You are here

Article Text

Article menu

Download PDFPDF

Research report
Cause-specific mortality by partnership status: simultaneous analysis using longitudinal data from England and Wales
  1. Sebastian Franke1,2,
  2. Hill Kulu3
  1. 1 Department of Geography and Planning, School of Environmental Sciences, University of Liverpool, Liverpool, UK
  2. 2 Health Services Management, Ludwig-Maximilians-Universität München, Munich, Germany
  3. 3 School of Geography and Sustainable Development, University of St Andrews, St Andrews, UK
  1. Correspondence to Dr Hill Kulu, School of Geography and Sustainable Development, University of St Andrews, St Andrews KY16 9AL, UK; Hill.Kulu{at}st-andrews.ac.uk

Abstract

Background This paper examines cause-specific mortality by partnership status. Although non-marital cohabitation has spread rapidly in industrialised countries, only a few studies have investigated mortality by partnership status and no recent study has investigated cause-specific mortality by partnership status.

Methods We use data from the Office for National Statistics Longitudinal Study and apply competing risks survival models.

Results The simultaneous analysis shows that married individuals have lower mortality than non-married from circulatory, respiratory, digestive, alcohol and accident related causes of deaths, but not from cancer. The analysis by partnership status reveals that once we distinguish premarital and postmarital cohabitants from other non-married groups, the differences between partnered and non-partnered individuals become even more pronounced for all causes of death; this is largely due to similar cause-specific mortality levels between married and cohabiting individuals.

Conclusions With declining marriage rates and the spread of cohabitation and separation, a distinction between partnered and non-partnered individuals is critical to understanding whether and how having a partner shapes the individuals’ health behaviour and mortality. The cause-specific analysis supports both the importance of selection into partnership and the protective effect of living with someone together.

  • mortality
  • marital status
  • longitudinal studies

https://doi.org/10.1136/jech-2017-210339

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

    Research shows that married people in industrialised countries have lower mortality levels than non-married individuals.1–9 However, it is far from clear whether mortality differences by marital status are explained by the protective effects of marriage or the selection into marriage.10–12 Many studies suggest that a combination between both protection and selection plays a role.4 9 13 Health influences selection into marriage at younger ages, but mortality differences at older ages are also observed due to the cumulation of the protective effects of marriage.9 14–16

    Although there is a large body of literature on mortality differences by marital status, mortality by causes of death has received much less attention. Previous studies show lower mortality rates for married individuals from circulatory and respiratory diseases as well as from accidents and self-harm. For cancer, research demonstrates the higher mortality risk for divorced in comparison to married people.2 17–21 However, most research on mortality by partnership status does not distinguish between individuals who are single and those who are cohabiting treating them as one group. These studies are likely to underestimate mortality differences between partnered and non-partnered individuals and increasingly misrepresent the changing demographic reality.16

    Marriage rates have significantly declined in many industrialised countries, with data showing  the postponement of marriages combined with some decline of the percentage who marry.22 In Britain, the median age of marriage for women born in the 1940s and 1950s was in the low 20s, whereas the same figure for cohorts born in the 1970s was in the low 30s; the share of married individuals declined from 90% among women born in the 1940s and 1950s to 80% among women born in the 1960s.23 In contrast, premarital and postmarital cohabitation have rapidly increased over recent decades. While only 10% of individuals who were born in the 1940s ever cohabited by age 45, more than 50% of women who were born in the 1960s cohabited by the age of 30. For younger cohorts, the percentage of individuals having ever cohabited by age 30 is about 70%.4 23 24 Divorce and separation levels have also increased. One-fifth of the marriages that were formed in the 1965–1974 period ended in divorce before their 15th anniversary, whereas more than one-third of marriages have experienced separation in the marriage cohorts from 1995 onwards. Separation levels have been even higher among cohabitants.23 These significant changes in partnership patterns in industrialised countries suggest that any analysis of mortality differences by partnership should distinguish non-marital cohabitants from single and divorced individuals and treat them as separate groups.

    The aim of this study is to investigate cause-specific mortality by partnership status. We build on previous research and extend it in two ways. First, we move beyond the non-married–married dichotomy and treat cohabitants as a separate group. Distinguishing cohabitants from other non-married groups (ie, single and divorced) in the cause-specific analysis of mortality is the first novelty of the study. Previous studies have either examined cause-specific mortality by marital status (eg, 2 3 17 25–31) or all-cause mortality by partnership status (eg, 14 16 32–34). There are very few (if any) recent studies that investigate cause-specific mortality by partnership status.

