Abstract

Purpose: We examined whether the presence of a spouse and the frequency of interaction with children, relatives, and friends significantly influence the risk of dying in late life. We assessed these effects separately by gender, controlling for self-reported health. In addition, we examined whether interaction with the co-twin has a different impact on mortality for identical and fraternal twins. Design and Methods: The data set consists of 2,147 Danish twins aged 75 years and older, who were followed prospectively from 1995 to 2001. We modeled the effect of social ties on mortality by using event history analysis. Results: Survival is extended by having a spouse and close ties with friends and the co-twin. However, contact frequency with friends and the co-twin is significant, respectively, only for women and identical twins. Implications: Investigating social relations sheds light on the life span of individuals older than 75 years of age. We stress the importance of social relations beyond the presence of the spouse for survival even at very old ages.

It is common to conceptualize social relations in terms of their structure and function. Although these two dimensions are equally relevant, research usually focuses on the first one. Questions on the structure of social relations normally cover the type of relationships (e.g., kinship or friendship), marital status, frequency of contact, duration of the relationship, and number of social referents. When the function of social relations is included, it normally refers to the resources provided by the relationship, labeled as “social support.” This includes emotional, instrumental, appraisal, and informational support, and it can be difficult to assess because the same tie can provide several types of support (House, Umberson, & Landis, 1988; Litwin, 2001).

Although it is widely recognized that social relations strongly influence health outcomes and longevity, less is known about the underlying mechanisms.

On the one hand, clinical research into the relationship between loneliness and biological mechanisms has found that lonely individuals have impaired cellular immunology, which predicts infectious-disease susceptibility. This effect is exacerbated later in life by the reduction in immune function arising from aging (Hawkley & Cacioppo, 2004). On the other hand, survival may be improved by social relations through better compliance with medical regimens or motivation to engage in healthy behaviors. For example, individual behavior that is detrimental to health seems to be regulated by the presence of a spouse (Antonucci, 1990).

Our study analyzes the association between structural measures (the presence of a spouse and the frequency of contact with children, other family members, friends, and the co-twin) and the risk of dying at older ages. We have not included measures of social support because they were not collected at follow-up. Although items on social support are useful as they provide information on the underlying meaning of a relationship, structural measures are more relevant when one is studying the pattern of contact frequency at different times. In addition, through the linkage with mortality data, they will show whether frequency of contact is significantly associated with the risk of dying. The following sections highlight the literature on social relations in later life from which our research hypotheses were derived, including conjugal ties and relations with children, relatives, and friends, and then twinship ties.

Presence of a Spouse and Relations Beyond the Household

Literature on social relations has devoted much attention to the conjugal tie, one of the most essential bonds between humans. Marriage supposedly promotes well-being by providing a sense of meaning in life and a sense of feeling cared for and loved, by reducing risky behavior, by aiding early detection of illness, and by helping recovery after illness. It is well known that married people have lower mortality rates than unmarried people (Lillard & Waite, 1995). Although marriage generally protects and improves health for both genders, men tend to derive more benefits from marriage than women. This is because men are less likely to have intimate social ties beyond the household and because women place a higher value on health and therefore provide more health benefits to their spouses than men do (Ross, Mirowsky, & Goldstein, 1990).

Past research on relationships beyond the household has focused on the frequency of contact between friends and family members, suggesting that kin and friendship relationships are different, though at times they overlap. Family members (spouse, children, and relatives) play an important role in determining the security of older people by performing helpful tasks and providing emotional assistance. For instance, Lopata (1973) found that children meet the tangible and emotional needs of parents when they are most required. With regard to friendship, past research shows that it is often characterized through reciprocity, the feeling of being needed, and the voluntary nature of the tie (Powers & Bultena, 1976). According to Adams and Blieszner (1989), the possibility of having significant relations beyond the household sphere can allow the older person to feel a sense of competence in the ability to reciprocate without the sense of obligation that may affect the assistance given and received within the family. Furthermore, friends may protect against negative thoughts by making older adults feel competent, liked, and needed (Pearlin, Menaghan, Lieberman, & Mullan, 1981).

When comparing the effect of friendship and family relations on mortality risk, most studies have found that the risk of dying is more often associated with contact frequency between friends rather than family members (Sabin, 1993; Wood & Robertson, 1978). A study on social relations and mortality must also take gender differences into account, because patterns of social relationships have been shown to vary by gender (Connidis & Davies, 1990; Shye, Mullooly, Freeborn, & Pope, 1995). The literature indicates that men tend to maintain close social relationships with only a few people, primarily their wives. Consequently, “the few people to whom they are closest” can be expected to have the greatest impact on them (Shye et al., p. 937). In contrast, women are more relationship oriented than men, and their well-being is defined as being contingent on the maintenance of social relations beyond the household. Regarding friendships, most studies have found that contact with friends may be relevant to the well-being of elders but it is differentially important to the well-being of women and men. In particular, past research has found that it is more meaningful for women (Powers & Bultena, 1976).

