Background Little is known about the effects of long-term marital history on mortality, and the relative importance of using marital history instead of baseline marital status in mortality analyses. No previous comparative studies on the associations of marital history and mortality exist.
Methods Longitudinal data from England & Wales and from Finland were used to assess the effects of marital history, constructed from census records from years 1971, 1981 and 1991, on all-cause mortality in 1991–2004 among men and women aged ≥50 years. Data from England & Wales include 57 492 deaths; data from Finland include 424 602 deaths. Poisson regression analysis was applied.
Results Adding marital history into models including baseline marital status was statistically significant when explaining male mortality, while it was generally not important for female mortality. Adjusted for socio-demographic covariates, those consistently married with no record of marital break-up had the lowest mortality rates among both men and women aged 50–74 in both countries. Those never married, those divorced with a history of divorce and those widowed with a history of widowhood showed the highest mortality risks. Associations between marital history and mortality were weaker among those aged 75+.
Conclusions Consistent evidence in favour of both protective effects of long-lasting marriage and detrimental effects of marital dissolution were found. Studies would benefit from including marital history in the models instead of baseline marital status whenever possible, especially when studying male mortality.
- Marital status
- Marital history
- England and Wales
- old age
Statistics from Altmetric.com
The effects of marital status on mortality are well documented. Many studies have concluded that married persons have lower mortality rates than those never married, divorced or widowed in different societies.1–5 Women are less vulnerable to not being married than men, and older age groups are less vulnerable than younger age groups.6–8 In addition to baseline marital status, studies have shown detrimental effects of marital transitions, especially divorce and bereavement.9–12 Mortality differences by marital status may arise from the protective effect of marriage, from the detrimental effects of marital dissolution and from health-related selection into and out of marriage.6 8 13–18
Measurement of marital status cross-sectionally at one point of time or focusing on one marital transition only may hide important differences within the marital status groups since individuals have experienced unequal times of exposure to different marital statuses and several changes in their status, resulting in different levels of cumulative support and stressful experiences. The few studies that have assessed effects of marital history on mortality have shown that past marital experiences do indeed play a role over and above current marital status. The benefits of marriage seem to accumulate with increasing time spent married,15 19 20 past experiences of marital break-ups are a disadvantage to those currently married, and remarriage decreases the adverse effects of previous marital break-ups.8 20 21 In addition, number of years being divorced may have cumulative adverse effects on mortality among men.19 However, research evidence on the association of marital history and mortality is still scarce and more research is needed on the effects of exposure to different marital status groups, distinguishing more specific combinations of past and present marital statuses. Furthermore, it is not clear whether inclusion of marital history in place of baseline marital status actually improves the models explaining mortality.
The aim of this study is to assess the effects of baseline marital status and marital history on all-cause mortality in two European countries, England & Wales and Finland, using longitudinal census data from years 1971, 1981 and 1991 linked to mortality follow-up in 1991–2004. The study aims to find out whether including marital history improves the models explaining mortality risks compared to including only baseline marital status, and whether effects differ between England & Wales and Finland. Furthermore, we aim to determine the relative mortality risks in groups with different marital histories. The two comprehensive datasets offer a rare opportunity to compare the effects in the populations of two countries with a similar research design.
Datasets and participants
We use highly comparable datasets linking census data to death records. Data for England & Wales are from the Office for National Statistics Longitudinal Study (ONS LS), a record linkage study of 1% of the population of England & Wales in 1971, with subsequent linkages of census data from years 1981, 1991 and 2001, and linked vital event data. Finnish data are register data compiled by Statistics Finland, based on the longitudinal individual-level data file of employment statistics for years 1987–2004 and linked to longitudinal census records from years 1970, 1975, 1980 and 1985 and to the death register. The dataset consists of an 11% sample of the Finnish population and an over-sample of all persons who died during the follow-up so that in total 80% of all deaths in Finland are covered. The over-representation of dead persons is taken into account by appropriate weighting of observations in all statistical analyses. Our analyses focus on persons aged 50 and over in 1991, living in private households and whose census records could be found for each time point 1971, 1981 and 1991. In the following, for presentation purposes, we also refer to the years 1971, 1981 and 1991 when referring to the Finnish census years 1970, 1980 and 1990.
Follow-up of deaths
The baseline for the mortality follow-up is the census dates which were 21 April 1991 in England & Wales and 31 December 1990 in Finland. Thus, mortality follow-up and calculation of person-days begin at 22 April 1991 and at 1 January 1991, respectively. Deaths and person-years are followed until 31 December 2004.
