There is increasing recognition that education is not only an individual-level resource but can have spillover health effects on other household members. For example, there is a robust connection between higher maternal education and improved child health.1 More educated adult children have also been found to be associated with improved health outcomes for parents.2–4 Within married couples, a more educated spouse has been associated with better health.5–16
Married people enjoy better health in general,6,17 and given that socioeconomic characteristics such as education influence selection into marriage, they may contribute to this advantage. However, the socioeconomic resources at the level of the household can also impact health by influencing access to vital material and psychosocial resources.12,15 Examining the extent to which the educational attainment of partners confers protection against heart attack incidence and survival is of interest for understanding the mechanisms behind social inequalities in heart disease. Only a few studies have recently examined the association between partner education and coronary heart disease (CHD), and none have separated disease onset from disease progress. A Norwegian study found that higher educational attainment of wives was associated with lower risk of CHD mortality and lower levels of CHD risk factors in married men, and this association was particularly strong for men with higher education.8 Research from Lithuania and the Netherlands also found a protective association of wives’ higher education with men’s health.5 The consequences of husbands’ education on women’s CHD risk have been less studied. Earlier research on mortality indicates that the impact of spousal education may be weaker in women than in men in studies conducted in Israel, Norway, and Sweden.10,11,13,14 Wives’ education may in fact be more influential for men’s health and mortality than men’s own education.5–8,10,11,13 A possible explanation is the greater influence that the wife and her education exert on the adoption of healthier behaviors in the household.18 This could delay the onset of cardiovascular diseases in both partners.
In addition to influencing the onset of cardiovascular disease, spousal education may also have a significant influence on survival chances. For example, the partner’s high educational level may contribute to earlier detection of disease and access to care. A more educated partner may be better equipped to support the partner’s recovery, treatment adherence, and lifestyle changes. Despite the key importance of the stage after the onset of disease for overall patterns of population health, to our knowledge, no prior study has examined how spousal education is associated with fatality after myocardial infarction (MI).
In contrast to the evidence of education as a positive resource, several studies reported an adverse impact of wives having a higher level of education relative to their spouses. Studies conducted in the United States in the 1960s and 70s reported that being married to a more educated wife increased the risk of CHD at least for some subgroups of men such as those with type A personality.19–24 A few studies also suggested elevated risks associated with the higher education of the husband relative to the woman’s own education.19,25,26 These findings garnered considerable interest and debate on the potentially adverse impact of “status discrepancy” between partners on health through role conflict and stress.27,28
The high level of gender equality in Finland provides an interesting context to examine the effects of both partners’ education and possible educational discrepancy effects. In the early 1980s, own and spousal education and occupation were more or less equally predictive of the mortality risks of both sexes.7 It is expected that with increasing female education, labor force participation, and prevalence of dual-earner households, the resources of both partners have become increasingly important for the general socioeconomic standing of the household, and hence for the health experience of its members. We examine whether spousal education is protective for the incidence of and fatality from MI controlling for own education in more recent longitudinal Finnish register data. The data include follow-up data for a large nationally representative cohort both for the first incidence of MI and subsequent mortality. Both married and cohabiting individuals are included for the estimation of the effects of the partner’s education net of the individual’s own education and other socioeconomic resources. Importantly, we also compare their risk of MI incidence and fatality to that of individuals living without a partner, as it is an increasingly common type of living arrangement,29 and often found to be associated with worse health and higher mortality.30,31 A novel aspect of our study is our ability to separate the associations of spousal education with initial incidence of MI versus survival in the aftermath of MI. Uniquely, to our knowledge, we also differentiate between short-term and long-term fatality, as the former may be more strongly influenced by access to care, whereas the latter may be influenced by the extent of life style modification after surviving a heart attack. Our main research questions are:
- (1) Is the partner’s education associated with lower risks of MI incidence and fatality net of own education and other socioeconomic resources?
