WHAT IS ALREADY KNOWN ON THIS TOPIC

Poor or deteriorating mental health is associated with an elevated divorce risk. Less is known about the impact of physical illness on union instability, and evidence on variation between social context and by gender and socioeconomic resources is particularly scarce.

WHAT THIS STUDY ADDS

Neurological conditions are associated with an increased divorce risk throughout the Nordic region. Divorce risk is particularly high if both spouses are ill, and in all countries and educational groups, it is either similar or larger for husband’s compared with wife’s illness.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

More support is needed for couples afflicted by neurological illness, even in comprehensive Nordic welfare states. Future studies should investigate neurological conditions and subsequent divorce risk in contexts with less developed welfare state arrangements than the Nordic countries.

Introduction

Poor mental health has been shown to increase divorce risk,1 2 and poor physical health might have a similar effect.3–7 However, there seems to be variation between illnesses. While register-based studies have not found an association between cancer and divorce risk,8–10 other severe diseases such as multiple sclerosis (MS) and stroke, which often lead to functional disabilities, have been shown to increase union instability.11–13

Neurological conditions are a major cause of disability in Europe14 and the USA,15 and a recent register-based study among Finnish middle-aged couples observed an increase in divorce risk following any neurological illness.9 Given the high prevalence of neurological conditions, the impact on union stability may be considerable at the population level. This effect might be mediated through multiple mechanisms, including changes in psychological distress, family responsibilities, participation in paid work and financial strain.4 16–19 Men are often assumed to experience larger resource losses following illness and disability because of their traditional breadwinner role,18 20 and a husband’s illness has thus been expected to increase divorce risk more than a wife’s illness. Moreover, studies of caregiving suggest that the risk of divorce is higher when the wife is the caregiver.21 Yet, previous empirical findings are mixed. Some studies report that a husband’s neurological illness is more strongly associated with union instability,9 11 while other studies suggest that illness of the wife increases instability more.3 13

According to social exchange theory,22 the partner who has more socioeconomic resources might be more likely to initiate divorce at a time of marital discord, because the expected economic loss for him/her is smaller.23 Both education and illness may affect current earnings as well as future earnings potential, and a healthy spouse with intermediate education might thus be inclined to leave an ill partner with only compulsory education but be more likely to stay with an ill partner with higher education. If unhealthy partners are more likely to avoid divorce compared with healthy partners, couples with two ill partners might be more stable than couples with only one ill partner. However, if only the burden of illness drives the decision to leave, then couples with two ill partners should be particularly prone to divorce.

In this study, we use population-wide register data from four Nordic countries—Denmark, Finland, Norway and Sweden—to examine the association between neurological conditions and subsequent divorce risk and if it depends on gender and the educational resources of both spouses. All Nordic countries have relatively generous welfare state arrangements, where healthcare and social insurance (eg, income maintenance schemes) are mainly financed through taxes. The welfare state is intended to offer sufficient protection against the social consequences of illness,24 suggesting only weak associations between neurological conditions and divorce risk in the Nordic countries. The countries are culturally quite similar. Individualistic family values are common,25 which is highlighted by the no-fault divorce legislation that does not require consent of both spouses. The countries are also considered forerunners in gender equality,26 27 and education is free of charge and/or heavily subsidised. As the welfare state strongly emphasises equality,28 any potential effect of illness on divorce risk is expected to be similar following husband’s and wife’s illness, irrespective of their educational resources. Examining whether an association between neurological illness and divorce risk exists, and whether and how it is modified by gender and education, provides an opportunity to evaluate how well the Nordic welfare state model lives up to these expectations.

