U.S. Department of Health and Human Services
N I H Logo
Cognitive and Emotional Health Project: The Healthy Brain

Demographic and Social Factors

III. Social Connectednes

B. Measures

o Structural network characteristics, social integration, social engagement, perceived support, & received support represent different constructs, are at best moderately correlated, & may be related to health through different mechanisms (Cohen & Wills 1985; Heller & Lakey 1985; Barrera 1986; Cohen 1988; Dunkel-Schetter & Bennett 1990; Sarason, Sarason & Pierce 1990).

o The following outline relies heavily on several excellent reviews, particularly those of Brissette & Cohen (2000) for social integration/network measures, Wills & Shinar (2000) for social support measures, & Lochner et al. (1999) for social capital measures. Orth-Gomer & Unden (1987) also critically review 17 instruments used to measure social integration or support in population surveys; for each instrument, detailed information is presented on content, number of items, length of time to complete, applicability (e.g., suitability of wording) in general populations, psychometric properties, generalizability, & predictive capacity. Heitzmann & Kaplan's (1988) review of 23 support measures is also quite comprehensive; in addition to brief descriptions of each measure, information on the samples in which measures were developed, reliability, criterion validity (another support measure is used as the criterion), & predictive validity (correlation between support measure & self- or clinician-rated health, life events, biologic measures) is provided. Other useful references include McDowell & Newell (1996) and Payne & Jones (1987). Several existing reviews focus on instruments for use in specific populations - e.g., Oxman et al. (1992) critique instruments with an eye towards their use in the elderly. Wills & Shinar (2000) catalogue measurement approaches used to assess support in specific patient groups (e.g., arthritis, cancer, CVD, diabetes, HIV/AIDS, kidney disease, multiple sclerosis).

1. Social network measures

Traditional network measures

o Network size has been utilized in a number of studies involving health outcomes (Gallo 1982; Haines & Hurlburt 1992; Seeman, Berkman, Blazer, Rowe 1994). It is a relatively weak predictor of health (Cohen & Wills, 1985; Uchino, Cacioppio & Kiecolt-Glaser, 1996) & probably not responsible for effects of social integration (Cohen et al., 1997). However, researchers traditionally assess a narrow range of network ties & predict health outcomes from differences in networks containing between 5-20 people. Few investigators have assessed extended network ties as they relate to health outcomes; thus, we know little about how properties of people's broad social networks affect health (Haines & Hurlburt 1992).

· Network density may be more relevant than network size to understanding how integration operates. High-density networks may be helpful in maintaining one's social identity & promoting flow of support resources from network members. Low-density networks may be valuable during life transitions (e.g. divorce, unemployment, geographic relocation). Few studies test the relationship between fine-grained network measures & health, however.

· Convoy measure, a.k.a. Social Networks in Adult Life Questionnaire (SNAL; Kahn & Antonucci 1980; Antonucci 1985; Antonucci & Akiyama 1987; Israel & Antonucci 1987; Antonucci, Fuhrer, Dartigues 1997). A measure based on affective network membership criterion, developed at the Survey Research Center at the University of Michigan. Respondents are first asked to name people to whom they feel so close that it would be difficult to imagine life without them & then are asked to name people with whom they have relationships that are less close. 3 concentric circles are used to represent degrees of emotional closeness. Using this list, measures of density, frequency of contact, proximity, emotional reciprocity, & tangible reciprocity are averaged across network members. Also, each network member is rated for the frequency of emotional & tangible aid support. Convoy measure is a straightforward technique to determine network size based on affective closeness. An advantage is that it allows for comparison of the relative effectiveness of definitions of ties based on different levels of emotional closeness.

o Antonucci et al. (1997). PAQUID Study. 3777 community-dwelling French elderly. Increased network size (across all 3 concentric circles) was inversely associated with depression, but social support variables accounted for more variance in depressive symptoms than social network variables.

· Social Network List (SNL; Hirsch 1979, 1980) measures network size, number of confidants, number & percentage of relatives in network, network density, number & type of main helpers. One of the few "traditional" measures used in studies examining the relationship between a formal network measure (density) & health (mental health) outcome.

· Hirsch (1980, 1981), Wilcox (1981). For women undergoing major life events (e.g. divorce, widowhood, returning to school), dense networks, esp. networks in which friends have relationships with members of nuclear family, were associated with poor adjustment on self-ratings of symptoms, mood, & self-esteem. Low-density networks, involving more relationships outside family, presumably allowed the women to develop new social roles appropriate to their changed statuses.

2. Social integration meausres

a. Role-based measures assess number of recognized social positions or social identities.

· Thoits' role measure (1983; 1986). An index assessing participation in 8 social roles. Respondents are given 1 point for each role (spouse, parent, worker, student, group member, church member, neighbor, friend) in which they actively participate, for a total of up to 8 points. More roles were associated with less psychological distress in cross-sectional analyses. Thoits (1995) also developed a role-based measure that includes 7 additional roles (lover, son/daughter, son/daughter-in-law, relative, hobbyist, athlete, stepparent); respondents are asked to select up to 3 most important role-identities. No information is provided as to whether expanded instrument is better at predicting mental health outcomes than original instrument.

· Cohen's Social Network Index (SNI; 1991; 1997). Assesses participation in 12 types of social relationships (relationships with spouse, parents, parents-in-law, children, other close relatives, close neighbors, friends, workmates, schoolmates, fellow volunteers, members of groups without religious affiliations, members of religious groups). Respondents are defined as participating in a relationship if they report talking to a person at least once every 2 weeks. One point is assigned for each type of relationship in which a person participates, for a total of 12 points. The total number of persons with whom a respondent speaks to provides an estimate of network size.

o Bolger & Eckenrode (1990). SNI was associated with positive mental health outcomes (reduced anxiety) in response to a stressful exam.

o Cohen et al. (1991)o In cross-sectional analysis, SNI was associated with increased positive affect, self-esteem, personal control, less smoking & drinking, better diet, sleep, & more exercise

o Cohen et al. (1997). Experimental study tested the hypothesis that diversity of network ties is related to susceptibility to the common cold. Subjects were given nasal drops containing one of two rhinoviruses & monitored for the development of a common cold. Those reporting more types of social ties on the SNI were less susceptible to common colds, produced less mucus, fought infection more efficiently, & shed less virus even after controlling for pre-challenge virus-specific antibody, virus type, age, sex, season, body mass index, education, & race. Moreover, susceptibility to infection decreased in a dose-response manner with increased diversity of social network.

b. Social participation measures assess extent or frequency of social activities. This approach allows investigators to compare the importance of different categories of activity.

· Welin Activity Scale (WAS; Welin et al. 1992). Asks respondents how often they engaged in certain activities in the past year. 3 categories of activity: (1) Social activities (8 items); (2) home activities (10 items); (3) outside home activities (14 items). Response options: 0=never; 1=occasionally; 2=often/regularly. Responses in each category are summed to create 3 distinct activity scores. Subscale scores are moderately correlated with each other (0.44-0.53) but reflect distinctive constructs.

· Welin et al. (1992). Increased social activity predicted less CV mortality in a cohort of 989 Swedish men followed for 9 yr.

· Social Participation Scale (SPS; House et al. 1982). Derived from questionnaire used in the population-based Tecumseh Community Health Study. Participation in 4 social activity categories: (1) intimate social relationships (e.g., marital status, visits with friends & family), (2) formal organizational involvements outside of work (e.g., going to church meetings), (3) active & relatively social leisure (e.g., going to movies, fairs, & museums), (4) passive & relatively solitary leisure (e.g., watching TV, reading). Parallel items tap respondents' satisfaction with activities. Items from the last 3 subscales require respondents to estimate the frequency with which they engaged in various types of activities in past yr. Items from first 3 subscales can be summed to create a cumulative participation index.