Finally, although research findings are consistent about the different effects of social relations on men's and women's well-being, it is unclear whether men and women differ in their level of contact with family members and friends. In fact, research has produced conflicting findings; some studies have found that men have more frequent contact with friends (Powers & Bultena, 1976), whereas others have revealed that women meet with both friends and relatives more often than men (Turner & Marino, 1994) and still others have indicated the same rates of social interaction with family and friends (Kohen, 1983). Consequently, it is essential to test whether men and women differ in terms of level of social contact in order to verify the association between social relationships and mortality risk, as a significant association could depend on the differences in contact frequency.

A Particular Sibling Tie: The Twin Relationship

By virtue of its uniqueness, siblingship appears to be particularly relevant in research on social relations in late life. In fact, siblingship represents the longest family tie and involves sharing an early sense of family belonging (Connidis, 1992). Scholars have also argued that siblingship aids in the reminiscence of life events, which is especially relevant at old age when individuals undergo their life-review process (Goetting, 1986).

Some research suggests that sibling ties are more comparable with friendship than kinship or parenthood ties (Lee, Mancini, & Maxwell, 1990). This comes from the observation that, like friends, siblings are typically age peers, which makes their tie more egalitarian and less based on a sense of obligation than other family ties. Cumming and Schneider (1961) have described the sibling bond as an association between equals hinging on personal choice and providing a sociability function. In contrast, Cicirelli (1980) argued that, with adulthood, “sibling contact becomes voluntary except on certain ritual occasions” (p. 455). However, other scholars have found that there is an obligation to maintain contact with siblings through all stages of the life course (Allan, 1977).

Among sibling ties, twinship deserves special consideration. In fact, twins differ from nontwin siblings in that they are together from conception and are typically exposed to the same environments at the same time (such as being in school or starting a career), which enhances their closeness. A U-shaped pattern has been found in twins' pattern of contact frequency over the life span and emotional closeness, with the highest intensity occurring during childhood until the twins leave home and again when they are retired. However, identical twins constantly show higher rates of contact than fraternal twins, and they feel emotionally more closely related to one another. The greater attachment between identical twins is generally explained on the grounds of their similar appearance, which leads them to feel a stronger sense of belonging to each other (Macdonald, 2002; Neyer, 2002). Given this greater closeness, identical twins' well-being should be more affected than fraternal twins' well-being by contact frequency with the co-twin.

Nevertheless, genetic studies on longevity (Herskind et al., 1996) indicate that caution is needed when one is analyzing the mortality risk of twins. Indeed, genetic studies have found that genes explain approximately 25% of the variance in longevity in Northern Europe and thus could explain the increasing mortality risk of the surviving twin (Yashin & Iachine, 1997). It follows that, in longitudinal studies on old twins, it is important to check the vital status of the co-twin, whose death during follow-up may be associated with an increasing risk that the surviving twin will die in the coming period.

Study Purpose

Our study aims to assess whether having social contact beyond the household and having a spouse determine different survival trajectories among Danish twins aged 75 and older. Our study is distinctive in the literature on social relations and mortality, as we analyze the effect of contact frequency with the co-twin in later life, which has been neglected in research on the mortality risk of identical and fraternal twins. Lack of research in this field is probably attributable to the difficulty in obtaining follow-up data, including information on social relations and the vital status for both twins in the pair. Furthermore, most previous studies were based on cross-sectional data, whereas we apply a longitudinal design to examine how the variables of interest affect mortality over time, using the Longitudinal Study of Aging Danish Twins (LSADT). This follow-up, covering a 6-year period (1995–2001), provides a rare opportunity to study prospectively the effects of social relations in late life. We are also able to determine the effect of contact frequency on mortality, net of the health effect across the study period. We can achieve this because the LSADT includes health status, which is a key variable in studies on social relations at older ages, at each assessment during the follow-up. If this factor is not taken into account, the association between social ties and mortality might be spurious because health can affect the level of engagement in social relations. In prior longitudinal research, only data on health status at baseline was available.

Finally, previous studies on social relations in Denmark have produced ambiguous results. Whereas Olsen, Olsen, Gunner-Svensson, and Waldstrom (1991) found no significant association between social relations and mortality, Avlund, Damsgaard, and Holstein (1998) found that people with fewer contacts had significantly shorter survival times than people with more frequent contacts. These studies focused on an index based on several types of relations. Compared with previous research, our study takes into account a number of social relations separately, which may resolve some of these inconsistencies.

On the basis of findings from previous research, we address the following hypotheses. First, conjugal and friendship ties will significantly decrease the risk of dying at very old ages. The second hypothesis summarizes the general findings derived from studies on social relations in men and women. It indicates that male mortality is largely affected by the presence or absence of the spouse. Conversely, women's mortality also should be affected by relationships outside the home because their well-being appears to be dependent on a wider range of social ties. The third hypothesis focuses on the linkage between the twin relationship and mortality. Because the literature suggests that identical twins are more closely connected than fraternal twins, it is expected that contact frequency with the co-twin affects the mortality of identical twins more than it does the mortality of fraternal twins.