- (2) Does the association of partner’s education with the risk of MI incidence and fatality vary according to own education?
METHODS
Data
The data originate from an 11% random sample of the registered adult population in Finland in 1987–2007 together with an 80% oversample of the population who died in the period. Sample weights were constructed from known sampling probabilities and were taken into account in all descriptive statistics and analyses. The data derive from several registers covering the whole Finnish population, including the labor market data file, cause of death records, registers on hospital discharge, and special drug reimbursement rights. Data linkage was approved by Statistics Finland’s ethical committee and was performed using unique personal identification numbers and then anonymized for research purposes. Permission to use the data for research was granted by Statistics Finland (TK-53-1519-09).
The sample for the study included 40- to 69-year-old men and women who were living in private households at baseline in 1990 (weighted n = 378,001). CHD status prior to, or at baseline, was identified by CHD diagnoses in hospital discharge records (International Classification of Diseases, 8th revision [ICD-8] and ICD-9 codes 410–414) or special rights to drug reimbursement for CHD between 1970 and 1990. Individuals with CHD (6%) were excluded to identify the first incidence of MI (n = 354,100 after exclusions). MI incidence and mortality (ICD-9 code 410 for 1991–1995 and ICD-10 codes I21-I22 for 1996–2007) in the follow-up between 1991 and 2007 were identified from the hospital discharge records and the cause of death register. In addition to the first MI incidence (fatal or nonfatal), we examined two fatality outcomes: First, to capture short-term fatality from MI, we constructed the outcome of 28-day fatality, which includes MI deaths occurring before any hospitalization for MI as well as all-cause mortality in the first 28 days starting from the date of the MI hospitalization. Second, all-cause mortality during the follow-up of the MI patients who were alive after the 28-day period was examined as long-term fatality, which had an average of 1710 follow-up days. We focused on all-cause mortality as the outcome, as this provides an integral view of potential spillover effects, including deaths from comorbidities related to heart disease. However, analyses on long-term fatality were repeated also by using cardiovascular disease mortality as the outcome.
Variables
The socio-demographic characteristics of the sample were identified using register-based information. Information on the highest educational qualification of the individuals was grouped into categories of (1) tertiary (13 years or more); (2) secondary (up to 12 years); and (3) basic level (up to 9 years) education. The basic level category includes individuals who have not obtained qualifications beyond the basic level, but may also include those whose qualifications are unknown, for example, recent immigrants. The group of nonregistered higher educational qualifications is likely to be small, as Finland had not experienced high levels of immigration in 1990. Education of the marital or cohabiting partner was constructed from yearly information on whether the person was living with a marital or cohabiting partner and that partner’s education, and was categorized in the same manner as own education. Cohabitation was identified in the data when an individual was living together with another adult of the opposite sex who was not a sibling and who had an age difference of less than 16 years. In the main effect analyses, we included an additional category that identified those without a married or cohabiting partner. For interaction analyses, dummies were constructed for own and spousal basic level education.
To control for possible confounders of partnership status, we included three additional indicators of the index individuals’ socioeconomic position at the baseline in 1990. Occupation was grouped into six categories: upper nonmanual, lower nonmanual, specialized manual, nonspecialized manual, farmer or entrepreneur, and other or unknown. Employment status was assessed by main activity in 1990 and differentiated between employment, unemployment, retirement, and other. The information was derived from previous censuses if not available for that year. Information on individual taxable income originated from the tax records and incorporated the wages, capital income, and taxable income transfers (no missing information). Income was divided into quintiles calculated for men and women separately. A covariate on the year of the first MI categorized into 3-year groups was also included for the fatality analyses.
