Article Text

Download PDFPDF

Marital status, living arrangement and mortality: does the association vary by gender?
  1. Katharina Staehelin1,2,
  2. Christian Schindler1,2,
  3. Adrian Spoerri3,
  4. Elisabeth Zemp Stutz1,2,
  5. for the Swiss National Cohort Study Group
  1. 1Department of Epidemiology and Public Health, Swiss Tropical and Public Health Institute, Basel, Switzerland
  2. 2University of Basel, Basel, Switzerland
  3. 3Institute of Social and Preventive Medicine, University of Bern, Bern, Switzerland
  1. Correspondence to Dr Katharina Staehelin, Swiss Tropical and Public Health Institute, associated Institute of the University of Basel, Switzerland, Socinstrasse 57, 4051 Basel, Switzerland; katharina.staehelin{at}unibas.ch

Abstract

Background Men appear to benefit more from being married than women with respect to mortality in middle age. However, there is some uncertainty about gender differences in mortality risks in older individuals, widowed, divorced and single individuals and about the impact of living arrangements.

Methods Longitudinal data with 1990 census records being linked to mortality data up to 2005 were used (Swiss National Cohort). The sample comprised all residents over age 44 years in Switzerland (n=2 440 242). All-cause mortality HRs for marital status and living arrangements were estimated by Cox regression for men and women and different age groups with adjustment for education and socio-professional category.

Results The benefit of being married was stronger for men than for women; however, mortality patterns were similar, with higher mortality in divorced and single individuals compared with widowed individuals (<80 years). After adjustment for living arrangements, the gender difference by marital status disappeared. Stratification by living arrangement revealed that mortality risks were highest for 45–64-year-old divorced (HR 1.72 (95% CI 1.67 to 1.76)) and single men (HR 1.67 (95% CI 1.63 to 1.71)) who lived alone. In women of the same age, the highest mortality risk was observed for those who were single and living with a partner (HR 1.70 (95% CI 1.58 to 1.82)). In older age groups, the impact of marital status decreased.

Conclusions Evaluation of living arrangements is crucial for identifying and explaining gender differences in mortality risks by marital status. The impact of living alone and living with a partner seems to be different in men and women.

  • Gender
  • marital status
  • living arrangement
  • mortality
  • census
  • cohort ME
  • epidemiology FQ
  • gender studies SI

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

The association of marital status with mortality has been investigated repeatedly, and studies have consistently shown a beneficial effect of being married.1–10 Also living arrangements have been found to be related to mortality with a longer survival of individuals living with a partner or not living alone, even though not consistently.2 11–15 Marital status and living arrangement are measures of social relations.2 The protective effect on survival mainly refers to the provision of social and economic support and to the social control of health behaviour (social causation theory).2 10 16–18 Furthermore, the selection of healthier individuals into marriage or partnership plays a role (selection theory).10 17–19

Previous research indicates that the protective effect of marriage and living with a partner as well as the harmful effect of living alone differ between men and women. Middle-aged men seem to be more strongly protected by marriage than women.3 5 6 9 10 Gender differences in the elderly population and in mortality patterns across the subcategories of marital status are less clear. There are conflicting findings across studies which unmarried marital status incurs the greatest mortality disadvantage in men and women and whether there are gender differences.1 3–5 7–10 20 21 27 The advantage of being married seems to decrease in older age groups as well as the gender differences.4 8–10 Inconsistent associations have also been described for living arrangements. Living alone and not living with a partner have been found to be risk factors for death only in men12 14 15 or in both genders with a stronger effect in men.13 However, other studies could not find an effect of living alone on mortality.2 11

In the last decades, substantial transformations in family forms have occurred with increasing proportions of single and divorced individuals, of individuals living alone and of consensual unions.22 23 As a consequence, discrepancies between official marital status and actual living situation have grown. Therefore, a classification that incorporates both marital status and living arrangement is appropriate to more accurately reflect the real social situation of an individual.13

However, to date, these variables mostly have been investigated separately.14 15 Little is known about the impact of living arrangements on mortality of men and women across the different marital status groups. For example, it is not clear whether the effect of living with a partner or living alone differs in widowed, divorced and single men and women and whether gender differences in the impact of living arrangement alter mortality differentials by marital status between men and women. Using a combination of these two variables allows to detect gender-specific mortality patterns according to living arrangements within the subcategories of marital status and to identify special groups at risk.