Methods

Study Population

Study data are from the LSADT. The follow-up was based on all registered Danish twins aged 75 and older who were alive in January 1995; the participation rate was 78%. Because a study by Christensen, Holm, McGue, Corder, and Vaupel (1999) showed that the responders and nonresponders were similar in terms of age distribution, gender, and health, variance adjustments for nonresponse did not appear to be required in our study.

The assessment of the LSADT began in February 1995 and was repeated between February and April every 2 years (1997, 1999, and 2001). At intake, 2,147 twins (764 identical twins and 1,383 fraternal twins) were interviewed. Afterward, 1,459, 921, and 639 were reinterviewed in 1997, 1999, and 2001, respectively. Overall, the numbers lost to follow-up consisted of people who died between two waves or refused to participate at the next assessment.

To investigate factors affecting decreases in the sample size at each time point, we considered the mean age and the frequency distribution of health status among dropouts and participants. Results indicate that older respondents and those whose health declined over the study period were overrepresented among those who died between two waves. In other words, individuals aging successfully were more likely to participate throughout the study period. This outcome usually occurs in longitudinal studies of old persons, so care is required when one is extending results to the entire older population. We also found that dropouts because of refusal were very similar in terms of health and age to those who continued their participation.

Measures

At each wave, a standardized questionnaire provided information on social relations and sociodemographic measures. Additional information, such as date of birth, date of death, and survival status of the co-twin, was provided by the Danish Twin Registry.

Marital status and contact frequency with children, relatives, friends, and the co-twin are the measures of social relationships. Interviewers asked follow-up participants about social relationships by using the following question: “How often do you meet … ?” We used the question separately for children, other family members, friends, and co-twins. We grouped the answers, based on a six-item scale, into three categories (often, sometimes, and rarely) to avoid the dispersal of information when we were running survival models.

The sociodemographic variables include age, gender, health status, and twinship status (fraternal or identical). For the measures of social relations and health, which are time-varying covariates, we have taken into account changes from one state to another during the study time. We did not include measures of education and social status in the models because they were not statistically significant and, consistent with statistical literature, the most parsimonious models are reported here.

Information on health status was provided by a measure of subjective health, which we assessed through the response to the following question: “How do you consider your health in general?” The rating scale was based on five items (excellent, good, acceptable, poor, and very poor). Participants who responded “excellent” or “good” were considered to be in good health; those who responded “acceptable” were considered to be in fair health; and those who responded “poor” or “very poor” were considered to be in bad health. We used logistic regression to examine the relationship between perceived health and mortality risk. The risk of dying appeared to be significantly related to the health categories, indicating that self-rated health is a strong predictor of death in our study.

Design

A key analysis feature is the use of longitudinal data, which allows us to investigate the link between the frequency of social contact and mortality by using event history analysis. To examine the linkage between contact frequency and mortality, we have prospectively followed respondents interviewed at baseline in 1995. Therefore, the age at intake represented the starting time of the analysis. We calculated survival time from the time of interview to the time at death or at last interview (i.e., censoring). The ending time of survival analysis for participants lost at follow-up because of death was the age at death. For individuals lost at follow-up because of refusal and for those still alive when the follow-up terminated, the ending time of survival analysis was the age at last interview.

In order to choose the best-fitting model for our data, we computed the Cox–Snell residuals (Cox & Snell, 1968) that were provided by three distributions (the Gompertz, Weibull, and log-logistic distributions). The fitting procedure was required because the Gompertz model, generally used to model the transition to death, may not appear correctly after a participant reaches the age of 85, when the mortality rate decelerates. To address this issue, we have compared the fit provided by the Gompertz distribution and by two distributions whose rates slow down for advancing ages (i.e., the Weibull and the log-logistic distribution; see Blossfeld & Rohwer, 2002). The model that best fits the observed curve was that provided by Gompertz. In light of these results, we regressed survival time on social relations and sociodemographic measures through the Gompertz model. We performed all analyses by using the STATA and TDA statistical packages.

Results

Descriptive Statistics

The sample composition, in terms of gender and age, is shown in Table 1. The majority of respondents are women, which is consistent with the higher life expectancy of female individuals at older ages in developed countries. Regarding changes in health status, results indicate that self-reported health is quite stable over the study period. In 2001, 65% of respondents gave the same report as in 1995. Self-perceived health improved for about 26% of the respondents and worsened for about 9%.

We observed a slight decline in the percentage of people who assessed their health as poor at each time point (Table 2). This finding is attributable to the selection process, which determined the sample size at each time point. In fact, as mentioned earlier, the respondents with poor health were overrepresented among the dropouts.

