Social Health in the National Social Life, Health, and Aging Project

Abstract

Objectives

In this article, we present the theoretical framework that guided the development of the National Social Life, Health, and Aging Project (NSHAP) including the measures of social health. We discuss the literature that links social measures to other outcomes, and we discuss in detail how researchers might construct common measures of social health, including those that reflect social relationships, sexuality, social networks, social resources, and social participation.

Methods

The NSHAP includes multiple detailed measures of social health, collected in the rounds of data collection carried out in 2005, 2010, and 2015, allowing for study of changes over time and as people age among a nationally representative sample of the community-dwelling population of older adults in the United States.

Results

We define indicators of social health, describe measures of each in the 2015 round of NSHAP, and show the distribution of the measures by gender and age. We present scales of dimensions of social health that have been developed elsewhere and describe their properties.

Discussion

We briefly discuss the distribution of these measures by age and gender in the 2015 round of NSHAP. Simple analyses of these categorized measures reveal differences by age and gender that deserve closer attention in future investigations using the NSHAP data.

In 1947, the World Health Organization (WHO) defined health as “… a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity” (Glenn & Weaver, 1979). Here we define “social health” as synonymous with the social well-being included in the WHO definition of health or with social functioning. Good social health on any measure should reflect some positive social characteristic or outcome for an individual or group. Similar to physical or mental health, an ideal measure of social health marks the presence of a social characteristic valued by the individual, family, or society and indicates good functioning and/or the absence of dysfunction. As such it has a normative component and might differ across social contexts.

The National Social Life, Health, and Aging Project (NSHAP) was designed with a focus on social health and its relationship to other dimensions of health at older ages. In this article, we focus on five dimensions of social health including social relationships, sexuality, social networks, social resources, and social participation. Four of these dimensions of social health are well represented in the research literature. But sexuality has only recently gained recognition as a social component of a healthy life, including at older ages, with measures of sexual attitudes and behavior included in a number of important surveys. During the design of NSHAP, specific measures within each dimension were chosen from well-validated and widely used questions or instruments when these were available (Hughes et al., 2004; Waite et al., 2009) or developed specifically for NSHAP when they were not (Cornwell et al., 2009, 2014; Galinsky et al., 2014; Kim & Waite, 2014). We point out that NSHAP has many measures of social life that are not included here. Here we present the measures in NSHAP that have been most widely used and are, in our opinion, most theoretically important and most central to assessing social health at older ages. All of these measures have been included in each round of NSHAP data collection since its inception, allowing for study of change over time and as people age, for a nationally representative sample of the community-dwelling population of older adults in the United States. In spite of the care used in constructing and selecting measures of social health for inclusion, many of these measures are not simple to construct, code, or use in analyses.

We discuss the theoretical model developed at the beginning of NSHAP to guide the choice of the measures to be included in the questionnaire that informs each dimension of social health mentioned above. This model may aid those seeking to use specific measures in their own research. For social relationships, we include current marriage or cohabitation and happiness with the relationship. For sexuality, we include partnered sexual activity and sexual interest. Within social networks, we include social network size and network diversity. Within the dimension of social resources, we include measures of loneliness, strained by partner, strained by friends, strained by family, supported by partner, supported by family, supported by friends, and supported by neighbors. In social participation, we include religious services, volunteering, attending group meetings, socialize with friends and family, socialize with neighbors, plus a scale of community participation, and a socializing scale. For each measure, we provide the exact wording of the question(s) used to construct the measure, an example of possible categorization of the measure, and Stata code for its construction with data collected in NSHAP Round 3 (henceforth, R3) which includes Cohort 1 Wave 3 (C1W3) and Cohort 2 Wave 1 (C2W1), collected 2015–2016. We present the distribution of each of these measures by age and gender in that round and discuss differences in the frequency of each measure. Then we discuss scales that can be constructed using multiple items and their characteristics.

Theoretical Underpinnings of NSHAP

NSHAP was designed to provide measures of all the components of the WHO definition of health including physical, mental, and social health. As the first step in this process, we developed the Interactive Biopsychosocial Model of Health (Lindau et al., 2003). We also drew heavily on the Berkman et al. (2000) conceptual model of the role of social networks in the production of health. The Interactive Biopsychosocial Model (IBM) provides the overall framework for our approach, and Berkman et al. give details on the mechanisms that underlie the connections between physical, mental, and social health. We review each of these perspectives below.