The aim of this study therefore was to investigate differences in the association of marital status with mortality between men and women in consideration of the effective living arrangement across different age groups.

Methods

Study population

This study was based on data from the Swiss National Cohort, a longitudinal national database with linked census and deaths records.24 25 The core cohort consists of 6.874 million residents of Switzerland who participated in the 1990 census. Participation in the census is mandatory in Switzerland, and enumeration is nearly complete.26 Deterministic and probabilistic record linkage was performed to link a 1990 census record to either a 2000 census record, a death record or an emigration record. No satisfactory link could be found for 6.9% of the census and 6.7% of the mortality records, being mostly young individuals (<30 years), which are not included in the present study.24 25

At the time of our analysis, the database included follow-up data until 31 December 2005. The population for this study was restricted to men and women of at least 45 years of age at the time of the 1990 census.

The Swiss National Cohort was approved by the cantonal ethics committees of Bern and Zurich.

Variables

The outcome variable was time to death from all causes. The exposure variables marital status and living arrangement were obtained from the census in 1990 as baseline information. Marital status was used as a binary (married vs non-married) and as a categorical variable: married (including separated), widowed, divorced and single. Living arrangement was categorised as follows: living with the partner, living with others but not the partner (children, parents, relatives, non-relatives) and living alone. Individuals living in a collective household (eg, nursing homes, prisons, monasteries; n=93 085) and individuals for who no information on the household was available (administrative household; n=2518) were included in the analyses but not always displayed in the results.

The following variables were used as covariates. Education was classified into primary education only (compulsory school or less, other education, education not mentioned), vocational education, tertiary education and graduate school (university). Socio-professional category refers to profession and employment status (upper management, self-employed, graduates, intermediate professions, qualified non-manual workers, qualified manual workers, unskilled workers, not allocable, unemployed, domestic work, retired). Further covariates were nationality (Switzerland, Europe, others), language region (German-, French-, Italian-speaking cantons) and religious affiliation (Protestant, Catholic, no denomination, others, not specified).

Statistical methods

Cox proportional hazard regression models were computed separately for men and women. The survival analysis was based on the time interval between 5 December 1990 and 31 December 2005. All analyses were stratified by age either in 5-year or three broad age groups (45–64 years, 65–79 years and 80+ years) and controlled for the level of education and socio-professional status (only <65 years).

To test the extent to which age (as a continuous variable), socioeconomic factors (education, socio-professional category), cultural factors (nationality, language region, religious affiliation) and living arrangement account for mortality differences by marital status in men and women, different models were considered with the introduction of these factors in a stepwise fashion.

To examine mortality differentials by living arrangement across the different marital status categories in men and women, HRs for a combined variable were calculated with the married individuals living with a partner as the reference group. We also explored interaction between marital status and living arrangement separately by gender and age classes using the category of married subjects of the respective gender as reference. Some interaction terms were statistically significant, but according to the Bayesian information criterion, the models with interaction terms were not better than the ones without. Therefore, the results of the interaction analyses are not presented in the paper.

The proportional hazards assumption was not fulfilled for age, education, socio-professional category, nationality and living arrangement, as tested by the Schoenfeld residuals. These variables were therefore controlled by stratification of the baseline hazard function. The proportional hazards assumption was also violated for marital status, meaning that the HRs for widowed, divorced and single as compared with married individuals changed during the observation time. Therefore, the results for this factor must be interpreted as the mean HRs over the observation period.

Data were analysed using Stata V.10 (StataCorp LP) software. The results are presented as HRs with 95% CIs.

Results

The study population consisted of 1 114 945 men (46%) and 1 325 297 women (54%). The women were on average older than the men (63 vs 60 years). The age range was 45 to 109 years. The mean observation and survival time were 12.0 and 7.3 years for men and 12.5 and 7.5 years for women, respectively. The total amount of person years was 13 383 974 with 388 887 recorded deaths for men and 16 566 825 with 415 929 recorded deaths for women.

