Abstract

Background Living arrangements have changed markedly in recent decades, so we wanted to provide an up-to-date assessment of mortality as a function of marital status and cohabitation status in a complete population.

Methods We studied mortality in a national cohort of 6.5 million Danes followed for 122.5 million person-years during 1982–2011, using continuously updated individual-level information on living arrangements, socio-demographic covariates and causes of deaths. Hazard ratios (HRs) estimated relative mortality in categories of marital status, cohabitation status and combinations thereof.

Results HRs for overall mortality changed markedly over time, most notably for persons in same-sex marriage. In 2000–2011, opposite-sex married persons (reference, HR = 1) had consistently lower mortality than persons in other marital status categories in women (HRs 1.37–1.89) and men (HRs 1.37–1.66). Mortality was particularly high for same-sex married women (HR = 1.89), notably from suicide (HR = 6.40) and cancer (HR = 1.62), whereas rates for same-sex married men (HR = 1.38) were equal to or lower than those for unmarried, divorced and widowed men. Prior marriages (whether opposite-sex or same-sex) were associated with increased mortality in both women and men (HR = 1.16–1.45 per additional prior marriage).

Conclusion Our study provides a detailed account of living arrangements and their associations with mortality over three decades, thus yielding accurate and statistically powerful analyses of public health relevance to countries with marriage and cohabitation patterns comparable to Denmark’s. Of note, mortality among same-sex married men has declined markedly since the mid-1990s and is now at or below that of unmarried, divorced and widowed men, whereas same-sex married women emerge as the group of women with highest and, in recent years, even further increasing mortality.

Background

It has long been known that mortality is higher in unmarried, divorced and widowed than married people.1–5 Over the past decades, however, living arrangements have changed markedly in Western societies. We used national demographic data in Denmark6 (population 5.6 million in 2011) to create a complete day-by-day record of each adult citizen’s marital status history between 1968 and 2011. For the same period, we developed an address-based algorithm to also identify each citizen’s complete day-by-day cohabitation status history. Taking socioeconomic confounders available since 1982 into account, we here address how living arrangements were linked with overall and cause-specific mortality among Danish men and women between 1982 and 2011.

Materials and Methods

The cohort

From the Civil Registration System (CRS),7 we identified a cohort of 6.5 million adult persons (18 years or older) who resided in Denmark for any period between 1 January 1982 and 30 September 2011. Each cohort member’s unique personal identification number enabled identity-secure linkage of data between registers.

Marital status

For each day under observation, the CRS provided data on the marital status for all cohort members. Between 1 January 1982 and 30 September 1989, there were four marital status categories: (i) unmarried (never married), (ii) married, (ii) divorced or (iv) widowed. Since 1 October 1989, three additional categories appeared, namely homosexually married, divorced or widowed, following the implementation of the world’s first national law on registered same-sex partnerships;8 to gain statistical power we combined these categories in (v) current or former same-sex marriage. In analyses of prior marriages, we included day-by-day information about marital status since 1 April 1968, when the CRS was established.

Cohabitation status

From CRS data we created a variable capturing each cohort member’s continuously updated cohabitation status between 1982 and 2011 in five categories. Periods when cohort members lived with one or both parents were categorized as (i) living with parents. Periods when cohort members lived at the same address as ≥9 unrelated adults were categorized as (ii) living in multi-adult household. Periods when cohort members lived at an address that was not shared with a spouse or any unrelated adult person with an age difference less than +/− 15 years were categorized as (iii) living as a single person. Periods when cohort members lived at a specific address with only one other adult person (a spouse regardless of age difference or any unrelated adult person aged +/− 15 years) were categorized as (iv) living in opposite-sex cohabitation or (v) living in same-sex cohabitation, as appropriate. When cohort members shared address with 2–8 unrelated adult persons aged +/− 15 years, we used a two-step algorithm to define periods of living in opposite-sex cohabitation, same-sex cohabitation or multi-adult household, as specified in the Appendix. In analyses of prior cohabitations of at least 1 year’s duration, we included cohabitation data since 1 April 1968.

Mortality data

From the CRS we obtained dates of death between 1 January 1982 and 30 September 2011. We obtained information about specific causes of death until 31 December 2010 from the Danish Register of Causes of Death9 to study cause-specific mortality from (i) cardiovascular diseases, (ii) malignant neoplasms, (iii) respiratory tract diseases, (iv) suicide, (v) acquired immunodeficiency syndrome (AIDS) and (vi) other causes. Causes 1, 2, 3 and 4 represent four of 14 specific causes used in national mortality statistics. AIDS (cause 5) was defined from specific mortality codes (ICD-8 code 079.7 through 1993; ICD-10 groups B20-B24 since 1994). Cause 6 represents all causes other than causes 1–5. We included AIDS mortality to evaluate mortality in subgroups of same-sex cohabiting men, because AIDS mortality is markedly elevated in homosexual men.10,11

Covariates

From the CRS and Statistics Denmark we obtained daily or annually updated individual-level information about a number of potential socio-demographic confounders. Examined covariates included:

  • country of birth [three categories: Denmark (accounting for 86% of the observation time between 1982 and 2011), other Western countries (2% of observation time) or non-Western countries (12% of observation time)];

  • municipality (98 categories representing the municipalities present since 2007; during 1982–2006, Denmark’s then 271 to 275 municipalities were categorized geographically to resemble the 98 municipalities present since 2007; time-dependent covariate, 1-day intervals);

  • population density (five categories: 1–25, 26–350, 351–1000, 1001–2000 or >2000 persons per km2, as described elsewhere,12 time-dependent covariate, 1-day intervals);

  • children in household (persons aged <18 years; four categories: 0, 1, 2 or 3+; time-dependent covariate, 1-day intervals);

  • adults in household other than the index person and the spouse or cohabiting partner (in analyses of marital status and cohabitation status, respectively) (two categories: 0 or 1+; time-dependent covariate, 1-day intervals);

  • educational level (eight categories representing the index person’s highest obtained level of education: basic school, high school, vocational education, short higher education, medium higher education, long higher education, university education with academic degree or unknown level of education, time-dependent covariate, 1-year intervals); and

  • relative income 2 years before the actual year (in quartiles, calculated on the basis of income levels for persons of the same sex and birth year; time-dependent covariate, 1-year intervals). As in other studies13,14, we used income 2 years before to reduce possible misclassification due to disease-associated changes in socio-economic conditions.

Statistical analysis

Women and men were analysed separately. Cox proportional hazards regression analyses yielded hazard ratios (HRs) with 95% confidence intervals (CIs) as our measure of relative mortality. Unless otherwise specified, we used married persons and persons in opposite-sex cohabitation as reference categories. We used attained age (1-day intervals) as the underlying time scale in all analyses, except one specifically aimed to illustrate HRs as a function of calendar time, in which we used calendar time (1-day intervals) as the underlying time scale. In all analyses, we stratified for birth year (1-year intervals) thereby also taking main effects of calendar time (or attained age in the calendar time analysis) into account. We evaluated the above-mentioned potential confounders one by one in a Cox model stratified only for birth year. To be included as a stratification variable, we required a covariate to produce a change of ≥10% in the HR estimate for overall mortality in the period 1982–2011 for at least one examined marital status or cohabitation category compared with the corresponding estimate obtained in the simpler model stratified only for birth year. Four covariates met this criterion. Consequently, in addition to inherent control for age, birth year and calendar period, we stratified for actual values of municipality, population density, educational level and relative income 2 years before the actual year, unless otherwise specified.

Initial analyses of secular trends in HRs for overall mortality were based on data for the period 1982–2011. Most subsequent analyses were restricted to the period after 1 January 2000 to focus on data with greatest relevance for society today. Because serious illness may lead to changes in marital status or cohabitation status, we examined to what extent HRs for overall mortality for the period 1982–2011 were influenced by alternative definitions, using marital status and cohabitation status 6, 12, 24 or 60 months before the day of observation. Specifically, we compared the Akaike information criterion (AIC) obtained in these supplementary models with corresponding AIC scores in the main model to identify the best fitting model.15 Reassuringly, the preferred definition of cohabitation status, as judged by the highest AIC score, was cohabitation status on the actual day of observation. Analyses of marital status suggested optimal model fit when using marital status 6 months (men) or 12 months (women) before the actual day, but HR estimates in these analyses were generally similar to those based on actual marital status (not shown). Consequently, all presented results are based on actual marital status and actual cohabitation status.