Methods

We used longitudinal linked register data from four Nordic countries: Denmark, Finland, Norway and Sweden. The data included the total population aged 30–64 years old during the years 2007–2016 for Denmark and Sweden, and during the years 2008–2017 for Finland and Norway. Younger couples were excluded as many of them are still studying and living in short-term non-marital unions, whereas those over 64 years old are mainly retired and divorces after prime working age are still quite uncommon.29 Information on marriages and divorces, as well as sociodemographic factors, was derived from population registers, recorded annually. Data on neurological conditions were derived from patient registers. The use of these data for research purposes was approved by the register holders in each country (see online supplemental table 1 for specific data sources and permissions).

We followed a cohort of couples who were married and living together at the end of year 2006 (Denmark and Sweden) or 2007 (Finland and Norway) and observed neurological conditions and end of marital cohabitation—henceforth referred to as divorce—until the end of year 2016 (Denmark and Sweden) or 2017 (Finland and Norway). We did not assess whether the union was later re-established. The data permission for Finland required the exclusion of same-sex couples. Furthermore, the number of same-sex couples with neurological conditions was too low for statistical analyses in Denmark, Norway and Sweden. In the Finnish data, only couples with both spouses living in Finland since the age of 16 were included because information on educational qualifications obtained abroad was mainly missing.

We used Cox proportional hazards (PH) regression models to examine how neurological conditions were related to divorce risk. Censoring occurred at emigration, death of either spouse or the end of follow-up. The year of the first outpatient visit to specialised care or inpatient care episode with an International Classification of Diseases 10th Revision code G00–G99 was identified as the first incidence of a neurological condition. Among men, sleep disorders (G47) were the most common diagnosis in Denmark (24.9%), Finland (41.5%) and Norway (45.2%) (online supplemental table 2). For Sweden, we did not have information on the specific diagnoses, but within the age group included in this study, sleep disorders were the most common neurological diagnosis among men (22.7%) in publicly available data (online supplemental table 3). Among women, mononeuropathies of the upper limb (G56) were the most common diagnosis in each country (from 16.6% of diagnoses in Norway to 29.3% in Denmark).

Neurological conditions of both spouses were used as a time-varying explanatory variable, categorised as none, wife only, husband only and both spouses. The results are presented as HRs with 95% CI. In the first model, we adjusted for the mean age (in years) and mean age squared of both spouses, the absolute and squared age difference of spouses (in years), and the duration of marriage (categorised as 0–4 years, 5–9 years and 10 years or more) to allow for non-linear changes in divorce risk by mean age, spousal age difference and marital duration. In the second model, we also adjusted for the presence of coresident children (categorised as no children, youngest aged 0–3 years, 4–6 years, 7–15 years and 16 years or older) and educational resources of the couples. Education was categorised in two steps. First, we identified the level of education of each spouse in three levels: compulsory (International Standard Classification of Education (ISCED) categories 0–2), intermediate (ISCED 3–4) and high (ISCED 5+). Second, we constructed three categories based on both spouses’ education: husband has higher education than the wife; both spouses have the same level of education; wife has higher education than the husband. In the third model, we further adjusted for household disposable income (operationalised as combined disposable income of all household members divided by the Organisation for Economic Co-operation and Development’s modified scale of household consumption units and classified into annual quintiles within the sample).

Next, in Denmark, Finland and Norway, we repeated these analyses for three specific neurological conditions: sleep disoders (G47), mononeuropathies of upper limb (G56) and MS (G35), that is, three common conditions that also differ in expected levels of disability. Finally, we analysed all neurological conditions stratified by the educational resources of the couples, classified as described above. All covariates were measured annually and treated as time-varying in the Cox models. The PH assumption was assessed both graphically and using Schoenfeld residual tests. Our main variable of interest—neurological conditions—satisfied the PH assumption, but there were some violations for other covariates. In these cases, we verified the robustness of our findings using models stratified by the covariates in question.

To formally compare results between countries, we used meta-analysis under the assumption that our analyses in four countries define the entire Nordic population of interest, and that the effect sizes describing the magnitude of associations between illness and divorce risk differ between countries.30 Since the study designs were identical in each country, statistical heterogeneity—the variation in effect size between countries that exceeds sampling variability—was assumed to result from true differences in effect sizes between countries.31 Heterogeneity was tested with Cochran’s Q.30 Due to national data protection legislation, individual-level register data from different countries could not be directly combined.

