Continued and New Personal Relationships in Later Life

Abstract

Objectives: The aim of this study is to increase our understanding of declining network size with aging by differentiating between processes of loss and gain and studying the associations with various health problems. Methods: Six observations of the Longitudinal Aging Study Amsterdam (LASA) across a time period of 16 years are used to study detailed network changes in a large sample of Dutch older adults aged 55 to 85 at baseline. Results: Results from multilevel regression analyses show that network size declines with aging, in particular for the oldest old. The decline in network size is to a large degree due to a lack of replacement of lost relationships with new relationships. Results show differential effects of health. Discussion: The older old and people in poor health have limited possibilities to compensate for network losses and may have a serious risk of declining network size in later life.

Keywords

personal network change compensation older adults longitudinal study health changes

Maintaining a large personal network is important for older adults. Having a substantial number of personal relationships reduces the risk of social isolation (Victor, Scambler, Bond, & Bowling, 2000; Wenger, Davies, Shahtahmasebi, & Scott, 1996), and personal networks provide support needed to secure autonomy and well-being in old age. Despite the importance of personal networks in old age, it is well-known that network size and social contact decline as people age (Barnes, Mendes de Leon, Bienias, & Evans, 2004; Krause, 1999; Lang, 2000; Shaw, Krause, Liang, & Bennett, 2007). In fact, the focus in later life is generally on network losses, resulting in a lack of support and loneliness. However, a focus solely on losses disregards the dynamic nature of personal networks in which losses may coincide with gains (Van Tilburg, 1998). For example, work relationships may after retirement be replaced by friendships with comembers of a voluntary organization. So, also in later life people might replace lost network members with new network members (Zettel & Rook, 2004). As a result, a declining network size may not primarily be caused by an increasing number of network losses but may also be the result of the incapacity to replace lost relationships, for example, due to poor physical and mental health. To date, little is known about the capacity of older adults to develop new network members and to compensate for network losses.

The aim of this study is to increase our understanding of declining network size with aging by differentiating between processes of loss and gain and studying the association with various health problems. As to our knowledge, these processes have not been distinguished in previous studies. The Longitudinal Aging Study Amsterdam (LASA; Huisman et al., 2011) provides a unique opportunity to study detailed network changes in a large sample of Dutch older adults aged 55 to 85 at baseline. The study consists of 6 observations across a time period of 16 years. Our data structure and measurements enable us to observe which network members are continued and which network members are new since the previous observation. Using all LASA observations, we examine changes over time in network size and processes of loss and gain. Besides describing changes with aging, we also attempt to explain why the networks of some people decrease more than others in old age. An important resource affecting capacities to maintain and replace lost relationships is health (Van Tilburg & Broese van Groenou, 2002). We add to existing knowledge by using longitudinal data to investigate the role of health in shaping the personal network in later life. Moreover, we explore the possible differential effects of various types of health problems. Previous studies analyzed cross-sectional data or focused on one dimension of health (e.g., Aartsen, Van Tilburg, Smits, & Knipscheer, 2004; Cornwell, 2009; Cornwell, Laumann, & Schumm, 2008; Penninx et al., 1999), but none of the studies have looked in more detail at effects of various health indicators on processes of loss and gain of personal relationships.

Aging and the Personal Network

There are several theoretical perspectives regarding social functioning in later life. Some of these focus on losses. For example, the detrimental or deficit model states that old age is characterized as period of decrease in social interaction due to a loss in societal roles. This was central in the disengagement theory, an influential theory in the early days of social gerontology (Cumming & Henry, 1961). More recently, the socioemotional selectivity theory (Carstensen, 1992) took an individual perspective and states that increasing network losses with aging are the result of a selection process. Older adults tend to narrow their personal networks as they age, by focusing on the most intimate personal relationships and discontinuing the mere instrumental relationships within the network. Others point out that network losses result from diminishing resources such as health and role changes due to for example, retirement or widowhood (Kahn & Antonucci, 1980). According to these perspectives, age and network size are strongly negatively correlated. Yet other theories point at compensation mechanisms that may strengthen and extend the personal network. An example is the social compensation model that considers a wide range of adaptive responses to loss and change in social relationships (Ferraro & Farmer, 1995). According to this model, the primary strategy for dealing with loss is compensation. People strive to maintain continuity in their personal networks by replacing relationships that have been lost. This compensation mechanism may reduce the negative correlation between age and network size. Although older adults may generally strive to maintain continuity in their personal networks as they age, changes in socioemotional objectives (according to the socioemotional selectivity theory) and in roles and resources (according to the convoy model) may affect both their capacity and disposition to maintain and to replace lost relationships. This may result in an age-specific decline in network size; the oldest age groups experience stronger decline than younger age groups (Barnes et al., 2004; Van Tilburg, 1998) due to changes in multiple roles and resources in late life or selective choices regarding who to maintain contact with.

Regarding age effects, two hypotheses are formulated:

Hypothesis 1a: With increasing age, a decrease in network size is expected, that is a decline in the number of continued and the number of new relationships.

Hypothesis 1b: We expect a stronger decrease in network size among the oldest old than among the youngest old, due to a relatively large number of lost relationships and insufficient new relationships.

