| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Rush Institute for Healthy Aging, Rush–Presbyterian–St. Luke's Medical Center, Chicago, Illinois
b Departments of Psychiatry and Behavioral Sciences, Duke University, Durham, North Carolina
c Departments of Sociology, Duke University, Durham, North Carolina
d Center for the Study of Aging and Human Development, Duke University Medical Center, Durham, North Carolina
e Department of Epidemiology and The Center on Aging and Health, Johns Hopkins School of Hygiene and Public Health, Baltimore, Maryland
f Rush Alzheimer's Disease Center, Rush–Presbyterian–St. Luke's Medical Center, Chicago, Illinois
Carlos F. Mendes de Leon, Rush Institute for Healthy Aging, 1645 W. Jackson Boulevard, Suite 675, Chicago, IL 60612. Email: cmendes@rush.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. Data are from a population-based sample of 4,136 African Americans and Whites aged 
65 living in North Carolina. Disability data were collected during seven consecutive yearly interviews and summarized in two outcome measures. Measures of social relationships included five measures representing network size, extent of social interaction, and specific type of relationships, as well as instrumental and emotional support. Weighted proportional odds models were fitted to model disability as a function of baseline social network and support variables, and the interaction of each variable with follow-up time.
Results. Network size and social interaction showed significant negative associations with disability risks, which did not vary by race, or as a function of time. Social interaction with friends was associated with a reduced risk for disability, but social interaction with children or relatives was not related to disability. Instrumental support was associated with a significantly increased disability risk, with a greater adverse effect among Whites than African Americans. Emotional support was not associated with disability, but a protective effect for ADL disability was found after controlling for its intercorrelation with instrumental support.
Discussion. The findings provide further evidence for the role of social relationships in the disablement process, although not all types of social relationships may be equally beneficial. Furthermore, these associations may be more complex than simple causal effects. There were few racial differences in the association of social relationships with disability, with the possible exception of instrumental support, which may allude to possible sociocultural differences in the experience of instrumental support exchanges.
AGE-RELATED disability is generally thought of as a common endpoint, through impairment in physical, cognitive, or sensory domains, of the chronic diseases and conditions that affect aging persons (
Fried and Guralnik 1997
). According to prevailing models of the disablement process, disability results when these diseases and conditions, via specific impairments and functional limitations, lead to limitations in the ability to perform basic social roles (
Pope and Tarlov 1991;
Verbrugge and Jette 1994). Hence, disability is usually defined as the degree of difficulty or inability to independently perform basic activities of daily living (ADLs) or other tasks essential for independent living. It is generally recognized, however, that disability is not merely a function of underlying pathology and impairment, but involves an adaptative process, which is subject to a host of individual (psychosocial) and ecologic (environmental) factors (
Glass 1998;
Verbrugge and Jette 1994). The present study focused on one such extraneous factor, social relationships, and considered the degree to which race, as a marker of the broader sociocultural context, influences its association with disability, using longitudinal data from a population-based study of older African Americans and non-Hispanic Whites.
In spite of the growing interest in the presumed beneficial health effects of social relationships in general, and social support in particular, much remains unclear about how these factors influence the disablement process, perhaps partly due to the complexity of the process itself. The course of disability may be decomposed into two distinct, but interlinked, components. The first of these involves the effect of slowly progressing chronic disease processes that gradually deteriorate a person's functional abilities over time. The second component involves a more episodic process, consisting of acute periods of sudden deterioration in function due to acute clinical events, which may be followed by complete or partial recovery (e.g.,
Ferrucci, et al. 1996). The present study focused on the gradual aspect of the disability process, (i.e., as long-term change in disability).
In formulating specific hypotheses about the role of social relationships in disability, we will, in accordance with others (e.g.,
Antonucci and Akiyama 1987;
Lynch 1998;
Steinbach 1992), distinguish here broadly between two sets of components of this multidimensional construct. The term structural will be used primarily to refer to characteristics of the social network itself, such as the number of ties (size) and frequency of contact, while reserving the term functional for actual exchange of resources between network members, or social support. Thus, we do not employ social support as a generic term for all social interactions, but only use it to refer to the functional aspects of social relationships.