Table 1 describes the characteristics of the study population by marital status, age and gender. In the youngest age class (45–64 years), 83% of men and 74% of women were married at the baseline year. These proportions were substantially lower in the higher age groups, especially among women (80+ years: 15% of women married vs 59% of men). Crude death rates were higher among men than among women in all subcategories of marital status and all age groups. The educational level was generally higher in men than in women. High education was most prevalent among married men and among single women. Gender differences were also found for employment status (45–64 years): married men and divorced women were most frequently employed.

Table 1

Characteristics of the study population at baseline 1990, by marital status, gender and age

In all three age groups, most married men and women lived with their partner (87%–97%) (figure 1). Among widowed, divorced and single men and women, the most prevalent living arrangement was living alone. Among the non-married individuals, women more often lived alone compared with men, whereas men more often lived with a partner compared with women. Considerably more non-married than married individuals (all age groups) and more women than men (80+ years) lived in collective households.

Figure 1

Living arrangements within marital status categories at baseline 1990, by gender and age.

Mortality differentials by marital status

In figure 2, the HRs for non-married compared with married men and women are displayed in 5-year age groups. An advantage of being married was observed in men and women. With increasing age, the impact of marital status decreased. In non-married men, the HR was 1.87 (95% CI 1.80 to 1.92) for ages 45–49 and it decreased to HR 1.09 (95% CI 1.07 to 1.11) for ages 80–84. In non-married women, the respective HRs were 1.65 (95% CI 1.57 to 1.72) and 1.00 (95% CI 0.99 to 1.02). The beneficial effect persisted up to the highest age group in men, while in women above 80 years, there was no difference between the mortality risk of married and non-married individuals. The excess mortality of not being married was significantly higher in men than in women, except for the highest age group (90+ years).

Figure 2

Mortality HRs for non-married men and women compared with married men and women, in 5-year age groups. Controlled for age, education (all age groups) and socio-professional category (only age groups <64 years).

A benefit of marriage could also be seen in comparison with all other subcategories of marital status up to the age of 80 years (table 2). When adjusting for age and socioeconomic factors (model 3), excess mortality among widowed, divorced and single relative to married individuals was considerably higher in men than in women under the age of 80 years. However, when additionally controlling for cultural factors and living arrangements, gender differences in mortality by marital status disappeared, except for 45–65-year-old widowed men and women (model 5). Furthermore, the relative mortality risks of non-married individuals compared with married ones decreased substantially in men and women, but marital status remained an independent predictor of mortality in both sexes. The highest mortality risks in 45–64-year-old individuals were observed for divorced men and women (HR 1.47 (95% CI 1.44 to 1.51) and HR 1.46 (95% CI 1.41 to 1.52), respectively) and single men and women (HR 1.41 (95% CI 1.38 to 1.45) and HR 1.42 (95% CI 1.36 to 1.47), respectively). The risk was considerably less elevated among widowed men and women (HR 1.32 (95% CI 1.26 to 1.37) and HR 1.20 (95% CI 1.15 to 1.25), respectively). In older age groups, divorced individuals (65–79 years) as well as widowed men and widowed and divorced women (80+ years) had the highest mortality risks. Cultural factors had only marginal effects on mortality risks of men and women in the different marital status categories (model 4).

Table 2

Unadjusted and adjusted analyses for the association between marital status and mortality

Mortality differentials by living arrangement within the subcategories of marital status

When exploring the combined variable, large mortality variations by living arrangement within the different marital categories could be observed (figure 3). This was especially pronounced among non-married individuals and younger age groups. For example, mortality differentials within 45–64-year-old divorced men ranged from HR 1.41 (95% CI 1.36 to 1.47) for those living with a partner to HR 1.72 (95% CI 1.67 to 1.76) for those living alone.

Figure 3

Mortality HRs for different living arrangements, by marital status, gender and age. Controlled for age, education (all age groups) and socio-professional category (only 45–64 years).