The Interactive Biopsychosocial Model of Health

The IBM, which was developed as the first stage of the design of NSHAP (Lindau et al., 2003), conceptualizes health as produced in intimate, family, or social relationships from available resources (Figure 1). These resources, or capital, are divided in this model into biophysical, psychocognitive, and social domains. In the IBM, biophysical capital includes, for example, genetic endowment, physiology, physical functioning, sensory functioning, presence or absence of disease, and health behaviors. Investments in biophysical capital might include exercise, nutrition, sleep, medical care, and reduction of stress. Psychocognitive capital includes such factors as personality, cognitive function, emotional well-being, self-esteem, and resilience. Investments in psychocognitive capital might include exercise, nutrition, sleep, social support, medical care, and reduction of stress. Social capital consists of social ties with others, including intimate relationships, family, friends, neighbors and others, social networks, participation in social activities, social support and demands, social connections, and satisfaction with social relationships. Investment in social capital might include time spent volunteering or attending religious services, contact with friends and neighbors, offering help or support to others, or maintaining close relationships. Figure 1 shows the IBM, with the two individuals in a dyad or family, in this example, represented by the two circle segments on the left and right of the figure. Each partner’s overall health is represented in the semicircle segments on each side of the figure and their endowments of each of the three types of resources—biophysical, psychocognitive, and social—are shown in the circles. The double-headed arrows represent the interaction between the three types of inputs to health. They also show a key feature of this model, which is the fungibility of resources across domains and between members of the intimate dyad or family. One person’s psychocognitive resources can be used to improve that person’s own biophysical health, for example, by gathering information on medical treatments for a new chronic condition. But these resources can also be invested in the partner’s treatment and so improving that person’s health. Social or psychocognitive resources of either person can be used for the benefit of either partner or the dyad. In the same way, poor-quality resources or inadequate resources may take a toll on either or both partners or the dyad. Some social resources belong to the dyad rather than the individual. And, all of this takes place within a sociocultural context.

Figure 1.

Interactive biopsychosocial model of health in a social dyad.

NSHAP was designed explicitly to allow the testing of key hypotheses derived from the IBM. This required the inclusion of detailed and extensive measures of the three domains—biophysical, psychocognitive, and social—in one survey (see the work of McClintock et al. (2017) for details on key measures of biophysical and psychocognitive health in NSHAP).

Within the domain of social health, we turn to the work of Berkman et al. (2000) and their model of the role of social networks in the production of health which points to key concepts that reflect social health. In their model, macrostructural forces condition the extent, shape, and nature of social networks, broadly defined, which comprise the mezzo level. Berkman et al. focused on social network structure, such as size, density, and range, and on characteristics of network ties, such as frequency of contact and intimacy. They also include a measure described as “frequency of organization participation (attendance)” (Berkman et al., 2000, p. 847; Figure 1).

Berkman et al. called the next level of their model “psychosocial mechanisms at the microlevel.” These mechanisms include social support, social influence, social engagement, person-to-person contact, and access to resources and material goods, all of which affect health through health behaviors and through psychological and physiological pathways.

We apply the Berkman et al. focus on mechanisms to argue that social health affects and is affected by other dimensions of health through access to resources, such as time, advice, caregiving, housing, expertise, and money (York Cornwell & Waite, 2012); through emotional support (Warner & Kelley-Moore, 2012; Warner & Adams, 2016); through stress reduction and management (Thoits, 2011); through shared social environments such as social networks (Cornwell, 2012), households (Schafer et al., 2017), and neighborhoods; and through physiological processes (Sbarra, 2009) that lead to chronic disease (Das, 2013; Liu & Waite, 2014; Liu et al., 2016) or social outcomes (Das, 2017).

The IBM, together with the Berkman et al. (2000) model, provides the conceptual foundation for all of NSHAP, including measures of physical health and functioning, psychological well-being, health behaviors, and medical care. The latter are discussed in detail elsewhere. We focus here on key measures of social health: social relationships including marital and intimate partnerships, sexuality, social network characteristics, social resources including loneliness and social support and strain, and social participation. Next, we discuss measures of social health on each of these dimensions, including theories that explain the mechanisms through which they affect and are affected by domains of health. NSHAP is unique in the range and scope of measures of the social world it includes, especially social health as defined above. These measures taken together with information on physical and mental health allow us and others to situate the individual in a complex setting. We present the NSHAP questions that can be used to create measures of each dimension of social health and suggest coding of these into variables or scales (Table 1).

Table 1.