Statistical Analysis
We analyzed MI incidence with Cox’s proportional hazards regression models with age as the underlying time. The baseline date was set at December 31st 1990 with a follow-up until the end of the year 2007. Information on own and partner’s education was updated for each year of follow-up from information from the previous year, allowing for changes in education and partnership status. Censoring occurred at MI, death, emigration (when the individual is no longer in the annual records), whichever took place first. For fatality analyses, we used information on education from the year prior to MI incidence. Twenty-eight-day fatality was examined with log binomial regression to estimate risk ratios (RRs). Long-term fatality was analyzed with Cox regression with the follow-up beginning on the 28th day after hospitalization and ending at the end of 2007 and censored for death or emigration.
The analyses were conducted separately for men and women, with the highest category used as the reference group in each socioeconomic variable. Model 1 shows total effects of own and partner’s education by including each of these variables of interest separately with age at baseline (1990 or at incidence) in 5-year age groups and the year of MI (for fatality). Model 2 incorporates both own education and partner’s education to distinguish independent effects, and to control for factors that may influence partnership status; model 2 additionally adjusts for own occupation, employment status, and individual taxable income. Analyses were performed using Stata 14.0 (StataCorp 2015, College Station, TX).
We examined whether the level of own education modified the effect of partner’s education using the dummy variables for own and spousal basic education in the married and cohabiting population in model 2. Deviations from additivity were assessed in model 2 by calculating the relative excess risk due to interaction using the method provided by Andersson et al,32 with 95% confidence intervals (CIs) computed using the delta method. In the risk due to interaction analyses, the null of 0 corresponds to exact additivity.
RESULTS
The majority of the sample was married (64% of women, 72% of men), and a small proportion was cohabiting (5% women, 6% men). More women had no partner (31%) than men (22%). Table 1 describes the distribution of the socioeconomic characteristics of the sample by gender and whether the person was in a married or cohabiting relationship at baseline or not. A greater proportion of men than women were tertiary educated. In general, those with partners had greater socioeconomic resources than those without partners, but the differences were greater in men. The correlation between an individual’s own education and that of their partner was 0.43 in women and 0.41 in men. eTable 1, https://links.lww.com/EDE/B296 (Supplemental Digital Content 1) describes the crude outcome rates by own and partner’s education in 1990, which are higher in the lower educational categories, but highest for those living without a partner.
;)
TABLE 1: Baseline Descriptive Characteristics by Gender and Partnership Status of Sample of Adults Aged 40–69 Years in 1990
The results presented in Table 2 demonstrate that both own and partner’s education had “dose–response” associations with MI incidence (model 1). When mutually adjusted and controlled for other indicators of socioeconomic resources, the estimated effects of own and partner’s education were quite similar in women (model 2). In men, MI incidence risk associated with partner’s basic education was somewhat higher than that of own education. For both sexes, the hazard associated with having no partner was higher than having a partner with only basic level education.
TABLE 2: Main Effects of Own and Partner’s Education for MI Incidence, and Post-MI Short-term and Long-term Fatality in 1991–2007
We examined short-term fatality in those who had an MI during follow-up (Table 2). Twenty-nine percent of incident MI events in women and 32% in men were fatal in the first 28 days. In women, partner’s education was as strong a predictor of 28-day fatality as own education, whereas in men own education was associated with greater risks of fatality (model 1). An increased fatality risk remained particularly for women with basic educated partners when adjusted for individual socioeconomic characteristics (model 2: RR, 1.18; 95% CI, 1.02–1.36). However, having no partner was associated with the highest risk of fatality, with RR of 1.39 (95% CI, 1.21–1.60) for women and RR of 1.46 (95% CI, 1.34–1.59) for men compared with those with tertiary level educated partners (model 2).