Effects of Health

It is well documented that health problems become more prominent in later life and have major consequences for daily functioning. Good health is a resource needed to maintain social connections (Ertel, Glymour, & Berkman, 2009; Morgan, 1988). Health may also influence the needs with regard to personal relationships. If health problems arise, there are several mechanisms that may affect the maintenance and development of personal relationships. First, poor physical health may decrease opportunities to be socially active because older adults’ mobility is affected. Decreased mobility makes it more difficult to travel to network members, and therefore face-to-face contact will be reduced and may finally lead to loss of relationships. Furthermore, mobility decline restricts people in the performance of roles and participation in organizations. Examples are the limited opportunities to participate in volunteer organizations (Li & Ferraro, 2006) or participation in religious social life (Kelley-Moore & Ferraro, 2001). Second, cognitive or mental health problems may change the ability to communicate with others (Strawbridge, Wallhagen, Shema, & Kaplan, 2000), which is essential for maintaining and developing relationships. For example, without proper communication, it is difficult to share feelings or to ask for help. Third, health problems may cause imbalance in the exchange of support. According to social exchange theory, people tend to keep the support exchanges in their social relationships in balance. This is known as the norm of reciprocity (Gouldner, 1960). A decline in health makes it difficult to give support and thereby to reciprocate the support received. If there is an imbalance in the exchange of support, a relationship is likely to end (Klein Ikkink & Van Tilburg, 1999). Fourth, poor health increases older adults’ need for support. People with health problems are more dependent on others and may adapt their network to meet their needs. Health decline triggers helpers in the network to provide support (Schwarzer & Leppin, 1991). During aging the personal network may increasingly consist of people who provide the majority of the informal support, which may result in loss of network members not involved in helping with everyday activities (Stoller & Pugliesi, 1991). At the same time, new people may enter the personal network, such as neighbors who take the role of caregiver (Barker, 2002). Following these four mechanisms of reduced opportunities, reduced ability to communicate, increase of imbalanced support exchange, and increased need of support, we conclude that losses in the personal network are likely after health decline. New support givers may enter the network, but full replacement of lost relationships is unlikely. Our second hypothesis is:

Hypothesis 2: Poor health is associated with a small network size and a small number of continued and new relationships.

Processes of loss and gain in personal networks are, at least partly, determined by the type of health problems experienced by an older adult. We distinguish between physical, sensory, cognitive, and mental impairment. Physical health decline may lead to functional limitations (Guralnik, Fried, & Salive, 1996). As a consequence, people have fewer opportunities to meet others predominantly due to reduced mobility. Sensory impairment alters a person’s ability to communicate with others. Personal relationships will be affected because social interaction relies heavily on the availability of information of the senses (Slawinski, Hartel, & Kline, 1993; Wang & Boerner, 2008). With both cognitive and mental health problems, there may be an accumulation of mechanisms. Cognitive and mental health problems often radically change the social behavior of individuals and make it difficult to keep functioning in everyday life (Kennedy, Foy, Sherazi, McDonough, & McKeon, 2007). There is not only a reduced ability to communicate but also the opportunities to meet others are restricted. Cognitive and mental health problems may be more disruptive for personal networks than physical health problems. First, this may be explained by the mentioned accumulation of mechanisms. Second, because physical and sensory impairments can be partly overcome by the use of helping devices (Hoenig, Taylor, & Sloan, 2003; Strawbridge et al., 2000), there may be more variation in the social consequences of physical and sensory decline than of cognitive and mental decline. There is some evidence for this specificity by types of health problems. In a study among Dutch older adults, Aartsen et al. (2004) investigated both the effects of physical and cognitive decline on personal networks and observed that especially cognitive decline was associated with smaller network size. The third hypothesis is:

Hypothesis 3: The effects of cognitive and mental impairment on the number of continued and new relationships are expected to be stronger than the effects of physical and sensory impairment.

Method

Study Sample

Data were derived from LASA, an ongoing study on physical, emotional, cognitive, and social functioning of older adults (Huisman et al., 2011). In 1992-1993 (Time 1) a survey was conducted among 3,107 respondents born between 1908 and 1937. The oldest respondents, particularly the oldest men, were overrepresented in the sample. The LASA sample was initially recruited for the Living Arrangements and Social Networks of Older Adults research program (Knipscheer, De Jong Gierveld, Van Tilburg, & Dykstra, 1995). Data were collected by means of computer-assisted personal interviewing on physical, emotional, and social functioning and medical interviews with clinical observations. For the first observation in 1992, the cooperation rate was 62%. The sample was stratified by sex and age, and respondents were randomly selected from the population registers of 11 municipalities in the west, northeast, and south of the Netherlands.

Follow-ups were carried out in 1995-1996 (Time 2, n = 2,545), 1998-1999 (Time 3, n = 2,076), 2001-2002 (Time 4, n = 1,691), 2005-2006 (Time 5, n = 1,257) and 2008-2009 (Time 6, n = 985). For each follow-up, on average 80% of the respondents was re-interviewed, 15% had died, 1% was too ill or too cognitively impaired to be interviewed, 3% refused to be reinterviewed, and less than 1% could not be contacted due to residential relocation to another country or an unknown, sometimes temporary, destination. For each observation, interviewers received a 4-day training and were intensively supervised. The interviews were taperecorded to monitor and enhance the quality of the data obtained. The interviews lasted between one and a half and two hours.

Data on personal network were available for 85% of the respondents across the six observations. Reasons for missing data were premature termination of an interview (1%), the use of an abridged version of the questionnaire (6%), or a telephone interview with the respondent or a proxy (9%) for respondents who were too physically or cognitively frail to be interviewed with the full questionnaire. The final data set consists of 9,852 observations of 2,960 respondents, ranging from 1 to 6 observations and on average 3.3 (SD = 1.9). Respondents were followed for a maximum of 16.9 years (M = 7.7, SD = 6.2). The age of the 1,434 men and 1,526 women varied between 54 and 100 (M = 73.4, SD = 8.5) at the various observations. The far majority lived independently; at each observation about 3% were institutionalized.