We also performed three robustness checks. First, to take into account that divorce risk might depend on time since illness onset (ie, divorce risk among couples diagnosed before follow-up could differ from those diagnosed during it), we followed couples who were married and living together at the end of the year 2008/2010 (Denmark and Sweden) or 2009/2011 (Finland and Norway) where neither spouse had received hospital care for neurological conditions during the preceding two/four previous years. Second, because non-marital cohabitation is increasingly common in the Nordic countries, we also included non-married cohabiting couples in Denmark, Finland and Norway and repeated the analyses. Information on cohabitation was not available for Sweden. Third, we used Norwegian data to examine whether excluding immigrants changed the results since we could only include non-immigrant couples in Finland. All analyses were performed using Stata (StataCorp. 2019. Stata Statistical Software: Release V.17, StataCorp).

Results

We followed altogether 2 809 209 married couples for neurological conditions and subsequent divorce (table 1, see also online supplemental table 4). During the 10-year follow-up period, about 25% of couples in Norway and Sweden, 23% in Finland and 15% in Denmark had at least one spouse with a neurological condition. The cross-national differences are partly due to variation in the prevalence of sleep disorders, the most common diagnosis among men in each country (online supplemental tables 2 and 3). In all four countries, neurological conditions were almost as common in women (7%–11%) as in men (7%–11%) and the proportion of couples where both spouses had a neurological condition was low (1%–2%) (table 1). The proportion of couples who divorced during the 10-year follow-up period varied from 10.6% in Sweden to 13.4% in Denmark.

Table 1

Characteristics of all married couples*

Neurological conditions were associated with an increased risk of divorce in all four Nordic countries (table 2). The excess divorce risk was particularly pronounced if both spouses were ill, with HRs from model 1 (adjusted for age and duration of marriage) ranging from 1.32 (95% CI 1.27 to 1.37) in Sweden to 1.58 (95% CI 1.44 to 1.73) in Denmark. In Sweden, there was no gender difference in excess divorce risk, while in Finland and Norway, a husband’s illness was associated with somewhat higher divorce risk compared with wife’s illness. In Denmark, this gendered pattern was more pronounced and couples with an ill husband had clearly higher divorce risk (HR 1.40, 95% CI 1.36 to 1.45)) compared with couples where the wife was ill (1.22 (95% CI 1.19 to 1.26)). Adjustment for coresident children and educational resources did not substantially change the main results. Further adjustment for family disposable income did not alter the results either, except for somewhat attenuated HRs in Sweden and for Danish couples where both spouses fell ill.

Table 2

HR of divorce by incidence of neurological conditions; all married couples*

For the four Nordic countries pooled, the weighted average effect sizes were 1.38 (95% CI 1.34 to 1.42) for illness of both spouses, 1.27 (95% CI 1.25 to 1.29) for illness of the husband and 1.21 (95% CI 1.20 to 1.23) for illness of the wife in model 1 (adjusted for age and duration of marriage) (figure 1). There was evidence of significant variation between countries in the effect size of both spouses having a neurological condition (Q=14.49, p=0.00). The effect size of husbands’ illness also varied between countries (Q=56.32, p=0.00), whereas the effect size of wives’ illness seemed to be more similar across countries (Q=10.59, p=0.01).

Figure 1

Weighted average divorce risk (HR) across the four Nordic countries by incidence of neurological conditions.

In Denmark, Finland and Norway, we could examine the associations between specific neurological conditions and divorce risk. Sleep disorders, mononeuropathies of the upper limb and MS were all associated with the risk of divorce, but the magnitude of the associations varied (table 3). For all three conditions, the risk of divorce was either similar or larger following husband’s illness as compared with wife’s illness. For sleep disorders, no gender difference could be observed. For MS, the risk of divorce was clearly larger following husband’s than wife’s illness, particularly in Denmark. For mononeuropathies of the upper limb, the excess divorce risk was smaller compared with the other two specific conditions.