Our address-based algorithm almost inevitably linked some individuals together as cohabiters notwithstanding that they might share address for a number of impersonal reasons. This consideration seems particularly important for periods of same-sex cohabitation, because such living arrangements should not be uncritically regarded as homosexual cohabitation.16 A non-trivial proportion of people living in same-sex cohabitation were probably non-homosexual persons living with a same-sex friend or tenant. We analysed AIDS mortality to explore this issue, knowing that 70% of AIDS cases in Danish men until December 2010 occurred in men who have sex with men.11

Proportional hazards assumption

We used age as the time scale in all but one analysis, because mortality is strongly age-dependent. Associations between the two explanatory variables, marital status and cohabitation status, and mortality also vary markedly by age, thus violating the proportional hazards assumption. To overcome this, we modelled HRs as a function of age using restricted cubic splines with knots every 10 years starting at age 25 years. Likewise, when analysing HRs as a function of calendar time we used restricted cubic splines with knots every 2, 4 or 5 years, depending on the particular marital status or cohabitation status category.17 To determine if sufficient knots had been introduced we plotted the martingale-based residuals as a function of the underlying time, which revealed visually satisfactory fit in all models.18 All analyses were performed using SAS version 9.2.

Results

Overall, 3.25 million men and 3.29 million women contributed at least 1 day of adult observation time between 1 January 1982 and 30 September 2011. A total of 1 709 850 deaths occurred in this period (853 919 in women, 855 931 in men).

Distribution of person-years in marital status and cohabitation status categories 1982–2011

By marital status, approximately half of the altogether 122.5 million person-years of observation were in married persons (51.3% for women; 53.6% for men) (Table 1). The remaining observation time was among unmarried (26.0%; 34.6%), divorced (9.8%; 8.0%), widowed (12.8%; 3.7%), and same-sex married (0.1%; 0.1%) women and men, respectively. By cohabitation status, most of the observation time was in opposite-sex cohabiters (61.1% in women; 63.9% in men), followed by single persons (30.9%; 22.7%), persons living with parents (4.0%; 7.5%), persons in multi-adult households (3.0%, 4.0%) and same-sex cohabiters (1.0%, 1.9%), respectively. The cross tabulation in Table 1 illustrates that, for example, for opposite-sex married women 94.0% of the time under observation was lived in opposite-sex cohabitation, whereas for women in opposite-sex cohabitation only 78.9% of the time under observation was lived in opposite-sex marriage. This is roughly equivalent to saying that around 94% of married women lived with a man (presumably the spouse), whereas 79% of women living with a man were married (presumably to that man).

Table 1

Distribution of person-years by marital status and cohabitation status among persons aged 18 years or older, Denmark 1982–2011

  Marital status  
  Person-years (column %; row %)
 
 
 Cohabitation status Unmarried persons Married persons (opposite sex) Divorced persons Widowed persons Same-sex married persons Total 
Women Persons living with parents 2 411 130 93 373 24 182 1362 463 2 530 509 
  (14.8%; 95.3%) (0.3%; 3.7%) (0.4%; 1.0%) (0.02%; 0.05%) (0.8%; 0.02%) (4.0%; 100%) 
 Single persons 6 159 325 1 530 454 4 318 147 7 315 570 10 459 19 333 956 
  (37.9%; 31.9%) (4.8%; 7.9%) (70.2%; 22.3%) (90.9%; 37.8%) (19.1%; 0.05%) (30.9%; 100%) 
 Opposite-sex cohabiting persons 6 074 147 30 194 930 1 608 955 385 507 869 38 264 408a 
  (37.3%; 15.9%) (94.0%; 78.9%) (26.2%; 4.2%) (4.8%; 1.0%) (1.6%; 0.002%) (61.1%; 100%) 
 Same-sex cohabiting persons 513 382 25 239 51 690 24 205 41 425 655 941b 
  (3.2%; 78.3%) (0.08%; 3.8%) (0.8%; 7.9%) (0.3%; 3.7%) (75.6%; 6.3%) (1.0%; 100%) 
 Persons in multi-adult households 1 114 241 279 397 149 021 320 188 1546 1 864 393 
  (6.8%; 59.8%) (0.9%; 15.0%) (2.4%; 8.0%) (4.0%; 17.2%) (2.8%; 0.08%) (3.0%; 100%) 
 Total 16 272 225 32 123 394 6 151 995 8 046 831 54 762 62 649 207 
  (100%; 26.0%) (100%; 51.3%) (100%; 9.8%) (100%; 12.8%) (100%; 0.09%) (100%;100%) 
        
Men Persons living with parents 4 328 150 123 525 49 349 703 490 4 502 217 
  (20.9%; 96.1%) (0.4%; 2.7%) (1.0%; 1.1%) (0.03%; 0.02%) (0.7%; 0.01%) (7.5%; 100%) 
 Single persons 7 614 203 1 315 541 2 810 421 1 842 362 20 320 13 602 847 
  (36.7%; 56.0%) (4.1%; 9.7%) (58.6%; 20.7%) (84.0%; 13.5%) (27.6%; 0.1%) (22.7%; 100%) 
 Opposite-sex cohabiting persons 6 267 299 30 205 625 1 538 462 251 961 1060 38 264 408a 
  (30.2%; 16.4%) (94.1%; 78.9%) (32.1%; 4.0%) (11.5%; 0.7%) (1.4%; 0.002%) (63.9%; 100%) 
 Same-sex cohabiting persons 842 458 76 567 135 857 9129 49 448 1 113 459b 
  (4.1%; 75.7%) (0.2%; 6.9%) (2.8%; 12.2%) (0.4%; 0.8%) (67.1%; 4.4%) (1.9%; 100%) 
 Persons in multi-adult households 1 690 082 365 505 264 449 89 966 2 360 2 412 363 
  (8.1%; 70.1%) (1.14%; 15.2%) (5.5%; 11.0%) (4.1%; 3.7%) (3.2%; 0.1%) (4.0%; 100%) 
 Total 20 742 193 32 086 764 4 798 539 2 194 120 73 679 59 895 294 
  (100%; 34.6%) (100%; 53.6%) (100%; 8.0%) (100%; 3.7%) (100%; 0.1%) (100%;100%) 
  Marital status  
  Person-years (column %; row %)
 
 
 Cohabitation status Unmarried persons Married persons (opposite sex) Divorced persons Widowed persons Same-sex married persons Total 
Women Persons living with parents 2 411 130 93 373 24 182 1362 463 2 530 509 
  (14.8%; 95.3%) (0.3%; 3.7%) (0.4%; 1.0%) (0.02%; 0.05%) (0.8%; 0.02%) (4.0%; 100%) 
 Single persons 6 159 325 1 530 454 4 318 147 7 315 570 10 459 19 333 956 
  (37.9%; 31.9%) (4.8%; 7.9%) (70.2%; 22.3%) (90.9%; 37.8%) (19.1%; 0.05%) (30.9%; 100%) 
 Opposite-sex cohabiting persons 6 074 147 30 194 930 1 608 955 385 507 869 38 264 408a 
  (37.3%; 15.9%) (94.0%; 78.9%) (26.2%; 4.2%) (4.8%; 1.0%) (1.6%; 0.002%) (61.1%; 100%) 
 Same-sex cohabiting persons 513 382 25 239 51 690 24 205 41 425 655 941b 
  (3.2%; 78.3%) (0.08%; 3.8%) (0.8%; 7.9%) (0.3%; 3.7%) (75.6%; 6.3%) (1.0%; 100%) 
 Persons in multi-adult households 1 114 241 279 397 149 021 320 188 1546 1 864 393 
  (6.8%; 59.8%) (0.9%; 15.0%) (2.4%; 8.0%) (4.0%; 17.2%) (2.8%; 0.08%) (3.0%; 100%) 
 Total 16 272 225 32 123 394 6 151 995 8 046 831 54 762 62 649 207 
  (100%; 26.0%) (100%; 51.3%) (100%; 9.8%) (100%; 12.8%) (100%; 0.09%) (100%;100%) 
        