Attrition during the study period may have contributed to a selective study sample. Of all 6,685 observations, 809 observations concerned respondents who had died at either one of the follow-ups, and 899 observations belonged to respondents who were left out of either one of the follow-ups due to incomplete data or refusal. In order to examine the specific characteristics of the longitudinal study sample we conducted a logistic regression on attrition due to incomplete data or refusal and a logistic regression on mortality for all observations. Attrition due to incomplete data or refusal (compared to nonattrition) occurred significantly (p < .05) more often among males, older persons, those with more physical limitations, lower cognitive functioning, more depressive symptoms, and living in more urbanized areas (results not shown). Mortality (compared to nonmortality) occurred significantly (p < .05) more often among males, older persons, those with smaller network sizes, more physical limitations, poor vision, lower cognitive functioning, and more depressive symptoms. These results implied that our study participants were relatively healthy compared to those who were excluded from the sample. In order to take account of this selective study attrition, we included a variable indicating whether the respondent was deceased at the next follow-up. Analysis including a variable indicating attrition due to missing data or refusal at next follow-up provided no additional information beyond the mortality status, so we excluded these variables from our final analyses.

Measures

Personal networks were identified using the domain-contact method (Van Tilburg, 1998). Network members were identified in seven domains: household members (including the partner if there was one); children and their partners, other relatives, neighbors, colleagues from work (including voluntary work) or school; fellow members of organizations (e.g., athletic clubs, churches, political parties); and others (e.g., friends and acquaintances). With respect to the domains, the question was asked: “Name the people with whom you have frequent contact and who are important to you.” People could be named once. The design of the measurements for all observations was the same, thus giving equal chances to network members identified in a previous observation and to others to be identified in later observations. The network size was computed as the number of individuals identified. The number of network members identified was between 0 and 75. Furthermore, we constructed two variables that indicated the number of continued and new relationships at each observation. The names of network members identified in different observations were compared and, if possible, linked in order to identify whether a person was new in the network (i.e., not identified at the previous observation) or continued to be network member. The network size at follow-up observations equals the number of new and continued relationships.

Health indicators included physical functioning, sensory impairment, cognitive functioning, and mental health. Physical functioning was assessed by asking whether respondents have difficulty performing six common daily activities: walking up and down a 15-step staircase without resting, getting dressed and undressed, sitting down and getting up from a chair, cutting their own toenails, walking 5 min outdoors without resting and driving or using public transport (Katz, Ford, Moskowitz, Jackson, & Jaffe, 1963). The response categories were: unable to do that, only with help, with a great deal of difficulty, with some difficulty, and without difficulty. The scales range from 6 to 30, with higher scores indicating better functional capacity. Cronbach’s alpha across the six observations was .83.

Sensory impairment was assessed using questions on hearing and vision from the Organization for Economic Cooperation and Development long-term disability indicator (McWhinnie, 1979). Hearing was assessed by asking three questions on whether they could follow a conversation with one or four persons (using a hearing aid if needed) and were able to use a normal phone (Kramer, Kapteyn, Kuik, & Deeg, 2002). Vision was assessed by asking respondents on whether they could read the small print in a newspaper and could recognize someone’s face at a distance of four meters (wearing glasses or contact lenses if needed). All questions were scored on a scale: (1) no I cannot, (2) with much difficulty, (3) with some difficulty, and (4) without difficulty. The average of items scores were computed to assess hearing and visual capacities.

Cognitive functioning was assessed with the Mini Mental State Examination (MMSE; Folstein, Folstein, & McHugh, 1975), which involves recall, orientation, registration, attention, language, and construction. Scores range from 0 to 30, with higher scores indicating better cognitive performance. As indicator of mental health, depressive symptoms were measured using the Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977). This is a 20-item self-report scale developed to measure depressive symptoms in the community. Respondents were asked how often they experienced each symptom during the previous week. Scores range from 0 to 60, with higher scores indicating more depressive symptoms. Cronbach’s alpha across the six observations was .86.

As both health status and network size are known to vary by various sociostructural individual characteristics (e.g., Cornwell et al., 2008; Kahn & Antonucci, 1980; Van Tilburg & Broese van Groenou, 2002), we controlled for sex (0 = male, 1 = female), age, time (years since Time 1), partner status (0 = no partner, 1 = having partner in or outside the household), level of education attained (in years, ranging from 5 to 18), urbanization (1 = low, 5 = high), and employment status (0 = not employed, 1 = employed), in addition to being deceased at follow-up (0 = no, 1 = yes). The characteristics of the sample are presented in Table 1. Note that the proportion of deceased at Time 4 is somewhat higher than at the other follow-ups, which is due to the fact that there the time interval between Time 3 and Time 4 is 4 years, whereas it is 3 years at all other follow-ups. When the text refers to “mortality status at follow-up” this implies a mortality status within 3 to 4 years.

Table 1.

Descriptives for study variables.

	Time 1		Time 2		Time 3		Time 4		Time 5		Time 6
	2,885		2,203		1,723		1,351		935		755
N	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD
Time since baseline (years)	0	0	3.1	0.2	6.0	0.1	9.0	0.1	13.1	0.1	16.1	0.1
Age (years)	70.3	8.7	72.3	8.5	73.8	8.0	75.3	7.5	77.3	6.3	79.6	6.0
Female (vs. male)	51%	—	—	—	—	—	—	—	—	—	—	—
Deceased at follow-up (vs. not)	12%		12%		11%		15%		11%		12%
Network size (0-75)	13.8	8.3	14.5	8.6	14.6	8.8	14.9	8.6	15.3	8.8	15.1	9.3
Having partner (vs. no partner)	68%		63%		61%		60%		60%		56%
Educational level (years)	8.8	3.3	—	—	—	—	—	—	—	—	—	—
Employed (vs. not employed)	11%		7%		4%		3%		2%		2%
Level of urbanization (1-5)	2.9	1.5	3.0	1.5	3.0	1.4	3.0	1.4	3.0	1.5	3.0	1.4
Functional capacity (6-30)	27.4	4.5	27.0	4.7	26.8	4.8	26.6	4.8	26.2	4.7	25.6	5.4
Visual capacity (1-4)	2.7	0.6	2.7	0.7	2.8	0.6	2.8	0.6	2.8	0.6	2.8	0.7
Hearing capacity (1-4)	3.8	0.4	3.8	0.4	3.8	0.4	3.8	0.4	3.7	0.4	3.7	0.4
Cognitive capacity (0-30)	27.1	2.7	26.9	3.0	27.1	2.8	27.1	2.8	27.1	2.6	27.1	2.8
Depressive symptoms (0-60)	7.7	7.6	8.0	7.7	8.3	7.3	9.0	7.5	8.6	7.0	8.4	7.2