Definitions of NSHAP’s Social Health Measures

NSHAP measure	CAPI question	Dichotomous categories
Social relationships
Marital/partnered status	Are you currently married, living with a partner, separated, divorced, widowed, or have you never been married? Do you currently have a romantic, intimate, or sexual partner?	Unmarried and unpartnered Married or partnered
Very happily partnered	Taking all things together, how would you describe your (marriage/relationship) with the partner on a scale from 1 to 7 with 1 (being very unhappy) and 7 (being very happy)?	Less “Very happy” relationship
Sexuality
	Constructed from NSHAP’s partner history module (Section 3a, Boxes A and B)
Frequency of sexual activity	During the last 12 months, about how often did you have sex with partner?	No At least once
	(0 = Never (volunteered), 1 = Less than once a month, 2 = Once to a few times a month, 3 = Once to a few times a week, 4 = Every day, 5 = Several times a day)
Frequency of sexual ideation	About how often do you think about sex?	Less often than once a month At least once a month
Social networks
	Constructed from NSHAP’s network module (RosterA, RosterB, RosterC)
Sizeable network	Three or more total people listed on network rosters	No (size 0–2) Yes (size ≥3)
Diverse network	Three or more different types of relationships listed	No (relationships 0–2) Yes (relationships ≥3)
Social resources
	UCLA Loneliness Scale composite score (Likert scale: 0 = Never, 1 = Hardly ever, 2 = Some of the time, 3 = Often)
Loneliness	(a) How often do you feel that you lack companionship? (b) How often do you feel left out? (c) How often do you feel isolated from others?	Not lonely (UCLA scale = 0) Felt lonely (UCLA scale = 1–6)
Social strain
	(0 = Never, 1 = Hardly ever, 2 = Some of the time, 3 = Often)
Strained by partner	(a) How often does partner criticize you? (b) How often does partner make too many demands on you? (c) How often does partner get on your nerves?	Not often Partner strain on one or more items
Strained by friends	(d) How often do friends criticize you? (e) How often do friends make too many demands on you? (f) How often do friends get on your nerves?	Not often Friends strain on one or more items
Strained by family	(g) How often does family criticize you? (h) How often does family make too many demands on you? (i) How often does family get on your nerves?	Not often Family strain on one or more items
Social support
	(1 = Hardly ever, 2 = Some of the time, 3 = Often)
Supported by partner	(a) How often can you open up to partner if you need to talk about your worries? (b) How often can you rely on partner for help if you have a problem? (c) How often does partner really understand the way you feel?	Not often Often supported on all 3
Supported by friends	(d) How often can you open up to your friends if you need to talk about your worries? (e) How often can you rely on friends for help if you have a problem? (f) How often do friends really understand the way you feel?	Not often Often supported on all 3
Supported by family	(g) How often can you open up to members of your family if you need to talk about your worries? (h) How often can you rely on family for help if you have a problem? (i) How often does family really understand the way you feel?	Not often Often supported on all 3
	(0 = Never, 1 = Rarely, 2 = Sometimes, 3 = Often)
Supported by neighbors	(j) How often do you and other people in this area do favors for each other? (k) How often do you and other people in this area ask each other for advice about personal things?	Never or rarely Sometimes/often on both items
Social participation
	(Likert Scale: 0 = Never, 1 = About once or twice a year, 2 = Several times a year, 3 = About once a month, 4 = Every week, 5 = Several times a week)
Religious services	Thinking about the past 12 months, about how often have you attended religious services?	Less About once a month or more (3–5)
	(Likert Scale: 0 = Never, 1 = Less than once a year, 2 = About once or twice a year, 3 = Several times a year, 4 = About once a month, 5 = Every week, 6 = Several times a week)
Volunteering	In the past 12 months, how often did you do volunteer work for religious, charitable, political, health-related, or other organizations?	Less About once a month or more (4–6)
Attending group meetings	In the past 12 months, how often did you attend meetings of any organized group? (Examples include a choir, a committee or board, a support group, a sports or exercise group, a hobby group, or a professional society.)	Less About once a month or more (4–6)
Socialize with friends and family	In the past 12 months, how often did you get together socially with friends or relatives?	Less Once a month or more (4–6)
	(Likert Scale: 0 = Never, 1 = Rarely, 2 = Sometimes, 3 = Often)
Socialize with neighbors	How often do you and people in this area visit in each other’s homes or when you meet on the street?	Rarely or never Often or sometimes (2–3)
Social participation scales
Community participation scale (0–6)	Volunteer + attend + atndserv2a	Score 0–6
Socializing scale (0–5)	Visit + socialb	Score 0–5

Note: CAPI = computer-assisted personal interview (In-person); NSHAP = National Social Life, Health, and Aging Project; UCLA = University of California, Los Angeles.