· House et al. (1982). Tecumseh Community Health Study. 2754 men & women aged 35-69 yrs followed for 9-12 yr. Men reporting higher levels of activities on the cumulative index had lower mortality. Women reporting higher levels of activities on the cumulative index also had lower mortality, but this effect did not persist when other risk factors were controlled. There were no reliable associations between satisfaction with activities & mortality among either gender, suggesting that weighting based on respondents' ratings of satisfaction may not be useful. It is not reported whether the 4 categories were individually associated with mortality.

c. Perceived integration measures assess individuals' own view of their communality (see also psychological sense of community, described under social capital)

· Heidrich & Ryff (1993a, 1993b). Set of scales reflecting 3 psychological dimensions of integration in the elderly: (1) presence of normative guidelines, (2) possession of meaningful roles (10 items), (3) presence of appropriate reference groups (8 items). "Roles" subscale taps extent to which individuals perceive themselves as holding important, satisfying, & meaningful roles (e.g., "There are a lot of important things for people to do after they retire"). "Reference groups" subscale assesses belongingness - i.e., whether individuals view themselves as members of social group, sharing values & attitudes with other elders (e.g., "I feel a sense of shared values with other senior citizens in this country.") Participants rate extent to which they agree with each statement on a scale ranging from 1-6. Higher scores indicate higher degree of integration. Subscales are scored separately.

· Heidrich & Ryff (1993). In cross-sectional analysis, subscales were associated with decreased psychological distress & increased life satisfaction.

· Malmo Influence, Contact and Anchorage Measure (MICAM; Hanson et al. 1997). Social anchorage (5 items) describes to what degree the person belongs to & is anchored within formal & informal groups, &, in a more qualitative sense, the feeling of membership in these groups. (e.g. "Are you rooted & feel a strong affinity to your residential area?") Social participation (13 items) describes how actively the person takes part in the activities of formal & informal groups in society. [Two types of social support, instrumental (1 item) & emotional (3 items), are also measured. (See 'Support Measures' below).] Psychometric properties were investigated in a cohort of 12,009 45-65-yr-old residents of Malmo, Sweden & discussed extensively by Hanson et al. (1997).

· Hanson et al. (1988). Correlational study of social anchorage, contact frequency, & social participation in relation to CV function. Controlling for social class, marital status & health habits, social anchorage predicted lower blood pressure in sample of 500 men born in 1914.

· Hanson et al. (1997) summarizes his research linking different aspects of networks & support, as measured by various versions of this scale, to smoking, diet, physical activity, alcohol use, nervous problems, blood pressure, CV morbidity, & all-cause mortality.

d . Complex indicators blend the above approaches.

· Berkman's Social Network Index (SNI; Berkman & Syme 1979). 4 components: (1) marital status, (2) contact with friends & family, (3) church membership, & (4) group membership. The SNI considers both number & relative importance of social ties across the 4 categories & combines this information into a summary measure ranging from 0-4. In Berkman's empirically-based weighting system, an index of intimate contacts (marital status, friends, relatives) is given nearly 4 times the weight as group membership & 2 times the weight of church membership.

· Social Network Questionnaire from New Haven EPESE (Seeman & Berkman 1988; Glass et al. 1997). Developed for NIA's Established Populations for the Epidemiologic Study of the Elderly (EPESE) program. Items originated from Berkman's SNI. Availability & frequency of social contact are assessed. Also included are 7 questions about types & adequacy of social support, but focus is on structural measures. Allows quantification of providers by available number or by frequency of contact (face-to-face or non-visual contacts). Measures can be classified into kin vs. nonkin & by geographic proximity.

· Berkman & Syme (1979), Seeman et al. (1987). Alameda County Study. 6928 adult residents of Alameda County, California. Low SNI scores were associated with greater total mortality in 9- & 17-year follow-ups.

· Schoenbach et al. (1986). Evans County Cardiovascular Epidemiologic Study. 2059 men & women followed for 13 yr. The Evans County index, a slight modification of SNI, predicted mortality, but only among white males.

· Reynolds & Kaplan (1990). Alameda County Study. 17-year follow-up found that integrated women were less likely to die of cancer.

· Seeman et al. (1993). EPESE. 3 cohorts aged 65+ from East Boston MA (n=3809), New Haven CT (n=2812), & 2 rural counties in Iowa (n=3673). Low SNI scores predicted greater 5-year mortality.

· Kawachi et al. (1996). Health Professionals Follow-up Study. 32,624 male health professionals aged 42-77. High SNI scores were associated with lower total mortality by reducing deaths from CVD & accidents/suicides. High SNI scores were associated with reduced incidence of stroke. SNI scores were not associated with incidence of CHD.

· Seeman et al. (1994). MacArthur Studies of Successful Aging. High-functioning men & women aged 70-79. High SNI scores are associated with lower urinary cortisol, norepinephrine, & epinephrine for men, but not women.

· Kaplan et al. (1987). Alameda County Study. Low network participation (a modification of the SNI, based on extent & frequency of contact with friends & relatives) was associated with increased risk of depressive symptoms at 9-yr follow-up.

· Goodenow et al (1990). Evidence regarding the association between SNI & mental health outcomes is inconsistent.

· Mendes de Leon et al. (1999). New Haven EPESE. Total social networks were associated with reduced risk of developing ADL disability over 9-yr follow-up.

o Social Disengagement Index (Bassuk et al. 1999). An extension of Berkman's SNI. 6 indicators: presence of spouse, monthly visual contact with at least 3 relatives or close friends, yearly nonvisual contact (telephone calls or letters) with at least 10 relatives or close friends, frequent attendance (at least once per month) at religious services, membership in other groups, & regular participation in recreational social activites.

· Bassuk et al. (1999). New Haven EPESE. Greater social disengagement was associated with cognitive decline over 12-yr follow-up.

· Social Connections Index (Kaplan et al. 1988). 5 questions concerning extent & frequency of social interaction - i.e., planned visits with friends & relatives, meetings with clubs & societies, number of daily interactions, & marital status. Easy to score. Marital status, a dichotomous variable, is weighted 1 (unmarried) or 6 (married) to equalize its contribution to the summary measure. Other questions have response categories ranging from 1 to 6. Sum across all items to get summary score.

· Kaplan et al. (1988, 1994). In studies in North Karelia & Kuipio, Finland, men with more social connections had lower mortality from all causes, CVD, & ischemic heart disease during 5-year follow-ups, but these relationships were not seen in women.

·Rand Social Health Battery (Donald & Ware 1982). Developed for Rand Health Insurance Experiment & administered to 4603 individuals. "Social contacts" index contains items asking about visits with friends & relatives, & home visits by friends. "Social resources" index contains items asking about attendance at religious services, number of neighborhood acquaintances & number of close friends & relatives. There is also a summary "social well-being" index.. This scale may not be optimal for studying the elderly: it has not previously been used with this age group; it combines close relatives & friends; & it does not separately assess interactions with spouse & children, who are important to health care of the elderly.

·In cross-sectional analyses, social well-being was associated with better mental health regardless of stressful life events, explaining 12% of the variance in mental health. In longitudinal analyses, high levels of social well-being were found to predict subsequent improvements in mental health. Social resources were better predictors of mental health than social contacts

3. Social support measures

Social relationships may provide emotional, instrumental, informational, or appraisal support, as well as companionship & validation. These functions may be differentially useful for various types of problems or stressors ("matching hypothesis"; Cohen & McKay 1984; Cutrona & Russell 1990). Support measures may be most appropriate for studying processes through which social resources contribute to coping with stressors (Wills & Shinar 2000). Supportive functions are not irrelevant for generally non-stressed populations -- many studies show that social support is associated with health at both low & high levels of stress. However, the buffering model suggests that support has greater effects among persons currently facing stressors & challenges.

a. Perceived support measures

    i. Brief unidimensional scales

· 1-3 item measures

· EPESE support measures (Seeman & Berkman, 1988). Availability of confidant: "Is there any one special person you know that you feel very close & intimate with?" Emotional support: "Can you count on anyone to provide you with emotional support - talking over problems and helping you make a difficult decision?" Instrumental support: "Can you count on anyone to help you with daily tasks like grocery shopping, housecleaning, cooking, telephoning, giving you a ride?" The 1st item has a Yes/No response format; the other 2 items are followed by a list of 11 potential persons who could provide support & a count is made of affirmative responses. No psychometric data exists.

· Berkman et al. (1992). Perceived emotional support is associated with increased survival post-MI.

· Seeman et al. (1994). Perceived emotional support is associated with lower urinary cortisol, norepinephrine, & epinephrine for men, but not women. Perceived instrumental support is not associated with these endocrine parameters.