    Second, we conduct simultaneous analysis of cause-specific mortality allowing us to explicitly distinguish the contribution of different causes to the all-cause mortality by partnership status. Simultaneous analysis is the second novelty of the paper; most previous studies on mortality by marital status have conducted separate analysis for each cause of death. We calculate cause-specific mortality rates by partnership status separately for men and women. We standardise the mortality rates to socioeconomic characteristics of individuals to examine mortality differences between non-married cohabiting and non-married not-cohabiting people. We show how all-cause mortality differences by partnership status in Britain translate into cause-specific mortality differences.

    Data and methods

    Data

    We use the Office for National Statistics Longitudinal Study (ONS LS), which is a 1% sample of the population of England and Wales (see Hattersley et al 35 for more detail). We study cause-specific mortality of the population aged 30–85 between the 2001 and 2011 census. The ONS LS provides demographic and socioeconomic characteristics from the 2001 census, linked with yearly death and embarkation events. Our analysis was conducted for men (158 061) and women (171 706) separately. Individuals who leave the study at a known or unknown embarkation date before 2011 are censored after 3.9 years based on the sensitivity analysis by Franke and Kulu.16

    Methods

    We use competing risks survival analysis. The cause-specific hazard function, hk (t), is defined as follows:

    Embedded Image (1)

    where D denotes the cause of death with k as the number of different causes and T represents the duration of an episode or an individual’s age. We define a cause-specific proportional-hazards regression model to study mortality by partnership status and cause of death.

    Embedded Image (2)

    where hk (t) denotes an individual’s hazard of mortality and hk ,0(t) is the baseline mortality risk for cause k at age t; x(t) is a variable measuring individual socioeconomic characteristics and β is the parameter estimate for this variable, with l variables; ɣk represents the effect of variable z (partnership status) on mortality from cause k.

    The advantage of the model defined in equation 2 is that the effect of age and all (other) covariates on mortality can vary by cause; however, the contribution of each cause to mortality by partnership status (our covariate of interest) is not easy to grasp from separate models. We extend this model to also measure the relative contribution of each cause to mortality variation by partnership status:

    Embedded Image (3)

    The model defined by equation 3 is similar to the one defined by equation 2, but assumes a common baseline for all causes and the same effect of control variables across the causes. Only the effect of partnership status is allowed to vary by cause; ɣk is a cause-specific parameter for variable z, partnership status. The partnership status differences obtained are now directly comparable with the equivalent differences for all-cause mortality.

    We use a Cox proportional-hazards model to estimate relative risks by partnership status without having to define the baseline hazard. We need to extend our data, so that we have k records per person with k being the number of possible causes. If an individual died from cause k, then our event variable has value 1 for this record (eg, cardiovascular disease, CVD); for all other records of this individual, the variable has value 0 (eg, cancer). We interact the cause of death variable with partnership status to simultaneously model cause-specific mortality among the population. We can then present mortality rates from each cause by partnership status (eg, cancer for singles) relative to a reference category (eg, CVD for married people). This approach has become common in some areas of mortality research,36 but has not yet been used to study mortality by partnership status.

    Causes of death

    Ideally, we would prefer to analyse causes of death in as much detail as possible. However, this is not feasible due to (1) a small number of events per partnership status when using a detailed cause classification and (2) the independence rule of competing risks. When analysing cause-specific mortality within a competing risk setting the independence of those causes needs to be ensured. This means that the probability of two causes of death between (Embedded Image ) should be negligible. One way to ensure this is to use broad disease categories.37

    We used WHO’s 10th revision of the International Classification of Diseases (see table 1 for more detail) to group our causes into the following categories—circulatory diseases, cancer (or malignant neoplasm), respiratory diseases, digestive disease, and nervous system diseases, which comprise 85% of all deaths for the age group 30–85. We fitted additional models including accidents and self-harm and alcohol-related causes, but we discuss these only briefly.