^aSTATA: recode volunteer (0/2 = 0 “Never to once or twice a year”) (3/4 = 1 “Once a month or less”) (5/6 = 2 “Several times a week or less”), recode attend (0/2 = 0 “Never to once or twice a year”) (3/4 = 1 “Once a month or less”) (5/6 = 2 “Several times a week or less”), and recode atndserv2 (0/2 = 0 “Never to several times a year”) (3/4 = 1 “Every week or less”) (5/6 = 2 “Several times a week”).

^bSTATA: recode visit (1 = 0 “Never or rarely”) (2 = 1 “Sometimes”) (3 = 2 “Often”) and recode social (0 = 0 “Never”) (1/2 = 1 “Once or twice a year or less”) (3/4 = 2 “Once a month or less”) (5/6 = 3 “Several times a week or less”).

Method

Dimensions of Social Health

Marriage, partnership, and relationship happiness

By social relationships, we mean any ongoing connection between two or more people. Among the most fundamental of these is the intimate partner relationship. Finding and keeping a mate in an intimate partnership is one of the key developmental tasks of adulthood (Kaufman, 2018), and a successful partnership, some argue, leads to better health of both members of the dyad across all dimensions of health (Waite & Gallagher, 2000). So we begin with an indicator that a person is married or has a cohabiting partner. We also often want to assess the quality of the marriage or partnership. A good quality intimate partnership appears to act differently than one evaluated more negatively (Liu & Waite, 2014). A poor-quality marriage or partnership, or none at all, might be used as an indicator of negative social health.

We measure current marital or partnership status with a direct question: Are you currently married, living with a partner, separated, divorced, widowed, or have you never been married? We measure happiness with the intimate relationship with the question: Taking all things together, how would you describe your (marriage/relationship) with [CURRENT/RECENT PARTNER] on a scale from 1 to 7, with 1 (being very unhappy) and 7 (being very happy)? We consider the person to be very happily partnered if they answer 7 (Table 1). Consideration can also be made to categorize responses of 1–4 as “not happy” and responses 5–7 as “happy.”

Sexuality

Sexuality is an important component of health and well-being throughout the life course. A 2001 report of the U.S. Surgeon General pointed to sexuality as essential to well-being, with calls to attend to sexual health (Office of the Surgeon General, 2001). But serious research consideration of sexual behavior and attitudes, especially among older adults, is relatively recent. Sexuality can be conceptualized as a component of well-being, as a social indicator, and as a predictor or consequence of other dimensions of health (Galinsky & Waite 2014; Galinsky et al., 2014; Lee et al., 2016; Liu et al., 2016; Waite et al., 2009). Because of the increasing recognition by researchers of the importance of understanding sexuality, detailed measures of sexual behavior throughout the life course, attitudes, beliefs, functioning, and well-being have been included recently in important national surveys of health at older ages, including the English Longitudinal Study and NSHAP, and new measures are appearing in other health surveys, including the National Health and Nutrition Examination Survey. The inclusion of both partners in some longitudinal surveys of older adults, together with questions on sexuality asked of each individually, has allowed researchers to study the contribution of each partner to the sexuality of the dyad (Galinsky & Waite, 2014; Kim & Waite, 2014; Waite et al., 2017). However, the development of theories of sexuality at older ages has lagged behind descriptions of this dimension of social health. Sexuality has been linked to self-rated physical health, especially of the male partner (Lindau et al., 2007), to marital quality in the face of health decline (Galinsky & Waite, 2014), and to perceived subjective well-being (Lee et al., 2016). Sexual problems have been shown to be more likely among those with poor mental health (Laumann et al., 2008).

We focus here on sexual interest and partnered sexual activity. We build on the Neuroendocrine Perspective, in which the motivation to have sex manifests itself in various behaviors and attitudes that may be either proceptive (i.e., sex-seeking) or receptive (i.e., being open to sexual activity) in nature. Iveniuk and Waite (2018) argue that a general interest in sex may underlie these varieties of behavior and thoughts. Sexual interest is measured most simply by asking a person how often he or she thinks about sex, although alternative measures appear in various data sets (Iveniuk & Waite, 2018). We measure sexual interest with the following question: How often do you think about sex? Response categories include 0 (never [volunteered]), 1 (less than once a month), 2 (once to a few times a month), 3 (once to a few times a week), 4 (every day), and 5 (several times a day). Levels of sexual interest differ by gender and change with age. Here we follow the work of Waite et al. (2009) to code sexual interest as high if the person reports thinking about sex at least once a month. This choice is illustrative only, and other categorizations may be preferable for specific research questions.