Some evidence is emerging that social relationships, when defined in terms of their structural components, provide a protective effect against long-term change in disability (
Boult, Kane, Louis, Boult, and McCaffrey 1994;
Kaplan, Strawbridge, Camacho, and Cohen 1993;
Mendes de Leon et al. 1999;
Strawbridge, Cohen, Shema, and Kaplan 1996;
Unger, MacAvay, Bruce, Berkman, and Seeman 1999). For example,
Boult and colleagues 1994 reported that frequent contact with friends and colleagues was associated with a reduced risk of developing disability 4 years later in the Longitudinal Study of Aging (LSOA). In the Alameda County Study, having five or more personal contacts was found to be predictive of "successful aging," defined as the absence of disability (
Strawbridge et al. 1996). However, the evidence is far from conclusive, as the MacArthur studies have generally failed to find a protective effect against disability for social relationships (
Seeman et al. 1995;
Seeman, Bruce, and McAvay 1996). One reason for the inconsistent results may be that most studies have been limited to single-state analyses, that is, based on two waves of data only, which tend to have less precision due to the inability to distinguish between within-subject sources and between-subject sources of variability. Another reason may be the failure of most previous research to discriminate between type of social relationship, that is, defined on the basis of a role relationship (children, friends, relatives). We have reported previously that only ties with relatives and friends are associated with episodic changes in disability status, whereas other types of ties were unrelated to such changes (
Mendes de Leon et al. 1999). The present study serves in part to replicate those findings; in addition, it will test whether overall social networks and interaction or role-specific ties are associated with disability when conceived as a process of gradual change over time, rather than an episodic process involving development of or recovery from disability.
It remains unclear whether functional aspects of social relationships affect long-term change in disability. There is a fair degree of evidence that social support promotes recovery from disability following acute medical conditions (
Glass, Matchar, Belyea, and Feussner 1993;
Mutran, Reitzes, Mossey, and Fernandez 1995;
Oxman and Hull 1997;
Wilcox, Kasl, and Berkman 1994), but reports on its effect on more gradual changes in disability have been mostly negative (
Mendes de Leon et al. 1999;
Seeman et al. 1995,
Seeman et al. 1996). Moreover, the beneficial effect of social support appears to depend on the type of support; although emotional support has been linked with positive recovery outcomes, instrumental support may be less effective in that setting (
Glass and Maddox 1992;
Glass et al. 1993;
Wilcox et al. 1994). In fact, some studies suggest that instrumental support may have a harmful effect on long-term change in disability (
Hays, Saunders, Flint, Kaplan, and Blazer 1997;
Seeman et al. 1996).
The literature reviewed so far suggests that the role of social relationships in the disablement process is complex, and that a better understanding of this role requires specific attention to the various facets that characterize these relationships. However, consideration of specific structural and functional characteristics of social relationships alone may be insufficient to fully comprehend their beneficial, or harmful, effects. Other research (e.g.,
Rushing, Ritter, and Burton 1992;
Umberson 1989) suggests that additional insight into their effect on health may be gained from the broader sociocultural context in which social relationships exist. This context may be understood as the setting that, through prevailing norms, values, and expectations, constructs the meaning and function of social interactions between individuals. An individual's experience or interpretation of social interactions, shaped by these broader social influences, may in turn modify the degree to which these interactions affect health states. For the purpose of this analysis, we will follow the work by
Rushing and colleagues 1992 and use race as a marker of the broader sociocultural context. Specifically, we will examine whether the impact of social relationships on disability is modified by race, that is, whether the effect is different among African Americans than among Whites.
With regard to size of social networks, African Americans are often assumed to have larger networks, particularly with regard to extended family (
Gibson and Jackson 1987;
Taylor, Chatters, Tucker, and Lewis 1990). However, more recent studies suggest that the network and support systems of African Americans are changing, possibly due to the changing family structure in African American communities (
Bengtson, Rosenthal, and Burton 1996;
Burton, et al. 1995;
Silverstein and Waite 1993). African Americans are also believed to maintain a more active flow of supportive resources between network members compared with Whites. Elderly African Americans receive more support from network members, especially adult children, and provide more assistance to others in their network (
Lee, Peek, and Coward 1998;
Mindel, Wright, and Starrett 1986;
Miner 1995;
Mutran 1985). Other research suggests that racial differences in informal assistance are not merely attributable to socioeconomic factors, but reflect real sociocultural differences between racial groups (
Tennstedt and Chang 1998). There is also some evidence that important racial differences exist in the perceived meaning of social interactions, insofar that such interactions, and the exchange of supportive resources, are considered more normative, and possibly more highly valued among African Americans than their White counterparts (
Groger and Kunkel 1995). This is consistent with the notion that African Americans have higher expectations of filial responsibility and attach greater value to providing and receiving informal support than do older Whites (
Lee et al. 1998).
Maintaining active social relationships and involvement within the family and community may foster a sense of fulfilling important social roles and boost feelings of personal control and competency (
Baltes, Wahl, and Schmid-Furstoss 1990;
Krause, Herzog, and Baker 1992). Other research has shown that social role involvement and personal control contribute to more successful adaptation to age-related declines in physical health and are associated with reduced disability (
Kempen, van Sonderen, and Ormel 1999;
Mendes de Leon, Seeman, Baker, Richardson, and Tinetti 1996;
Seeman, Unger, McAvay, and Mendes de Leon 1999). Thus, the extent to which social relationships and exchanges of social support play a more important role in the daily lives of African Americans raises the possibility that these relationships confer a greater protective effect against disability than they do among Whites.