    View this table:
    Table 1

    Cause of death categories and deaths by marital status

    Covariates

    For all our models the first covariate is the interaction between marital status and cause of death. Marital status is hereby defined as ‘married’, ‘single’, ‘divorced/separated’ or ‘widowed’. In the model which includes cohabitation, partnership status is defined as ‘married’, ‘single’ (not cohabiting), ‘divorced/separated’ (not cohabiting), ‘widowed’ (not cohabiting), ‘premarital cohabitation’ (cohabiting singles) or ‘postmarital cohabitation’ (cohabiting divorced/separated and widowed).

    Control variables are as follows: country of birth (‘England and Wales’, ‘Scotland or Northern Ireland’, or ‘Others’); ethnicity (‘White’, ‘White mixed’, ‘Asian’, ‘Black’, ‘Chinese’ or ‘Other’); education (‘No qualification’, ‘Low qualification’, ‘Medium qualification’, ‘High qualification’ or ‘Missing’); socioeconomic status (‘Higher managerial and professional occupations’, ‘Lower managerial and professional occupations’, ‘Intermediate occupations (clerical, sales, service)’, ‘Small employers and own account workers’, ‘Lower supervisory and technical occupations’, Semi-routine occupations’, ‘Routine occupations’ or ‘Others’). We also include in the analysis two variables on living arrangements: household size (‘one person’, ‘two persons’, ‘three persons’, ‘four persons’, ‘five and more persons’ or ‘Unknown’) and the presence of dependent children (‘No children’, ‘Dependent children’, ‘Non-dependent children’ or ‘Not applicable’).

    Results

    Mortality by marital status by five causes

    We analysed mortality by marital status by five main causes: circulatory diseases, cancer, respiratory diseases, digestive diseases and nervous system diseases, and observed the following patterns. For both sexes, circulatory diseases and cancer are the leading causes of death, followed by respiratory and other diseases. The risk of dying from one of the latter two is only half of the risk of dying from a circulatory disease. Non-married individuals have higher mortality from all causes, with the exception of cancer for singles and widowed and nervous system diseases for divorced and widowed (figure 1 and online supplementary table A1). Although there is some variation between men and women, overall the patterns are relatively similar.

    Supplementary file 1

    Figure 1
    Figure 1

    Comparative cause-specific mortality risk for men and women aged 30–85 years (description: risk for five causes compared with married-circulatory disease, source: ONS LS, authors’ own calculation).

    In the next step, we extended the analysis in two ways. First, we added accidents (including self-harm) and alcohol-related causes. Second, we fitted one model with marital status and one model with partnership status, with both fitted separately for men and women. Comparing the results of the marital status model from the seven-cause analysis (dots in figure 2) with those of the five-cause analysis (dots in figure 1), we observe the following: the inclusion of alcohol-related diseases in the analysis leads to a decline in the effect of digestive system diseases. The inclusion of accidents results in a decrease in the impact of the ‘Other causes’ category. For both of these causes of death, single and divorced individuals have a significantly higher risk of dying than married people.

    Figure 2
    Figure 2

    Comparative cause-specific mortality risk by marital and partnership status for men and women aged 30–85 years (description: risk for seven causes compared with married-circulatory disease, source: ONS LS, authors’ own calculation).

    Next, we compared the results of the models for marital status and partnership status. For most causes, the differences between partnered and non-partnered individuals (triangles in figure 2) are bigger than those between married and non-married people (dots in figure 2). This is largely due to distinguishing cohabitants from singles in our analysis. For all causes but cancer and alcohol-related diseases, we find for men a significantly lower mortality risk for premarital cohabitants compared with non-cohabiting singles as well as lower risks for postmarital cohabitants than for non-cohabiting divorced/widowed individuals (online supplementary table A2). The results for women support a clear mortality difference between partnered and non-partnered individuals for the circulatory, respiratory and the ‘Other causes’ category .

    The effects of covariates were largely as expected. Mortality levels were lower among individuals with higher educational level and higher social class, and among ethnic minorities and migrants from outside the UK. There were minor mortality differences by household size or number of dependent children (table 2). Mortality levels of control variables were similar to the effects found in the all-cause-mortality models by Franke and Kulu.16 The similarity to the overall-mortality model suggests independence of the causes used in the competing risks model.