Men Persons living with parents 4 328 150 123 525 49 349 703 490 4 502 217 
  (20.9%; 96.1%) (0.4%; 2.7%) (1.0%; 1.1%) (0.03%; 0.02%) (0.7%; 0.01%) (7.5%; 100%) 
 Single persons 7 614 203 1 315 541 2 810 421 1 842 362 20 320 13 602 847 
  (36.7%; 56.0%) (4.1%; 9.7%) (58.6%; 20.7%) (84.0%; 13.5%) (27.6%; 0.1%) (22.7%; 100%) 
 Opposite-sex cohabiting persons 6 267 299 30 205 625 1 538 462 251 961 1060 38 264 408a 
  (30.2%; 16.4%) (94.1%; 78.9%) (32.1%; 4.0%) (11.5%; 0.7%) (1.4%; 0.002%) (63.9%; 100%) 
 Same-sex cohabiting persons 842 458 76 567 135 857 9129 49 448 1 113 459b 
  (4.1%; 75.7%) (0.2%; 6.9%) (2.8%; 12.2%) (0.4%; 0.8%) (67.1%; 4.4%) (1.9%; 100%) 
 Persons in multi-adult households 1 690 082 365 505 264 449 89 966 2 360 2 412 363 
  (8.1%; 70.1%) (1.14%; 15.2%) (5.5%; 11.0%) (4.1%; 3.7%) (3.2%; 0.1%) (4.0%; 100%) 
 Total 20 742 193 32 086 764 4 798 539 2 194 120 73 679 59 895 294 
  (100%; 34.6%) (100%; 53.6%) (100%; 8.0%) (100%; 3.7%) (100%; 0.1%) (100%;100%) 

aOf the total observation time among opposite-sex cohabiting women and men (38 264 408 person-years in each sex), 98.3% were for cohabitants living together with no other unrelated adult persons in the household; the remainder were for cohabitants living with 2-8 other unrelated adults linked together by opposite-sex marriage (1.3%), shared children (0.2%), or history of prior simultaneous relocation (0.3%).

bOf the total observation time among same-sex cohabiting women (655 941 person-years) and men (1 113 459 person-years), 97.9% (women) and 97.8% (men) were for cohabitants living together with no other unrelated adult persons in the household; the remainder were for cohabitants living with 2-8 other unrelated adults linked together by same-sex marriage (0.2%, women; 0.2% men), shared children (0.0%, women; 0.0%, men), or history of prior simultaneous relocation (1.9%, women; 2.0%, men).

Secular trends in living arrangements 1982–2011

Marked changes in living arrangements took place during 1982–2011. Gradual declines were seen in proportions of married (women: from 58% to 49%; men: from 61% to 51%) and widowed persons (women: from 14% to 11%; men: from 4% to 3%), with corresponding increases in proportions of unmarried (women: from 21% to 30%; men: from 30% to 37%) and divorced persons (women: from 7% to 11%; men: from 6% to 9%) (Figure 1, upper panels). Changes in cohabitation status during 1982–2011 included decreasing proportions of people living in opposite-sex cohabitation (women: from 64% to 59%; men: from 68% to 61%) and corresponding increases in proportions of single persons (women: from 28% to 32%; men: from 19% to 26%). Proportions living with parents, in multi-adult households or in same-sex cohabitation were rather stable (Figure 1, lower panels).

Figure 1

Observation time between 1 January 1982 and 30 September 2011 (122.5 million person-years) divided in categories of marital status (upper panels) and cohabitation status (lower panels), Danish population age 18 years and older. The category of same-sex married persons is hardly visible, as this category accounted for only 0.1% of the total observation time (see Table 1)

Figure 1

Observation time between 1 January 1982 and 30 September 2011 (122.5 million person-years) divided in categories of marital status (upper panels) and cohabitation status (lower panels), Danish population age 18 years and older. The category of same-sex married persons is hardly visible, as this category accounted for only 0.1% of the total observation time (see Table 1)

Secular trends in HRs for overall mortality 1982–2011

Marital status

Being opposite-sex married (reference, HR = 1) was associated with consistently lower mortality than all other marital status categories (Figure 2, upper panels). Among women, HRs for widows (around 1.4) and divorcees (around 1.6) were rather stable, whereas HRs for unmarried women increased from around 1.5 to 1.7 between 1982 and 2011. Among men, increasing HRs over time were observed for widowers (from around 1.2 to 1.4), divorcees (from around 1.3 to 1.7) and, most markedly, for unmarried men (from around 1.2 to 1.7). Mortality was markedly elevated among persons in same-sex marriage in the first decade after its introduction in 1989. Since the year 2000, mortality among same-sex married women has remained higher than in all other marital status categories, with a tendency towards increasing HRs in recent years. In contrast, mortality among same-sex married men has reached a level below that of unmarried and divorced men and similar to that of widowed men (HR around 1.4 in 2011).

Figure 2

Secular trends in hazard ratios for overall mortality in women and men by marital status (upper panels) or cohabitation status (lower panels), Danish population age 18 years and older, 1 January 1982 through 30 September 2011. Calendar time in 1-day intervals used as underlying time scale in Cox proportional hazards regression model

Figure 2

Secular trends in hazard ratios for overall mortality in women and men by marital status (upper panels) or cohabitation status (lower panels), Danish population age 18 years and older, 1 January 1982 through 30 September 2011. Calendar time in 1-day intervals used as underlying time scale in Cox proportional hazards regression model

Cohabitation status.

Persons in opposite-sex cohabitation (reference, HR = 1) consistently had the lowest mortality (Figure 2, lower panels). HRs were markedly elevated for persons in multi-adult households (women: 4.4–5.9; men: 2.7–4.0). Also, persons living with parents had increased HRs (women: 1.6–2.3; men: 1.5–1.9). Among women, increasing HRs were seen for those living as a single (from 1.4 to 1.8 between 1982 and 2011) and those living in same-sex cohabitation (from 1.6 to 2.2). An even more pronounced increase in mortality was seen among single men (from 1.3 to 2.1), whereas HRs were rather stable (1.5–1.7) for same-sex cohabiting men throughout the study period.

Age-specific HRs for overall mortality 2000–2011

We next studied overall mortality between 1 January 2000 and 30 September 2011 (n = 652 159 deaths).

Marital status

Age-specific HR patterns for unmarried and divorced persons were rather similar, with mortality rates exceeding those of married persons in all age groups above 30 years (Table 2; Figure 3, upper panels). Among women, peaks were in unmarried (HR = 1.9) and divorced (HR = 1.7) 50–69 year-olds, whereas among men, peaks occurred around 20 years earlier in unmarried (HR = 2.0) and divorced (HR = 2.1) 30–49 year-olds. In all age groups, women in same-sex marriage had markedly increased mortality (HRs between 1.7 and 3.5). For men, widowers and those in same-sex marriage had the highest HRs below age 50 years. However, above that age same-sex married men experienced the lowest mortality among all groups of men who were not in opposite-sex marriage.

Figure 3

Age-specific hazard ratios for overall mortality in women and men by marital status (upper panels) or cohabitation status (lower panels), Danish population age 18 years and older, 1 January 2000 through 30 September 2011. Attained age in 1-day intervals used as underlying time scale in Cox proportional hazards regression model. The starting point for each graph was determined by the youngest age below which at least five deaths had occurred within the particular marital status or cohabitation status category. For persons living with parents we show hazard ratios for the age interval 18–49 years, because only few people live with their parents after age 50 years. For all other marital status and cohabitation status categories, graphs are shown for all ages below 90 years

Figure 3

Age-specific hazard ratios for overall mortality in women and men by marital status (upper panels) or cohabitation status (lower panels), Danish population age 18 years and older, 1 January 2000 through 30 September 2011. Attained age in 1-day intervals used as underlying time scale in Cox proportional hazards regression model. The starting point for each graph was determined by the youngest age below which at least five deaths had occurred within the particular marital status or cohabitation status category. For persons living with parents we show hazard ratios for the age interval 18–49 years, because only few people live with their parents after age 50 years. For all other marital status and cohabitation status categories, graphs are shown for all ages below 90 years

Table 2

Overall mortality. Hazard ratios with 95% confidence intervals by age in different marital status and cohabitation status categories, Denmark 2000–2011

  Marital status Cohabitation status 
  Hazard ratio (95% confidence interval) N = No. Deaths
 