Overall, the present study has two broad sets of objectives: the first is to examine the role of social relationships in long-term change in disability, with particular emphasis on distinguishing between specific types of social ties, and specific types of social support; and second, to explore the role of social relationships in disability within a broader sociocultural context as defined by race. More specifically, we hypothesize that measures that emphasize the structural components of social relationships are associated with a protective effect against long-term changes in disability, but that measures that emphasize the functional aspects (social support) are not. Furthermore, in order to cross-validate our previous findings in another population (
Mendes de Leon et al. 1999), we hypothesize that only ties with relatives and friends are associated with a protective effect against disability, but that ties with children are not. We further hypothesize that social relationships have a stronger protective effect against disability among African Americans compared with Whites.
We note that the present study differs in three important ways from a previous report addressing social support and disability with the same data (
Hays et al. 1997). First, the present study differs from the Hays study by addressing more specific hypotheses with regard to the role of social relationships and social support in disability, in particular by distinguishing between type of relationship and type of social support. Second, in contrast to the Hays study, the present study explicitly focuses on racial differences in the association of social relationships and disability. Third, the present analysis uses seven yearly waves of data on disability status instead of the two waves analyzed in the previous report, which should afford far greater precision in estimating the effect of social relationships on long-term changes in disability.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
). Of all EPESE sites, the North Carolina site obtained the largest proportion of African Americans in its sample. The cohort was derived from a stratified random sample of five counties in the Piedmont area of North Carolina, and was designed to yield a sample that was representative of the noninstitutionalized elderly population. One county was primarily urban, and four mostly rural. Housing units were selected through a three-stage cluster sampling design within each county, and one person was randomly selected from each household containing one or more persons aged 65 years and over. Housing units were stratified by race, and African Americans were oversampled so that they would constitute at least half of the sample. The response rate at baseline was 80%. The cohort included 4,162 men and women, of whom 2,260 were (non-Hispanic) African American (55%) and 1,876 (non-Hispanic) White (45%). The remaining 26 subjects of other ethnic origin were omitted from analysis (total N = 4136). Survey data were collected at baseline in 1986, and each subsequent year until 1992, for a total of seven interviews. The baseline, fourth, and seventh interviews were conducted face-to-face at the participant's home, with the intervening yearly interviews conducted by telephone. All surviving subjects were reinterviewed during follow-up, including those who were admitted to nursing homes. All seven waves of data were used for the present analysis. Vital status was ascertained by monitoring obituary notices in local newspapers, by information from relatives, and by obtaining the death certificates of all deceased subjects. Information on vital status was virtually complete (> 99%); 1,281 of the 4,136 (31.0%) had died by the seventh interview. Attrition due to loss to follow-up was 4.4 % (125 out of 2,862; 2.9% among African Americans and 6.1% among Whites).
Measures
Disability status.
Disability is typically defined in terms of the degree of difficulty or inability to perform basic activities of daily living (ADLs), or other tasks essential for independent living, without assistance. For the purpose of this study, we used two common measures of disability that represent relatively distinct aspects of the disablement process. The first measure is derived from the work of Katz (
Branch, Katz, Kniepmann, and Papsidero 1984), and focuses on basic ADLs, including bathing, dressing, eating, using the toilet, walking across a small room, and transferring from bed to chair. Consistent with previous studies (
Beckett et al. 1996;
Manton, Corder, and Stallard 1993;
Mendes de Leon et al. 1999), task-specific disability is defined as the inability to perform a task independently, without help from a person or special equipment or device. A summary score is constructed by summing across items, yielding a range from 0–6. The second measure, based on the work of
Rosow and Breslau 1966, focuses on tasks requiring mobility and strength. Decline in these basic physical functions is generally thought to precede the development of limitations in ADLs (
Jette, Assmann, Rooks, Harris, and Crawford 1998;
Verbrugge and Jette 1994). The Rosow–Breslau measure of disability contains three items that assess the ability to do heavy work around the house, to walk up and down the stairs, and to walk half a mile. Summing across items yields a total score ranging from 0–3. Both measures have shown adequate levels of short-term stability (
Smith et al. 1990;
Tager, Swanson, and Satariano 1998).
Social network and support variables.