Both own and partner’s education were associated with long-term fatality when adjusted for age at incidence and year of MI (Table 2, model 1). When additionally adjusted for other socioeconomic characteristics, women with partners with secondary or basic level education continued to show greater hazards compared with those with tertiary educated partners (model 2), such as the hazard ratio (HR) of 1.53 (95% CI, 1.22–1.92) for women with a partner with basic level education. For men, the estimated effect of partner’s education was smaller. The strongest association with fatality was found in having no partner in both women (HR, 1.82; 95% CI, 1.46–2.27) and men (HR, 1.68; 95% CI, 1.46–1.95) even after controlling for other socioeconomic characteristics. Supplementary analyses of long-term mortality focusing exclusively on cardiovascular deaths showed similar or slightly higher effect estimates (eTable 2, https://links.lww.com/EDE/B296, Supplemental Digital Content 2). Similar results were expected, as 65% in the long fatality period were cardiovascular disease deaths.
Figures 1–3 describe the results for the interactions for dummy variables of own and partner’s basic education focusing exclusively on individuals with a partner. The individuals with both basic education themselves and basic educated partners have the highest risks of MI incidence and short- and long-term fatality. However, the CIs overlap with couples in which only one has basic education. The MI risks for those who have basic education themselves, but more highly educated partners, and vice versa, appear similar. For MI incidence, there was evidence for a deviation from the additive effects of own and partner’s low education for men, which indicated that the hazard associated with partner’s basic education was less strong than expected in men with basic education (relative excess risk due to interaction for men −0.13, 95% CI, −0.24, −0.02; for women −0.04, 95% CI, −0.20, 0.12). The relative excess risk for short- and long-term fatality did not indicate substantial deviations from additivity (short-term fatality relative excess risk due to interaction for men −0.01, 95% CI, −0.13, 0.10; for women −0.07, 95% CI, −0.32, 0.18; long-term fatality relative excess risk due to interaction for men −0.07, 95% CI, −0.63, 0.49; for women 0.03, 95% CI, −0.36, 0.41), but the estimates for the interactions had low precision.
FIGURE 1: Myocardial infarction incidence by own and partner’s education. Note: analyses in partnered individuals, and model additionally adjusted for age, individual income, occupation, and employment status.
FIGURE 2: Post-myocardial infarction (MI) short-term fatality by own and partner’s education. Note: analyses in partnered individuals, and model additionally adjusted for age, year of MI, individual income, occupation, and employment status.
FIGURE 3: Post-myocardial infarction long-term fatality by own and partner’s education. Note: analyses in partnered individuals, and model additionally adjusted for age, year of MI, individual income, occupation, and employment status.
DISCUSSION
This study examined how the level of education of the marital or cohabiting partner—the closest and most intimate social connection that many people will have in their lives—was associated with MI incidence and survival. The results suggest important spillover benefits of high educational levels within couple dyads. In the Finnish context of the study, we found that a high educational level of a partner was associated with substantial health advantages irrespective of the level of one’s own education. The partner’s lack of qualification beyond basic education was associated with higher MI incidence and fatality risk net of an individual’s own education and other socioeconomic characteristics. Those without a partner had, however, even greater risks of MI incidence and shorter survival, a finding that emphasizes the public health needs of this population group.
An advantage of the large sample size of the study was that it enabled studying educationally heterogamous couples. There is a strong tendency toward educational homogamy (i.e., the tendency for people to marry others with similar educational backgrounds) in most societies including Finland.30,31 In our sample, 42% of individuals at baseline had incongruent education relative to their partner, translating to relatively few couples in specific combinations of different educational levels. The results indicated that there was some variation in the effects of partner’s education according to level of own education. For MI incidence, there was evidence of deviation from additive effects, which revealed that the gradient by partner’s education was greater particularly for men with higher education, similar to findings for CHD mortality in a Norwegian study.8 However, the effect in our study was small. Thus, the educational discrepancy hypothesis19–28 received some weak support in our data, but the estimated effects of one’s own and one’s spouse’s education seemed to be mostly additive. On the whole, the advantages of greater household-level educational capital seem to outweigh possible negative effects. The results are consistent with findings from other recent studies of CHD incidence5,33 and mortality.10–14 Earlier studies indicating associations with status discrepancy were situated in 1960–70s in United States when having a more educated wife was more unusual28 and more traditional gender role attitudes are likely to have prevailed. The cultural significance of differing educational levels within couples can be expected to have changed as women’s educational levels have begun to surpass those of men in many countries.
