health increases with relationship duration.²¹ This implies that assortative mating at younger ages (when couples are established) may have a cumulative effect on health as a consequence of behavioural factors (smoking, diet, alcohol consumption, etc.).

There are also gender implications from having a partner in poor health. When the husband is in poor health, the wife's health is likely to deteriorate due to the increase in their domestic workload.²⁸ This might explain why, according to our results, living with a much older husband penalised female health. Conversely, the additional workload is lower for husbands when their wives have poor health as this is usually shared with other, mainly female, relatives, particularly daughters or daughters-in-law.²⁹

Women's health was slightly more sensitive to their own and partner's educational level than men. First, their health was significantly (p <0.01) better if their educational level was higher than their partner's, a result that was less significant (p <0.1) in the case of men. Second, in the context of the known educational gradient in health, there is some suggestion that risk factors leading to health deterioration among women are also mediated through the educational status of their partner. However, this is mainly observed among women with medium education (the probability of having less than good health declines as the educational level of the partner becomes higher). The lower level of association for men in this case is consistent with results for the Netherlands⁵ and accords with the gendered social roles that typifies this Spanish generation where few women contributed financially to household income³⁰ and were therefore dependent on the socioeconomic status of their partner.

Repeating the analysis for the two other health indicators included in EU-SILC led to similar results, thus confirming the importance of combining the educational attainments from both partners to better explain health inequalities among partnered older Spaniards. The only exception was the male model for chronic conditions and limitations, as results were not statistically significant. This seems to support results from elsewhere that show an educational gradient in chronic conditions and limitations among older women but not among men where the effect of education is less strong or significant in fewer chronic conditions.^{31 32} It would certainly be interesting for future research to explore these sex differences further as well as consider different types of chronic diseases.

Our study has shown that within an apparently advantageous population group in terms of health status, namely married or cohabiting couples, there are significant socioeconomic differences as higher educated individuals and those without economic problems have better heath than their lower educated and poorer counterparts. Therefore, a relevant contribution for public policy is that not only typical vulnerable groups like single-person, single-parent or low-income households should be targeted, but also retired low-educated partnered individuals, even if their partner is higher educated. This is not only because their probability of not having good health does not improve if the latter is the case, already 44% of 65-81 year-olds do not live with a partner, among whom health is significantly worse.³ Better targeted health education and health promotion³³ and ensuring adequate income in old age for low-educated individuals are thus pertinent in the quest to reduce social health inequalities. Also noteworthy is the effect of partner's health on own health. This result is particularly relevant for health professionals who are in daily contact with patients as health inequalities are larger at the household level than at the individual level.²¹ The Spanish Society for Family and Community Medicine already recommends through their regularly updated Programme of Preventive and Health Promotion Activities (PAPPS in its Spanish acronym) general practitioners to assess the motivation and availability of the patient and his or her partner or other family members to change health-damaging behaviour and establish a joint action plan.³⁴ Our result therefore reinforces the importance of involving the patient's partner as this would ease implementing specific interventions to achieve moderate habit changes (and may also benefit from it).

One weakness of our study is the lack of long-term longitudinal information about individuals' trajectories to study possible mechanisms of selection prior to partner co-residence. The data source (EU-SILC) also lacks information on health behaviour, including fruit, vegetable, smoking and alcohol consumption, and physical activity, that may provide additional clues as to why higher educated individuals have better health than their lower educated counterparts. Additionally, combining individual and partner categories reduced the number of cases in each combination of educational level, resulting in fewer significant associations, an issue that would be resolved with a larger sample size. We must also keep in mind possible long-term influences of the economic crisis in Spain (2008-2014) at the time of the study (2015), as this may have amplified the sex-differences we found in the influence of the partner's socio-economic profile on individual's health. Finally, we cannot disclose whether the results for Spain are transferable to other countries as only one country was analysed. A comparative study for this age cohort between European countries characterised by different levels of gender inequality is therefore another recommendation for future study.