Questions on the structural and functional aspects of social relationships were derived from the Duke Social Support Index (DSSI), developed for use in the Epidemiologic Catchment Area Program (see
George, Blazer, Hughes, and Fowler 1989). These items were designed to provide a reasonable representation of the social contacts older people have, and the types of help they commonly receive, without the assumption that they refer to a common "underlying" latent construct. Consequently, there are no specific assumptions about the intercorrelation between the items (
DeVellis 1991). From these data, two overall variables representing the structural dimension of social networks were constructed. The first variable emphasizes the size of the total social network (Network Size), and is assessed by the number of family members you have (parents, spouse, siblings, children), and the number of friends and other relatives you feel close to. The second variable (Social Interaction) focuses on active social participation, and is based on a count of the number of children, relatives, and friends seen once a month (maximum of 10 for each category); the number of different phone contacts per week; and membership in clubs or organizations. Three additional variables were constructed representing specific types of relationships, based on the role relationship to the respondent (hereafter referred to as "role-specific" relationships). These variables included the Children Network, Relatives Network, and Friends Network, based on the number of children, relatives, and friends seen per month. Because these three variables had very skewed distributions, they were recoded into approximate tertiles, with the lowest tertile serving as the referent level. We will refer to these five measures that pertain to the structural aspects of social relationships as social network variables.
The functional aspects of social relationships were measured by a series of questions on whether family or friends had provided help for specific needs during the last year. Based on the content of the questions (i.e., face validity), we grouped the items into ad hoc measures of instrumental support and emotional support. Seven of the items assessed assistance with specific tasks (e.g., prepare or provide meals, fix things around the house, provide transportation), which were grouped to form a measure of instrumental support, while three items (listen to your problems, give advice about life's problems, provide companionship) were combined to form a measure of emotional support. Summary scores were created by adding individual items, yielding total scores of 0–7 for instrumental support, and 0–3 for emotional support. As described above, there were no specific assumptions about the inter-item correlations for each measure, and internal reliability coefficients were relatively modest, being alpha = .72 for instrumental support and alpha = .62 for emotional support. The intercorrelation between the two support measures was r = .41. Because of the skewed distribution of the emotional support variable, a dummy variable was created differentiating those with low support (scores 0–1, 15% of the sample) from high support (scores 2–3, 85% of the sample).
Control variables.
Social relationships may be related to disability because of their association with other factors known to be involved in the disability process. To reduce the possibility of spurious associations, we selected a set of control variables for the analysis guided by previous research in this area (e.g.,
Boult et al. 1994;
Hays et al. 1997;
Mendes de Leon et al. 1999;
Seeman et al. 1996), taking into consideration factors that have been shown to be empirically related to both social relationships and disability. We considered two broad sets of variables—sociodemographic characteristics and health-related variables—and a series of preliminary analyses was done to identify a parsimonious set of variables that were most likely to confound the association of social relationships with disability. Sociodemographic variables considered in this analysis included age (coded in single years), sex, marital status (married vs nonmarried), education (years of schooling completed), and income. Total household income was ascertained by asking respondents to select one of 11 categories, reflecting either yearly or monthly income. As is typical in survey research, income was subject to frequent and nonrandom missing data (18.9%, unweighted). In order to reduce their influence on data analysis and inference, the Duke EPESE has constructed an imputation method for missing income data (
Landerman, Land, and Pieper 1997). After imputation, the income variable was recoded into three categories, representing income levels of < $4,500/year; $4,500–$8,500/year, and > $8,500/year (referent category). Race was ascertained through self-report. Health-related variables included relative weight, overall physical health, and cognitive function. Relative weight was measured by body mass index (BMI, height/weight2). Physical health was assessed using information on five self-reported chronic conditions and their symptoms: myocardial infarction, hypertension, diabetes mellitus, stroke, and cancer. Thirty-six physicians rated the medical impact of each of the conditions and symptoms, after which the ratings were averaged for each condition. A summary measure of overall physical health status was derived by adding the average rating for each condition that was reported (see
Fillenbaum, et al. 1993;
Fillenbaum, Leiss, Pieper, and Cohen 1998). Cognitive function was assessed by the 10-item Short Portable Mental Status Questionnaire (
Pfeiffer 1975), with poor cognitive function defined as > 3 incorrect responses (
Fillenbaum 1980). Of these variables, marital status and BMI had negligible effects on the association of social network and support variables with disability outcomes, and were not included in the final set of control variables in multivariate models. We also reran all models after excluding all respondents with poor cognitive function (14%), which yielded essentially the same results (data not shown).
Statistical Analysis
The first part of the analysis consisted of comparing racial differences in social network and support variables, using t tests and contingency table (chi-square) analysis. Generalized Linear Modeling (GLM) was used to estimate the effect of each social network and support variable on disability. GLM models with a logistic link function for ordinal response data, rather than a linear link function, were fitted, due to skewed distribution of the disability outcome data. The Rosow–Breslau measure was modeled in its original scale (0–3), but the Katz variable was recoded to three levels (0, 1 or 2, or 3 limitations), as the number of subjects at the more impaired end of the scale was very small, leading to sparse or empty cells, with subsequent problems in estimation. The overall analytic approach was aimed at modeling disability as a function of baseline social variables, using disability outcome data from all seven waves, including baseline, and regardless of initial (baseline) disability level.