That a partner’s education—as well as having a partner at all—was more clearly associated with long-term than short-term fatality may suggest that a partner’s resources are of greater importance for general lifestyle modification and support, rather than accessing health care in the immediate aftermath of the MI. The results also suggested that partner’s education was more influential than own education for women’s long-term fatality. These findings run somewhat contrary to the suggestion that women have a greater role in shaping health-related behaviors within the household.18 The man’s educational resources may be associated with other mechanisms beneficial for the health of the female partner after MI incidence, which we could not study directly. Greater educational capital in the household may increase the instrumental support available for navigating the health care system and following medical treatments. It is possible that a highly educated man is in a better position to provide greater social support and exert more social control for making necessary behavioral adjustments. The man’s socioeconomic resources, such as income, may also be an important factor for women’s health due to disadvantages that women face in the labor market.6,14,34 This effect may be amplified if illness precludes participation in the labor market, as the partner’s high income can improve the financial stability of the household. We adjusted for the study subject’s own socioeconomic resources at baseline as a potential confounder of the association between partner’s education and the outcomes, but not the partner’s income, which may be an important mechanism.
We hypothesize that the educational resources of both partners influence the shared environment of the couple, and the material and psychosocial resources available in the household, which may in turn shape the health-related behaviors of both partners. A limitation of the study was that the specific psychosocial or behavioral mechanisms explaining differentials in MI incidence or post-MI survival could not be investigated. Further research is needed to understand the hypothesized mechanisms in detail. The greatest threat to causal inference is the unobserved factors preceding socioeconomic attainment that result in partner selection based on health-related behaviors or cognitive characteristics. Furthermore, we had to restrict the analyses of fatality to those in the sample who experienced a MI in the follow-up and, as such, there may be a risk of collider stratification bias compromising the identification of true causal effects of education for fatality.35
The tendency to educational homogamy and assortative partner selection is, however, likely to reinforce health inequalities. On the one hand, the resources of those closely socially connected can provide additional advantages for health, whereas lacking close social contacts is likely to put a person at a distinct disadvantage. Individuals living alone may experience cumulative disadvantage as they are excluded from the health-enhancing effects of marital and cohabiting relationships as well as access to the additional socioeconomic resources of partners18 and possibly those of adult offspring, who may provide additional material and nonmaterial resources.3,4 Intergenerational transmission of education from parents to offspring would be a relevant pathway in this respect and deserves attention in future studies. On the other hand, partnered households with low socioeconomic resources may also lack health-enhancing resources, and possibly experience a diffusion of risk behaviors and strain from health problems of other household members.36 In our study, the gradient of MI incidence when comparing educationally homogamous couples was around twice the size of that when only using own education, which echoes findings for self-rated health in Norway37 and the Netherlands.9 The strong accumulation of advantage in some households and disadvantage in others is likely to lead to diverging socioeconomic and health trajectories, and greater population level social inequalities in health.
The strengths of the study lie in the data, with which we could follow large, nationally representative cohorts for first MI incidence and the survival thereafter both in the short-term and the long-term. The Finnish hospitalization data for MIs has been found to be of good quality.38 Registers provide reliable information on socioeconomic characteristics with a low level of missing information. The strong correlation of own education and spousal education needs to be borne in mind when interpreting the mutually adjusted models as their collinearity can inflate standard errors and make estimates unstable. However, our estimates do not fluctuate much from the more simple models to the fully adjusted model and the CIs remain quite narrow. We also calculated Cramér’s V’s and variance inflation factors and found that their values were acceptable. Nevertheless, these models do not account for the effects of own and spousal education that are shared. The possibility of using detailed time-varying measures of partnership status and education and take into account changes in partnership situation was, however, an advantage. We were also able to include a measure of the partner’s education among individuals who were identified as cohabiting partners within the registers. However, as the proportion of cohabiters in these cohorts is still relatively small, we were not able to examine whether the effects of partner’s education differed between the married and the unmarried cohabiters. Nor was it possible to consider whether partnership duration modified the effects. We also could not identify same-sex couples in the data.