Procedure

Observations are nested within respondents and multilevel regression analysis was applied (MLwiN; Rasbash, Steel, Browne, & Prosser, 2004). In Model 1 we regressed personal network size, the number of continued and of new relationships on time, age at baseline, and the interaction between age and time, adjusted for gender and mortality status at next follow-up. In Model 2, control variables and health characteristics are added to Model 1, to explore the associations between health status and the dependent variables. This also allows estimating the net effect of health status and changes in health on the continuation and development of network relationships over time. Gender and level of education were entered in the models as time-fixed variables; all other variables were entered as time-varying variables. The number of new and continued relationships was only computed for Time 2 and following observations (leaving n = 2,280 with 6,967 observations) and when there were two consecutive observations of the network (leaving n = 2,210 with 6,685 observations). All independent variables were centered around the mean. Unstandardized coefficients are presented in Table 3 allowing comparison of effects between the models of the three dependent variables.

Results

Descriptive Results

Table 1 presents the means and standard deviations for personal network, control variables, and health characteristics. Comparison over time is difficult as the figures combine the effects of sample attrition and the changes in the lives of people. For example, at Time 6 respondents are 16 years older than at Time 1; however, the average age has increased with 9.3 years. At baseline the average network size is 13.8 persons with a large variation (SD = 8.3) within the sample.

Detailed figures on changes in the network are presented in Table 2. The numbers of continued, new, and lost relationships at each observation show that there is a lot of turnover at the level of network members. From Time 1 to Time 2 on average relationships with 8.8 network members are continued and 5.7 are new. The network size of the 2,136 respondents included in this analysis (67 respondents are excluded because data on their Time 1 networks are missing) is therefore 14.5. On average, 5.4 relationships were lost from a network size of 14.2 (8.8 + 5.4 = 14.2) at Time 1. The breakdown of these figures according to relationship type is presented in the table. For example, there were more respondents who lost their partner (7%) than respondents who started a new partner relationship (1%), and relationships with other kin were more often started (M = 0.9) than lost (M = 0.6). On average, across the six observations 62% of the network members (9.2 persons) were continued from the previous observation; 5.6 persons were new, meaning that they were not identified at the previous observation. It is however likely that some changes occur in the network periphery. For example, newly started relationships with a sibling-in-law might be related to marriage of a sibling, however, will often be due to increasing contact frequency or importance with the sibling-in-law. New relationships might also be renewed, that is, persons returned in the network; for example, 38% of the relationships new at Time 6 were in the network at one of the Time 1 to Time 4 observations. The average of lost network members was 5.8 reflecting that lost network members generally were replaced.

Table 2.

Average Number of Continued, New, and Lost Relationships.

	Time 2			Time 3			Time 4			Time 5			Time 6
	2,136			1,672			1,282			888			707
N	Continued	New	Lost	Continued	New	Lost	Continued	New	Lost	Continued	New	Lost	Continued	New	Lost
Total	8.81	5.70	5.41	9.24	5.35	5.90	9.49	5.58	5.78	9.38	6.08	6.30	9.59	5.67	6.31
Partner	0.63	0.01	0.07	0.61	0.01	0.05	0.59	0.01	.06	0.59	0.00	0.06	0.56	0.01	0.07
Child	2.40	0.20	0.19	2.34	0.16	0.25	2.37	0.23	0.15	2.37	0.13	0.20	2.34	0.16	0.18
Child-in-law	1.22	0.43	0.41	1.26	0.38	0.41	1.28	0.39	0.41	1.27	0.33	0.40	1.18	0.34	0.43
Sibling	0.98	0.41	0.43	1.01	0.35	0.44	1.06	0.38	0.41	1.07	0.32	0.45	1.04	0.33	0.43
Sibling-in-law	0.76	0.83	0.72	0.95	0.77	0.77	0.99	0.70	0.93	0.91	0.65	0.91	0.86	0.70	0.79
Other kin	0.50	0.87	0.63	0.61	0.87	0.76	0.62	0.92	0.75	0.62	1.00	0.83	0.68	1.12	0.84
Friend	0.72	0.62	0.87	0.69	0.57	0.75	0.68	0.61	0.73	0.69	0.78	0.77	0.83	0.64	0.84
Neighbor	0.87	0.85	0.77	0.91	0.80	0.82	1.00	0.92	0.83	0.98	1.10	0.97	1.14	0.86	0.99
Other nonkin	0.73	1.48	1.32	0.86	1.44	1.64	0.90	1.43	1.51	0.88	1.75	1.70	0.96	1.51	1.74

Table 3.

Multilevel Regression of Personal Network Size, and Number of Continued and New Relationships; Unstandardized Coefficients (Respondents, n = 2,210; Observations, n = 6,685).