· Williams et al. (1992). Individuals who were married &/or had a confidant had reduced mortality over 9-yr follow-up; however, a composite score was used, so it is unclear whether the confidant item alone or marital status alone predicted mortality.

· Cohen & Wills (1985). 1- to 3-item measures of confidant relationships show buffering relationships in samples with 800-2300 respondents. Such measures may be useful in large-scale epidemiologic research.

· Intimacy Scale (Hobfoll & Leiberman 1987; Hobfoll & Lerman 1988). Items assess perception that important thoughts & feelings can be shared with others & will be accepted; essentially a measure of emotional support. 6-item scale taps emotional support from spouse. Other versions have 10-15 items & tap emotional support from closest friend & from closest family member. Scores predict psychiatric symptoms (Hobfall & Leiberman 1987).

· OARS Social Support Scale (Fillenbaum & Smyer 1981; Blazer 1982). Developed for elderly persons but is also appropriate for general populations. 6-item scale of emotional & instrumental support. The items are: "Lonely, even with people," "Seems like no one understands," "Difficulty speaking to new people," "Someone cares what happens to you," "Enough contact with confidant," "Someone would help if ill or disabled."

· Blazer (1982). Scale predicted 30-month mortality in community sample of 331 participants aged 65+ in Durham, North Carolina.

· Malmo social support measure (Hanson & Ostergren, 1987; Hanson et al. 1997). The 9-item version of this measure contains 4 items on emotional support & 5 items on instrumental/ informational support. Similar to OARS support measure.

· Hanson et al. (1989). Emotional support predicted 5-yr mortality in 621 men.

· LaRocco et al. (1980). 12 items. 3 parallel scales tap support from worksite supervisor, coworkers, & home. Each scale indexes a combination of emotional & instrumental support. Scale scores are not highly correlated.

· All 3 scales have shown buffering effects with respect to work-related stress in sample of 636 male workers from a variety of occupational groups (LaRocco et al. 1980) & sample of 2800 hourly manufacturing workers (House et al. 1979).

· ENRICHD Social Support Instrument (ESSI; ENRICHD Investigators 2000). 5-item scale constructed from other well-validated support scales; contains items that were individually predictive of death in cardiac patients in other studies (Williams et al. 1992; Berkman et al. 1992; Gorkin et al. 1993; & the MOS Social Support Survey). 3 items measure emotional support; 2 items measure instrumental support. Response categories were modified to follow a consistent format, ranging from 1 (none of the time) to 5 (all of the time). Cronbach's alpha=0.86. ESSI correlates positively with ISSB & Perceived Social Support Scale, & negatively with Beck Depression Inventory.

· ENRICHD Investigators (2000). ENRICHD is a large multicenter clinical trial investigating the effect of psychosocial intervention on reinfarction & death in acute MI patients who are depressed or lack social support. A low ESSI score is used to determine eligibility for enrollment. To be classified as a successful completer of the trial, patients enrolled on the basis of low ESSI score must report at least 1 sustainable supportive relationship & adequate social support as measured by the Perceived Social Support Scale.

    ii. Broad-based scales of close and diffuse support measure support from a primary relationship with one close person & also from a broader network of alliances with friends & coworkers (strong vs. weak ties).

· Interview Schedule for Social Interaction (ISSI; Henderson et al. 1980). Designed to delineate how social relationships protect against psychiatric disorder under adversity. The original 52-item interview, which takes 30 minutes to complete, first determines existence of a primary confidant relationship by asking "At present, do you have someone you can share your most private feelings with (confide in) or not?", followed by probe items about whether the relationship provides confiding, comfort, & closeness, & is likely to endure. This score is termed "Attachment." A score for existence of friendship or coworker relationships is based on items asking about persons (aside from confidant) who provide a sense of reliable alliance & reassurance of worth; this score is termed "Integration" (but assesses a different construct than the integration scales discussed above). After items assessing availability of attachment & integration, further items explore perceived adequacy of each type of relationship - i.e., whether the respondent wanted more or less of it. The ISSI is one of the few instruments that assesses unwanted ("too much") support. Scores include availability of attachment (8 items), adequacy of attachment (12 items), availability of integration (16 items), adequacy of integration (17 items). An abbreviated 12-item version suitable for use in large-scale epidemiologic studies was developed by Unden & Orth-Gomer (1987). A drawback of both the original & abbreviated ISSI is that they do not assess specific support functions. The attachment scale appears to index primarily emotional support, but it & the integration scale may also tap other functions such as instrumental support & companionship

. Henderson et al. (1980). 52-item ISSI demonstrated stress-buffering relationships in an Australian community sample of 756 people, with depression as the outcome. Significant effects were found for attachment & integration scales.

· Orth-Gomer, Rosengren et al. (1993). Rosengren, Orth-Gomer, et al. (1993). Gothenburg Study. 736 50-year-old men followed for 7 yr. 12-item ISSI predicted incident CHD & exhibited stress-buffering relationships with all-cause mortality.

· Kristenson et al. (1998). LiVicordia study. Cross-sectional survey comparing 150 50-year-old men in Lithuania & Sweden in order to examine why CHD mortality is 4 times higher in Lithuanian than in Swedish men. Lithuanian men had lower attachment & lower social integration as assessed by 12-item ISSI. (Note: A broad range of psychosocial characteristics was assessed in this study: Lithuanian men also had lower coping, self-esteem, sense of coherence, higher vital exhaustion, & depression than did Swedish men.)

· Knox et al. (1985). Correlational study of social support & CV function (HR, BP) & immune function (resting plasma epinephrine). 56 hypertensive, 27 normotensive, 23 hypotensive men. Lack of attachment (intimate contacts) & low number of regular contacts with acquaintances were associated with high resting plasma epinephrine levels, which in turn were associated with higher resting diastolic pressure.

· Perceived Social Support from Family and Friends (PSS-FA, PSS-FR: Procidano & Heller 1983). 20-item self-report scale on friend support. Scale includes items on availability of closeness & confiding & on getting ideas about how to solve problems; thus, functions tapped are emotional & informational support. There are also two items on social companionship. There is a parallel 20-item scale for family support. Scales are not highly correlated. Developed in college students, but reliability has been demonstrated in chronic psychiatric patients & diabetic outpatients. Instrument predicts a wide range of mental health outcomes, including anxiety, depression, drug use, & suicidal ideation. Advantages include simple Yes/No response pattern. Disadvantages include a focus on adequacy of support providers rather than on support function.

· Social Support Questionnaire (SSQ; Sarason et al. 1983). Based on Bowlby's theory of attachment & Sarason's theory of sense of acceptance. Respondents are asked to list up to 9 potential support providers for each of 27 hypothetical types of support situations -- e.g., "Whom can you really count on to listen when you need to talk?", "Who do you feel really appreciates you as a person?", "Whom can you count on to console you when you are very upset?" Respondents then rate their overall satisfaction with support for each situation (rather than for each provider). The mean of the 27 scores is the support satisfaction score. The SSQ yields a total score for number of supporters (SSQN; alpha=0.97, 4 week test-retest r=0.90) & support satisfaction (SSQS; alpha=0.94, 4 week test-retest r=0.83). SSQ has been used in several methodologic studies involving correlations with other support measures & with personality measures (Sarason et al., 1983; Sarason et al., 1987) & has been employed in studies with a considerable range of samples. As a global measure of perceived support, the SSQ has a reasonable track record. However, it does not distinguish between different types of support. The instrument has not been used in elderly populations; some situations may not be relevant for this age group.

· Malcolm & Janisse (1991). Correlational study of social support & CV function. 64 men aged 30-54. SSQS associated with lower systolic blood pressure. SSQN unrelated to blood pressure.

· McNaughton et al. (1990). Correlational study of social support & immune function. 33 women. Mean age=74. Perceived emotional support negatively related to CD8+ cells; social support unrelated to CD4+ cells.

· Snyder et al. (1990, 1993). 8-week prospective study of social support & immune function. 89 women aged 18-24. Social support was unrelated to IgG antibody levels at preimmunization, 3 weeks postimmunization, or 8 weeks postimmunization. Social support was associated with stronger proliferative response to keyhole limpet hemocyanin (KLH) at baseline but not at 3- or 8-wk assessments.

· Friedmann & Thomas (1995). Cardiac Arrhythmia Suppression Trial (CAST). Greater amount of social support, measured by 6-item SSQ, was associated with reduced mortality in 1-yr follow-up of 369 post-MI patients.