Regarding social relationships, the frequency of contact with children, relatives, friends, and the co-twin during the study period is shown in Table 3. An increase can be seen in the proportion of respondents with frequent contacts at next assessments over the 6-year time span of the study. This may be because people aging successfully were more likely to complete the follow-up and healthier respondents were more likely to have frequent contacts.

Regarding the variation in contact frequency, Table 4 illustrates that most respondents maintained the same frequency of contact over time, especially for twin contacts (about 72%). Contact frequency with relatives, children, and friends increases over time, which is consistent with the data shown in Table 3.

To verify whether the frequency of contact differs by gender, we applied the z test to the difference between two independent proportions. We computed a two-sided hypothesis test on the difference between male and female contact frequency with children, relatives, and friends. The test is based on data at intake, which provides a reasonable sample size. Using baseline data, we tested the null hypothesis, which assumed no significant difference by gender in frequency of social contacts with family members and friends (i.e., H0 : πM − πW = 0). The results presented in Table 5 show a value always greater than p =.05, indicating that there is no significant gender difference in frequency of contacts with family and friends. This result is extremely relevant, as it suggests that any difference in the association between contact frequency and mortality is not influenced by gender differences in the amount of contacts.

The Impact of Marital Status and Social Relations on Mortality

We analyzed the impact of contact frequency on mortality, controlling for other factors independently associated with the risk of dying (i.e., age, gender, and health). In particular, we modeled the effects of social relations on mortality by using marital status and frequency of contact with children, relatives, and friends. Given the collinearity between marital status and children (91% of never married respondents were childless), we excluded the never married individuals from the analysis whereas we included the married and previously married (widowed or divorced) individuals. This selection left a sample of 1,734 observations from the previous 2,147.

To verify the effects of marital status and frequency of contact on mortality risk, we tested three models that sequentially included all explanatory variables (Table 6). The first model included gender, age, marital status, and contact frequency. A set of dummy variables for all types of social relations compared individuals who report infrequent contact with those with frequent contact (i.e., rarely vs sometimes and often). The comparison was restricted to infrequent–more frequent contacts because this dichotomy had the strongest association with mortality. Model 2 then adjusted for self-reported health so that we could obtain net estimates. The full model (Model 3) added the interaction term with gender to test whether the effect of being married and frequency of contact with friends varies by gender. Table 6 also shows the Likelihood Ratio Test for each model to see the fit improvement from one model to the next.

Model 1 shows that, consistent with the literature, married people have lower death rates than do previously married people. With regard to contacts beyond the household, only the frequency of contact with friends is significantly associated with mortality. The model reveals that individuals who report rarely meeting with friends have a significantly greater risk of dying than those who meet friends more frequently.

When self-reported health is adjusted for (Model 2), the coefficients for being married and meeting with friends are still significant. These estimates suggest that marital status and friendship contacts are independent predictors of mortality risk. A comparison of the size of the coefficients in Model 1 and Model 2 reveals that only the coefficient for friendship shows a greater reduction from Model 1 to Model 2, indicating that without a measure of health the estimate of effect of contacts with friends would be spurious. Additionally, it should be noted that the coefficient for friendship is larger (0.52) than the coefficient for being married (–0.36), suggesting that research on older adults' social relations and mortality should examine the full range of their social contacts in addition to marital status.

To address gender differences in the relation between frequency of social contacts and mortality, Model 3 adds the interaction terms for gender and the coefficient for being married and for infrequent contacts with friends. The interaction for being married and male is not significant, indicating that the effect of marital status on mortality is similar for both genders. However, the interaction for infrequent relations with friends and females is significant, thus supporting our hypothesis regarding the greater relevance of social contacts with persons beyond the household for women.

The Impact of Twin Relationship on Mortality

To examine the mortality risk associated with the twin relationship, we used a different data set than used in prior analyses. Among respondents interviewed in 1995 and reinterviewed in subsequent waves, we selected those with the co-twin still alive during their participation in the follow-up. Through this procedure, we found it possible to obtain all the information available on the relationship of twin pairs.

Using the Gompertz model, we tested whether the frequency of contact in twin relationships is more highly related to identical twins' mortality risk. This hypothesis is verified through three models (Table 7), which progressively include the control and the key variables. Control variables refer to age, gender, twinship status, and health status. Key variables are represented by the frequency of contact with the co-twin and the interaction term between twinship status and frequency of twin contact.

Model 1, which includes age, gender, twinship status, and the frequency of contact with the co-twin, shows that infrequent contact with the co-twin does not predict the risk of dying (coefficient = 0.25). When adding the interaction between infrequent contacts and being an identical twin (Model 2), the interaction is significant (coefficient = 1.08**). Furthermore, the coefficient for interaction continues to be significant in the model adjusted for health (Model 3). The magnitude and significance of this interaction illustrates that twinship is a mortality risk factor for identical twins, a finding that has not been reported previously.