Preliminary analysis revealed that the effects of the major predictor variables were reasonably consistent across the range (from one level to the next) of each disability measure, and proportional odds models were considered appropriate for these outcome data. Given the correlated structure of the outcome data within sampling clusters as well as within subjects across repeated measures, regression parameters were estimated using a generalized estimating equations approach (GEE) with a robust variance estimation (
Liang and Zeger 1986). The regression coefficients represent the change in the logit of the probability of having the same or a higher level of disability per unit increase in the predictor variable. These coefficients represent the average effect across all seven years of observation, including baseline. The antilog of these coefficients, in turn, represents the odds of having the same or higher level of disability per unit change in the predictor variable. To provide a sense of the magnitude of the effect, odds ratios for the Social Network and Social Interaction variables were calculated for the interquartile range (75th–25th percentile). We first tested the effect of each social network and support variable by fitting models for each variable with adjustments for age, sex, and race. Consistent with our previous analysis of these data, there was only a small and nonsignificant Age by Race interaction for disability outcomes (
Mendes de Leon et al. 1997). Adjustment for this interaction did not affect the results, and therefore models without this interaction are presented here. Formal tests of racial differences were tested by fitting an interaction term between race and each social variable. However, actual odds ratios were computed by constructing race-specific variables for each network or support measure, and fitting these variables within the same model.
The next step in the analysis consisted of adding the control variables to the multivariate models to estimate the degree to which these variables accounted for the relationship between social variables and disability outcomes. Finally, we tested whether the effect of each social variable varied as a function of (follow-up) time by fitting interaction terms of each variable with time. All analyses were weighted, with weights being computed on the basis of the sampling probabilities and with postsampling adjustment for nonparticipation. The longitudinal models were fitted using the MULTILOG procedure of SUDAAN (
Shah, Barnwell, and Biegler 1996).
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
). Briefly, the average age in the overall cohort at baseline was 73.4 years, and 62.3% were female, with both age and sex distributed approximately equally across race. Blacks had a lower level of education compared with Whites (7.5 vs 10.2 years, p < .001), and were less likely to have an annual income greater than $10,000 (17.5% vs 52.1%, p < .001). As Table 1 shows, there was virtually no difference in terms of overall network size between Blacks and Whites, but Whites reported higher levels of social interaction (14.4 vs 13.2 for Blacks, p < .001). With regard to role-specific relationships, Blacks reported seeing on average more children (2.0 vs 1.6, p < .001) on at least a monthly basis, likely due to the fact that they also had a greater number of living children compared to Whites (4.0 vs 2.7, p < .001). Blacks also had visual contact on at least a monthly basis with more relatives (2.4 vs 2.2, p < .001), whereas Whites had monthly contact with a greater number of friends (4.9 vs 4.0, p < .001). Blacks reported significantly higher levels of instrumental support (5.0 vs 4.6, p < .001), but there was no difference by race in emotional support (2.5. vs 2.5, p = .89).
|
), the excess disability risk among Blacks was reduced to nonsignificant levels after adjustment for education and income for both disability measures.
|
Analysis of the role-specific variables revealed differential patterns of association with disability outcomes (see Table 3 ). Children Network, in particular monthly visual contact with three or more children, was associated with slightly increased disability risks in Model 1, but these risks were no longer significant after controlling for SES and physical health. Relatives Network was not related to disability risks, but Friends Network was inversely related to disability risk. This association attained statistical significance among both Blacks and Whites, and was largely unaffected by adjustment for SES and physical health status.
|
|
Decomposition of the effect of Instrumental Support revealed a significant positive baseline effect as well as a significant positive interaction with Time among Blacks and Whites, and for both disability outcomes (see Table 4 ). This means that subjects with high levels of Instrumental Support not only had higher disability levels at baseline, but also showed larger increases in disability risks over time. Although the baseline effects of Instrumental Support on disability were smaller among Blacks than Whites, the time-varying effects were larger, suggesting that disability risks due to Instrumental Support increased more rapidly among Blacks. However, differences in time-varying effects between Blacks and Whites were not statistically significant.