Overall, the study shows that partners and their educational resources are associated with health advantages that delay heart disease onset and improve survival. Moreover, due to increasing tendencies to marital homogamy particularly by characteristics such as education,39,40 these education spillover effects may be contributing to the widening of socioeconomic gaps in CHD incidence and survival.
REFERENCES
1. Aizer A, Currie JThe intergenerational transmission of inequality: maternal disadvantage and health at birth. Science. 2014;344:856–861.
2. Zimmer Z, Martin LG, Ofstedal MB, et alEducation of adult children and mortality of their elderly parents in Taiwan. Demography. 2007;44:289–305.
3. Torssander JFrom child to parent? The significance of children’s education for their parents’ longevity. Demography. 2013;50:637–659.
4. Friedman EM, Mare RDThe schooling of offspring and the survival of parents. Demography. 2014;51:1271–1293.
5. Bosma H, Appels A, Sturmans F, et alEducational level of spouses and risk of mortality: the WHO Kaunas-Rotterdam Intervention Study (KRIS). Int J Epidemiol. 1995;24:119–126.
6. Lillard LA, Waite LJ‘Til death do us part: marital disruption and mortality. Am J Sociol. 1995;100:1131–1156.
7. Martikainen PSocioeconomic mortality differentials in men and women according to own and spouse’s characteristics in Finland. Sociol Health Illn. 1995;17:353–375.
8. Egeland GM, Tverdal A, Meyer HE, et alA man’s heart and a wife’s education: a 12-year coronary heart disease mortality follow-up in Norwegian men. Int J Epidemiol. 2002;31:799–805.
9. Monden CW, van Lenthe F, de Graaf ND, et alPartner’s and own education: does who you live with matter for self-assessed health, smoking and excessive alcohol consumption? Soc Sci Med. 2003;57:1901–1912.
10. Jaffe DH, Eisenbach Z, Neumark YD, et alDoes one’s own and one’s spouse’s education affect overall and cause-specific mortality in the elderly? Int J Epidemiol. 2005;34:1409–1416.
11. Jaffe DH, Eisenbach Z, Neumark YD, et alEffects of husbands’ and wives’ education on each other’s mortality. Soc Sci Med. 2006;62:2014–2023.
12. Kravdal OA broader perspective on education and mortality: are we influenced by other people’s education? Soc Sci Med. 2008;66:620–636.
13. Skalická V, Kunst AEEffects of spouses’ socioeconomic characteristics on mortality among men and women in a Norwegian longitudinal study. Soc Sci Med. 2008;66:2035–2047.
14. Torssander J, Erikson RMarital partner and mortality: the effects of the social positions of both spouses. J Epidemiol Community Health. 2009;63:992–998.
15. Huijts T, Monden CWS, Kraaykamp GEducation, educational heterogamy, and self-assessed health in Europe: a multilevel study of spousal effects in 29 European countries. Eur Sociol Rev. 2010;26:261–276.
16. Brown DC, Hummer RA, Hayward MDThe importance of spousal education for the self-rated health of married adults in the United States. Popul Res Policy Rev. 2014;33:127–151.
17. Ross CE, Mirowsky J, Goldsteen KThe impact of the family on health: the decade in review. J Marriage Fam. 1990;52:1059–1078.
18. Umberson DGender, marital status and the social control of health behavior. Soc Sci Med. 1992;34:907–917.
19. Bruhn JG, Chandler B, Lynn TN, et alSocial characteristics of patients with coronary heart disease. Am J Med Sci. 1966;251:629–637.