	Network size			Number of continued relationships			Number of new relationships
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
	B	B	B	B	B	B	B	B	B
Constant	14.52***	14.61***	14.63***	8.99***	9.05***	9.06***	5.58***	5.59***	5.59***
Female (vs. male)	−0.07	.20	1.01**	−0.01	0.16	0.83***	−0.09	0.00	0.22
Age at baseline (54.8-85.6 years)	−0.24***	−0.17***	−0.12***	−0.17***	−0.13***	−0.08***	−0.07***	−0.04***	−0.03*
Time since baseline (0-16.8 years)	−0.08***	−0.04	−0.01	−0.05***	−0.02	0.01	−0.02	−0.01	0.00
Age at baseline * Time since baseline	−0.014***	−0.012***	−0.011***	−0.009***	−0.007***	−0.006***	−0.004*	−0.004	−0.004*
Deceased at follow-up (vs. not)	−1.08***	−0.74**	−0.68*	−0.57***	−0.38*	−0.34*	−0.58**	−0.41	−0.39
Functional capacity (6-30)	—	0.03	0.03	—	0.04*	0.03	—	0.00	0.00
Visual capacity (1-4)	—	0.40*	0.37*	—	0.03	0.02	—	0.25*	0.23*
Hearing capacity (1-4)	—	0.60*	0.61*	—	0.24	0.24	—	0.38*	0.38*
Cognitive capacity (0-30)	—	0.25***	0.24***	—	0.14***	0.13***	—	0.12***	0.11***
Depressive symptoms (0-60)	—	−0.06***	−0.04**	—	−0.04***	−0.03**	—	−0.02*	−0.01
Having partner (vs. no partner)	—	—	1.55***	—	—	1.50***	—	—	0.27
Educational level (years)	—	—	0.16***	—	—	0.10**	—	—	0.07**
Employed (vs. not employed)	—	—	1.05*	—	—	0.48	—	—	0.90*
Level of urbanization (1-5)	—	—	−0.61***	—	—	−0.29***	—	—	−0.29***

Note. Values of all explanatory variables are centered around the mean.

p < .05. **p < .01. ***p < .001.

Changes in Network Size

We regressed network size on sex, age at baseline and time, and controlled for mortality at follow-up, to reveal whether network size was stable across the observations taking into account sample attrition. The unstandardized regression coefficient (B) for time is –.08. Congruent with Hypothesis 1a the significance of this effect indicates that there is a decline of network size from Time 2 to Time 6. However, a much stronger effect (B = –.24) is observed for age differences at baseline indicating that at Time 2 the oldest respondents had small networks compared to younger respondents. Furthermore, we observed an interaction of changes over time and age at baseline (B = –.014) indicating that decline in network size was larger for respondents older at baseline, congruent with Hypothesis 1b. Respondents who were deceased at the follow-up observation were estimated to have 1.1 network members less than those alive at the follow-up observation.

To illustrate these observations Figure 1 presents point estimates for the years 1992-2009 for two age groups: 57-year-olds and 77-year-olds at baseline (aged 60 and 80 at Time 2), adjusted for gender and mortality at follow-up. The figure shows that a 57-year-old on average has a network size of 16.1 at baseline, which is almost stable over time taking into account both lost and new relationships. A 77-year-old starts out with on average 14.2 network members at baseline, which decreases to 10.0 network members at Time 6 when he or she is 93, an estimated net loss of 4.2 network members. This loss is estimated as 5.3 when this person deceases after Time 6, about 1 per 3 years of aging.

Figure 1.

Estimated change over 16 years in network size, and number of continued and new relationships for 60-year-olds and 80-year-olds at Time 2, adjusted for gender and mortality at follow-up.

Specifying the net changes into continued and new relationships, the analyses also show interesting effects of age at baseline and time. Similar to network size, differences in the number of continued and new relationships were observed according to age at baseline (B = –.17 and –.07, respectively). The number of continued and, not significantly, of new relationships decreased in the course of the longitudinal study (B = –.05 and –.02, respectively), and these changes were stronger among the oldest at baseline, as indicated by the interaction effects (B = –.009 and –.004, respectively). Figure 1 illustrates that there is no change in the number of continued and new relationships for a 60-year-old over 13 years. An 80-year-old is estimated to have a decrease in the number of continued relationships (2.1) as well as in the number of new relationships (1.0). Both numbers are 0.6 higher when this person deceases after Time 6; taken together this makes a loss of 4.3, that is, about 1 per 3 years of aging. It can be concluded that among older age groups network size decreases significantly over time, in particular due to discontinuation of relationships that are not replaced in sufficient numbers by new relationships. However, there is no sign that replacement ends at higher ages.

Impact of Health on Network Changes

We regressed the five health indicators on time since baseline to reveal whether health changed over respondent’s course within the study. Controlled for gender, health decline was observed for physical (B = –.35), hearing (B = –.15), and cognitive capacities (B = –.23), and for depressive symptoms (B = .11). Visual capacities increased (B = .06), probably due to changing glasses or surgery. All effects are statistically significant (p < .001).

Results for Model 2 presented in Table 2 show for all three dependent variables to what degree the effects of age at baseline and time are associated with health. Comparing the effects of age at baseline and time on network size in Models 1 and 2, we see that these are decreased considerably in Model 2, suggesting that the negative effects of age and time on network size are to an important part due to age differences in health- and time-related decline in health. Adding the sociostructural variables in Model 3 hardly changes the effects of the health factors, showing the independent effects of health on network size. Supporting Hypothesis 2 we observed significant effects for all health characteristics in Model 3. In particular, congruent with Hypothesis 3 a better cognitive capacity and less depressive symptoms add to a larger network size, whereas functional limitations do not. A better visual and hearing capacity also contributes to a larger network size. It can be concluded that age differences in network size as well as the decline in network size over time are to a large degree due to differences in sensory, cognitive, and emotional health.