Table 3

HR of divorce by incidence of sleep disorders, mononeuropathies of upper limb and multiple sclerosis; all married couples*

Turning to the results stratified by education, the majority of couples had the same level of education in all four countries (table 1). In Finland and Sweden, it was more common that the wife had a higher level of education than the husband, whereas in Denmark and Norway the proportions were roughly similar for more educated wife and husband, respectively. In Finland, husbands’ illness and wives’ illness were similarly associated with divorce risk when the husband had a higher level of education than the wife (figure 2). If both spouses had the same level of education, husbands’ illness was associated with somewhat higher divorce risk than wives’ illness. Moreover, when the wife had a higher level of education than the husband, husbands’ illness was associated with clearly higher divorce risk than wives’ illness. In the other Nordic countries, a similar pattern was not found. In Denmark and Norway, husbands’ illness was more strongly associated with divorce risk than wives’ illness irrespective of the educational resources of the couple. However, the excess divorce risk following husbands’ illness appeared slightly smaller among Danish couples where the husband had a higher educational level than the wife. In Sweden, no gender differences were observed within any educational group. Adjustment for coresident children and family disposable income had negligible impact in Denmark, Finland and Norway (online supplemental table 5). In Sweden, the HRs of illness were reduced by about 50% in each educational category after adjustment for household income.

Figure 2

Divorce risk (HR) in the four Nordic countries by incidence of neurological conditions and educational level.

We conclude this section with robustness checks. When only couples healthy at baseline were included, all results remained similar except for a slightly smaller excess divorce risk among Norwegian couples where both spouses developed a neurological condition during the follow-up period (online supplemental tables 6–8). Including non-married cohabiting couples in the samples in Denmark, Finland and Norway had negligible impact on the results (online supplemental tables 9 and 10). Finally, in Norway, the exclusion of immigrant couples from the sample did not change the results, except that when both spouses were ill, the excess divorce risk was slightly higher (online supplemental table 11). However, due to the small number of observations in or this illness category, the CIs were quite wide.

Discussion

Using population-wide register data on 2 809 209 couples in 4 Nordic countries, this study examined the association between neurological conditions and marital divorce risk during a 10-year follow-up period. We found that in all countries—Denmark, Finland, Norway and Sweden—the risk of divorce was elevated among couples where at least one spouse had a neurological condition, and even more so if both spouses were ill. The weighted average effect sizes (HRs) were 1.27 (95% CI 1.25 to 1.29) for husbands’ and 1.21 (95% CI 1.20 to 1.23) for wives’ illness, and 1.38 (95% CI 1.34 to 1.42) for couples where both spouses had a neurological condition. In the Nordic region, divorce risk was thus slightly higher following husbands’ than wives’ illness. The effect sizes are similar to the findings of an earlier study among middle-aged Finnish marital and non-marital couples.9

A husband’s illness is expected to result in more substantial losses in household income,16 18 which may lead to the husband’s illness having a stronger influence on divorce risk.4 Some previous research supports this expectation, as work disability increased divorce risk among US middle-age men18 but not women.20 However, another study did not find any association between men’s work disability and divorce risk,16 and among older US couples, disability of wives but not husbands predicted increased risk of divorce.5 In the current study, husband’s illness was more strongly associated with divorce risk than wife’s illness following MS, which is a severe neurological condition associated with substantial functional disability. For less severe conditions such as sleep disorders, gender differences were smaller or non-existent. However, sleep disorders among husbands and wives were quite strongly associated with an elevated divorce risk. Thus, despite the expectation that effective treatment of sleep disorders benefits both spouses, they still seem to adversely affect relationship quality.32 33

Interestingly, the association between neurological conditions and divorce risk remained nearly unchanged after adjustment for household income, suggesting that changes in income did not explain the association. The exception to this pattern was Sweden where adjustment for income attenuated the associations by up to 50%. Sweden offers easier access to specialised healthcare than other Nordic countries,34 but there is a lower net replacement rate for earnings in Sweden than in Denmark, Finland and Norway during shorter sickness periods,35 and income in Sweden may, therefore, in part reflect variation in disability among those with a neurological condition.