The additional analysis also revealed a significant positive time-varying effect for Emotional Support, suggesting that this type of support, like Instrumental Support, tends to have an increasingly adverse effect on disability risks over time. Because this finding was contrary to the beneficial effect of emotional support observed in many other studies, we wondered whether this adverse effect was partly due to the degree of co-occurrence (r = .41) between Instrumental Support and Emotional Support. We therefore reestimated the effect of Emotional Support after controlling for Instrumental Support and its interaction with Time (data not shown). This resulted in a markedly reduced and no longer significant Emotional Support x Time interaction terms for both ADL disability (ß = 0.0287; p = .29) and Rosow–Breslau disability (ß = –0.0014; p = .94). Moreover, adjustment for Instrumental Support led to a significant and negative main effect of Emotional Support on ADL disability (ß = –0.3047; p = .02), indicating that, when averaged across all years, persons with high emotional support were at reduced risk for ADL disability relative to those with low emotional support, with an Odds Ratio (OR) = 0.74 (95% confidence intervals: 0.57–0.95). The overall effect of Emotional Support on Rosow–Breslau disability also became negative, that is, protective, but was nonsignificant (ß = –0.1343; p = .17).
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
With regard to our first hypothesis, the results of our primary analysis provide evidence that structural aspects of an older person's social relationships, including overall network size and social interaction, are negatively associated with level of disability. This effect was largely unaffected by differences in education and income, or a summary measure of physical health status. In contrast to our analysis of the New Haven EPESE (
Mendes de Leon et al. 1999), however, the association was not restricted to self-care (ADL) functions, but also evident for the Rosow–Breslau measure of disability. Performance of self-care tasks as specific, goal-oriented behaviors constitutes an adaptive response balancing the disabling effects of physical disease and impairment with psychosocial and environmental conditions that shape those behaviors (
Glass 1998). We have argued previously that ADL disability is more likely to be affected by psychosocial conditions than disability outcomes that focus on more generic physical functions (e.g., mobility), such as the Rosow–Breslau, which may depend more directly upon the physical impairments imposed by underlying disease processes (
Mendes de Leon et al. 1999). However, our findings, as well as those of others (
Unger et al. 1999), leave open the possibility that the influence of psychosocial conditions, such as those exemplified by social relationships extends beyond ADL functions, and may affect the disability process more broadly.
In partial confirmation of our hypothesis regarding specific types of ties, we found that only contact with friends, but not with relatives, appears to confer a protective effect against disability. As we have speculated previously (
Mendes de Leon et al. 1999), social contact with others may sustain a feeling of social embeddedness, providing a sense of meaning and belonging in an older person's life. Such perceptions, in turn, may confer certain health benefits through a variety of psychological and physiological mechanisms (
House, Landis, and Umberson 1988;
Glass, Mendes de Leon, Marottoli, and Berkman 1999;
Seeman, Berkman, Blazer, and Rowe 1994). Although contact with children may well perform a similar function (that is, provide a sense of belonging) older adults are more likely to turn to their children when they start needing assistance with daily tasks, or anticipate this need in the near future, due to declining health (
Stoller and Pugliesi 1988). Indeed, our analysis suggests that receiving assistance with common daily tasks, or instrumental support, increases disability risk. The fact that most of the adverse effect disappeared after adjustment for physical health further corroborates this explanation. It is also consistent with socioemotional selectivity theory, which posits that older people who sense having arrived at the final years of life increasingly shift the focus of their social relationships to members of their nuclear family because they may be able to provide them with more immediate benefits (
Carstensen 1995).
Additional analysis revealed that the effect of social relationships on disability was relatively constant over time, hence raising doubts about the causal order of this association. In other words, our findings, while clearly suggesting a robust relationship, do not necessarily imply that smaller social networks, or lower levels of social interaction, are also associated with new or worsening disability. It is possible that the lack of a clearer longitudinal effect, which would have permitted stronger causal inferences, is due to selection bias. This may have occurred if the failure to participate in follow-up interviews was associated with both greater decline in function and reporting lower social network and interaction levels at baseline. However, given the uniformly high retention rates (98% for telephone interviews, and 95% for face-to-face interviews) during follow-up, this bias, if present, is likely to be small. A more likely explanation may be that the role of social relationships in the disablement process is more complex than a "simple" causal effect. As suggested previously (
Verbrugge, Reoma, and Gruber-Baldini 1994), social contacts and participation may have reciprocal effects on disability, which may well produce a pattern of an apparent constant difference over time. For example, participation in social relationships may serve as a motivation to preserve high function and prevent disability, which, in turn, enables continued social engagement (
House et al. 1988). The Duke EPESE study lacked the yearly ascertainment of social relationships to investigate such a reciprocal causation in more detail, and this interpretation requires validation in future studies.
The likely complex nature of this relationship may also explain some of the inconsistencies in previous research, with some studies reporting a protective effect for larger social networks (
Boult et al. 1994;
Kaplan et al. 1993;
Mendes de Leon et al. 1999;
Strawbridge et al. 1996), but others failing to detect such an effect (e.g.,
Harris, Kovar, Suzman, Kleinman, and Feldman 1989;
Seeman et al. 1995). As noted above, most studies have employed statistical methods that are relatively inefficient to detect complex relationships (e.g., by relying on only two waves of data to estimate change over time). Other sources of imprecision involve the use of relatively crude disability categories—for example, by restricting analyses to "at-risk" individuals (including only the "nondisabled" in an analysis) or using dichotomous outcomes (no disability vs any disability). Such analyses risk substantial misspecification of the actual relationship and mask residual heterogeneity within crude disability categories, both of which may result in spurious (causal) associations. We recognize that residual heterogeneity may have affected the presented findings as well, given the relatively small number of individual disability categories in each measure.