· Work Relationships Index, Family Relationships Index (WRI & FRI; Moos 1981; Moos & Moos 1981). 27-item composite measure is based on sum of 3 subscales that index (1) cohesion, the extent to which a person perceives group members are supportive of each other, (2) expressiveness, the extent to which group members are encouraged to express their feelings directly, & (3) conflict, the extent to which open expression of anger is not characteristic of the group. Item wording is adapted for work or family environment. Measures have been widely used in the psychological literature (e.g. Billings & Moos 1982; Holahan & Moos 1987). It is unclear what supportive functions these scales measure, & despite some relatively strong tests, they typically do not produce buffer effects (Holahan & Moos 1987).

· Multidimensional Scale of Perceived Social Support (MSPSS; Zimet et al. 1988). An early version is known as the Perceived Social Support Scale (PSS; Blumenthal et al. 1987). Measure of perceived adequacy of support. 12 items divide into 3 factor groups (family, friends, significant other); there is also a total support score. Respondents answer each question using a 7-point rating scale ranging from very strongly disagree to very strongly agree. Internal consistency & test-retest reliability have been demonstrated. Focuses on subjective evaluation of providers rather than on support function.

o Blumenthal et al. (1987). Examined relationship between Type A personality, social support, & CAD. Social support, as measured by PSS, buffered the relationship between personality & CAD. Type A individuals with high perceived support had less CAD than Type A individuals with low perceived support. Among Type B individuals, however, there was no association between perceived social support & CAD.

o Oxman et al. (1995). Social support, as measured by MSPSS, was not a predictor of mortality in 232 post-cardiac surgery patients followed for 6 months.

o Irvine et al. (1999). Social support, as measured by MSPSS, was not related to sudden cardiac death (SCD) in 671 post-MI patients followed for 2 yr; social network size is positively assoc with SCD (odd finding); social participation is negatively assoc with SCD.

o ENRICHD Investigators (2000). Social support intervention for acute MI patients with low social support or depression. The MPSSS is used as a primary measure of social support outcome. Results of trial have not yet published.

· Self-Evaluation and Social Support Schedule (SESS; Brown & Harris, 1978; O'Connor & Brown, 1984; Brown et al. 1986; Harris et al. 1999). A structured interview schedule to assess availability of confidants. Contains detailed questions about relationship with primary confidant, including closeness, confiding, intimacy, dependency, & negative interactions. 3-4 hours administration time, extensive interviewer & rater training required. Not appropriate for large-scale epidemiologic studies.

o Brown & Harris (1978). An intimate relationship with a husband or boyfriend protected women from clinical depression following serious life events. Intimate relationships with mother, sister, or friend did not offer the same protection.

   iii. Multidimensional inventories

· Interpersonal Support Evaluation List (ISEL; Cohen & Hoberman, 1983; Cohen et al., 1985). 40-item general population version & 48-item college student version. 40-item version has 10-item subscales for perceived availability of 4 separate functions of social support: (1) 'appraisal' (emotional) support, (2) 'tangible' (instrumental) support, (3) 'belonging' (companionship support), & (4) 'self-esteem' maintenance through social comparisons. Items are counterbalanced for desirability. Subscales are moderately intercorrelated. Widely used in health research, the ISEL has a consistent record for showing stress-buffering effects (Wills, 1991) & predicts recovery from physical illness (King et al., 1993). Suitable for a variety of populations. A limitation is that the ISEL does not include a separate scale for advice & guidance; supplementation of content might be desirable for settings where this function is believed to be important.

o Jackson & Adams-Campbell (1994). Correlational study of social support & CV function. 162 black men, 259 black women (mean age = 19). Appraisal support was associated with lower systolic blood pressure in women.

o Uchino, Cacioppio et al. (1995). Correlational study of social support & CV function. 45 healthy young women (mean age = 19); 20 healthy older women (mean age=67). Appraisal support was associated with lower blood pressure.

o Glaser et al. (1992). Prospective study of social support & immune function. 48 second-year medical students (mean age=23) who received 3 injections of HepB vaccine. Total perceived support was positively related to a summary index of Ab titers & T-lymphocyte response during third inoculation.

o Perrry et al. (1992). Prospective study of social support & immune function. 221 HIV+ patients without AIDS (mean age=35) followed for 12 months. Total perceived support was unrelated to CD4+ counts at baseline or at 12 mos.

o Horsten et al. (1999). Correlational study of social support & CV function. Low social support (15-item version, assessing belongingness, tangible support, & appraisal support) was associated with decreased heart rate variability.

o Uchino et al. (1999). 67 normotensive men & women. Mean age = 38. Social support moderated age-related differences in resting SBP & DBP. Age predicted higher resting BP among those with low social support but was unrelated to BP among those with high social support. These results were not mediated by various health-related variables, personality factors, or psychological processes.

· Social Support Behaviors Scale (SS-B; Vaux & Harrison 1985; Vaux et al. 1987). 45-item measure, with 5 subscales: emotional support, practical assistance, financial assistance, advice-guidance, socializing. For each item, subjects rate the likelihood that family members & friends would engage in this behavior in a time of need. Separate scores for overall perceived support from family & from friends. Subscale correlations are not clearly reported, but it appears from a confirmatory analysis that the subscales are only moderately intercorrelated (Vaux et al. 1987, Study 5). A 23-item version provides a global score for appraised (perceived) emotional support (SS-A; Vaux et al. 1986). Another version has availability of the 4 functions rated for up to 10 specific network members, together with structural indices (Vaux & Harrison 1985). The scale has received some usage in community psychology. The 45-item inventory may also be administered as a received support measure, with instructions to rate whether each item was recently received (see below).

· Social Provisions Scale (SPS; Cutrona & Russell 1987). 24-item instrument with subscales termed (1) attachment, (2) social integration, (3) reassurance of worth, (4) reliable alliance, (5) guidance, & (6) opportunity for nurturance. Subscale correlations are rather high (r=0.55-0.99). Has been used with variety of adult populations, including new mothers & elderly community residents, & in intervention research. Conceptual basis for the scales is mixed, some being clearly functional & others more structural. Instrumental support is not well represented. High subscale correlations indicate that this measure may be used to assess a higher-order construct of perceived support. Reis & Collins (2000) describe the scale as a "global measure of perceived support."

o RS Baron et al. (1990). Correlational study of social support & immune function parameters. 23 spouses of cancer patients (mean age=48). Total social support & all components were positively related to phytohemagglutinin & natural killer cell activity, even after controlling for life events & depression. Social support unrelated to concanavalin A.

· Medical Outcomes Study Social Support Survey (MOS-SSS; Sherbourne & Stewart 1991). Self-administered questionnaire developed for patients with chronic medical conditions, ages 18-98. 19 items measuring 4 dimensions: (1) emotional support (the expression of positive affect, empathetic understanding, encouragement of expression of feelings; 4 items) & informational support (the offering of advice, information, guidance, or feedback; 4 items), (2) tangible support (provision of material aid or behavioral assistance; 4 items), (3) positive social interactions (availability of others to do fun things with you; 4 items), (4) affectionate support (involving expressions of love & affection; 3 items). Affectionate support has not been emphasized in the literature as a distinct type of support, but it was felt that this type of support would be beneficial to health outcomes of the chronically ill. Did not assess self-esteem support (positive comparison between self & others), as this may be more relevant for populations other than the chronically ill (e.g. college students). To decrease respondent burden, various types of support are measured without regard to the source. For each item, patients are asked to indicate how often each kind of support was available to them if they needed it (none of the time; a little of the time; some of the time; most of the time; all of the time). Subscale alphas 0.91-0.96, 1-year stability coefficients 0.72-0.76; overall support index alpha=0.97, 1-year stability coefficient 0.78. All items correlated highly (at least 0.72) with their hypothesized scales. Although the support subscales were highly correlated (0.69-0.82), all items in the 4 functional support subscales met criteria for discriminant validity (i.e. correlated higher by 2 standard errors with their own scale than with any other support measures). They also discriminated well from the validity measures, supporting their distinction from measures of loneliness or feelings of belonging, mental health, current health perceptions, other aspects of family & social functioning, & structural network measures.

o Sherbourne et al. (1992). Medical Outcomes Study. 1402 chronically ill patients followed for 2 yr. Social support was beneficial for physical function & emotional well-being in longitudinal analysis. Low social support was particularly damaging for physical functioning of older respondents.