Hazard ratio (95% confidence interval) N = No. deaths
 
  Married persons (opposite sex) Unmarried persons Divorced persons Widowed persons Same-sex married persons Opposite-sex cohabiting persons Persons living with parents Single persons Same-sex cohabiting persons Persons in multi-adult households 
Women 
Age (years) 18–29 1 (ref) N = 186 1.09 (0.90–1.33) N = 965 1.37 (0.86–2.19) N = 27 7.32 (1.52–35.2) N = 3 3.54 (1.19–10.5) N = 5 1 (ref) N = 374 1.55 (1.22–1.96) N = 223 1.94 (1.65–2.28) N = 427 1.50 (1.02–2.21) N = 32 2.00 (1.56–2.57) N = 130 
 30–49 1 (ref) N = 4899 1.60 (1.52–1.69) N = 3054 1.58 (1.49–1.68) N = 2012 2.40 (2.06–2.78) N = 234 2.01 (1.41–2.87) N = 38 1 (ref) N = 5352 2.59 (2.17–3.09) N = 194 2.16 (2.06–2.27) N = 4122 1.47 (1.14–1.89) N = 77 3.09 (2.73–3.49) N = 492 
 50–69 1 (ref) N = 30 466 1.89 (1.82–1.95) N = 5355 1.71 (1.67–1.76) N = 13 480 1.54 (1.50–1.59) N = 9305 1.92 (1.51–2.44) N = 87 1 (ref) N = 30 713 1.75 (1.25–2.45) N = 47 1.90 (1.86–1.94) N = 26 332 1.61 (1.39–1.86) N = 226 4.63 (4.32–4.95) N = 1375 
 70+ 1 (ref) N = 42 511 1.47 (1.44–1.51) N = 18 335 1.44 (1.41–1.48) N = 24 119 1.28 (1.25–1.30) N = 178 362 1.72 (1.26–2.35) N = 57 1 (ref) N = 36 193 1.09 (0.15–8.15) N = 1 1.65 (1.62–1.68) N = 236 030 2.19 (2.03–2.36) N = 1024 5.95 (5.80–6.11) N = 20 136 
 Overall 1 (ref) N = 78 062 1.62 (1.59–1.65) N = 27 709 1.57 (1.55–1.60) N = 39 638 1.37 (1.36–1.39) N = 187 904 1.89 (1.60–2.23) N = 187 1 (ref) N = 72 632 2.12 (1.87–2.41) N = 465 1.76 (1.74–1.78) N = 236 911 2.04 (1.92–2.18) N = 1359 5.87 (5.74–6.00) N = 22 133 
Men 
Age (years) 18–29 1 (ref) N = 205 1.29 (1.09–1.53) N = 3051 1.27 (0.79–2.03) N = 27         − N = 0 2.23 (0.54–9.26) N = 2 1 (ref) N = 486 2.00 (1.74–2.31) N = 1039 2.69 (2.38–3.03) N = 1156 2.20 (1.79–2.71) N = 145 2.95 (2.52–3.46) N = 459 
 30–49 1 (ref) N = 5521 2.00 (1.92–2.08) N = 9224 2.09 (1.99–2.20) N = 3127 2.73 (2.24–3.33) N = 129 2.44 (1.93–3.07) N = 92 1 (ref) N = 5971 2.47 (2.26–2.71) N = 746 3.18 (3.06–3.30) N = 9347 2.43 (2.20–2.69) N = 542 3.06 (2.84–3.31) N = 1487 
 50–69 1 (ref) N = 46 524 1.78 (1.74–1.81) N = 15 830 1.86 (1.82–1.89) N = 20 838 1.69 (1.64–1.75) N = 5054 1.43 (1.24–1.65) N = 231 1 (ref) N = 45 780 1.49 (1.27–1.76) N = 192 2.26 (2.22–2.30) N = 38 900 1.67 (1.56–1.80) N = 934 2.85 (2.71–2.99) N = 2671 
 70+ 1 (ref) N = 105 887 1.46 (1.43–1.49) N = 16 171 1.45 (1.43–1.48) N = 18 457 1.28 (1.26–1.30) N = 68 132 1.11 (0.93–1.32) N = 157 1 (ref) N = 90 862 0.88 (0.12–6.66) N = 2 1.71 (1.69–1.73) N = 106 802 1.63 (1.47–1.81) N = 538 4.50 (4.36–4.64) N = 10 600 
 Overall 1 (ref) N = 158 137 1.63 (1.61–1.65) N = 44 276 1.66 (1.64–1.68) N = 42 449 1.37 (1.35–1.38) N = 73 315 1.38 (1.25–1.53) N = 482 1 (ref) N = 143 099 1.79 (1.68–1.90) N = 1979 1.95 (1.94–1.97) N = 156 205 1.70 (1.62–1.79) N = 2159 3.70 (3.61–3.78) N = 15 217 
  Marital status Cohabitation status 
  Hazard ratio (95% confidence interval) N = No. Deaths
 
Hazard ratio (95% confidence interval) N = No. deaths
 
  Married persons (opposite sex) Unmarried persons Divorced persons Widowed persons Same-sex married persons Opposite-sex cohabiting persons Persons living with parents Single persons Same-sex cohabiting persons Persons in multi-adult households 
Women 
Age (years) 18–29 1 (ref) N = 186 1.09 (0.90–1.33) N = 965 1.37 (0.86–2.19) N = 27 7.32 (1.52–35.2) N = 3 3.54 (1.19–10.5) N = 5 1 (ref) N = 374 1.55 (1.22–1.96) N = 223 1.94 (1.65–2.28) N = 427 1.50 (1.02–2.21) N = 32 2.00 (1.56–2.57) N = 130 
 30–49 1 (ref) N = 4899 1.60 (1.52–1.69) N = 3054 1.58 (1.49–1.68) N = 2012 2.40 (2.06–2.78) N = 234 2.01 (1.41–2.87) N = 38 1 (ref) N = 5352 2.59 (2.17–3.09) N = 194 2.16 (2.06–2.27) N = 4122 1.47 (1.14–1.89) N = 77 3.09 (2.73–3.49) N = 492 
 50–69 1 (ref) N = 30 466 1.89 (1.82–1.95) N = 5355 1.71 (1.67–1.76) N = 13 480 1.54 (1.50–1.59) N = 9305 1.92 (1.51–2.44) N = 87 1 (ref) N = 30 713 1.75 (1.25–2.45) N = 47 1.90 (1.86–1.94) N = 26 332 1.61 (1.39–1.86) N = 226 4.63 (4.32–4.95) N = 1375 
 70+ 1 (ref) N = 42 511 1.47 (1.44–1.51) N = 18 335 1.44 (1.41–1.48) N = 24 119 1.28 (1.25–1.30) N = 178 362 1.72 (1.26–2.35) N = 57 1 (ref) N = 36 193 1.09 (0.15–8.15) N = 1 1.65 (1.62–1.68) N = 236 030 2.19 (2.03–2.36) N = 1024 5.95 (5.80–6.11) N = 20 136 
 Overall 1 (ref) N = 78 062 1.62 (1.59–1.65) N = 27 709 1.57 (1.55–1.60) N = 39 638 1.37 (1.36–1.39) N = 187 904 1.89 (1.60–2.23) N = 187 1 (ref) N = 72 632 2.12 (1.87–2.41) N = 465 1.76 (1.74–1.78) N = 236 911 2.04 (1.92–2.18) N = 1359 5.87 (5.74–6.00) N = 22 133 
Men 
Age (years) 18–29 1 (ref) N = 205 1.29 (1.09–1.53) N = 3051 1.27 (0.79–2.03) N = 27         − N = 0 2.23 (0.54–9.26) N = 2 1 (ref) N = 486 2.00 (1.74–2.31) N = 1039 2.69 (2.38–3.03) N = 1156 2.20 (1.79–2.71) N = 145 2.95 (2.52–3.46) N = 459 
 30–49 1 (ref) N = 5521 2.00 (1.92–2.08) N = 9224 2.09 (1.99–2.20) N = 3127 2.73 (2.24–3.33) N = 129 2.44 (1.93–3.07) N = 92 1 (ref) N = 5971 2.47 (2.26–2.71) N = 746 3.18 (3.06–3.30) N = 9347 2.43 (2.20–2.69) N = 542 3.06 (2.84–3.31) N = 1487 
 50–69 1 (ref) N = 46 524 1.78 (1.74–1.81) N = 15 830 1.86 (1.82–1.89) N = 20 838 1.69 (1.64–1.75) N = 5054 1.43 (1.24–1.65) N = 231 1 (ref) N = 45 780 1.49 (1.27–1.76) N = 192 2.26 (2.22–2.30) N = 38 900 1.67 (1.56–1.80) N = 934 2.85 (2.71–2.99) N = 2671 
 70+ 1 (ref) N = 105 887 1.46 (1.43–1.49) N = 16 171 1.45 (1.43–1.48) N = 18 457 1.28 (1.26–1.30) N = 68 132 1.11 (0.93–1.32) N = 157 1 (ref) N = 90 862 0.88 (0.12–6.66) N = 2 1.71 (1.69–1.73) N = 106 802 1.63 (1.47–1.81) N = 538 4.50 (4.36–4.64) N = 10 600 
 Overall 1 (ref) N = 158 137 1.63 (1.61–1.65) N = 44 276 1.66 (1.64–1.68) N = 42 449 1.37 (1.35–1.38) N = 73 315 1.38 (1.25–1.53) N = 482 1 (ref) N = 143 099 1.79 (1.68–1.90) N = 1979 1.95 (1.94–1.97) N = 156 205 1.70 (1.62–1.79) N = 2159 3.70 (3.61–3.78) N = 15 217 

Hazard ratios obtained in Cox regression models with age as the underlying time scale stratified for birth year and socioeconomic confounders (municipality, population density, educational level, and relative income two years before the actual year).