Consistent with our hypothesis, functional aspects of social relationships were not associated with a protective effect against gradual changes in disability. In fact, our findings indicate that one type of support, instrumental support, appears to have a clear adverse effect on disability status, as persons with more instrumental support showed a more rapid increase in disability risk over time. An indirect explanation for this finding is that our measure of instrumental support captures not only availability, but also need for support, arising from declining health and physical capacity. In other words, receiving instrumental support may partly reflect a state of "preclinical" disability (
Fried, Herdman, Kuhn, Rubin, and Turano 1991), which is thought to precede the development of overt disability in mobility and self-care tasks. In fact, most of the instrumental support items (5 out of 7) assess support with tasks that are very similar to those commonly used in measures of instrumental activities of daily living (IADLs), with the distinction that IADL measures typically focus on need for help, rather than availability of support. IADL disability itself has been found to be strongly predictive of developing more severe disability (
Anderson, James, Miller, Worley, and Longino 1998).
However, instrumental support may also be more directly involved in increasing disability risk by a process of deconditioning (
Siegler, Stineman, and Maislin 1994); that is, receiving assistance with common daily tasks may gradually weaken an older person's ability to perform these tasks and related physical functions by him- or herself. Reliance on others may also foster a sense of dependency and undermine an older person's confidence or self-efficacy to perform these daily tasks. Self-efficacy beliefs have been shown to be an important mechanism in forestalling the disabling consequences of declining physical health (
Kempen et al. 1999;
Mendes de Leon et al. 1996;
Seeman et al. 1999). Somewhat paradoxically, dependence on others may be reinforced by having close social ties, as significant others, based on a sense of obligation, may feel a need to provide care and assistance when the health of an older person is starting to fail.
Although emotional support was associated with increased disability risks, this apparent adverse effect was found to be due to the substantial degree of co-occurrence between emotional support and instrumental support. After statistically accounting for this co-occurrence, emotional support showed an overall beneficial effect on ADL disability. This indicates that emotional support may not only promote recovery from disability in the specific context of acute medical illnesses (
Glass et al. 1993;
Mutran et al. 1995;
Oxman and Hull 1997;
Wilcox et al. 1994), but also may have a more general protective effect on the disablement process. These findings may have particular relevance to caregivers, as they imply that care that emphasizes instrumental assistance at the expense of emotional support may be detrimental to progression of disability. On the other hand, care that emphasizes empathy and encouragement (i.e., emotional support), may enable an older adult to mitigate the disabling consequences of age-related chronic conditions, perhaps in part by fostering a sense of self-efficacy (
Mendes de Leon et al. 1996).
While cautioning that we only examined crude (i.e., unadjusted) population levels, our data show few substantive differences in overall size of social networks between African Americans and Whites, which is consistent with other reports (
Bengtson et al. 1996;
Burton et al. 1995;
Silverstein and Waite 1993). African Americans had slightly lower levels of social participation, due, in part, to less frequent membership in social organizations, which
Miner and Tolnay 1998 have attributed to historical racial barriers in such organizations. In accordance with other studies (
Burton et al. 1995;
George 1988;
Lee et al. 1998;
White-Means 1993), we also found that African Americans in this population report more extensive ties with children and extended family and receive more social support, especially in the form of assistance with common daily tasks, although absolute differences with Whites were fairly small.
We found little support for our second hypothesis concerning racial differences in the protective effect of social relationships against disability. This may indicate that the mechanisms by which social relationships are related to disability are not dependent on characteristics of the sociocultural context in which these relationships exist. Another explanation is that the sociocultural context is important, but that race as a marker of this context is not sufficiently sensitive to detect such influences. For example, differences in the meaning and function of social interaction may vary at least as much within each racial group as they do across race. The only exception to this pattern was the differential race effect for instrumental support, which may be suggestive of the influence of broader sociocultural factors. Although clearly in need of replication, a possible interpretation of this finding could be that the exchange of instrumental support, both giving and receiving, is more normative in Black families and communities, and that accepting such support may not represent as much a specific need for assistance as it might in White families (
Lee et al. 1998). Older African Americans may also be more accustomed to giving and receiving instrumental support throughout the life course (
Lee et al. 1998;
Miner 1995;
Taylor and Chatters 1986), and accepting support from others may not be perceived as much of a threat to autonomy as among Whites. In addition, older African Americans may have more opportunities to provide support to others in old age, and experience less of an imbalance between giving and receiving support (
Groger and Kunkel 1995). Thus, receiving instrumental support may be less directly linked to specific care needs that arise from declining health among elderly African Americans, and indicate a greater sense of social engagement; this might explain part of the less adverse effect on disability in African Americans compared with Whites.