Specifying the net changes into continued and new relationships (Model 3) again the importance of good health is visible. Cognitive capacities and absence of depressive symptoms are more important for the maintenance of relationships while visual and hearing capacities are more important for the start of new relationships.

Women and respondents with a partner are advantaged in maintaining relationships, but differences in gender and partner status are not related to starting new relationships. Having a high educational level and living in rural areas is equally important for maintaining and starting new relationships. Being employed facilitates starting new relationships.

Discussion

The first objective of this study was to examine changes in personal network size with aging. We showed that network size declines with age, in particular for the oldest old. More specific, the decline in network size is to a large degree due to a lack of replacement of lost relationships with new relationships, again in particular among the oldest old. The results indicate a large turnover in personal relationships over time: The average number of lost and new relationships is pretty comparable at all observations (Table 2), but it was relatively high among other nonkin relationships and relatively low among relationships with children. Still, with aging the replacement of lost relationships became more difficult and we estimated a net loss of 1 relationship every 3 years on average among the older old. Figure 1 illustrated that young old are better able to replace lost relationships resulting in a stable network size over time. It is concluded that replacement of network relationships is rather common among the young old but insufficient among the older old resulting in network losses in later life. The compensation mechanism seems to work fine until a decline in health distorts the process of replacing lost relationships.

The age-related decline in our findings may in part reflect cohort differences in social and personal resources (as education, employment, partner status, and health) that are due to the different societal circumstances of the young old and the oldest old earlier in life. In the present study we controlled for sociostructural factors, which may in part take care of these cohort differences. More important, our results suggest that the age-related decline may be postponed in the long run among the young old, although the present study cannot provide proof for such a favorable development. In previous studies we observed cohort differences in age-related decline in social outcomes, as volunteering (Broese van Groenou & Van Tilburg, 2012) and having friends in the social network (Stevens & Van Tilburg, 2011). The current and the previous studies were based on a longitudinal data set with a time span of 17 years or less and revealed that between the ages of 55 and 75 no decline in network size, friendship, or volunteering is notable, in contrast to developments among older people. The young old have to be observed for a longer period to establish whether age-related decline in network size is actually postponed until after the age of 75 for later cohorts. Such long-term studies may underscore that the notion of “active aging” indeed replaces the general view of “social disengagement.” We can at least be optimistic because in particular the young old are able and capable to engage in new forms of social engagement, resulting in the development of new personal relationships and stable network sizes.

The second objective of this study was to explore the effects of health on processes of loss and gain in personal networks. Poor health was associated with lower network size and lower numbers of continued and new relationships, supporting Hypothesis 2. We also found evidence for differential effects of health, in line with Hypothesis 3. We observed that mental and cognitive health were more important for the numbers of continued and new relationships than functional limitations that only have a minor effect on the continuation of relationships and do not impact the gain of relationships. Sensory capacity proved important for establishing new relationships only. Especially, cognitive health seems to play an important role with regard to maintaining and developing relationships. The interpretation of these results might be sought in two directions. First, physical limitations primarily affect the mobility of a person. With reduced mobility, there are still alternatives to stay in contact with other people (e.g., mobile phones and internet) and there are helping devices such as wheelchairs to keep functioning in social activities. Second, cognitive and mental decline changes both the ability to communicate and reciprocate support. The support needs of older adults with mental and cognitive health problems are complex, which may limit the number of potential providers of support. It seems that the ability to communicate is more fundamental for developing and maintaining personal relationships than mobility. It can be concluded that health, in particular cognitive and mental health, is an important resource for maintaining a stable network size in old age.

The limited replacement of lost relationships among the oldest old may also be due to a changed disposition toward the maintenance of specific relationships. According to the socioemotional selectivity theory the more social-emotionally close relationships are likely to be continued and the more instrumental-based relationships are likely to be discontinued. A true test of such explanation would require information on the support exchanges within network relationships and on individual standards regarding emotion-based and instrumental-based relationships. Information about support exchanges is only available for the nine most frequently contacted members in the network in the LASA data. Using this limited data set would have restricted our study on the replacement of relationships as this occurs more often in the network periphery. Besides support exchanges, information about the proximity of relationships may also have added insight in how health impacts personal relationships. Proximity of relationships could be important for those who develop physical mobility limitations and they may lose relationships that are difficult to contact face to face. This may be the case for contacts in sport organizations, long-distance friends, and so on. Further research into the dynamics of network relationships in later life should take more information about the supportive nature, the distance, and the contact modus of network relationships into account.

A limitation of longitudinal studies is the attrition of respondents over time due to frailty or mortality, which may lead to a relatively healthy sample and, in our study, an underestimation of the network loss over time. Our attrition analyses did show important differences between those who dropped out and those who remained in the study sample. Yet controlling for mortality status at follow-up took care of possible sample biases and did, in fact, hardly change the effects of age and time on the loss and continuation of network relationships. Although it was beyond the focus of the study, it is interesting to note that mortality status has, regardless of health factors, a direct negative effect on the continuation of relationships but not on new relationships. This suggests that for those who died within the next 3 or 4 years, a decline in health limited them to the development of new relationships but that additional explanations are needed why they lost relationships with network members. More research at the network relationship level may show which relationships are likely to be lost in the years before dying and whether the degree of interaction, level of support, or domain of the relationship may be at stake here.