Cohabitation status

At all ages the reference category of opposite-sex cohabiters had lowest mortality (Table 2; Figure 3, lower panels). Highest HRs were seen for single men (peak around age 40 years) and persons in multi-adult households (peaks around age 35 years in women and 45 years in men and, most notably, above age 70 years in both sexes). Among women below age 70 years, lowest HRs, although consistently above 1.3, were in same-sex cohabiters. Among men, same-sex cohabiters also had the lowest HRs, although HRs in this category were not below 1.6 at any age.

HRs for overall mortality as a function of prior marriages and prior cohabitations 2000–2011

So far, we have studied associations of actual marital status and actual cohabitation status with overall mortality without considering the possible additional role of prior marriages or cohabitations. Because information about marriages and cohabitations was only available since 1 April 1968, we studied the impact of prior marriages and cohabitations in a subcohort of 3 484 537 persons born 1 April 1950 or later who were 18–61 years old during follow-up between 2000 and 2011 and for whom we could establish complete marital status and cohabitation status records since age 18 years (Table 3). There were strong positive associations between the number of prior opposite-sex marriages and overall mortality in both women and men (27% and 16% increase in mortality among women and men, respectively, per additional prior opposite-sex marriage). Likewise, although numbers were sparse, mortality increased in both women and men (by 45% and 42%, respectively), with each additional prior same-sex marriage. Mortality also increased (by 11% in women and 4% in men) with each additional prior opposite-sex cohabitation of at least 1 year’s duration, whereas mortality was not (women) or only marginally (men) associated with the number of prior same-sex cohabitations of at least 1 year’s duration (Table 3).

Table 3

Overall mortality. Hazard ratios with 95% confidence intervals according to number of prior marriages and cohabitations of at least 1 year's duration since 1968 among 18-61 year-old persons born 1 April 1950 or later, Denmark 2000–2011

graphic 
graphic 

Hazard ratios for the number of prior opposite-sex marriages (and prior same-sex marriages) obtained in Cox regression models with age as the underlying time scale with adjustment for actual marital status and number of prior same-sex (opposite-sex) marriages, and stratified for birth year and socioeconomic confounders (municipality, population density, educational level, and relative income two years before the actual year). Hazard ratios for the number of prior opposite-sex cohabitations (and prior same-sex cohabitations) of at least 1 year's duration are calculated with adjustment for actual cohabitation status and number of prior same-sex (opposite-sex) cohabitations of at least 1 year's duration with similar stratifications for birth year and socioeconomic confounders.

HRs for overall mortality as a function of both marital status and cohabitation status 2000–2011

Having explored patterns of mortality among categories of marital status and cohabitation status separately, we next combined them to examine, within each marital status, mortality among specific cohabitation status categories (Figure 4), restricting the focus to single persons, opposite-sex cohabiters and same-sex cohabiters to reduce complexity. Patterns were similar in women and men. Whether unmarried, married, divorced or widowed, compared with opposite-sex cohabiters, HRs were consistently higher among those living as a single (women: 1.4–2.5; men: 1.6–2.5) and those in same-sex cohabitation (women: 1.2–2.9; men: 1.6–2.5). Among same-sex married persons, a category comprising both currently and formerly same-sex married persons, HRs were higher (1.4 in both women and men) among those currently living as a single than among those in same-sex cohabitation.

Figure 4

Hazard ratios for overall mortality in women (upper panel) and men (lower panel) by cohabitation status in different categories of marital status, Danish population age 18 years and older, 1 January 2000 through 30 September 2011. Attained age in 1-day intervals used as underlying time scale in Cox proportional hazards regression model. WS = single women, WM = opposite-sex cohabiting women, WW = same-sex cohabiting women, MS = single men, MW = opposite-sex cohabiting men, MM = same-sex cohabiting men

Figure 4

Hazard ratios for overall mortality in women (upper panel) and men (lower panel) by cohabitation status in different categories of marital status, Danish population age 18 years and older, 1 January 2000 through 30 September 2011. Attained age in 1-day intervals used as underlying time scale in Cox proportional hazards regression model. WS = single women, WM = opposite-sex cohabiting women, WW = same-sex cohabiting women, MS = single men, MW = opposite-sex cohabiting men, MM = same-sex cohabiting men

We also examined, within each cohabitation status, mortality among specific marital status categories (Figure 5). Among persons living as a single, widows/widowers had low HRs, and married persons had high HRs, whereas divorced persons had mortality rates comparable to the reference group of unmarried persons. Among single-living men, those in current or former same-sex marriage had slightly lower mortality than unmarried men. Conversely, among single-living women, those in current or former same-sex marriage had higher mortality than unmarried women.

Figure 5

Hazard ratios for overall mortality in women (upper panel) and men (lower panel) by marital status in different categories of cohabitation status, Danish population age 18 years and older, 1 January 2000 through 30 September 2011. Attained age in 1-day intervals used as underlying time scale in Cox proportional hazards regression model. U = unmarried persons, M = opposite-sex married persons, D = opposite-sex divorced persons, W = opposite-sex widowed persons, S = same-sex married, divorced or widowed persons

Figure 5

Hazard ratios for overall mortality in women (upper panel) and men (lower panel) by marital status in different categories of cohabitation status, Danish population age 18 years and older, 1 January 2000 through 30 September 2011. Attained age in 1-day intervals used as underlying time scale in Cox proportional hazards regression model. U = unmarried persons, M = opposite-sex married persons, D = opposite-sex divorced persons, W = opposite-sex widowed persons, S = same-sex married, divorced or widowed persons

Among opposite-sex cohabiters, mortality patterns in marital status subgroups differed noticeably between the sexes. Among opposite-sex cohabiting women, lowest mortality was in married women, with successively higher HRs among unmarried, widowed and divorced women. In contrast, among opposite-sex cohabiting men, mortality was lower among unmarried or widowed men than among married or divorced men.

Among same-sex cohabiters, mortality was highest among those who were opposite-sex married or widowed; among same-sex cohabiting men those in same-sex marriage had lowest mortality.

HRs for cause-specific mortality 2000–2010

We analysed HRs of dying from cardiovascular diseases, malignant neoplasms, respiratory tract diseases, suicide, AIDS and ‘other causes’ between 1 January 2000 and 31 December 2010 (n = 613 380 deaths) (Table 4). Unmarried, divorced and widowed persons had consistently elevated HRs for each of the six specified causes of death. Same-sex married women also had increased mortality, notably from suicide (HR = 6.4), cancer (HR = 1.6) and ‘other causes’ (HR = 1.5). Among same-sex married men, HRs for cardiovascular disease and cancer mortality, the two most common causes of death in Denmark, were inconspicuous, whereas deaths from suicide (HR = 4.1), AIDS (HR = 356) and ‘other causes’ (HR = 1.6) were increased.

Table 4

Cause-specific mortality. Hazard ratios with 95% confidence intervals according to actual marital status and actual cohabitation status among persons aged 18 years or older, Denmark 2000–2010

graphic 
graphic 
graphic 
graphic 

Hazard ratios obtained in Cox regression models with age as the underlying time scale stratified for birth year and socioeconomic confounders (municipality, population density, educational level, and relative income two years before the actual year).

HRs for cause-specific mortality varied considerably more across cohabitation status categories. Without exception in either sex, cause-specific mortality in each of the studied cohabitation status categories was higher than that observed for the reference category of opposite-sex cohabiters. Particularly high HRs were seen for persons in multi-adult households. Likewise, AIDS mortality was markedly elevated among same-sex cohabiting men (HR = 49.8), men in multi-adult cohabitation (HR = 23.6) and single men (HR = 15.1), as well as among women in multi-adult households (HR = 18.4).