In conclusion, although our findings provide further evidence for the importance of social relationships in long-term changes in age-related disability, the nature of this association is likely to be complex and may evolve along a path of reciprocal causation. The findings also suggest considerable diversity in how specific social relationships affect the disability process, and support the notion that not all social relationships provide unambiguous health protection. Race does not appear to substantially modify the role of social relationships in disability. However, the adverse impact of instrumental support on disability was found to be weaker among African Americans than Whites. It is possible that this is related to sociocultural differences in the role and interpretation of this type of support among groups defined by race or ethnicity.
|
Acknowledgments |
|---|
Received for publication March 30, 2000. Accepted for publication December 11, 2000.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  T. M. Gadalla Sense of Mastery, Social Support, and Health in Elderly Canadians J Aging Health, June 1, 2009; 21(4): 581 - 595. [Abstract] [PDF]
|
|
|
|
|
|
B. Cornwell Network Bridging Potential in Later Life: Life-Course Experiences and Social Network Position J Aging Health, February 1, 2009; 21(1): 129 - 154. [Abstract] [PDF]
|
|
|
|
 |
|
 D. K. Miller, F. D. Wolinsky, E. M. Andresen, T. K. Malmstrom, and J. P. Miller Adverse Outcomes and Correlates of Change in the Short Physical Performance Battery Over 36 Months in the African American Health Project J. Gerontol. A Biol. Sci. Med. Sci., May 1, 2008; 63(5): 487 - 494. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Mavandadi, K. S. Rook, and J. T. Newsom Positive and Negative Social Exchanges and Disability in Later Life: An Investigation of Trajectories of Change J. Gerontol. B. Psychol. Sci. Soc. Sci., November 1, 2007; 62(6): S361 - S370. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. V. Ostir, Y.-F. Kuo, I. M. Berges, K. S. Markides, and K. J. Ottenbacher Measures of Lower Body Function and Risk of Mortality over 7 Years of Follow-up Am. J. Epidemiol., September 1, 2007; 166(5): 599 - 605. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. R. Krueger, R. S. Wilson, R. C. Shah, Y. Tang, and D. A. Bennett Personality and incident disability in older persons. Age Ageing, July 1, 2006; 35(4): 428 - 433. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. W. Li Predictors of ADL Disability Trajectories Among Low-Income Frail Elders in the Community Research on Aging, November 1, 2005; 27(6): 615 - 642. [Abstract] [PDF]
|
|
|
|
 |
|
 C. F Mendes de Leon Why do friendships matter for survival? J Epidemiol Community Health, July 1, 2005; 59(7): 538 - 539. [Full Text] [PDF]
|
|
|
|
 |
|
 S. A. Everson-Rose, K. A. Skarupski, J. L. Bienias, R. S. Wilson, D. A. Evans, and C. F. M. de Leon Do Depressive Symptoms Predict Declines in Physical Performance in an Elderly, Biracial Population? Psychosom Med, July 1, 2005; 67(4): 609 - 615. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. L. Barnes, C. F. Mendes de Leon, R. S. Wilson, J. L. Bienias, and D. A. Evans Social resources and cognitive decline in a population of older African Americans and whites Neurology, December 28, 2004; 63(12): 2322 - 2326. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. R. Peng, M. Navaie-Waliser, and P. H. Feldman Social Support, Home Health Service Use, and Outcomes Among Four Racial-Ethnic Groups Gerontologist, August 1, 2003; 43(4): 503 - 513. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. F. Mendes de Leon, T. A. Glass, and L. F. Berkman Social Engagement and Disability in a Community Population of Older Adults: The New Haven EPESE Am. J. Epidemiol., April 1, 2003; 157(7): 633 - 642. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Jang, W. E. Haley, B. J. Small, and J. A. Mortimer The Role of Mastery and Social Resources in the Associations Between Disability and Depression in Later Life Gerontologist, December 1, 2002; 42(6): 807 - 813. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Bukov, I. Maas, and T. Lampert Social Participation in Very Old Age: Cross-Sectional and Longitudinal Findings From BASE J. Gerontol. B. Psychol. Sci. Soc. Sci., November 1, 2002; 57(6): P510 - 517. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. F. M. de Leon, J. M. Guralnik, and K. Bandeen-Roche Short-Term Change in Physical Function and Disability: The Women's Health and Aging Study J. Gerontol. B. Psychol. Sci. Soc. Sci., November 1, 2002; 57(6): S355 - 365. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