· Duke Social Support Index (DSSI; Landerman et al. 1989). 35-item questionnaire concerning respondent's social network & support provided by that network. Used in Epidemiologic Catchment Area study (Piedmont region of North Carolina). Factor analyses led to creation of 5 support indices: satisfaction with support (4 items), perceived social support (7 items), frequency of social interaction (4 items), size of social network (4 items), & instrumental support (13 items). Koenig et al. (1993) developed 11-item version of the DSSI, with 2 subscales: social interaction (4 items) & support satisfaction (7 items). Goodger et al. (1999) report strong evidence for validity & reliability of the 11-item DSSI in sample of 565 community-dwelling elderly Australians.

o DSSI is being used in 2 large Australian studies: The Preventive Care Trial (4-yr clinical trial, 1500 veterans) & the Australian Longitudinal Study on Women's Health (12,000 elderly women followed for 20 yr).

    iv. Network-based inventories employ a 2-stage process. Respondents first identify network members perceived as support providers & then rate availability & adequacy of support from the support providers. These instruments generally required a skilled interviewer and are not self-administered.

· Arizona Social Support Interview Schedule (ASSIS; Barrera 1980, 1981). Requires a skilled interviewer. Assesses 6 functions: (1) intimate interaction, (2) material aid, (3) physical assistance, (4) guidance, (5) social participation, (6) positive feedback. The respondent is first read a general description of the support function (e.g., "If you wanted to talk to someone about things that are very personal & private, who would you talk to?") & is then asked to give initials of all persons who would fit the description as providing that function & whom the respondent had talked to in the last 30 days. There is no limitation on source of support ("friends, family members, teachers, ministers, doctors, or other people you might know") or on the number of supporters who can be listed. Ratings of support satisfaction are obtained in a similar manner as in the ISSI. Protocol includes a score for negative interactions, based on the question "Who are the people that you can expect to have some unpleasant disagreements with, or people that you can expect to make you angry & upset?" The ASSIS yields a measure of total network size, defined as the number of people providing at least 1 supportive function, as well as conflicted network size. Measure has been used extensively, mainly for predicting psychiatric symptoms. Scores for individual support functions are not usually derived from this measure.

· Social Support Questionnaire (SSQ; Schaefer, Coyne & Lazarus 1981). Covers both actual support & perceived adequacy of tangible support, guidance, emotional support. For tangible support, respondents respond to 9 hypothetical incidents in which he/she might count on support. For guidance & emotional support, each specific type of network relationship (e.g. spouse, friend, relative) is rated on 1 question of guidance support adequacy & 4 highly correlated questions of emotional support adequacy.

· Social Relationship Scale (SRS; McFarlane et al. 1980,1981,1983). Developed to measure the effects of social relationships on illness responses in 6 areas in which subject may have experienced life changes: work, money, home & family, personal & social, health, & general social issues. Respondent asked to identify each person that he or she talked to about a problem; the type of relationship (e.g. kin, nonkin, professional); the helpfulness of the discussion on 7-point scale; & the reciprocity with each person. Perceived helpfulness of network is estimated from average of 7-point helpfulness ratings ("makes things a lot worse" to "helps things a lot"). Negative interactions are thus assessed. 3 scores: quality of network, extent of network, reciprocity. Provides overall measure of perceived adequacy (helpfulness) regardless of type or source of support. Asks about support adequacy in context of actual rather than hypothetical problems. Response bias has been examined by changes in wording of questions, & validity has been assessed in small sample of couples with or without family problems & in larger sample of family practice patients. Disadvantages are that it does not specify type of support that is given. Has not been used with elderly. Scale can be self-administered, but use of trained interviewer is recommended. Construction may be too complicated for use in population surveys; discussions of "personal & social things" or "society in general" may be too abstract to elicit meaningful answers (Orth-Gomer & Unden, 1987).

· UCLA Social Support Interview (UCLA-SSI; Dunkel-Schetter, Folkman, & Lazarus 1987; Dunkel-Schetter & Bennett 1990). 70-item interview protocol focused on specific stressors. Respondent is asked to identify recent stressful situation & then asked questions about persons who may provide support relevant to that situation. In the basic instrument respondent is probed about support from 3 specific persons (e.g. parent, friend, partner) & asked to rate the extent to which each person provides emotional, instrumental, & informational support. In medical settings, the support sources can be varied (e.g. spouse, family member, physician). In a variant, respondents are given a list of 15 individuals or groups & are asked whether each of them helped in that situation. In addition to obtaining info about types of support functions & sources of support, the instrument is designed to provide ratings about several aspects of support, including quantity, satisfaction, & reciprocation. Interview includes items on negative aspects of social relationships thru questions about sources of stress in one's relationships. Scores for total number of support sources, or for amounts of emotional, instrumental, & informational support averaged over 3 sources. Wording can be altered to tap perceived availability of support ("To what extent do you feel that you can turn to X for information & advice about health-related concerns?") or received support ("During the past 3 months, how often did X give you information & advice about health related concerns?")

· Close Persons Questionnaire (Stansfeld & Marmot 1992). Developed for Whitehall II Study. Respondent is asked to name up to 4 "closest" network members. 3 support scales were derived from answers to questions about the person nominated as closest network member: (1) confiding/emotional support (7 items) measures wanting to confide, confiding, sharing interests, boosting self-esteem & reciprocity; (2) practical support (4 items) measures major & minor practical help received, (3) negative aspects of close relationships (4 items) measures negative interaction & perceived inadequacy of support from the closest person.

o Stansfeld et al. (1998). Whitehall II Study. 7697 British civil servants followed for 1-3 yr. Low confiding/emotional support in men & high negative aspects of close relationships in men & women associated with greater risk of psychiatric morbidity (General Health Questionnaire), after adjustment for baseline psychiatric morbidity.

b. Received support measures

These measures ask respondents to indicate how often they have received specific support actions in the recent past.

· Inventory of Socially Supportive Behaviors (ISSB; Barrera et al. 1981). 40-item self-report inventory, recent transactions that involved receipt of same functions assessed by ASSIS - i.e., emotional, instrumental, informational, & companionship. Respondents are asked to indicate the extent to which each of 40 support behaviors was received (from anyone) during the past month. A 5-point response scale is used. Factor analyses resulted in 3 types of support: emotional, tangible aid, & guidance. A score for each type of support can be derived by adding the appropriate item scores from the identified factors. However, the authors recommend scoring with a single total score for received support. Widely used in etiologic research but not in intervention studies. Findings include some failures to show buffering effects or predicted main effects. Questions are relatively complex, requiring more than average schooling to be understood - e.g. "How often do people assist you in setting a goal for yourself?', 'How often do people express esteem or respect for a competence or personal quality of yours?" (Orth-Gomer & Unden 1987). Krause (1987) modified ISSB for use with community-dwelling elderly populations; he reports that internal consistency is "adequate" for emotional & guidance support but "lower" for tangible support.

o Glass et al. (1993). 46 stroke patients in Durham, North Carolina were followed for 6 months after stroke. Social support, using a short version of ISSB, was measured at 1, 3, & 6 months. High levels of social support were associated with faster & more extensive recovery of functional status (Barthel ADL) after stroke.

o Oxman et al. (1995). Social support, as measured by ISSB, was not a predictor of mortality in 232 post-cardiac surgery patients followed for 6 months.

· Social Support Behaviors Scale , described above, is a questionnaire measure that can be used to assess received support by changing the instructional set. Other than the original development studies, Wills & Shinar (2000) found no usages of this scale as a received support measure.

· UCLA-Social Support Interview , described above, has been used as a received support measure. Helgeson (1993) found that scores for perceived & received support were related to outcomes in opposite directions. Received support was positively correlated with symptomatology whereas perceived support was inversely related to symptomatology. It is the only interview measure of received support in the literature.