Finally, we examined to what extent same-sex cohabitating men might serve as a proxy for cohabiting homosexual men (Table 5). When compared with men in opposite-sex cohabitation, HRs for AIDS mortality between 1982 and 2010 were rather extreme for men whose actual (HR = 74.3) or cumulative period in same-sex cohabitation since 1968 (HR = 62.4) exceeded 60 months. Also, regardless of actual cohabitation status, men with a cumulative period of ≥ 60 months in same-sex cohabitation since 1968, had markedly increased risk (HR = 17.9) compared with men who never cohabited with another man. Combining information about marital status, cumulative period in same-sex cohabitation and municipality yielded greater specificity; in a Cox model with age as the underlying time scale and stratified for birth year, educational level and relative income, unmarried or divorced men in Copenhagen with ≥60 months of same-sex cohabitation since 1968 were markedly more likely to die from AIDS (113 AIDS deaths during 110 132 person-years) than married or widowed men in other parts of the country who had never cohabited with another man (161 AIDS deaths during 26 155 610 person-years; HR = 139; 95% CI: 108–179).

Table 5

AIDS mortality. Hazard ratios with 95% confidence intervals among men aged 18 years or older, Denmark 1982-2010

 Person-years Number of AIDS deaths Hazard ratio (95% confidence interval) 
Actual cohabitation status 
 Opposite-sex cohabiting men 37 267 097 250 (ref) 
 Men living with parents 4 392 268 82 7.34 (5.33–10.1) 
 Single men 13 189 584 1135 9.36 (7.99–11.0) 
 Men living in multi-adult households 2 349 313 125 4.38 (3.36–5.70) 
 Same-sex cohabiting men (duration of actual same-sex cohabitation) 
  <6 months 423 607 35 8.26 (5.61–12.2) 
  6–11 months 192 980 25 9.93 (6.24–15.8) 
  12–23 months 170 012 40 18.9 (13.0–27.4) 
  24–59 months 140 224 79 32.1 (23.9–43.0) 
  ≥60 months 152 077 174 74.3 (58.2–94.8) 
     
 Opposite-sex cohabiting men 37 267 097 250 (ref) 
 Men living with parents 4 392 268 82 7.35 (5.33–10.1) 
 Single men 13 189 584 1135 9.41 (8.04–11.0) 
 Men living in multi-adult households 2 349 313 125 4.37 (3.35–5.69) 
 Same-sex cohabiting men (duration of same-sex cohabitation since 1 April 1968) 
  <6 months 252 263 3.10 (1.14–8.43) 
  6–11 months 172 069 10 6.99 (3.62–13.5) 
  12–23 months 211 408 27 9.37 (5.99–14.6) 
  24–59 months 236 108 83 18.9 (14.3–25.2) 
  ≥60 months 207 054 228 62.4 (50.1–77.7) 
     
Cumulative period of same-sex cohabitation since 1 April 1968 
  Never 43 412 580 543 ref 
  <6 months 5 969 779 213 1.67 (1.39–1.99) 
  6–11 months 2 861 744 147 2.27 (1.86–2.78) 
  12–23 months 2 805 803 235 3.12 (2.61–3.71) 
  24–59 months 2 502 723 385 5.50 (4.72–6.41) 
  ≥60 months 724 534 422 17.9 (15.3–21.0) 
 Person-years Number of AIDS deaths Hazard ratio (95% confidence interval) 
Actual cohabitation status 
 Opposite-sex cohabiting men 37 267 097 250 (ref) 
 Men living with parents 4 392 268 82 7.34 (5.33–10.1) 
 Single men 13 189 584 1135 9.36 (7.99–11.0) 
 Men living in multi-adult households 2 349 313 125 4.38 (3.36–5.70) 
 Same-sex cohabiting men (duration of actual same-sex cohabitation) 
  <6 months 423 607 35 8.26 (5.61–12.2) 
  6–11 months 192 980 25 9.93 (6.24–15.8) 
  12–23 months 170 012 40 18.9 (13.0–27.4) 
  24–59 months 140 224 79 32.1 (23.9–43.0) 
  ≥60 months 152 077 174 74.3 (58.2–94.8) 
     
 Opposite-sex cohabiting men 37 267 097 250 (ref) 
 Men living with parents 4 392 268 82 7.35 (5.33–10.1) 
 Single men 13 189 584 1135 9.41 (8.04–11.0) 
 Men living in multi-adult households 2 349 313 125 4.37 (3.35–5.69) 
 Same-sex cohabiting men (duration of same-sex cohabitation since 1 April 1968) 
  <6 months 252 263 3.10 (1.14–8.43) 
  6–11 months 172 069 10 6.99 (3.62–13.5) 
  12–23 months 211 408 27 9.37 (5.99–14.6) 
  24–59 months 236 108 83 18.9 (14.3–25.2) 
  ≥60 months 207 054 228 62.4 (50.1–77.7) 
     
Cumulative period of same-sex cohabitation since 1 April 1968 
  Never 43 412 580 543 ref 
  <6 months 5 969 779 213 1.67 (1.39–1.99) 
  6–11 months 2 861 744 147 2.27 (1.86–2.78) 
  12–23 months 2 805 803 235 3.12 (2.61–3.71) 
  24–59 months 2 502 723 385 5.50 (4.72–6.41) 
  ≥60 months 724 534 422 17.9 (15.3–21.0) 

Hazard ratios obtained in Cox regression models with age as the underlying time scale stratified for birth year and socioeconomic confounders (municipality, population density, educational level, and relative income two years before the actual year).

Discussion

Our study expands century-old knowledge that married people generally have lower mortality than unmarried and divorced persons.1 Over the years, authors have investigated the role of marital status2–5 but no prior study has explored overall and cause-specific mortality in an entire country using complete day-by-day individual-level information about actual living arrangements over almost 30 years with additional 14 years of marriage and cohabitation history to refine cumulative exposures. Our study has several advantages. With a total of 6.5 million adult persons followed for 122.5 million person-years, during which period a total of 1.71 million deaths occurred, our study had unprecedented power to provide precise HR estimates and even enabled meaningful analyses of persons in same-sex marriage. Another asset is that we did not rely on self-reports for any explanatory variables, covariates or outcomes. Rather, our study relied on high-quality national registers that use as key the unique personal identifier ascribed to Danish residents at birth or immigration,7 thus ensuring accurate linkage of data between registers. Ethnically and socially Denmark is a relatively homogeneous population with only 12% of the observation time between 1982 and 2011 in persons born in non-Western countries; we examined if country of birth was a potential confounder, finding no evidence that HRs for overall mortality were materially influenced by ethnicity.

We also studied mortality according to cohabitation status. Most married people (94%) lived with an opposite-sex partner, and the remainder lived alone, in multi-adult households, with parents or in same-sex cohabitation. By combining information about marital status and cohabitation status, our study revealed 2-fold or higher mortality in married persons not living with their spouse (Figure 4), a finding that has not been reported before. Another novel observation was that being married was not always protective. Among single-living persons and same-sex cohabiters, opposite-sex married men and women had noticeably higher mortality than unmarried and same-sex married persons, respectively (Figure 5).

In several situations, mortality patterns according to cohabitation status were parallel to findings for marital status. This is perhaps unsurprising considering that most (94%) married women and men lived in opposite-sex cohabitation and, conversely, most (79%) opposite-sex cohabiters were married. Of note, however, unmarried persons and single-living persons, the second largest categories, were far less overlapping. Observation time among unmarried persons was divided evenly between single-living persons and opposite-sex cohabiters whereas, among single-living women, widows constituted the largest marital status subcategory. Marital status and cohabitation status offer different subdivisions of the population’s observation time, and each measure has its strengths and limitations.

A recent study in Finland, England and Wales included, in weighted numbers, a total of 4.8 million person-years and 165 963 deaths.19 Like in prior studies and the present larger Danish cohort, the authors showed that married people have lower mortality than non-married peers. Moreover, adding marital history information to their regression model improved model fit considerably, implying that marital status history conveys important information beyond that captured by marital status at baseline. We took this a step further; whereas marital status history in the Finnish/British/Welsh study was approximated by census data on marital status from 1971, 1981 and 1991, we utilized day-by-day marital status information available in the Danish population since 1968. Our data showed that, after controlling for actual marital status and confounders, HRs for overall mortality increased by a remarkable 27% in women and 16% in men per additional prior opposite-sex marriage. To our knowledge, no prior study has provided estimates of this kind. Our results show that whereas Danish women and men are rather equally susceptible to the effects of actual marital status (Table 2), effects of prior marriages are stronger in women than men (Table 3). A noteworthy exception to the similarities between women and men with respect to actual marital status was for same-sex marriage, which was more strongly associated with mortality in women (HR = 1.89) than men (HR = 1.38). Lesbians may constitute a largely unnoticed high-risk population for suicide20–22 and breast cancer,23–25 so our findings call for efforts to identify responsible underlying factors and ensure access to basic health care in this minority population.