To further investigate this finding, we considered a confounding variable, which was the death of the co-twin between two waves. Because identical twins share a genetic background, the death of the co-twin may be followed by the death of the other twin because of the genetic limits of the life span. To verify whether the contact frequency with the identical co-twin is, in fact, a mortality risk factor, we computed new models based only on identical twins and then adjusted for the co-twin's vital status (Table 8). Results show that the coefficient for contact frequency with the co-twin is significant when co-twin vital status is adjusted for, which reveals that the frequency of contact with the identical co-twin is an independent predictor of mortality.

Discussion

This study, which examined the link between the frequency of social contact and mortality risk among elderly Danish twins, found that frequent contacts between identical compared with fraternal older twins has a beneficial effect on their survival. This result may be explained by the fact that identical twins are much closer than fraternal twins, supported by previous studies on this topic (Macdonald, 2002; Neyer, 2002).

Concerning the effect of marital status on mortality risk, findings from this study confirm those from past research, which indicate that having a spouse enhances the survival of individuals of both genders.

Similarly, the survival models shown here support the linkage between social relations and mortality risk. In line with most earlier studies (Sabin, 1993; Wood & Robertson, 1978), we found that frequency of contact with friends is more strongly associated with mortality risk than contacts with children and other family members. Moreover, although contact with children occurs as frequently as contact with friends, only frequency of contact with friends is significantly associated with mortality risk.

In addition, findings show that frequency of contact with friends enhances survival only for women, and this is consistent with findings reported in other studies (Connidis & Davies, 1990; House et al., 1988; Shye et al., 1995). Moreover, the frequency distribution of friendship contacts does not differ significantly by gender, suggesting different involvement of men and women in friendship ties. As noted earlier, men are less likely to have intimate social ties beyond the household. Furthermore, men's friendships emphasize sociability and task-activity orientation, whereas women's friendships focus on intimacy and self-disclosure, which has been found to be more related to well-being (Bell, 1981).

To date, there has been little empirical research in Denmark on the relationship between mortality and social relationships in old age. Given the ambiguities of previous Danish studies based on an index of social relations, findings from this research indicate that a broader range of types of social relations and frequency of contacts should be included to avoid misleading results.

In conclusion, our analysis could have suffered from several limitations because of selection bias. Older people reaching age 75 are survivors who have progressed successfully through all the previous stages of the life course, so the study sample used here could have limited the generalization of our results. Because the study does not include individuals who have never married, the generalizability of the results may be further limited. In addition, the assessment of health only included a self-reported health measure, and inclusion of a wider range and types of measures could have yielded different results. Lastly, the research lacked qualitative measures that could have provided insight into the differential effects of gender on the relationship between contact frequency with friends and mortality, and about the nature and meaning of later life friendships for men and women. Thus, future research investigations should include, when available, some qualitative aspects of older adults' social ties in later life in order to better examine the underlying dimensions of the social contact.

This research was completed while D. Rasulo was a research scientist at the Max Planck Institute of Demographic Research (Rostock, Germany). The project was supported by Grant P01-08761 from the National Institutes of Health–National Institute on Aging. We thank Professors James Vaupel, Graziella Caselli, and Francesco Billari, whose suggestions and comments were extremely useful, and Jo Tomlinson for her invaluable editorial help.

1

Department of Demographic Sciences, University of Rome “La Sapienza,” Italy.

2

Institute of Public Health, University of Southern Denmark, Odense.

3

Centre for Population Studies, London School of Hygiene and Tropical Medicine, London, United Kingdom.

Decision Editor: Linda S. Noelker, PhD

Table 1.

Gender Distribution of the Sample.

LSADT Wave Male
 
  Female
 
  
  Age
 
  Age
 
N M ± SD N M ± SD 
1995 762 35.49 81.13 ± 4.43 1,385 64.51 81.60 ± 4.70 
1997 506 34.68 82.73 ± 4.21 953 65.32 82.99 ± 4.33 
1999 322 34.96 84.38 ± 4.01 599 65.04 84.41 ± 3.99 
2001 217 33.96 85.90 ± 3.72 422 66.04 86.04 ± 3.81 
LSADT Wave Male
 
  Female
 
  
  Age
 
  Age
 
N M ± SD N M ± SD 
1995 762 35.49 81.13 ± 4.43 1,385 64.51 81.60 ± 4.70 
1997 506 34.68 82.73 ± 4.21 953 65.32 82.99 ± 4.33 
1999 322 34.96 84.38 ± 4.01 599 65.04 84.41 ± 3.99 
2001 217 33.96 85.90 ± 3.72 422 66.04 86.04 ± 3.81 

Notes: LSADT = Longitudinal Study of Aging Danish Twins.

Table 2.

Frequency Distribution of Self-Rated Health at Each Time Point.