    View this table:
    Table 2

    Relative mortality risks by seven causes of death for control variables

    Summary and discussion

    This study investigated cause-specific mortality by marital and partnership status distinguishing between partnered and non-partnered individuals. Distinguishing cohabitants from singles in this way was the first novelty of the study. Another novelty was the use of simultaneous analysis  of cause-specific mortality in order to distinguish the contribution of different causes to the all-cause mortality differences by partnership status. The analysis of mortality by marital status showed that married individuals have lower mortality from all causes with the exception of cancer compared to non-married individuals. Differences by marital status, and thus the relative contribution towards all-cause mortality differences, were highest for circulatory, respiratory, digestive, alcohol and accident related causes of deaths. The analysis by partnership status showed that once we distinguished cohabitants from other non-married groups, the differences between partnered and non-partnered individuals became even more pronounced for almost all causes of deaths. Increased differences between partnered and non-partnered individuals were largely due to similar cause-specific mortality between married and cohabiting individuals, although this is a topic that requires further investigation using a larger sample.

    The study demonstrates the importance of cause-specific analysis and of distinguishing between partnered and non-partnered individuals. With declining marriage rates and an increase in  cohabitation and separation, such a distinction is critical to understanding whether and how having a partner shapes the individuals’ health behaviour and mortality. Like previous studies, we observed significant mortality differences by partnership status for circulatory and respiratory systems for both men and women (eg, 2 17 20). Our study clearly showed that those differences are underestimated if cohabitants are not distinguished from single and divorced individuals. Additionally, we extended previous research by distinguishing alcohol-related deaths, which can be due to short-term causes such as alcohol poisoning or accidents, or long-term causes such as alcohol-related damage to the digestive or nervous systems. The majority of alcohol-related causes in our study was attributed to digestive system diseases. The analysis showed a clear marital/partnership advantage. Further, single and divorced individuals, especially men, had a higher risk of death due to accidents and self-harm. These differences may be attributed to differences in lifestyle and both short- and long-term health behaviour. Therefore, having a partner provides protection from health-damaging behaviour.

    The advantage of having a partner or a spouse was not found for malignant neoplasm (cancer). Cancer became the number one cause of death in the UK in 2011, mostly due to a larger decline in circulatory diseases.38 Thus, for cancer, our analysis showed no advantage of people living with a partner in comparison with those living without a partner. Lower cancer mortality found in singles was more likely due to higher prevalence of other causes rather than to singlehood being a protection for cancer. Although we have mostly emphasised the importance of the protective effect of having a partner, relatively high mortality from nervous system diseases observed among non-partnered individuals supports the idea of selection into partnership by health status.

    With circulatory diseases being the main cause of death, a decline in CVD mortality is expected to lead to a decline in overall mortality differences by partnership status. Previous research has suggested that mortality differences by partnership status may gradually disappear. Crucially, our findings show that there remain significant mortality differences by partnership status, for example, in respiratory diseases. Thus, a continued increase in life expectancy, as well as a continued decline in circulatory disease mortality, will not necessarily result in a declining mortality gap between partnered and non-partnered individuals.

    In this study, survival models used the partnership status as recorded in the 2001 and 2011 Censuses since there is no information on partnership changes in the ONS LS. Methodologically, the simultaneous analysis of causes used in this study assumes a cause-independent baseline hazard, which is the same as the all-cause mortality baseline hazard. It is possible that different causes of death have different baseline hazards. In our preliminary analysis, we fitted separate models for the main causes of death—the baseline and covariate effects were thus cause specific. Most importantly, we observed similar cause-specific mortality differences by partnership status between separate and simultaneous analysis confirming that the results of our simultaneous analysis are robust.

    In a competing risks setting with k competing causes, only the first failure is observed. We grouped the causes of death into seven groups under the assumption that they are independent of each other. The independence assumption has the advantage that the survival function of a specific cause derived from the joint-survival function or the multiple decrement function of all competing risks is reduced to a cause-specific survival function where we can simply censor failures due to different causes. This is because censored observations do not change the survival or hazard function as long as they are independent of the estimated failure. A limitation of this approach is that this independence hypothesis is not testable.39 Future studies could compare the results of different competing risks approaches for cause-specific mortality analysis.