Mortality differences according to marital status and cohabitation status may have a number of explanations, either working alone or in combination. A beneficial effect of marriage or, conversely, a negative effect of divorce or widowhood, has been proposed, which might reflect the influence of better financial status, healthier lifestyles and higher levels of social support among the married19,26 and it is reasonable to assume that such advantages may be shared to some extent by persons in unmarried cohabitation. Alternatively, marriage and cohabitation per se may not be all that important. Rather, personal traits associated with health status and subsequent mortality may influence the likelihood of entering and staying in marriage or cohabitation. If so, the lower mortality among the married or cohabiting than among the unmarried or single might reflect basic selection mechanisms that differentially lead greater proportions of healthy than unhealthy individuals into marriage and cohabitation.

A number of social factors may affect health and, thus, be associated with mortality in ways that might confound associations of living arrangements with mortality. To reduce the risk of spurious associations, we examined a number of socio-demographic factors as potential confounders; of seven covariates examined, only four met our confounder criteria, including educational level, relative income, municipality and population density. We cannot exclude the possibility that additional, unmeasured factors associated with social disadvantage or poor health were influential, for example in the observed high mortality among people in multi-adult households. However, by stratifying our analyses for the four identified confounding factors, we believe our HR estimates represent valid expressions of associations between living arrangements and mortality in Denmark.

With due caution our novel address-based cohabitation variable permits the study of minority groups of non-heterosexual people with unprecedented statistical power and what appears to be acceptably high levels of specificity. So far, the possibility to study demographically complete subsets of homosexuals has largely been restricted to countries where same-sex marriage is legal, as exemplified by studies of cancer incidence, suicide risk and overall mortality among same-sex married lesbians and gays.21,27,28 Here, we provide updated mortality estimates for same-sex married lesbians and gays, whom we followed for approximately 55 000 and 74 000 person-years, respectively, between 1989 and 2011. A particularly promising feature is that, in addition to studying these well-defined and complete groups of same-sex married persons, our cohabitation algorithm helped to identify another larger group of persons that most likely comprised a major proportion of cohabiting homosexual men. Specifically, we delineated 110 132 person-years of observation between 1982 and 2010 among unmarried or divorced men in Copenhagen who had lived ≥60 months in same-sex cohabitation since age 18 years. In this group, which was around 50% larger and presumably covered broader segments of homosexual cohabiters than those in same-sex marriage, AIDS mortality was 139-fold increased over that of married or widowed men in other parts of Denmark who never lived with another man. Although plausible, due to the lack of similarly specific outcomes for lesbians we cannot tell to what extent unmarried or divorced women in long-term same-sex cohabitation will serve as an equally good proxy for cohabiting lesbians. Future studies should examine these matters in more detail, because population-based investigations are needed to address understudied health needs in lesbians and gay men.25,29–32

From a public health viewpoint it is important to try to identify those underlying factors and mechanisms that explain the lower mortality among married and cohabiting persons. Several studies have shown that married and cohabiting persons generally have healthier lifestyles than unmarried and single persons.33–35 By including several socio-demographic variables as stratification variables in our study, notably education, income and municipality, we probably accounted partly for the variation in such lifestyle factors. However, since our findings are not cleaned of the influence of tobacco smoking, drug use, excessive alcohol consumption, overweight, sedentarism, unhealthy diets and other indicators of poor health or social disadvantage, such factors are likely to explain, at least in part, the increased mortality among persons not currently married or living in opposite-sex cohabitation.33-37

Overall, our study confirmed and expanded known associations between marital status and mortality, and we provide the first national assessment of links between cohabitation status and mortality. Among advantages of studying marital status is the likelihood of capturing people in strongly committed relationships. However, a non-negligible minority (around 6%) of opposite-sex married persons did not live with their spouse. Non-married cohabitation may more accurately divide persons in relevant living arrangements in countries where unmarried cohabitation is broadly accepted. On the other hand, cohabitation does not necessarily imply a close personal relationship, as people may share address for other reasons as well. Depending on the setting, some research questions may be best addressed using marital status whereas others may benefit from cohabitation algorithms like ours. For instance, future studies of health patterns among homosexual persons will gain statistical power, and presumably be more representative of the target population, when combining information about marital status (being unmarried, divorced or, if optional in the particular geographical setting, in same-sex marriage) and cohabitation status (having lived for a non-trivial period in same-sex cohabitation). Importantly, however, researchers should examine the relevance of such algorithms in their own setting, because marriage and cohabitation may have different social and health implications in different cultures.

Author Contributions

M.F. conceived the study, guided the analyses, wrote article drafts and is the guarantor of the study. J.S. created the cohabitation variable, conducted all data handling and statistical analyses, commented on article drafts and approved the final version of the manuscript before submission.

Conflict of interest: None declared.

Key Messages

  • Marriage has long been known to be associated with reduced mortality, but noticeable changes have occurred in the marital status distribution of Western populations over the past decades.

  • Our study assessed changes in marital status and cohabitation status in Denmark over a 30-year period (1982–2011) and their associations with mortality, using continuously updated individual-level information on living arrangements, confounders and deaths.

  • Compared with opposite-sex married persons, hazard ratios for overall mortality were consistently elevated in unmarried, divorced, widowed or same-sex married persons. Likewise, compared with opposite-sex cohabiting persons, hazard ratios for overall mortality were consistently elevated among persons living alone, with parents, in multi-adult households or in same-sex cohabitation.

  • The most marked changes in mortality were seen among same-sex married persons. Between 2000 and 2011, same-sex married Danish women emerged as a group with particularly increased mortality; in contrast, same-sex married Danish men now have mortality rates that are lower than those of unmarried and divorced men.

References

1
Farr
W
Hastings
GW
Influence of marriage on the mortality of the French people
Transactions of the National Association for the Promotion of Social Science 1858
 , 
1859
London
John W. Parker and Son, West Strand
(pg. 
504
-
20
)
2
Bliss
GI
The influence of marriage on the death-rate of men and women
Publications of the American Statistical Association
 , 
1914
, vol. 
14
 (pg. 
54
-
61
)
3
Ben-Shlomo
Y
Davey Smith
G
Shipley
M
Marmot
MG
Magnitude and causes of mortality differences between married and unmarried men
J Epidemiol Community Health
 , 
1993
, vol. 
47
 (pg. 
200
-
05
)
4
Ebrahim
S
Wannamethee
G
McCallum
A
Walker
M
Shaper
AG
Marital status, change in marital status, and mortality in middle-aged British men
Am J Epidemiol
 , 
1995
, vol. 
142
 (pg. 
834
-
42
)
5
Johnson
NJ
Backlund
E
Sorlie
PD
Loveless
CA
Marital status and mortality: the national longitudinal mortality study
Ann Epidemiol
 , 
2000
, vol. 
10
 (pg. 
224
-
38
)
6
Frank
L
Epidemiology. When an entire country is a cohort
Science
 , 
2000
, vol. 
287
 (pg. 
2398
-
99
)
7
Pedersen
CB
Gøtzsche
H
Møller
JO
Mortensen
PB
The Danish Civil Registration System. A cohort of eight million persons
Dan Med Bull
 , 
2006
, vol. 
53
 (pg. 
441
-
49
)
8
Anonymous
Act No. 372 of 7 June 1989 on registered partnerships
Annu Rev Popul Law
 , 
1989
, vol. 
16
 pg. 
56
 