A remark should be made regarding the assessment of loss and gain of network relationships. For the current study we only compared the names of network members with the observation 3 years earlier and did not use the information whether network members had been identified at previous waves. In fact, about 38% of the new network members had been identified at previous observations. This implies that “new” relationships may also reflect a revival of the contact with a network member, whereas “lost” network members may turn up again at a follow-up. Previous studies showed that the revival of relations with old friends and acquaintances may serve as a compensatory function (Lang, 2000). They take over the position of network members who are lost and our study shows that being still in good health may facilitate this replacement of more superficial relationships with old friends. For those with cognitive and mental health problems, the replacement of relationships may in part reflect difficulties with the recall of network members, and these recall problems may even increase over time. But, as the attrition analysis showed that those with poor cognitive and mental health are likely to drop out during the study period, it is not likely that recall issues contribute to many new relationships. Also, the majority of the gained relationships are included in the network for the first time. This suggests that, probably due to late life transitions such as retirement, grandparenthood, or widowhood, social domains have been (re)discovered from which new relationships are recruited.

In conclusion, this study shows that the personal network size is relatively stable in old age but with substantial changes in personal relationships. In general, the development of new relationships continues in later life and almost fully compensates for network losses; however, for some the replacement process does end at a certain time. Our results show that the older old, people in poor health, and those who decease within the next 3 to 4 years have limited possibilities to compensate for network losses and may have a serious risk of declining network size. In particular, aging adults with cognitive and mental health problems are at risk for relationship losses. With a decline in network size, the potential sources of support are diminished and older adults’ dependency on formal care is increased. The associations between health and personal networks will become increasingly important in the future with growing numbers of older adults. People will live longer with health problems and the prevalence of multimorbidity is increasing (Fortin, Soubhi, Hudon, Bayliss, & Van den Akker, 2007). Although this study gave some insights into the differential effects of health, more research is needed to understand the effects of specific health problems on personal network characteristics and processes.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Longitudinal Aging Study Amsterdam is largely supported by a grant from the Netherlands Ministry of Health Welfare and Sports, Directorate of Long-Term Care.

References

Aartsen

M. J.

Van Tilburg

T. G.

Smits

C. H. M.

Knipscheer

C. P. M.

(2004). A longitudinal study on the impact of physical and cognitive decline on the personal network in old age. Journal of Social and Personal Relationships, 21, 249-266. doi:10.1177/0265407504041386

Barker

J. C.

(2002). Neighbors, friends, and other nonkin caregivers of community-living dependent elders. Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 57, 158-167. doi:10.1093/geronb/57.3.S158

Barnes

L. L.

Mendes

Leon

C. F.

Bienias

J. L.

Evans

D. A.

(2004). A longitudinal study of black-white differences in social resources. Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 59, 146-153. doi:10.1093/geronb/59.3.S146

Broese van Groenou

M. I.

Van Tilburg

T. G.

(2012). Six-year follow-up on volunteering in later life: A cohort comparison in the Netherlands. European Sociological Review, 28, 1-11. doi:10.1093/esr/jcq043

Carstensen

L. L.

(1992). Social and emotional patterns in adulthood: Support for socioemotional selectivity theory. Psychology and Aging, 7, 331-338. doi:10.1037/0882-7974.7.3.331

Cornwell

(2009). Network bridging potential in later life: Life-course experiences and social network position. Journal of Aging and Health, 21, 129-154. doi:10.1177/0898264308328649

Cornwell

Laumann

E. O.

Schumm

L. P.

(2008). The social connectedness of older adults: A national profile. American Sociological Review, 73, 185-203. doi:10.1177/000312240807300201

Cumming

Henry

W. E.

(1961). Growing old: The process of disengagement. New York, NY: Basic Books.

Ertel

K. A.

Glymour

M. M.

Berkman

L. F.

(2009). Social networks and health: A life course perspective integrating observational and experimental evidence. Journal of Social and Personal Relationships, 26, 73-92. doi:10.1177/0265407509105523

10.

Ferraro

K. F.

Farmer

M. M.

(1995). Social compensation in adulthood and later life. In Dixon

R. A.

Backman

(Eds.), Compensating for psychological deficits and declines: Managing losses and promoting gain (pp. 127-145). Hillsdale, NH: Lawrence Erlbaum.

11.

Folstein

M. F.

Folstein

S. E.

McHugh

P. R.

(1975). Mini-Mental State: A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12, 189-198. doi:10.1016/0022-3956(75)90026-6

12.

Fortin

Soubhi

Hudon

Bayliss

E. A.

Van den Akker

(2007). Multimorbidity’s many challenges. British Medical Journal, 334, 1016-1017. doi:10.1136/bmj.39201.463819.2C

13.

Gouldner

A. W.

(1960). The norm of reciprocity: A preliminary statement. American Sociological Review, 25, 161-178.

14.

Guralnik

J. M.

Fried

L. P.

Salive

M. E.

(1996). Disability as a public health outcome in the aging population. Annual Review of Public Health, 17, 25-46. doi:10.1146/annurev.pu.17.050196.000325

15.

Hoenig

Taylor

D. H.

Jr. Sloan

F. A.

(2003). Does assistive technology substitute for personal assistance among the disabled elderly? American Journal of Public Health, 93, 330-337. doi:10.2105/AJPH.93.2.330

16.

Huisman

Poppelaars

Van der Horst

Beekman

A. T. F.

Brug

Van Tilburg

T. G.

Deeg

D. J. H.

(2011). Cohort profile: The Longitudinal Aging Study Amsterdam (LASA). International Journal of Epidemiology, 40, 868-876. doi:10.1093/ije/dyq219

17.

Kahn

R. L.

Antonucci

T. C.

(1980). Convoys over the life course: Attachment, roles and social support. In Baltes

P. B.

Brim

(Eds.), Life span development and behavior (pp. 253-286). New York, NY: Academic Publishers.

18.