    Future analysis on cause-specific mortality risks in England and Wales could also disaggregate the patterns by age (which we did in further analysis not reported here due to small sample size). Such analysis could also include measures of health differences like morbidity or social deprivation (eg, 40). The detailed analysis of mortality and health differences between cohabiting and married individuals is another important avenue to proceed. With the spread of premarital and postmarital cohabitation among younger cohorts, this is increasingly feasible and important. To conclude, using longitudinal data from England and Wales this study showed first that partnered individuals had lower mortality from all causes with the exception of cancer. Second, the differences between partnered and non-partnered individuals became pronounced for all causes of death, once premarital and postmarital cohabitants were distinguished from other non-married groups.

    What is already known on this subject

    • Married people have better health and lower mortality levels than non-married individuals.

    • Married individuals have lower mortality from circulatory and respiratory diseases as well as from accidents and self-harm.

    • Marriage rates have significantly declined in many industrialised countries over the last decades; in contrast, premarital and postmarital cohabitation have rapidly increased.

    • However, most research on mortality by partnership status does not distinguish between singles and cohabitants, instead treating them as one group. These studies are likely to underestimate mortality differences between partnered and non-partnered individuals and increasingly misrepresent the changing demographic reality.

    What this study adds

    • Married individuals have lower mortality from all causes with the exception of cancer.

    • Once we distinguish cohabitants from other non-married groups, the differences between partnered and non-partnered individuals become even more pronounced for all causes of deaths.

    • For most of the causes, patterns are similar for cohabiting and married individuals, suggesting that having a partner is an important indicator and determinant of health behaviour.

    • With declining marriage rates and the increase of cohabitation and separation, a distinction between partnered and non-partnered individuals is critical to understanding whether and how having a partner shapes the individuals’ health behaviour and mortality.

    Acknowledgments

    The authors are grateful to Mark Green and three anonymous referees for valuable comments and suggestions on a previous version of this article. The permission of the Office for National Statistics to use the Longitudinal Study is gratefully acknowledged, as is the help provided by staff of the Centre for Longitudinal Study Information and User Support (CeLSIUS). CeLSIUS is supported by the ESRC Census of Population Programme (Award Ref: ES/K000365/1).