Different mortality patterns were observed in men compared with women. Among 45–64-year-old widowed, divorced and single men, mortality risks increased gradually from men living with a partner over men living with others to men living alone. In women of the same age, this pattern was only seen for divorced women. Among widowed women, the living situation had no impact on mortality, whereas in single women, living with a partner was associated with the highest mortality risk. The overall highest mortality risks among 45–64-year-old individuals were observed for divorced and single men living alone (HR 1.72 (95% CI 1.67 to 1.76) and HR 1.67 (95% CI 1.63 to 1.71), respectively) and for single women living with a partner (HR 1.70 (95% CI 1.58 to 1.82)). Among 65–79-year-old men and women, these patterns were weaker, and over the age of 80 years, they were no longer demonstrated. The harmful effect of living alone was reversed and became protective, especially in women over the age of 80 years (HR for single women living alone 0.92 (95% CI 0.90 to 0.94)).

Discussion

In this large census-based study, we found a protective effect of marriage for men and women, with the largest impact in middle age. With increasing age, there was a decrease of the impact of marital status on mortality in men and women. The benefit of being married was stronger for men than for women, but patterns of mortality risks across the different subcategories of marital status were similar with divorced and single individuals having had the highest mortality up to the age of 80 years. However, the gender differences in mortality by marital status could largely be explained by the differential impact of living arrangement in men and women. A harmful effect of living alone compared with living with a partner could be shown for non-married middle-aged men. This penalty was weaker or even reversed in women of the same age. Additionally, different risk groups could be identified: Divorced and single men living alone and single women living with a partner were most likely to die in the age groups below 80 years.

The present study confirms the well-established finding of a beneficial effect of marriage on mortality in men and women.1–10 Our investigation also adds strong evidence to clarify previous inconsistent findings about gender differences in mortality across the subcategories of marital status1 3–5 7–10 20 21 27 and in the association of marital status with mortality in older age groups.4 8–10 The inconsistencies in previous studies may partly be due to differences in sample sizes, age ranges, adjustment factors and cultural environments. Studies performed with elderly populations8 and small sample sizes14 15 were less likely to find significant associations or gender differences. Our study comprised the largest sample by far and we considered middle-aged as well as elderly individuals.

The unique contribution of the present study is the detection of differences in the impact of living arrangement on mortality by marital status between men and women. We could show that the more detrimental effect of living alone in men than in women can largely explain the higher mortality of non-married men compared with non-married women. To our knowledge, this is a new finding since most studies so far did not investigate marital status and living arrangement in the same model.14 15

The highest mortality risks were observed for divorced and single men living alone and for single women living with a partner up to the age of 80 years. These findings are of public health relevance because it becomes more prevalent to remain single, to get divorced, to cohabit without formal marriage and to live alone in middle and older age.22 23

Underlying reasons for the high mortality risk of the mentioned groups possibly relate to social causation.2 6 10 13 16–18 28 29 In men, the lack of social control and social support may play a role. Unmarried men and men living alone have been shown to demonstrate poorer health habits and to be more likely to die from alcohol-related and external causes of death.1 3 5 9 13 16 Men also more often than women rely on their partner as their only confident and primary source of emotional and social support possibly resulting in a greater vulnerability to marital dissolution and to living alone.17 18 29 30 In women, on the other hand, economic support seems to be more important for the benefit of marriage.1 18 29 The negative effect of living with a partner in single women may therefore partly be explained by a lack of financial security compared with married women. Furthermore, single women are less likely to have children, which has been shown to be associated with mortality in women.13 20 Women may also benefit less from living with a partner because of their greater sensitivity to negative marital interactions.18

In elderly populations, the impact of health-promoting mediators such as social and economic support and social control of health behaviour may diminish, which might explain the weaker associations of marital status with mortality in older age.4 8–10 Conversely, the health status becomes more important with ageing. For example, living alone at older ages requires a certain level of health and functional ability.28 31 Nevertheless, despite the small relative differences in mortality in older age, there is a large impact in absolute terms,32 also with respect to gender. A greater number of older women are widowed and live alone than men.22 23 28

More research is needed to fully understand the underlying reasons for the differential impact of living arrangements in men and women. Particularly, the high mortality of middle-aged single women living with a partner was a surprising result and to our knowledge has not been observed so far. This group probably is very heterogeneous with regard to education, employment status, financial resources and social relationships. This result therefore should be interpreted with some caution and needs confirmation.