Katz

Ford

A. B.

Moskowitz

R. W.

Jackson

B. A.

Jaffe

M. W.

(1963). Studies of illness in the aged; the index of ADL: A standardized measure of biological and psychological function. Journal of the American Medical Association, 185, 914-919. doi:10.1001/jama.1963.03060120024016

19.

Kelley-Moore

J. A.

Ferraro

K. F.

(2001). Functional limitations and religious service attendance in later life: Barrier and/or benefit mechanism? Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 56, 365-373. doi:10.1093/geronb/56.6.S365

20.

Kennedy

Foy

Sherazi

McDonough

McKeon

(2007). Long-term social functioning after depression treated by psychiatrists: A review. Bipolar Disorders, 9, 25-37. doi:10.1111/j.1399-5618.2007.00326.x

21.

Klein Ikkink

C. E.

Van Tilburg

T. G.

(1999). Broken ties: Reciprocity and other effects on the ending of older adult’s relationships. Social Networks, 21, 131-146. doi:10.1016/S0378-8733(99)00005-2

22.

Knipscheer

C. P. M.

De Jong Gierveld

Van Tilburg

T. G.

Dykstra

P. A.

(1995). Living arrangements and social networks of older adults. Amsterdam, Netherlands: VU University Press.

23.

Kramer

S. E.

Kapteyn

T. S.

Kuik

D. J.

Deeg

D. J. H.

(2002). The association of hearing impairment and chronic diseases with psychosocial health status in older age. Journal of Aging and Health, 14, 122-37. doi:10.1177/089826430201400107

24.

Krause

(1999). Assessing change in social support during late life. Research on Aging, 21, 539-569. doi:10.1177/0164027599214002

25.

Lang

F. R.

(2000). Endings and continuity of social relationships: Maximizing intrinsic benefits within personal networks when feeling near to death. Journal of Social and Personal Relationships, 17, 155-182. doi:10.1177/0265407500172001

26.

Ferraro

K. F.

(2006). Volunteering in middle and later life: Is health a benefit, barrier or both? Social Forces, 85, 497-519. doi:10.1353/sof.2006.0132

27.

McWhinnie

J. R.

(1979). Disability indicators for measuring well-being (OECD Social Indicators Development Program, Special Studies No. 5). Paris: Organization for Economic Cooperation and Development.

28.

Morgan

D. L.

(1988). Age differences in social network participation. Journal of Gerontology, 43, 129-137. doi:10.1093/geronj/43.4.S129

29.

Penninx

B. W. J. H.

Van Tilburg

T. G.

Kriegsman

D. M. W.

Boeke

A. J. P.

Deeg

D. J. H.

Van Eijk

J. T. M.

(1999). Social network, social support, and loneliness in older persons with different chronic diseases. Journal of Aging and Health, 11, 151-168. doi:10.1177/089826439901100202

30.

Radloff

L. S.

(1977). The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385-401. doi:10.1177/014662167700100306

31.

Rasbash

Steel

Browne

Prosser

(2004). A user’s guide to MLwiN: Version 2.0. London: Centre for Multilevel Modelling.

32.

Schwarzer

Leppin

(1991). Social support and health: A theoretical and empirical overview. Journal of Social and Personal Relationships, 8, 99-127. doi:10.1177/0265407591081005

33.

Shaw

B. A.

Krause

Liang

Bennett

(2007). Tracking changes in social relations throughout late life. Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 62(2), 90-99. Retrieved from http://psychsocgerontology.oxfordjournals.org/content/62/2/S90.short

34.

Slawinski

E. B.

Hartel

D. M.

Kline

D. W.

(1993). Self-reported hearing problems in daily life throughout adulthood. Psychology and Aging, 8, 552-561. doi:10.1037/0882-7974.8.4.552

35.

Stevens

N. L.

Van Tilburg

T. G.

(2011). Cohort differences in having and retaining friends in personal networks in later life. Journal of Social and Personal Relationships, 28, 24-43. doi:10.1177/0265407510386191

36.

Stoller

E. P.

Pugliesi

K. L.

(1991). Size and effectiveness of informal helping networks: A panel study of older people in the community. Journal of Health and Social Behavior, 32, 180-191. https://www-jstor-org.web.bisu.edu.cn/stable/2137151

37.

Strawbridge

W. J.

Wallhagen

M. I.

Shema

S. J.

Kaplan

G. A.

(2000). Negative consequences of hearing impairment in old age: A longitudinal analysis. The Gerontologist, 40, 320-326. doi:10.1093/geront/40.3.320

38.

Van Tilburg

T. G.

(1998). Losing and gaining in old age: Changes in personal network size and social support in a four-year longitudinal study. Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 53, 313-323. doi:10.1093/geronb/53B.6.S313

39.

Van Tilburg

T. G.

Broese van Groenou

M. I.

(2002). Network and health changes among older Dutch adults. Journal of Social Issues, 58, 697-713. doi:10.1111/1540-4560.00041

40.

Victor

Scambler

Bond

Bowling

(2000). Being alone in later life: Loneliness, social isolation and living alone. Reviews in Clinical Gerontology, 10, 407-417. Retrieved from http://journals.cambridge.org/action/displayAbstract?fromPage=online&aid=66727

41.

Wang

S.-W.

Boerner

(2008). Staying connected: Re-establishing social relationships following vision loss. Clinical Rehabilitation, 22, 816-824. doi:10.1177/0269215508091435

42.

Wenger

G. C.

Davies

Shahtahmasebi

Scott

(1996). Social isolation and loneliness in old age: Review and model refinement. Ageing & Society, 16, 333-358. doi:10.1017/S0144686X00003457

43.

Zettel

L. A.

Rook

K. S.

(2004). Substitution and compensation in the social networks of older widowed women. Psychology and Aging, 19, 433-443. doi:10.1037/0882-7974.19.3.433