    References

    1. ↵
      2. Hu YR ,
      3. Goldman N
      . Mortality differentials by marital status: an international comparison. Demography 1990;27:233–50.
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      2. Ben-Shlomo Y ,
      3. Smith GD ,
      4. Shipley M , et al
      . Magnitude and causes of mortality differences between married and unmarried men. J Epidemiol Community Health 1993;47:200–5.doi:10.1136/jech.47.3.200
      OpenUrlAbstract/FREE Full Text
    3. ↵
      2. Johnson NJ ,
      3. Backlund E ,
      4. Sorlie PD , et al
      . Marital status and mortality: the national longitudinal mortality study. Ann Epidemiol 2000;10:224–38.
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      2. Murray JE ,
      3. Protection M
      . Marital protection and marital selection: evidence from a historical-prospective sample of American men. Demography 2000;37:511–21.doi:10.1353/dem.2000.0010
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      2. Brockmann H ,
      3. Klein T
      . Love and death in Germany: the marital biography and its effect on mortality. J Marriage Fam 2004;66:567–81.doi:10.1111/j.0022-2445.2004.00038.x
      OpenUrlCrossRefWeb of Science
    6. ↵
      2. Murphy M ,
      3. Grundy E ,
      4. Kalogirou S
      . The increase in marital status differences in mortality up to the oldest age in seven European countries, 1990–99. Popul Stud 2007;61:287–98.doi:10.1080/00324720701524466
      OpenUrlCrossRef
    7. ↵
      2. Dupre ME ,
      3. Beck AN ,
      4. Meadows SO
      . Marital trajectories and mortality among US adults. Am J Epidemiol 2009;170:546–55.doi:10.1093/aje/kwp194
      OpenUrlCrossRefPubMedWeb of Science
    8. ↵
      2. Blomgren J ,
      3. Martikainen P ,
      4. Grundy E , et al
      . Marital history 1971–91 and mortality 1991–2004 in England & Wales and Finland. J Epidemiol Community Health 2012;66:30–6.doi:10.1136/jech.2010.110635
      OpenUrlAbstract/FREE Full Text
    9. ↵
      2. Guner N ,
      3. Kulikova Y ,
      4. Llull J
      . Does marriage make you healthier? CEPR Discussion Paper No . 2014. DP10245 .
    10. ↵
      2. Lillard LA ,
      3. Panis CWA
      . Marital status and mortality: the role of health. Demography 1996;33:313.
    11. ↵
      2. Rendall MS ,
      3. Weden MM ,
      4. Favreault MM , et al
      . The protective effect of marriage for survival: a review and update. Demography 2011;48:481–506.doi:10.1007/s13524-011-0032-5
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      2. Su D ,
      3. Stimpson JP ,
      4. Wilson FA
      . Racial disparities in mortality among middle-aged and older men. Am J Mens Health 2015;9:289–300.doi:10.1177/1557988314540199
      OpenUrlCrossRefPubMed
    13. ↵
      2. Waldron I ,
      3. Hughes ME ,
      4. Brooks TL
      . Marriage protection and marriage selection—prospective evidence for reciprocal effects of marital status and health. Soc Sci Med 1996;43:113–23.doi:10.1016/0277-9536(95)00347-9
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      2. Koskinen S ,
      3. Joutsenniemi K ,
      4. Martelin T , et al
      . Mortality differences according to living arrangements. Int J Epidemiol 2007;36:1255–64.doi:10.1093/ije/dym212
      OpenUrlCrossRefPubMedWeb of Science
    15. ↵
      2. Ploubidis GB ,
      3. DeStavola BL ,
      4. Grundy E
      . Health differentials in the older population of England: an empirical comparison of the materialist, lifestyle and psychosocial hypotheses. BMC Public Health 2011;11:1655.doi:10.1186/1471-2458-11-390
    16. ↵
      2. Franke S ,
      3. Kulu H
      . Mortality differences by partnership status in England and Wales: the effect of living arrangements or health selection? European Journal of Population 2017;18:413.
    17. ↵
      2. Ebrahim S ,
      3. Wannamethee G ,
      4. McCallum A , et al
      . Marital status, change in marital status, and mortality in middle-aged British men. Am J Epidemiol 1995;142:834–42.doi:10.1093/oxfordjournals.aje.a117723
      OpenUrlPubMedWeb of Science
    18. ↵
      2. Maxwell R ,
      3. Harding S
      . Mortality of migrants from outside England and Wales by marital status. Popul Trends 1998;91:15–22.
      OpenUrlPubMed
    19. ↵
      2. Nativity TF
      . Marital status and mortality in Canada*. Can Rev Sociol 1998;35:65–91.
      OpenUrl
    20. ↵
      2. Cheung YB
      . Marital status and mortality in British women: a longitudinal study. Int J Epidemiol 2000;29:93–9.doi:10.1093/ije/29.1.93
      OpenUrlCrossRefPubMedWeb of Science
    21. ↵
      2. Wild SH ,
      3. Fischbacher CM ,
      4. Brock A , et al
      . Mortality from all cancers and lung, colorectal, breast and prostate cancer by country of birth in England and Wales, 2001–2003. Br J Cancer 2006;94:1079–85.doi:10.1038/sj.bjc.6603031
      OpenUrlCrossRefPubMedWeb of Science
    22. ↵
      2. Thomson E
      . Family complexity in Europe. Ann Am Acad Pol Soc Sci 2014;654:245–58.doi:10.1177/0002716214531384
      OpenUrlCrossRef