Item Wave
 
   
1995 1997 1999 2001 
Good health 59.81 60.05 59.07 60.88 
Fair health 24.03 25.70 29.21 29.26 
Poor health 13.88 12.06 10.42 9.55 
Missing 2.28 2.19 1.30 0.31 
Total 100.00 100.00 100.00 100.00 
Item Wave
 
   
1995 1997 1999 2001 
Good health 59.81 60.05 59.07 60.88 
Fair health 24.03 25.70 29.21 29.26 
Poor health 13.88 12.06 10.42 9.55 
Missing 2.28 2.19 1.30 0.31 
Total 100.00 100.00 100.00 100.00 
Table 3.

Frequency Distribution of Social Relations at Each Time Point.

Item Contact With Children (%)
 
   Contact With Relatives (%)
 
   Contact With Friends (%)
 
   Contact With Co-twin (%)
 
   
1995 1997 1999 2001 1995 1997 1999 2001 1995 1997 1999 2001 1995 1997 1999 2001 
Rarely 9.32 9.12 9.01 9.23 33.58 30.64 29.32 27.07 25.85 6.99 3.15 2.03 57.89 56.45 60.85 62.41 
Sometimes 19.33 18.23 19.11 9.38 33.12 32.42 33.44 23.95 20.77 26.59 29.32 30.83 21.15 24.20 24.34 24.81 
Often 53.14 54.97 56.13 67.14 32.93 36.94 37.24 48.98 52.91 66.35 67.42 67.14 20.96 18.06 13.76 11.28 
Without children 17.75 17.07 15.53 14.25 — — — — — — — — — — — — 
Missing 0.46 0.61 0.22 0.00 0.37 0.00 0.00 0.00 0.47 0.07 0.11 0.00 0.00 1.29 1.05 1.50 
Total 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 
Item Contact With Children (%)
 
   Contact With Relatives (%)
 
   Contact With Friends (%)
 
   Contact With Co-twin (%)
 
   
1995 1997 1999 2001 1995 1997 1999 2001 1995 1997 1999 2001 1995 1997 1999 2001 
Rarely 9.32 9.12 9.01 9.23 33.58 30.64 29.32 27.07 25.85 6.99 3.15 2.03 57.89 56.45 60.85 62.41 
Sometimes 19.33 18.23 19.11 9.38 33.12 32.42 33.44 23.95 20.77 26.59 29.32 30.83 21.15 24.20 24.34 24.81 
Often 53.14 54.97 56.13 67.14 32.93 36.94 37.24 48.98 52.91 66.35 67.42 67.14 20.96 18.06 13.76 11.28 
Without children 17.75 17.07 15.53 14.25 — — — — — — — — — — — — 
Missing 0.46 0.61 0.22 0.00 0.37 0.00 0.00 0.00 0.47 0.07 0.11 0.00 0.00 1.29 1.05 1.50 
Total 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 
Table 4.

Contact Frequency of the Respondents During the Study Period (1995–2001).

Item Children Relatives Friends Co-twin 
Same category 60.72 42.88 53.36 72.18 
Less frequent contact 5.79 22.07 17.53 21.06 
More frequent contact 19.41 35.05 29.11 5.26 
Missing 0.63 0.00 0.00 1.50 
Without children 13.46 — — — 
Total 100.00 100.00 100.00 100.00 
Item Children Relatives Friends Co-twin 
Same category 60.72 42.88 53.36 72.18 
Less frequent contact 5.79 22.07 17.53 21.06 
More frequent contact 19.41 35.05 29.11 5.26 
Missing 0.63 0.00 0.00 1.50 
Without children 13.46 — — — 
Total 100.00 100.00 100.00 100.00 
Table 5.

Hypothesis Test on Contact Frequency by Gender.

Items Contact With Relatives
 
 Contact With Children
 
 Contact With Friends
 
 
Test Statistics p Test Statistics p Test Statistics p 
Rarely 1.83 0.067 0.31 0.754 −0.72 0.472 
Sometimes −1.95 0.052 0.65 0.514 0.63 0.526 
Often 0.20 0.843 0.00 0.996 0.08 0.939 
Items Contact With Relatives
 
 Contact With Children
 
 Contact With Friends
 
 
Test Statistics p Test Statistics p Test Statistics p 
Rarely 1.83 0.067 0.31 0.754 −0.72 0.472 
Sometimes −1.95 0.052 0.65 0.514 0.63 0.526 
Often 0.20 0.843 0.00 0.996 0.08 0.939 
Table 6.

The Effect of Contact Frequency and Presence or Absence of the Spouse on Mortality: Gompertz Model.