In Denmark, divorce risk was particularly high following a husband’s illness. While husband’s severe illness was more strongly associated with divorce risk than wife’s severe illness also in Finland and Norway, this gender difference was by far most pronounced in Denmark. Furthermore, the proportion of couples afflicted by neurological conditions was the lowest in Denmark, suggesting that we might be measuring conditions that are on average more severe than in the other three countries. Sleep disorders, the most common neurological diagnosis among men, were clearly more common in Finland and Norway, as compared with Denmark. In all these countries, the association between sleep disorders and divorce risk seemed independent of the gender of the ill spouse.

According to social exchange theory,22 the excess divorce risk after husbands’ illness should have been especially large among couples with more educated wife since she in these cases would have more favourable options outside of the union. In other words, among couples with a higher educated wife than the husband, husbands’ neurological illness may lead to declining partnership rewards for her and make continuing in a union less likely. However, only findings for Finland were consistent with this expectation. In Denmark and Norway, husbands’ illness was consistently associated with higher divorce risk than wives’ illness in all educational categories. In Sweden, no clear gender differences could be observed. Although all Nordic countries rank high in gender equality, Sweden is the most egalitarian when it comes to wages and parental leave schemes.36 37 Swedish fathers are the most active in taking parental leave,37 suggesting a more equal division of responsibilities within families. Finland, on the other hand, has the largest wage gap between men and women,36 and parental leave use is highly gendered, combined with a relatively large proportion of mothers with young children staying at home.36 38 39 Thus, gender inequality in wages and parenting might not only be detrimental to women but also exacerbate the adverse social consequences of illness for men. Furthermore, the large risk of divorce following illness of both spouses contradicts the idea that spouses in poor health themselves would be particularly prone to avoid divorce. In contrast, it seems that the burden of illness within the couple drives divorce risk, a burden that is larger when both spouses are ill.

Although our study is unique in following all married couples in four countries using high-quality longitudinal register data, it is not without limitations. We have only examined opposite-sex couples, and the association between neurological conditions and divorce risk might be different among same-sex couples, particularly among female couples who have an overall higher divorce risk.40 To keep the study design similar across countries, we had to estimate relatively simple models with a limited number of variables. We had no information on severity of disease, associated disability or needs for caregiving, and in Sweden, we had no information on the detailed diagnoses. Nevertheless, our findings show a clear association between neurological conditions and the risk of subsequent divorce in comprehensive Nordic welfare states, a context expected to limit the adverse social consequences of illness.24

Data availability statement

Data may be obtained from a third party and are not publicly available. The data used in this study were collected by register authorities and are not publicly available. Those interested may apply for permission to use these data for scientific research from the register holders.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

This study involves human participants and was approved by Ethics Committee of Statistics Finland (TK/32/07.03.00/2020), Findata (THL/3154/14.06.00/2022), the Danish Data Protection Agency (UCPH reference number: 514-0230/18-3000), The Norwegian Agency for Shared Services in Education and Research (872223) and The Regional Committees for Medical and Health Research Ethics Norway (2017/1297), and the Central Ethical Review Board (Dnr Ö 25-2017) in Sweden. The use of register data for purposes of scientific research carried out in public interest did not require informed consent from participants.

Acknowledgments

We thank Riikka Sallinen at the Population Research Unit, Faculty of Social Sciences, University of Helsinki and Lise Kristine Højsgaard Schmidt at the Section of Epidemiology, University of Copenhagen, for assistance with data set-up during the initial phase of the project.