    23. ↵
      2. Hannemann T ,
      3. Kulu H
      . Union formation and dissolution among immigrants and their descendants in the United Kingdom. Demogr Res 2015;33:273–312.doi:10.4054/DemRes.2015.33.10
      OpenUrl
    24. ↵
      2. Ermisch J ,
      3. Francesconi M
      . Cohabitation in Great Britain: not for long, but here to stay. J R Stat Soc Ser A Stat Soc 2000;163:153–71.doi:10.1111/1467-985X.00163
      OpenUrl
    25. ↵
      2. Burgoa M
      . Mortality. Mortality by cause of death and marital status in Spain. The European Journal of Public Health 1998;8:37–42.doi:10.1093/eurpub/8.1.37
      OpenUrlCrossRefWeb of Science
    26. ↵
      2. Kravdal O
      . The impact of marital status on cancer survival. Soc Sci Med 2001;52:357–68.
      OpenUrlCrossRefPubMedWeb of Science
    27. ↵
      2. Martikainen P ,
      3. Martelin T ,
      4. Nihtilä E , et al
      . Differences in mortality by marital status in Finland from 1976 to 2000: analyses of changes in marital-status distributions, socio-demographic and household composition, and cause of death. Popul Stud 2005;59:99–115.doi:10.1080/0032472052000332737
      OpenUrlCrossRef
    28. ↵
      2. Ikeda A ,
      3. Iso H ,
      4. Toyoshima H , et al
      . Marital status and mortality among Japanese men and women: the Japan Collaborative Cohort Study. BMC Public Health 2007;7:949.doi:10.1186/1471-2458-7-73
      OpenUrl
    29. ↵
      2. Jaffe DH ,
      3. Manor O ,
      4. Eisenbach Z , et al
      . The protective effect of marriage on mortality in a dynamic society. Ann Epidemiol 2007;17:540–7.doi:10.1016/j.annepidem.2006.12.006
      OpenUrlCrossRefPubMedWeb of Science
    30. ↵
      2. Berntsen KN ,
      3. Kravdal O
      . The relationship between mortality and time since divorce, widowhood or remarriage in Norway. Soc Sci Med 2012;75:2267–74.doi:10.1016/j.socscimed.2012.08.028
      OpenUrlCrossRefPubMed
    31. ↵
      2. Metsä-Simola N ,
      3. Martikainen P
      . The short-term and long-term effects of divorce on mortality risk in a large Finnish cohort, 1990–2003. Popul Stud 2013;67:97–110.doi:10.1080/00324728.2012.746386
      OpenUrl
    32. ↵
      2. Lund R ,
      3. Due P ,
      4. Modvig J , et al
      . Cohabitation and marital status as predictors of mortality—an eight year follow-up study. Soc Sci Med 2002;55:673–9.doi:10.1016/S0277-9536(01)00219-2
      OpenUrlCrossRefPubMedWeb of Science
    33. ↵
      2. Scafato E ,
      3. Galluzzo L ,
      4. Gandin C , et al
      . Marital and cohabitation status as predictors of mortality: a 10-year follow-up of an Italian elderly cohort. Soc Sci Med 2008;67:1456–64.doi:10.1016/j.socscimed.2008.06.026
      OpenUrlCrossRefPubMedWeb of Science
    34. ↵
      2. Drefahl S
      . Do the married really live longer? The role of cohabitation and socioeconomic status. J Marriage Fam 2012;74:462–75.doi:10.1111/j.1741-3737.2012.00968.x
      OpenUrlCrossRef
    35. ↵
      2. Hattersley L ,
      3. Creeser R ,
      4. Study L
      . Longitudinal study 1971–1991. History, organisation and quality of data . London: HMSO, 1995.
    36. ↵
      2. Putter H ,
      3. Fiocco M ,
      4. Geskus RB
      . Tutorial in biostatistics: competing risks and multi-state models. Stat Med 2007;26:2389–430.doi:10.1002/sim.2712
      OpenUrlCrossRefPubMedWeb of Science
    37. ↵
      2. David HA ,
      3. Moeschberger ML
      . The theory of competing risks. London: Griffin, 1978.
    38. ↵
      Statistics ON. Deaths Registered in England and Wales (Series DR), 2011. Newport: Office for National Statistics, 2012.
    39. ↵
      2. Kalbfleisch JD ,
      3. Prentice RL
      . The statistical analysis of failure time data. 2nd edn. Hoboken: J. Wiley, 2002.
    40. ↵
      2. Norman P ,
      3. Boyle P
      . Are health inequalities between differently deprived areas evident at different ages? A longitudinal study of census records in England and Wales, 1991–2001. Health Place 2014;26:88–93.doi:10.1016/j.healthplace.2013.12.010
      OpenUrlCrossRefPubMed

    Footnotes

    • Contributors SF was the lead author for this paper, coauthored by HK. SF and HK both conceptualised and designed the study. SF planned the study, conducted data set-up and analysed and wrote the paper. HK significantly revised the draft manuscript.

    • Funding SF’s research was supported by the Economic and Social Research Council (ES/J500094/1) through the North West Doctoral Training Centre Social Statistics pathway (PhD project: ‘Health, Mortality and Partnership Status: Protection or Selection’). He also held an Advanced Quantitative Methods (AQM) enhanced stipend.

    • Disclaimer The authors alone are responsible for the interpretation of the data. This work contains statistical data from ONS, which is Crown Copyright. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets that may not exactly reproduce National Statistics aggregates.

    • Competing interests None declared.

    • Patient consent Not required.

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