4. Social conflict

o Shuster, Kessler, & Aseltine (1990). Negative interactions (demands, criticisms) are assessed separately for spouse/partner, relatives, and friends with 6 items (e.g. "How often does your spouse/partner make too many demands on you?") on a 4-point scale (1="never" to 4="often").

o Mickelson & Kubzansky (under review). This measure is used in the National Comorbdity Survey, a nationwide household survey of 8098 persons aged 15-54, designed to produce data on prevalence & correlates of psychiatric disorder.

o Seeman (1998). A modification of this measure is used in the MacArthur Midlife Survey.

· Test of Negative Social Exchange (TENSE; Ruehlman & Karoly, 1991)

· Social Network Interview (Finch et al. 1989). An abridged 14-item version of a scale developed by Fischer (1982) & revised by Zautra (1984). Items ask subjects to identify people who provided various types of instrumental support & emotional support. 4 items ask subjects to identify individuals who were sources of negative social experiences. Probes ask respondents to name individuals who caused problems by criticizing subjects' behavior, breaking promises to subject, or consistently provoking feelings of anger. Scores are created by summing the total number of individuals named as providing that particular type of support or negative social experiences. Scale is appropriate for elderly.

5. Measures of social capital/cohesion and psychological sense of community

a. Individual indicators

General Social Survey items (as used by Kawachi et al., 1997, 1999a, 1999b). Individual responses to the following items from the General Social Surveys conducted by the National Opinion Research Center were aggregated to the state level:

· Interpersonal trust: "Most people can be trusted or you can't be too careful in dealing with people."
· Norms of reciprocity: "Most of the time people try to be helpful - or are they mostly looking out for themselves?"
· Civic associations: Membership in wide variety of voluntary organizations, including church groups, sports groups, hobby   groups, fraternal organizations, labor unions.

o Kawachi et al. (1997a). Ecological analysis linking variations in social capital to mortality rates across U.S. states. Social capital indicators - interpersonal trust, norms of reciprocity, & associational membership - were obtained from the General Social Surveys conducted by the National Opinions Research Center between 1986 & 1990 & aggregated to the state level (Putnam 1993; 1995). Per capita group membership in each state was associated with age-adjusted mortality rates. Adjusted for state poverty rates, a 1-unit increase in average per capita group membership was associated with a lower age adjusted mortality rate of 66.8 deaths per 100,000 population. Similarly, variations in levels of trust & reciprocity were associated with mortality across states; an increase in level of trust by about 10% (1 S.D.) was associated with a 9% lower level of overall mortality. These social capital indicators were also associated with most major causes of death, including CHD, malignant neoplasms, & infant mortality. In addition, greater interpersonal mistrust at the state level was linked with higher rates of homicide (r = 0.82), assault (r=0.61), & robbery (r=0.45).

o Kawachi et al. (1999a). Ecologic analysis linking variations in social capital to violent crime rates across US states. Indicators of social capital, not indicators of socioeconomic deprivation, were strongest correlates of state violent crime rates.

o Kawachi et al. (1999b). Multilevel study of association between state social capital & individual self-rated health. Social capital indicators were those used in ecologic analyses just described. Unlike the ecologic studies, this study not only had individual-level data for the outcome variable but also had additional information on individual level confounds, such as health insurance coverage, health-related behaviors, & SES. After adjusting for these individual-level variables, individuals living in states with low levels social trust were at increased risk of poor self-rated health (OR=1.4, comparing states with lowest levels of trust to states with highest levels). The first study to demonstrate contextual effect of state-level social capital on individual well-being.

o Kawachi & Lochner (in progress). Project on Human Development in Chicago Neighborhoods. Aim of study is to examine association between neighborhood-level social capital indicators (trust, reciprocity, associational membership) & neighborhood mortality rates.

b. Composite indices

· Social capital index (Putnam 2000). 14 indicators that are sufficiently intercorrelated to justify averaging them into a single summary measure. Developed to assess social capital at state level.

Measures of community organizational life
      1. Served on a committee of local organization in the last year (percent)
      2. Served as an officer of some club or organization in the last year (percent)
      3. Civic & social organizations per 1,000 population
      4. Mean number of club meetings attended last year
      5. Mean number of group memberships
Measures of engagement in public affairs
      6. Turnout in presidential elections, 1988 and 1992
      7. Attended public meeting on town or school affairs in last year (percent)
Measures of community volunteerism
      8. Number of non-profit organizations per 1,000 population
      9. Mean number of times worked on community project in last year
      10. Mean number of times did volunteer work in last year
Measures of informal sociability
      11. Agree that "I spend a lot of time visiting friends"
      12. Mean number of times entertained at home in last year
Measures of social trust
      13. Agree that "Most people can be trusted"
      14. Agree that "Most people are honest"

o Social cohesion scale (Sampson et al. 1997). Project on Human Development in Chicago Neighborhoods. 8,700 residents in 343 Chicago neighborhoods. Residents were asked how strongly they agreed with the following items on a 5-point scale ranging from "strongly agree" to "strongly disagree". Responses were aggregated to form a neighborhood-level measure of social cohesion.      

      1. People around here are willing to help their neighbors
      2. This is a close-knit neighborhood
      3. People in this neighborhood can be trusted
      4. People in this neighborhood generally don't get along with each other
      5. People in this neighborhood do not share the same values

o Sampson et al. (1997). Project on Human Development in Chicago Neighborhoods. "Collective efficacy" as a determinant of neighborhood variation in violence in Chicago. Collective efficacy index is comprised of 2 sub-scales: social cohesion, which included questions on trust & reciprocity, & informal social control, which included questions on whether neighbors would intervene if children were engaged in deviant behavior, such as skipping school or vandalizing a building. Collective efficacy index was inversely associated with reports of neighborhood violence & violent victimization as well as with homicide rates.

o Sense of Community Index (SCI; MacMillan & Chavis, 1986). The most used & broadly validated measure of sense of community. A useful, reliable, & valid measure that has been applied in different communities, age groups, & cultures. Short form consists of 12 true/false items.

      1. I can recognize most of the people who live on my block.
      2. I feel at home on this block.
      3. Very few of my neighbors know me.
      4. I care about what my neighbors think of my actions.
      5. I have no influence over what this block is like.
      6. If there is a problem on this block, people who live here get it solved.
Fulfillment of needs
      7. I think my block is a good place for me to live.
      8. People on this block do not share the same values.
      9. My neighbors & I want the same thing from this block.
Emotional connection
      10. It is very important to me to live on this block.
      11. People on this block generally don't get along with one another.
      12. I expect to live on this block for a long time.

c. Other measurement issues

i. Aggregate vs. integral measurement approaches for assessment of ecologic variables

o Aggregate variables are those that result from aggregating individual responses to survey items.

o Most indicators of social capital used in health research are aggregate variables

o Integral variables are those that result from direct social observation of neighborhoods.

o Raudenbush & Sampson (1999). Project on Human Development in Chicago Neighborhoods. Used person-based & videotaped approach to collect systematic observations of neighborhood social & physical characteristics. Trained observers drove a sport utility vehicle at a rate of 5 mph down every street within the sample of Chicago neighborhoods. The unit of observation was the face-block (the block segment on 1 side of the street). As vehicle drove down the street, a pair of observers located on each side with video recorders captured social activities & physical features of both face-blocks simultaneously. Also, 2 additional trained observers recorded their observations onto a log for each face block.

ii. Social capital as an individual-level measure?

While generally conceptualized as a collective rather than individual characteristic, social capital is sometimes measured at the individual level (Brehm & Rahn 1997; Veenstra 2000).

o Veenstra (2000). Cross-sectional survey of randomly selected adults in Saskatchewan, Canada. Social capital indicators were individual responses to questions addressing civic participation, trust in government, trust in neighbors, trust in people from your community, & trust in people in general. These indicators were not related to health status.

C. Epidemiologic research linking social connectedness and health

Much evidence linking social integration/networks & support exists. This evidence is reviewed below, and distinctions between social integration/networks & social support are highlighted. By contrast, evidence linking social capital to health outcomes is sparse, reflecting the relatively recent application of the concept to the field of health. Existing evidence on social capital & health was reviewed above.