Marital status in 1991 was classified into married, divorced, widowed and never married. Marital status history was constructed as a combination of information of marital status from years 1971, 1981 and 1991, each year classifying the subjects into married, divorced, widowed or never married. The following 8-class classification was formed, the baseline marital status in 1991 being the starting point:
Consistently married (marital status was married in 1971, in 1981 and in 1991).
Late consistent marriage (marital status was never married in 1971 and married in 1981–91 or never married in 1971–81 and married in 1991, no records of previous divorce/widowhood).
Remarried (marital status was married in 1991, divorced or widowed in 1981 and/or in 1971).
Recently divorced (marital status was divorced in 1991, no previous records of divorce).
Divorced with a history of divorce (marital status was divorced in 1991 and in 1981 and/or in 1971).
Recently widowed (marital status was widowed in 1991, no previous records of widowhood).
Widowed with a history of widowhood (marital status was widowed in 1991 and in 1981 and/or in 1971).
Never married (marital status was never married in 1991).
To minimise the effects of selection, we adjusted for several socio-demographic covariates in 1991. All results are presented separately for men and women. The analyses were performed separately for two age categories: 50–74 and 75+, and within these, age was adjusted for in 5-year age groups. Household size was classified into one or two or more persons. Housing tenure was classified into owner-occupied and other. Last week's economic activity separates out those who were employed, unemployed or studying from those retired, homemakers or outside the labour force for other reasons. The level of education was divided into two classes: those with some post-18 education and those with lower education. Lastly, we included occupation-based social class in five categories: professional or intermediate; skilled non-manual; skilled manual; partly skilled or unskilled; other or unknown. Since social class in 1991 was not available for retired persons or those outside the labour force, it was retrieved retrospectively from years 1981 and 1971 if it was recorded.
Both datasets were first tabulated according to the classes of the covariates. We calculated the number of person-years lived and deaths experienced during the follow-up time in each combination of classes. In the Finnish data, weighting of observations was used in order to take into account the over-representation of dead persons in the data. First, we calculated age-adjusted mortality rates according to each covariate, using the total population of men and women aged 50+ in 5-year age classes in each country as the standard population. Second, all-cause mortality according to sets of variables was modelled with Poisson regression, using the logarithm of accumulated person-years as the offset variable. Analyses were performed for two age groups: those aged 50–74 and those aged 75+. The results of these analyses are presented as mortality RRs with their 95% CIs. Furthermore, the relevance of inclusion of marital history into models with only baseline marital status was tested with likelihood ratio tests.
Table 1 shows the distributions and age-adjusted mortality rates according to each covariate. The study population in England & Wales consists of about 1.6 million person-years and experienced 57 492 deaths in 1991–2004. The Finnish study population lived about 4.3 million person-years and experienced 424 602 deaths. Since dead persons were heavily oversampled in the Finnish data, we used appropriate weighting in the analyses (weights calculated by Statistics Finland). In weighted numbers, the Finnish study population lived about 3.2 million person-years and experienced 108 471 deaths.
Table 2 presents distributions of the marital status variables as well as relative mortality rate ratios for men and women in England & Wales and Finland in age groups 50–74 and 75+ separately according to baseline marital status and according to marital status history: first, adjusted for age; and second, adjusted for socio-demographic covariates.
Relative mortality rates by marital status and marital history among men
Among men aged 50–74, the non-married groups at baseline had higher mortality risks than those married, and the relative adjusted mortality risks were larger in Finland compared to England & Wales. Among men aged 75+, differences by baseline marital status were non-existent or weak.
Adding marital status history into the model, including baseline marital status and other covariates, improved the model statistically significantly among men aged 50–74 in both countries and among men aged 75+ in England & Wales. Among married men aged 50–74 in England, those with a later timing of first marriage had a lower mortality risk (RR=0.87) than those married since 1971, and those remarried had a higher mortality risk than those in late consistent marriage but lower risk than those with a history of widowhood (result based on contrasts analysis, details not shown here). Finnish remarried men had a higher mortality risk than those in both long-term and late consistent marriage, and lower mortality risks than the divorced or widowed groups. Furthermore, men with a long history of widowhood had higher mortality risks compared to those recently widowed in England & Wales. Recently divorced men differed significantly from those with a longer history of divorce in Finland. Among men aged 75+ in England & Wales, there was strong evidence of a protective effect of a late consistent marriage. However, note that among those aged 75+, there are only a few persons in the categories of late consistent marriage, remarried and recently divorced.