Variable Model 1 Model 2 Model 3 
Constant −5.38 −5.88 −5.90 
Males (vs females) 0.51** 0.55** 0.62** 
Aged 85+ (vs aged 75–84) 0.65** 0.61** 0.60** 
Married (vs previously married) −0.39** −0.36** −0.45** 
Kin: rarely (vs sometimes or often) 0.12 0.02 0.03 
Children: rarely (vs sometimes or often) −0.08 −0.09 −0.07 
Friends: rarely (vs sometimes or often) 0.70** 0.52** 0.23 
Fair health (vs good health) — 0.70** 0.70** 
Bad health (vs good health) — 1.70** 1.68** 
Married × Males — — 0.13 
Rarely meeting friends × Females — — 0.46* 
LRT 179.44 464.56 470.68 
df 10 
Variable Model 1 Model 2 Model 3 
Constant −5.38 −5.88 −5.90 
Males (vs females) 0.51** 0.55** 0.62** 
Aged 85+ (vs aged 75–84) 0.65** 0.61** 0.60** 
Married (vs previously married) −0.39** −0.36** −0.45** 
Kin: rarely (vs sometimes or often) 0.12 0.02 0.03 
Children: rarely (vs sometimes or often) −0.08 −0.09 −0.07 
Friends: rarely (vs sometimes or often) 0.70** 0.52** 0.23 
Fair health (vs good health) — 0.70** 0.70** 
Bad health (vs good health) — 1.70** 1.68** 
Married × Males — — 0.13 
Rarely meeting friends × Females — — 0.46* 
LRT 179.44 464.56 470.68 
df 10 

Notes: LRT = likelihood ratio test. For the table, the sample size is 1,734 (627 males, 1,107 females). Data are taken from the Longitudinal Study of Aging Danish Twins (LSADT) 1995–2001.

*p <.05; **p <.01.

Table 7.

The Effect of Contact Frequency With the Co-Twin on Mortality: Gompertz Model.

Variable Model 1 Model 2 Model 3 
Constant −5.45 −5.15 −5.59 
Male (vs female) 0.57** 0.56** 0.59** 
Aged 85+ (vs aged 75–84) 0.75** 0.74** 0.71** 
Identical twins −0.05 −0.73** −0.64* 
Co-twin: rarely (vs sometimes or often) 0.25 −0.17 −0.18 
Identical twins × Co-twin rarely — 1.08** 1.01** 
Fair health (vs good health) — — 0.63** 
Bad health (vs good health) — — 1.45** 
LRT 35.84 46.63 98.19 
df 
Variable Model 1 Model 2 Model 3 
Constant −5.45 −5.15 −5.59 
Male (vs female) 0.57** 0.56** 0.59** 
Aged 85+ (vs aged 75–84) 0.75** 0.74** 0.71** 
Identical twins −0.05 −0.73** −0.64* 
Co-twin: rarely (vs sometimes or often) 0.25 −0.17 −0.18 
Identical twins × Co-twin rarely — 1.08** 1.01** 
Fair health (vs good health) — — 0.63** 
Bad health (vs good health) — — 1.45** 
LRT 35.84 46.63 98.19 
df 

Notes: LRT = likelihood ratio test. For the table, the sample size is 520 twins (197 identical twins, 323 fraternal twins). Table data are taken from the Longitudinal Study of Aging Danish Twins (LSADT) 1995–2001.

*p <.05; **p <.01.

Table 8.

The Effect of Contact Frequency With the Identical Co-Twin on Mortality: Gompertz Model, Adjustment for Co-Twin Vital Status.

Variable Model 
Constant −6.04 
Male (vs female) 0.53* 
Aged 85+ (vs aged 75–84) 0.37 
Co-twin: rarely (vs sometimes/often) 0.77** 
Co-twin loss (vs co-twin alive) 1.95** 
Fair health (vs good health) 0.35 
Bad health (vs good health) 0.82* 
LRT 54.72 
df 
Variable Model 
Constant −6.04 
Male (vs female) 0.53* 
Aged 85+ (vs aged 75–84) 0.37 
Co-twin: rarely (vs sometimes/often) 0.77** 
Co-twin loss (vs co-twin alive) 1.95** 
Fair health (vs good health) 0.35 
Bad health (vs good health) 0.82* 
LRT 54.72 
df 

Notes: LRT = likelihood ratio test. For the table, the sample size is 197 identical twins. Table data are taken from the Longitudinal Study of Aging Danish Twins (LSADT) 1995–2001.

*p <.05; **p <.01.

References

Adams, R. G., & Blieszner, R., (
1989
). Older adult friendship: Structure and process. Newbury Park, CA: Sage.
Allan, G., (
1977
). Sibling solidarity.
Journal of Marriage and the Family,
 
39
,
177
-184.
Antonucci, T., (
1990
). Social supports and social relationships. In R. H. Binstock & L. K. George (Eds.), The handbook of aging and the social sciences (pp. 205–226). Orlando, FL: Academic Press.
Avlund, K., Damsgaard, M. T., & Holstein, B. E., (
1998
). Social relations and mortality. An 11-year follow-up study of 70-year-old men and women in Denmark.
Social Science and Medicine,
 