1. Epidemiologic evidence, by outcome

a. All-cause mortality

o Many large prospective cohort studies show that people who are isolated or disconnected from others are at increased risk of dying prematurely (Berkman & Glass, 2000). Many of these studies, & the measures of social integration that were used in these studies, are summarized in 'Measures' section, above.

b. Cardiovascular disease

o Incidence

o Low social integration & support generally not associated with CVD onset; however, available data are limited & conflicting

Null studies (with measure used):

o Reed et al. (1983). Network indices included physical proximity of parents, marital status, # children, # household members, frequency of church attendance & attendance at social organizations

o Vogt et al. (1992). Scope of network, frequency of contact, & size of network

o Kawachi et al. (1996). Berkman's SNI

Positive study:

o Orth-Gomer et al. (1993). Low integration (abbrev. ISSI) predicted incident MI

o Survival among persons with CVD & CVD prognosis

o Social ties, esp. intimate ties & emotional support provided by those ties, promote survival & better prognosis among post-MI patients or people with serious CVD

Positive studies (with social network/support measure used):

o Ruberman et al. (1984). Social isolation defined as 2 of 3 items: (1) patient did not talk with medical personnel about need for life changes, (2) patient was not a member of any club, social organization, or church, (3) patient "hardly ever" visited friends or relatives

o Orth-Gomer et al. (1988). Recreational social activities assessed with 5-item Daily Rounds of Life Questionnaire adapted from Cornell Heart Study

o Berkman et al. (1992). # sources of emotional support (EPESE)

oCase et al. (1992). Living alone

oWilliams et al. (1992). Unmarried & without confidant

o Oxman et al. (1995). 3 measures: (1) Social Network Questions from EPESE, (2) ISSB (received support), (3) MSPSS (perceived adequacy of support). Only lack of participation in organized groups was associated with mortality following cardiac surgery

o Krumholz et al. (1998). Lack of emotional support (EPESE)

o Farmer IF et al. (1996). Corpus Christi Heart Project. 292 Mexican Americans followed for 2 yr post MI. Low levels of social connectedness, defined as not married, living alone, not advised by neighbor, friend, or relative to seek help, predicted increased mortality.

o Friedmann & Thomas (1995). Cardiac Arrhythmia Suppression Trial. 1 yr follow-up of post-MI patients. Greater social support, measured by 6-item version of SSQ, was associated with reduced mortality.

o Social networks & support have greatest impact on prognosis/survival rather than onset of CVD.

c. Stroke

o Stroke incidence

A few studies have found that socially isolated individuals are at increased risk of fatal stroke (Berkman & Breslow 1983; Kawachi et al. 1996), but these studies lacked the strength to evaluate fully the associations. To date, no studies have reported a link between social isolation & incidence of nonfatal stroke. Kawachi et al. (1996) found a trend in the association between social networks & risk of nonfatal stroke among a cohort of 32,000 male health professionals, but inadequate statistical power precluded multivariate analyses.

o Survival among persons with stroke

Social networks, particularly social isolation, & social support are significantly associated with case fatality in the post-stroke period. Vogt et al. (1992) found that structural network measures (scope of network, frequency of contact, & size of network) were strong predictors of cause-specific mortality among persons with incident cerebrovascular disease. Morris et al. (1993) found that social isolation conferred an additional mortality risk among depressed newly-diagnosed stroke patients followed for 10 yr.

o Stroke prognosis

Aspects of social integration & support, particularly emotional support, influence stroke recovery both in terms of physical function & psychological adjustment (Evans et al. 1987; Friedland & McColl 1987; Glass et al. 1993; McLeroy et al. 1984; Morris et al. 1991). Friedland & McColl (1987) found that satisfaction with social support, measured with their Social Support Inventory for Stroke Survivors, was associated with less psychiatric symptomatology as measured by GHQ. Glass et al. (1993) found that social support, measured with a short version of the ISSB, was associated with faster & more extensive recovery of functional status, as measured by Barthel ADL, 6 months post-stroke. Social support predicts quality of life after stroke (Angeleri et al. 1993; Evans et al. 1994; King 1996). Lack of social support is associated with a variety of negative responses to stroke, including suicidal thoughts (Kishi et al. 1996) & post-stroke depression (Andersen et al. 1995). Social networks & support are also important predictors of hospital course, including length of stay & discharge disposition. In a study of 152 stroke survivors, Brosseau et al. (1996) found that the presence of social support predicted both discharges to rehabilitation & discharges to nursing homes. A larger social network, measured by Berkman's SNI, was associated with lower risk of institutionalization in hospitalized post-stroke patients (Colantonio et al. 1993), though contrary to expectation, total social support (availability of financial, instrumental, & emotional support) was not. Despite this negative finding, evidence regarding the impact of social support on stroke recovery is particularly robust. In a literature review that discarded studies not adhering to sound methodology, social support was the only psychosocial factor with consistent explanatory power (Kwakkel et al. 1996).

o Data supporting the hypothesis that social integration is associated with stroke recovery is more compelling than data suggesting that social integration is associated with stroke onset.

d. Cancer

o Few studies have examined the influence of social integration on cancer outcomes

o In general, network measures do not predict cancer mortality & incidence (Reynolds & Kaplan 1990; Vogt et al. 1992; Welin et al. 1992), although involvement in a range of social activities has been associated with reduced case fatality rates (e.g. Funch & Marshall 1983; Hislop et al. 1987; Waxler-Morrison et al. 1991).

o Even fewer studies have examined the influence of social support on cancer outcomes

o See Helgeson & Cohen (1996) & Helgeson et al. (1998) reviews

e. Physical function

" Seeman et al. (1995). MacArthur Study of Successful Aging. Of social network & support variables assessed, emotional support had the strongest association with change in physical performance (i.e., performance-based assessment) over 2-yr follow-up.

" Mendes de Leon et al. (1999). New Haven EPESE. Total social networks were associated with reduced risk of developing ADL disability over 9-yr follow-up. Emotional & instrumental support did not affect protective effect of social networks against disability but partially accounted for their effect on enhanced recovery.

" Strawbridge et al. (1996). Alameda County. Having close personal contacts was predictive of better physical function (self-report & performance-based measures) among 356 men & women aged 65-95 years followed 6 yr.

o Sherbourne et al. (1992). Medical Outcomes Study. 1402 chronically ill patients followed 2 yr. Social support was associated with better physical function, especially among elderly respondents

f. Cognitive decline

o Associations between a socially engaged lifestyle & higher scores on memory & intelligence tests have been observed among community-dwelling older persons (Gribbin et al. 1980; Arbuckle et al. 1986; Arbuckle et al. 1992; Hultsch et al. 1993).

o Short-term interventions fostering social & intellectual engagement have enhanced cognition among small samples of nursing home residents (Langer et al. 1979) & dementia patients (Koh et al. 1993).

o Few large scale prospective cohort studies have examined relationship between social networks/support & cognitive outcomes

o Bassuk et al. (1999). New Haven EPESE. Social disengagement (as measured by the Social Disengagement Index, a modification of Berkman SNI), but not lack of emotional or instrumental support, was associated with greater cognitive decline in 12-yr follow-up of 2412 community-dwelling elderly persons.

o Fratiglioni et al. (2000). Kungsholmen, Sweden study. A poor or limited social network increased risk of developing dementia in a 3-yr follow-up of 1203 community-dwelling elderly persons.

Potential pathways:

o Neuroendocrine. MacArthur Study of Successful Aging. A high score on Berkman SNI was related to lower urinary cortisol in a sample of healthy 70- to 79-year-olds (Seeman et al. 1994), & lower cortisol, in turn, was related to better cognitive performance in this group (Seeman et al. 1997)

o "Use it or lose it". A recent case-control study of 193 Alzheimer's disease patients & 358 controls, mostly in their early 70s, showed that those who were "mentally active" from age 40-60 were 3 times less likely to have developed AD. (Friedland et al. 2001)

g. Depression

o Studies of the impact of social ties on the risk of depression & psychological distress show consistent protective effects associated with greater social integration, especially when the presence of more intimate ties with spouse, children, or supportive significant others is examined (Broadhead et al. 1983; Barnett & Gotlib 1988; George 1989; Dean et al. 1990; Johnson 1991).

o The increased risks for psychological distress that result from disruption of such ties, particularly marital disruption (bereavement, divorce) have also been extensively documented (e.g. Stroebe & Stroebe 1987; Bowling 1987; Aseltine & Kessler, 1993).

o Oxman et al. (1992). New Haven EPESE. 3-yr changes in depressive symptoms were examined in relation to baseline social network characteristics. Baseline measures of social integration, including more contacts with children & with close friends, were associated with decline in depressive symptoms. Data on changes in social network ties also illustrate negative consequences of losing important ties. Respondents who lost their spouse during the 3-yr follow-up were more likely to experience higher levels of depressive symptoms.

o Emotional support from caregivers affects psychological health outcomes in children. Children exposed to deficient nurturing are at increased risk for depression (Kaslow et al. 1994; Lewinsohn et al. 1994). Studies of college cohorts from Johns Hopkins & Harvard suggest that parental relationships in early childhood characterized by warmth & closeness are predictive of mental (& physical) illness 35-50 yrs later (Thomas & Duszynski 1974; Russek et al. 1997).