Relative mortality rates by marital status and marital history among women
Among women, mortality rate ratios among most non-married groups increased when adjusting for other covariates, the effect being produced by adjustment for household size that had a different relationship with mortality among women and men (detailed results not shown here). The relative mortality risks associated with not being married were roughly similar among women in the two countries. Compared to men, women's adjusted mortality rate ratios were somewhat higher in England & Wales in both age groups but lower or rather similar in Finland.
Among women, adding the marital history information to the adjusted model was statistically significant only in the group aged 75+ in England & Wales. The adjusted association of marital history with mortality was strikingly similar among women aged 50–74 in the two countries. The contrast analyses revealed that in this age group in both countries, those remarried had statistically significantly lower mortality than those divorced with a history of divorce and there was some indication of difference between remarried and also other formerly married groups, but these differences did not quite reach statistical significance. Similarly to older English men, older English women who had married late had a significantly lower mortality rate ratio (RR=0.53) compared to those constantly married, and they also differed from those with previous marital break-ups (however, as was noted concerning men, there are only relatively few older women in the late married, remarried and divorced groups). Furthermore, English older women with a history of widowhood had significantly higher mortality risks than those recently widowed.
Mortality differences by marital status may arise from three separate processes. First, the causation hypothesis assumes a protective effect of marriage: married persons may have higher financial status and better health habits, and they may suffer less from stress and enjoy higher levels of social support and social integration compared to those not married.13–15 22 Second, marital dissolution may induce stress related to changes in financial circumstances, social relationships and health behaviour.8 16 23 Third, the association may be due to health-related selection into and out of marriage. While some studies have concluded that health selection does not play a major role in the association between marital status and mortality,6 studies on associations between marital status and health have estimated that the role of selection may be at least as important as the role of causal effects.17 18 24
While marital status differences in mortality are well known, cross-sectional measurement of marital status does not cover the possible effects of past partnership experiences. Most mortality analyses have investigated effects of only baseline marital status or single transitions and no study so far has specifically investigated whether including marital histories actually improves models explaining mortality. Furthermore, there are no comparative studies examining the associations between marital status history and mortality. This study focused on mortality effects of marital status history, constructed from measurements in 1971, 1981 and 1991, on mortality in 1991–2004 among men and women aged 50–74 and 75+ in England & Wales and in Finland. We found that adjusting for age, household size, housing tenure, economic activity, education and social class, taking into account marital history in addition to baseline marital status improved most models explaining male mortality, while the same pattern was not generally found among women. Thus, male mortality analyses would benefit from including the history information when available.
Our results confirm the beneficial effects of an enduring marriage received elsewhere in mortality and morbidity studies15 20 25 26: all those who were married at baseline with no records of previous divorce or widowhood had lowest mortality rates in all sub-groups studied even though marital history was not generally significant among women. We also found evidence that mortality risks may be even lower among those who marry relatively late in their life course. Relative mortality rates were especially low among older English men and women, whose first marriage was contracted after the age of 55. The result may indicate protective effects of marriage, evoking positive changes associated with reduced mortality risks. Selection into marriage at older age may also have played an important role. However, the general importance of this effect is small since only a few people belong to the group. Our classification of marital history may be more suitable in the younger age groups, and the older group might benefit from a follow-up extending further back in time to determine the timing of first marriage, and from a more detailed classification concerning duration of widowhood.
Being remarried and thus having past experiences of marital break-up (divorce or widowhood) increased the mortality risks among those currently married, providing further evidence of the longer term detrimental effects of marital dissolution reported in previous studies.8 On the other hand, remarriage after marital dissolution had a protective effect compared to those who remained divorced or widowed at baseline. In addition, we found evidence in favour of cumulative effects of divorce (among Finnish men aged 50–74 and to some extent among Finnish women in the same age group) and of widowhood (among English men aged 50–74 and English women aged 75+), also in line with previous results.19 We thus found support for both theories of protective effects of marriage and of detrimental effects of marital break-up.