47
,
635
-643.
Bell, R. R., (
1981
). Worlds of friendship. Beverly Hills, CA: Sage.
Blossfeld, H. P., & Rohwer, G., (
2002
). Techniques of event history modeling: New approaches to causal analysis. Mahwah, NJ: Lawrence Erlbaum Associates.
Christensen, K., Holm, N. V., McGue, M., Corder, L., & Vaupel, J. W., (
1999
). A Danish population-based twin study on general health in the elderly.
Journal of Aging and Health,
 
11
, (1),
49
-64.
Cicirelli, V. G., (
1980
). Sibling relationships in adulthood: A life span perspective. In L. W. Poon (Ed.), Aging in the 1980s: Psychological issues (pp. 445–462). Washington, DC: American Psychological Association.
Connidis, I. A., (
1992
). Life transitions and the adult sibling tie: A qualitative study.
Journal of Marriage and the Family,
 
54
,
972
-982.
Connidis, I. A., & Davies, L., (
1990
). Confidants and companions in later life: The place of family and friends.
Journal of Gerontology: Social Sciences,
 
45B
,
S141
-S149.
Cox, D. R., & Snell, E. J., (
1968
). A general definition of residuals (with definition).
Journal of the Royal Statistical Society, Series B,
 
30
,
248
-275.
Cumming, E., & Schneider, D. M., (
1961
). Siblings solidarity: A property of American kinship.
American Anthropologist,
 
63
,
498
-508.
Goetting, A., (
1986
). The developmental tasks of siblingship over the life cycle.
Journal of Marriage and the Family,
 
48
,
703
-714.
Hawkley, L, & Cacioppo, J. T., (
2004
). Stress and the aging immune system.
Brain, Behavior and Immunity,
 
18
,
114
-119.
Herskind, A. M., McGue, M., Holm, N. V., Sørensen, T. I. A., Harvald, B., & Vaupel, J. W., (
1996
). The heritability of human longevity: A population-based study on 2,872 Danish twin pairs born 1870–1900.
Human Genetics,
 
97
,
319
-323.
House, J. S., Umberson, D., & Landis, K. R., (
1988
). Structures and processes of social support.
Annual Review of Sociology,
 
14
,
293
-318.
Kohen, J. A., (
1983
). Old but not alone: Informal social supports among the elderly by marital status and sex.
The Gerontologist,
 
23
,
57
-63.
Lee, T. R., Mancini, J. A., & Maxwell, J. W., (
1990
). Sibling relationships in adulthood: Contact patterns and motivations.
Journal of Marriage and the Family,
 
52
,
431
-440.
Lillard, L. A., & Waite, L. J., (
1995
). Till death do us part: Marital disruption and mortality.
American Journal of Sociology,
 
100
,
1131
-1156.
Litwin, H., (
2001
). Social network type and morale in old age.
The Gerontologist,
 
41
,
516
-524.
Lopata, H. Z., (
1973
). Widowhood in an American city. Cambridge, MA: Schenkman.
Macdonald, A. M., (
2002
). Bereavement in twin relationships: An exploration of themes from a study of twinship.
Twin Research,
 
3
,
218
-226.
Neyer, F. J., (
2002
). Twin relationships in old age: A developmental perspective.
Journal of Social and Personal Relationships,
 
19
,
155
-177.
Olsen, R. B., Olsen, J., Gunner-Svensson, F., & Waldstrom, B., (
1991
). Social networks and longevity. A 14-year follow-up study among elderly in Denmark.
Social Science and Medicine,
 
33
,
1189
-1195.
Pearlin, L. I., Menaghan, E., Lieberman, M. A., & Mullan, J., (
1981
). The stress process.
Journal of Health and Social Behavior,
 
22
,
337
-356.
Powers, E. A., & Bultena, G. L., (
1976
). Sex differences in intimate friendship of old age.
Journal of Marriage and the Family,
 
38
,
739
-747.
Ross, C. E., Mirowsky, J., & Goldstein, K., (
1990
). The impact of family on health: The decade in review.
Journal of Marriage and the Family,
 
52
,
1059
-1078.
Sabin, E. P., (
1993
). Social relationships and mortality among the elderly.
The Journal of Applied Gerontology,
 
12
, (1),
44
-60.
Shye, D., Mullooly, J. P., Freeborn, D. P., & Pope, C. R., (
1995
). Gender differences in the relationship between social network support and mortality: A longitudinal study of an elderly cohort.
Social Sciences and Medicine,
 
41
,
935
-947.
Turner, R. J., & Marino, F., (
1994
). Social support and social structure: A descriptive epidemiology.
Journal of Health and Social Behavior,
 
35
,
193
-212.
Wood, V., & Robertson, J. F., (
1978
). Friendship and kinship interaction: Differential effect on the morale of the elderly.
Journal of Marriage and the Family,
 
40
,
367
-374.
Yashin A. I., & Iachine, I. A., (
1997
). How frailty models can be used for evaluating longevity limits: Taking advantage of an interdisciplinary approach.
Demography,
 
3/4
, (1),
31
-48.