2. Social support (vs. integration/networks) & health outcomes

o The strongest associations between social support, particularly emotional support, & health outcomes are seen in psychological outcomes.

o There is no evidence linking social support (as opposed to social networks/integration) to incidence of major physical health outcomes such as MI, stroke, or cancer.

o However, there is evidence linking social support to less extensive development of coronary atherosclerosis (Seeman & Syme 1987; Blumenthal et al. 1987). The strongest evidence linking support to physical health is research demonstrating better survival post-MI for those with emotional support (Berkman et al. 1992; Williams et al. 1992). Similar data indicate beneficial effects of support on post-stroke prognosis (Glass & Maddox 1992). In a comprehensive review of correlational, intervention, and laboratory studies, Uchino et al. (1996) conclude that the correlational evidence supporting a relationship with physiologic processes is similar for social integration/networks & for social support, although most of the evidence examined was not from large-scale epidemiologic studies.

o Epidemiologic evidence suggests that emotional support is protective with respect to physical function (Seeman et al, 1995) & enhances recovery from ADL disability (Mendes de Leon et al. 1998). However, Bassuk et al. (1999) found that emotional & instrumental support were not themselves associated with cognitive decline, nor did support account for the association between social engagement & cognitive decline.

o The generally weaker evidence linking social support (as opposed to social networks & integration) to health outcomes may reflect the greater variability of support over time, making it more difficult to predict disease outcomes that are themselves the results of developmental processes.

o Future studies should include both structural network & functional support measures in order to uncover relationships between them & also to compare their predictive ability with respect to health outcomes. Structural network & functional support measures are not highly correlated - e.g. in the New Haven EPESE study of community-dwelling elderly persons, 60% of respondents reporting no face-to-face contacts with relatives & friends still reported that instrumental & emotional support were available (Seeman & Berkman 1988). Furthermore, few studies have directly compared predictive ability of structural & functional measures with respect to health outcomes in the same population (Cohen 1988). The observational literature suggests that short-term outcomes are more sensitive to social support, while longer-term outcomes, including survival, are most powerfully influenced by social networks (Glass 2000).

o Future studies should include measures of both perceived & received support as existing discrepancies in results for perceived vs. received support have not been resolved (Dunkel-Schetter & Bennett 1990). Empirical studies generally show that perceived availability of support is related to less psychiatric symptomatology, whereas received support correlates positively with such symptomatology. Buffering effects have been observed for perceived but not received support (Dunkel-Schetter & Bennett 1990). [It has been generally assumed that buffering occurs through reducing perceived stress, but empirical evidence has provided only mixed support for this hypothesis (Uchino et al., 1996).]

3. Physiologic pathways

a. Neuroendocrine system

There are documented associations between social integration/support & patterns of neuroendocrine regulation, including the hypothalamic-pituitary-adrenal (HPA) axis, the sympathetic nervous system (SNS), & blood pressure (BP) regulation (Uchino et al. 1996; Seeman & McEwen 1996; Seeman 1996; Knox & Uvnas-Moberg 1998).

    i. CV parameters, i.e. BP & heart rate (Uchino, Cacioppio & Kiecolt-Glaser 1996):

Study types:

o Correlational studies. 19 studies used structural network measures & 14 studies used functional support measures. Network & support measures both predict beneficial effects; indeed, focused comparisons between structural & functional measures revealed no differences in significance level or effect size. Only 4 studies assessed multiple functional dimensions of social support - e.g. Strogatz & James (1986) found that tangible but not appraisal support was associated with reduced hypertension. More research on the multidimensional nature of support is needed.

o Intervention studies. Mixed results. One reason for lack of positive results may be that such studies were designed to see if tangible support would reduce blood pressure among hypertensive patients, yet tangible support was not specifically analyzed!

o Laboratory studies of social support. Social support, especially emotional support, reduces reactivity to acute psychological stress induced by lab-based challenge paradigms (e.g. public speaking). The importance of emotional support is perhaps not surprising, as most of the studies have used stressors in which emotional support may be salient. Some studies have also examined whether emotional reactions mediated the effects of social support on CV reactivity. Mediation was generally not observed. Whether lack of mediation reflects methodologic issues (retrospective assessment) or the possibility that other unmeasured psychological processes are important areas to be explored.

Potential mechanisms:

o Stress-buffering. All 5 intervention studies that examined family networks or support reported stress-buffering effects. Family relationships may be especially important.

o Health-related behaviors. Health-related behaviors do not appear to account for association between social networks/support & CV function, since controlling for health-related behaviors does not eliminate associations. More research is needed in this area.

o Psychological pathways. Perceived stress, feelings of controllability, intrusive or ruminative thinking, feelings of loneliness or depression & other emotional processes are potential psychological mechanisms in pathway between networks/support & CV function. More research is clearly needed in this area.

ii. Endocrine parameters (Uchino et al. 1996; Berkman 2000)

o Gunnar (1992), Gunnar et al. (1992). In children, the presence of a supportive caregiver lowers HPA responses (salivary cortisol levels) to maternal separation.

o Seeman et al. (1994). MacArthur Study of Successful Aging. Persons reporting more frequent emotional support excreted lower urinary levels of epinephrine, norephinephrine, & cortiso.

o Several studies have failed to find an association between social support & cortisol; however, inadequate measurement of cortisol might account for null findings.

o Little data exist on temporal stability of interindividual variations in endocrine function.

o Relationship between social networks/support & cortisol should be examined over full range of diurnal cycle of cortisol

o Distinction between plasma (s-term) vs. urine (l-term) endocrine measures must be kept in mind.

o More research is needed on the relationship between social support & endocrine function, including an examination of behavioral & psychological mechanisms.

b. Immune system

o Social integration has beneficial effects on primary immune system parameters that regulate host resistance (Glaser et al. 1992; Kiecolt-Glaser et al. 1994; Esterling et al. 1996; Uchino et al. 1996; Knox & Uvnas-Moberg 1998).

o Cohen et al. (1997) conducted an experiment to test the hypothesis that diversity of network ties is related to susceptibility to the common cold. Subjects were given nasal drops containing one of two rhinoviruses and monitored for the development of a common cold. Those reporting more types of social ties (e.g. spouse, parent, friend, workmate, etc.) were less susceptible to common colds, produced less mucus, fought infection more efficiently, and shed less virus even after controlling for pre-challenge virus-specific antibody, virus type, age, sex, season, body mass index, education, and race. Moreover, susceptibility to infection decreased in a dose-response manner with increased diversity of the social network.

o Theorell et al. (1995). Tracked decline in CD4 levels over 5 yrs among 48 HIV+ men in Sweden. Used the Availability of Attachment Scale of the Social Network Support Questionnaire (availability of emotional & instrumental support during difficult situations). CD4 counts of men who reported lower "availability of attachments" at baseline declined more rapidly.

4. Social conflict & health.

Seeman (1998): Research on the association between social conflict and health outcomes in adults is nonexistent except as such conflict relates to psychological distress (e.g. Shuster, Kessler, & Aseltine, 1990; Rook 1992; Burg & Seeman 1994); evidence for health effects among children is also largely related to psychological and/or behavioral outcomes, though there is a small literature relating family conflict to physical health outcomes.

o MacArthur studies show that greater social conflict is associated with greater psychological distress (Ryff & Seeman, unpublished) and may also be associated with greater HPA axis activity (Seeman et al. 1994).

o Other studies show relationships between social conflict & greater physiologic arousal - i.e., increased blood pressure (Ewart et al. 1991; Gerin et al. 1992) and neuroendocrine activity (Kiecolt-Glaser et al. 1994).