20. Shekelle RB, Ostfeld AM, Paul OSocial status and incidence of coronary heart disease. J Chronic Dis. 1969;22:381–394.
21. Haynes SG, Eaker ED, Feinleib MSpouse behavior and coronary heart disease in men: prospective results from the Framingham Heart Study. I. Concordance of risk factors and the relationship of psychosocial status to coronary incidence. Am J Epidemiol. 1983;118:1–22.
22. Suarez L, Barrett-Connor EIs an educated wife hazardous to your health? Am J Epidemiol. 1984;119:244–249.
23. Carmelli D, Swan GE, Rosenman RHThe relationship between wives’ social and psychologic status and their husbands’ coronary heart disease. A case-control family study from the Western Collaborative Group Study. Am J Epidemiol. 1985;122:90–100.
24. Eaker ED, Haynes SG, Feinleib MSpouse behavior and coronary heart disease in men: prospective results from the Framingham Heart Study. II. Modification of risk in type A husbands according to the social and psychological status of their wives. Am J Epidemiol. 1983;118:23–41.
25. Szklo M, Tonascia J, Gordis LPsychosocial factors and the risk of myocardial infarctions in white women. Am J Epidemiol. 1976;103:312–320.
26. Talbott E, Kuller LH, Detre K, et alBiologic and psychosocial risk factors of sudden death from coronary disease in white women. Am J Cardiol. 1977;39:858–864.
27. Vernon SW, Buffler PAThe status of status inconsistency. Epidemiol Rev. 1988;10:65–86.
28. Matthews KACommentary: is an educated wife hazardous to her husband’s heart? Never, always, or sometimes? Int J Epidemiol. 2002;31:806–807.
29. Demey D, Berrington A, Evandrou M, et alPathways into living alone in mid-life: diversity and policy implications. Adv Life Course Res. 2013;18:161–174.
30. Davis MA, Neuhaus JM, Moritz DJ, et alLiving arrangements and survival among middle-aged and older adults in the NHANES I epidemiologic follow-up study. Am J Public Health. 1992;82:401–406.
31. Schmaltz HN, Southern D, Ghali WA, et alLiving alone, patient sex and mortality after acute myocardial infarction. J Gen Intern Med. 2007;22:572–578.
32. Andersson T, Alfredsson L, Källberg H, et alCalculating measures of biological interaction. Eur J Epidemiol. 2005;20:575–579.
33. Strogatz DS, Siscovick DS, Weiss NS, et alWife’s level of education and husband’s risk of primary cardiac arrest. Am J Public Health. 1988;78:1491–1493.
34. McDonough P, Williams DR, House JS, et alGender and the socioeconomic gradient in mortality. J Health Soc Behav. 1999;40:17–31.
35. Glymour MOakes JE, Kaufman JSUsing causal diagrams to understand common problems in social epidemiology. In: Methods in Social Epidemiology. 2006.San Francisco, CA: Jossey-Bass;
36. Umberson D, Crosnoe R, Reczek CSocial relationships and health behavior across life course. Annu Rev Sociol. 2010;36:139–157.
37. Nilsen SM, Bjørngaard JH, Ernstsen L, et alEducation-based health inequalities in 18,000 Norwegian couples: the Nord-Trøndelag Health Study (HUNT). BMC Public Health. 2012;12:998.
38. Pajunen P, Koukkunen H, Ketonen M, et alThe validity of the Finnish Hospital discharge register and causes of death register data on coronary heart disease. Eur J Cardiovasc Prev Rehabil. 2005;12:132–137.
39. Schwartz CRTrends and variation in assortative mating: causes and consequences. Annu Rev Sociol. 2013;39:451–470.
40. Mäenpää E, Jalovaara MAchievement replacing ascription? Changes in homogamy in education and social class origins in Finland. Adv Life Course Res. 2015;26:76–88.