Also selection effects have to be considered. Individuals with higher than average risk of death have a lower likelihood to become or remain married, which might partly explain the lower mortality risk of married subjects (selection theory).10 17–19 This mechanism may differ between men and women. Indeed, single men in our sample were the least educated, whereas the situation was just the contrary for single women. This pattern points to education-based selection into marriage.9 Therefore, adjusting for socioeconomic factors partly accounts for such selection. Additionally, unhealthy married men appear to have larger divorce chances than respective women.19 However, based on the lack of health information in our data set, we could not account for gender differences regarding health selection into marriage.

Our study has the following limitations: information on marital status and living arrangement was measured at baseline. This may have involved a potential misclassification bias due to changes in marital status or living arrangement over time. In fact, divorced and widowed men were more likely to remarry during observation time; married women, on the other hand, more frequently became widowed (data not shown). These changes may have led to an underestimation of the gender differences in middle age in our study.

A further limitation refers to uncontrolled confounding. Wehad no information on health, health behaviour, income, children, social relationships or marital quality. These factors are associated with marital status and have been shown to have a differential impact on mortality in men and women.1 3 5 9 13 16–18 20 28–31 33 However, adjustment for health nd behavioural factors did not significantly alter the results of the association between living arrangement and mortality in several studies.12 14 15

We believe that these weaknesses are balanced by the strengths of this study, which are the large sample size, the long follow-up duration, the broad age range and the unselected census-based sample. Also individuals living in institutions were included, which eliminates the potential bias of a disproportionate selection of non-married individuals with good health in private households compared with married ones.

In conclusion, consideration of living arrangements is crucial for identification and explanation of differences in mortality risks by marital status between men and women. Groups at highest mortality risk in middle age are divorced and single men living alone and single women living with a partner. The present study therefore provides implications for research and policy in an area of increasing importance given the changing proportions of men and women living in consensual unions and living alone.

What is already known on this subject

  • Previous research on marital status and mortality indicates a beneficial effect of being married in men and women.

  • The association of marital status with mortality seems to differ between middle-aged men and women with a stronger protection of marriage in men.

  • Gender differences have also been described for the association of living arrangements with mortality, even though not consistently.

  • The role of living arrangements for mortality across the subcategories of marital status in men and women is largely unstudied.

What this study adds

  • This study provides evidence that both marital status and living arrangement are important predictors of mortality. While mortality patterns across the different marital status categories are similar in men and women with single and divorced individuals being most likely to die, the associations of mortality with living arrangements differ between men and women. In middle-aged men, living alone was associated with the highest mortality risk, while in women of the same age, this was the case for the ones living with a partner.

  • Furthermore, this study shows that the larger benefit of being married in men compared with women can largely be explained by this different impact of living arrangements on mortality in men and women.

  • Groups at highest mortality risk could be identified, which are divorced and single men who live alone and single women who live with a partner up to the age of 80 years.

Acknowledgments

We thank the Federal Statistical Office, whose support made the Swiss National Cohort and this study possible. The members of the Swiss National Cohort Study Group are Felix Gutzwiller (Chairman of the Executive Board), Matthias Bopp and David Faeh (Zurich, Switzerland); Matthias Egger (Chairman of the Scientific Board), Adrian Spoerri, Kurt Schmidlin and Marcel Zwahlen (Bern, Switzerland); Charlotte Braun-Fahrländer (Basel, Switzerland); Fred Paccaud (Lausanne, Switzerland) and André Rougemont (Geneva, Switzerland).

References

Footnotes

  • Funding This work was supported by the Swiss National Science Foundation (grant 3347C0-108806).

  • Competing interest None declared.

  • Ethics approval This study was conducted with the approval of the Cantonal Ethics Committees of Bern and Zurich, Switzerland.

  • Provenance and peer review Not commissioned; externally peer reviewed.