The study is set in two European welfare states with roughly similar population structures and population health levels measured by life expectancy at birth.27 However, the countries diverge in their health and social policy features. England & Wales belongs to the regime of Anglo-Saxon liberal market-led welfare states, with means tested services and minimal protection.28 In contrast, the Finnish social protection system is built on the Nordic model of a welfare state where social and health services and provision of social security are based on principles of universality and equality, delivery of services not typically being dependent on one's income or assets.29
Overall, the rough similarity of the effects of marital histories on mortality in the two country contexts is striking. However, we found that the differences between marital groups among men were somewhat larger in Finland than in England & Wales. Even though there are differences in the country circumstances, here as in most other comparative analyses, it is difficult to disentangle the specific reasons for the slightly divergent effects. These country differences in the association of marital status with mortality among men may partly be a result of different political, cultural and societal factors and differential selection into the marital states according to unmeasured individual characteristics. More specifically, cultural climates and social policies may partly explain why widowhood and accumulation of time spent widowed has pronounced detrimental effects on mortality in England & Wales, while in Finland this is the case for divorce and especially accumulation of time divorced. The factors behind these differences may be related to the differential propensity of the non-married groups in the two countries to engage in non-healthy behaviour, to be socially isolated, or face financial difficulties, all related to higher mortality risks. In addition, reactiveness of the social and healthcare systems to changes in marital status and household composition may play a role. Stronger selection into divorce in Finland may also explain some of the differences. These factors, however, could not be analysed in depth in this paper.
However, it has to be kept in mind that based on these data, selection into different marital states and marital histories cannot be ruled out as an explanation for positive effects of marriage and negative effects of marital dissolution. For example, the selection effect may be particularly salient for those who divorce and remain divorced or who divorce several times, while selection may play only a weak or no role among those bereaved. Furthermore, those who marry at older ages may have better health and health habits than those who remain never married.
Furthermore, our classification of marital history is a rough simplification of a complex reality. As we could not distinguish all changes in marital history during the life course, for example changes that happened between the measurements, and changes that had happened before 1971, misclassification of people into wrong categories may have occurred. For example, those who were married at one census but who had divorced or widowed and remarried before the next census appear here as consistently married. Moreover, those married in 1971 may already have experienced marital break-ups that are intractable with our data. While it would be possible to distinguish those remarried from first marriages in the ONS LS data in 1981 and 1991, this is unfortunately not possible in our Finnish data. However, due to the misclassification bias, our results are only likely to be underestimations of the true effects, and the observed associations would be even stronger using a more detailed classification. Another gap in our data was lack of information of cohabitation status, which unfortunately could not be taken into account due to incomplete Finnish data for years 1970 and 1980. Future studies may clarify the associations with a more refined classification of marital history and taking into account the cohabitation status of the unmarried groups.
Our cross-national comparison with similar data and research design enabled us to establish the relationships between baseline marital status, marital history and mortality. A major strength of the study lies in the possibility to use similar extensive longitudinal register-based and census data with information extending from the 1970s to the 1990s, and a 14-year mortality follow-up. In all, our study showed that studies on marital status and mortality may benefit from taking into account the lifelong marital history of an individual in addition to baseline marital status, especially when studying male mortality. Our analyses also raised questions to be tackled in further studies. Future studies should examine health behaviour, morbidity and cause-specific mortality since marital biographies may influence onset of different illnesses and causes of deaths differentially. Additionally, interactions between marital status and other socio-demographic factors deserve detailed analysis.
What is already known on this subject
Associations of baseline marital status with mortality are well known: never married, divorced and widowed persons have higher mortality risks than married persons.
Studies on the effects of marital histories are rarer and tend to focus on single transitions, not assessing the complete pattern of life course marital status.
There are no comparative studies on marital history and mortality, and no comparative study has documented whether adding marital history information with baseline status is significant in explaining mortality risks.
What this study adds
This study provided a rare opportunity to asses effects of marital histories with extensive comparative register- and census-based data in two countries.
Marital history information is significant in addition to measurement in one time point (the baseline of mortality follow-up) in explaining male mortality but not when explaining female mortality. Thus, baseline information captures enough information with regard to female but not male mortality.
There is increased mortality among those having past experience of marital break-up even though married at baseline; remarriage after marital dissolution has a protective effect.
Mortality risks may be lower among those who marry relatively late during their life course.
The permission of the Office for National Statistics to use the Longitudinal Study (project number 30021, clearance number 30021A), is gratefully acknowledged, as is the help provided by staff of the Centre for Longitudinal Study Information & User Support (CeLSIUS), especially help by Chris Marshall. CeLSIUS is supported by the ESRC Census of Population Programme (award ref: RES-348-25-0004). The authors alone are responsible for the interpretation of the data. Census output is Crown copyright and is reproduced with the permission of the Controller of HMSO and the Queen's Printer for Scotland. We also thank Statistics Finland for permission to use the data (permission TK-53-1783-96).
Funding This work was supported by a grant from the Academy of Finland (grant number 2960501). PM is supported by the Ministry of Health and Social Affairs and the Academy of Finland.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
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.