9
Helweg-Larsen
K
The Danish Register of Causes of Death
Scand J Public Health
 , 
2011
, vol. 
39(7 Suppl)
 (pg. 
26
-
29
)
10
Johansen
JD
Smith
E
Juel
K
Rosdahl
N
The AIDS epidemic in the city of Copenhagen, Denmark: potential years of life lost and impact on life expectancy
Scand J Public Health
 , 
2005
, vol. 
33
 (pg. 
222
-
27
)
11
Statens Serum Institut
Annual HIV and AIDS report 2010
Epi-News National Surveillance of Communicable Diseases
 , 
2011
, vol. 
45
 (pg. 
1
-
2
)
12
Simonsen
J
Frisch
M
Ethelberg
S
Socioeconomic risk factors for bacterial gastrointestinal infections
Epidemiology
 , 
2008
, vol. 
19
 (pg. 
282
-
90
)
13
Jørgensen
KT
Nielsen
NM
Pedersen
BV
Jacobsen
S
Frisch
M
Hyperemesis, gestational hypertensive disorders, pregnancy losses and risk of autoimmune diseases in a Danish population-based cohort
J Autoimmun
 , 
2012
, vol. 
38
 (pg. 
J120
-
28
)
14
Jørgensen
KT
Pedersen
BV
Nielsen
NM
Hansen
AV
Jacobsen
S
Frisch
M
Socio-demographic factors, reproductive history and risk of osteoarthritis in a cohort of 4.6 million Danish women and men
Osteoarthritis Cartilage
 , 
2011
, vol. 
19
 (pg. 
1176
-
82
)
15
Claeskens
G
Hjort
NL
Model Selection and Model Averaging
 , 
2008
Cambridge
Cambridge University Press
16
Phua
VC
Kaufman
G
Using the census to profile same-sex cohabitation: A research note
Popul Res Policy Rev
 , 
1999
, vol. 
18
 (pg. 
373
-
86
)
17
Harrell
FE
Jr
Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis
 , 
2001
New York
Springer
18
Lin
DY
Wei
LJ
Ying
Z
Checking the Cox model with cumulative sums of martingale-based residuals
Biometrika
 , 
1993
, vol. 
80
 (pg. 
557
-
72
)
19
Blomgren
J
Martikainen
P
Grundy
E
Koskinen
S
Marital history 1971-91 and mortality 1991-2004 in England & Wales and Finland
J Epidemiol Community Health
 , 
2012
, vol. 
66
 (pg. 
30
-
36
)
20
Plöderl
M
Kralovec
K
Fartacek
R
The relation between sexual orientation and suicide attempts in Austria
Arch Sex Behav
 , 
2010
, vol. 
39
 (pg. 
1403
-
14
)
21
Mathy
RM
Cochran
SD
Olsen
J
Mays
VM
The association between relationship markers of sexual orientation and suicide: Denmark, 1990-2001
Soc Psychiatry Psychiatr Epidemiol
 , 
2011
, vol. 
46
 (pg. 
111
-
17
)
22
King
M
Semlyen
J
Tai
SS
, et al.  . 
A systematic review of mental disorder, suicide, and deliberate self harm in lesbian, gay and bisexual people
BMC Psychiatry
 , 
2008
, vol. 
8
 pg. 
70
 
23
Case
P
Austin
SB
Hunter
DJ
, et al.  . 
Sexual orientation, health risk factors, and physical functioning in the Nurses' Health Study II
J Womens Health (Larchmt)
 , 
2004
, vol. 
13
 (pg. 
1033
-
47
)
24
Fish
J
Wilkinson
S
Understanding lesbians' healthcare behaviour: the case of breast self-examination
Soc Sci Med
 , 
2003
, vol. 
56
 (pg. 
235
-
45
)
25
Brown
JP
Tracy
JK
Lesbians and cancer: an overlooked health disparity
Cancer Causes Control
 , 
2008
, vol. 
19
 (pg. 
1009
-
20
)
26
Henretta
JC
Lifetime marital history and mortality after age 50
J Aging Health
 , 
2010
, vol. 
22
 (pg. 
1198
-
1212
)
27
Frisch
M
Smith
E
Grulich
A
Johansen
C
Cancer in a population-based cohort of men and women in registered homosexual partnerships
Am J Epidemiol
 , 
2003
, vol. 
157
 (pg. 
966
-
72
)
28
Frisch
M
Brønnum-Hansen
H
Mortality among men and women in same-sex marriage: a national cohort study of 8333 Danes
Am J Public Health
 , 
2009
, vol. 
99
 (pg. 
133
-
37
)
29
Health care needs of gay men and lesbians in the United States. Council on Scientific Affairs, American Medical Association
JAMA
 , 
1996
, vol. 
275
 (pg. 
1354
-
59
)
30
Health of lesbian, gay, bisexual, and transgender populations (editorial)
Lancet
 , 
2009
, vol. 
377
 pg. 
1211
 
31
IOM (Institute of Medicine)
The health of lesbian, gay, bisexual, and transgender people: building a foundation for better understanding
2011
Washington, DC
National Academies Press
(pg. 
1
-
347
)
32
Cochran
SD
Mays
VM
Physical health complaints among lesbians, gay men, and bisexual and homosexually experienced heterosexual individuals: results from the California Quality of Life Survey
Am J Public Health
 , 
2007
, vol. 
97
 (pg. 
2048
-
55
)
33
Nystedt
P
Marital life course events and smoking behaviour in Sweden 1980-2000
Soc Sci Med
 , 
2006
, vol. 
62
 (pg. 
1427
-
42
)
34
Prescott
CA
Kendler
KS
Associations between marital status and alcohol consumption in a longitudinal study of female twins
J Stud Alcohol
 , 
2001
, vol. 
62
 (pg. 
589
-
604
)
35
Lund
R
Due
P
Modvig
J
Holstein
BE
Damsgaard
MT
Andersen
PK
Cohabitation and marital status as predictors of mortality – an eight year follow-up study
Soc Sci Med
 , 
2002
, vol. 
55
 (pg. 
673
-
79
)
36
Lewis
CE
Saghir
MT
Robins
E
Drinking patterns in homosexual and heterosexual women
J Clin Psychiatry
 , 
1982
, vol. 
43
 (pg. 
277
-
79
)
37
Ryan
H
Wortley
PM
Easton
A
Pederson
L
Greenwood
G
Smoking among lesbians, gays, and bisexuals: a review of the literature
Am J Prev Med
 , 
2001
, vol. 
21
 (pg. 
142
-
49
)

Appendix

Among cohort members sharing address with 2-8 unrelated adult persons, we used a two-step procedure to categorize periods as living in opposite-sex cohabitation, living in same-sex cohabitation or living in multi-adult household, as follows:

First, we aimed to identify the possible cohabiting partner of each cohort member (index person):

  • If one of the 2-8 address-sharing adult persons was the index person’s spouse, that person was considered the index person’s possible cohabiting partner.

  • If none of the 2-8 address-sharing adult persons was the index person’s spouse, then the only adult person who had one or more shared children with the index person was considered the index person’s possible cohabiting partner. If two or more persons had shared children with the index person, no further attempt was made to identify the index person’s possible cohabiting partner.

  • If none of the 2-8 address-sharing adult persons was the index person’s spouse or had shared children with the index person, then the only unrelated adult person who had a history at or after age 18 years of at least one simultaneous relocation (ie. on the same date) from one shared address to a new shared address with the index person was considered the index person’s possible cohabiting partner. If no adult person or ≥2 adult persons had a history of simultaneous relocation to a shared address with the index person, no further attempt was made to identify the index person’s possible cohabiting partner.

Second, to be living in opposite-sex or same-sex cohabitation, we required that the index person was simultaneously the possible cohabiting partner’s own possible cohabiting partner according to the criteria described above. In other words, only in time intervals when two cohort members were each other’s possible cohabiting partner, both were considered as living in opposite-sex or same-sex cohabitation, as appropriate.

The requirement for reciprocity in the possible cohabiting partner status served to reduce the inclusion of periods of unclear cohabitation patterns in the living in opposite-sex cohabitation and living in same-sex cohabitation categories. For instance, in a household composed of three unrelated adults who were not married to each other, who had no shared children, and who had no history of simultaneous relocation, a 40 year-old woman might share address with a man aged 50 years and a woman aged 60 years. Following our criteria, both women would have the 50 year-old man as their possible cohabiting partner, due to the less than 15 years age difference. Consequently, the 50 year-old man would be both women’s possible cohabiting partner. However, for the 50 year-old man, neither woman would be his possible cohabiting partner, because we had no means to select one woman over the other. In such a household, therefore, the requirement for reciprocity with respect to the possible cohabiting partner status was not met. As a consequence, for as long as they lived together, all three persons would be categorized as living in multi-adult household. In a similar household of three unrelated adults, where the 50 year-old man had a child with the 40 year-old woman, these two unrelated persons would be each other’s possible cohabiting partner, and they would both be categorized as living in opposite-sex cohabitation. In that example, the 60 year-old woman would also have the 50 year-old man as her possible cohabiting partner, but the requirement for reciprocity would not be met, implying that the 60-year old woman would be categorized as living in multi-adult household.