Partnered sexual activity includes any behavior with another person that the individual considers sex. At older ages, it is important to define sexual activity with a partner broadly, as the activities that couples engage in shift away from vaginal intercourse toward touching, cuddling, and kissing (Waite et al., 2009), and sexual inactivity among those with a partner increases with age (Lindau et al., 2007). Assessment of sexual activity might include the specific activities that the person engaged in the last time he or she had sex, such as sexual touching or oral sex (Liu et al., 2019). We measure partnered sexual activity here as any sex with a current or recent partner in the past 3 months, using the question: During the last 3 months how often did you have sex with [CURRENT PARTNER]?

Partnered sexual activity is inherently dyadic and so much less available to those without a partner. Sexual interest, on the other hand, can be measured for all adults, regardless of partnership. Sexual desire is not social in the sense of taking place between or among separate persons (i.e., not in the same sense that social networks are social phenomena), but is rather an inclination to take part in social activity with others, and therefore a part of the psychosocial process leading to partnership and partnered sex. Partnered activity and the inclination toward partnered activity, therefore, give a clearer picture of social functioning on this dimension than either by itself.

Social networks

People are connected to others in a variety of ways, from kin relationships to socializing, to exchanges. Social networks are created by webs of connections among groups of people, so the social network of an individual includes that person’s connections to others and the connections of those other people to each other (Cornwell et al., 2009). Berkman et al. (2000) developed an elegant conceptual model of the links between macrolevel social forces, social networks, psychosocial factors, and pathways to health.

There are many ways to define social networks and many ways to measure them. NSHAP follows the General Social Survey by focusing on their discussion networks: the people with whom they talk about things that are important to them. NSHAP expanded and extended measures of discussion networks to include links between the alters. The respondent names these people, called “alters,” and then the relationship of each of them to the respondent (called “ego” in network research) is ascertained. Are they related and how? How old are they? Do they live with ego? And does ego talk to them about health? Then the respondent is asked in detail about the connection, if any, between each of the pairs of alters named. Did they know each other? Were they related? How often were they in contact? How close was their relationship? This innovation allows researchers to look closely at the links between all those in the network, including flows of information and affection (Cornwell et al., 2009).

We include here two of the most basic measures of social network characteristics: network size and network diversity. We consider network size to be “sizeable” if the individual names three or more people on the total of the three network rosters. We define network diversity as having at least three different types of relationships in one’s network. For discussion of NSHAP, social network measures generally and of other dimensions of social networks and changes in networks (Cornwell et al., 2009, 2014).

Social networks are not cast in stone; they change as the situations of the people in them change. Although we do not discuss network change in this article, the repeated measures of social networks allow researchers to study network change. The second and third rounds of NSHAP which include Waves 2 and 3 for the original NSHAP Cohort 1, obtain the current social network, compare it to the last network by names of alters, and ask specifically about losses and additions to the network and reasons for them (Cornwell et al., 2014). Network loss over 5 years has been found to be greater for older Blacks and those of low socioeconomic status than for others (Cornwell et al., 2014). Cornwell and Laumann (2015) find that those who add new confidants see improvements in functional, self-rated, and psychological health, net of baseline connectedness as well as any network losses that occurred during the same period. They find that those who lost network members overall showed poorer physical but not psychological well-being.

Social resources

Social resources flow through relationships with other individuals or groups. A large body of literature indicates that perceptions of the availability of these resources are much more consequential than objective flows (Haber et al., 2007). For this reason, NSHAP asked respondents a series of questions about perceptions of social support, social strain, and loneliness.

Social support.

The theoretical perspectives outlined earlier all include mention of social support as a mechanism or pathway through which one dimension of social health may affect other outcomes (Berkman et al., 2000). Social support is, quite broadly, any resource that flows between people. These resources can be exchanged within social dyads, such as between spouses or partners, and within social networks or larger social groups such as communities or neighborhoods. Anything that people can exchange can act as a social support resource, but we think most often of instrumental support (such as help with a home repair or picking something up at the store), emotional support, advice or information, financial support, provision of care (such as when one is sick), moral support in a crisis, and social connections to others (such as when a friend calls her sister, the physician, to ask if she can be seen today for that odd symptom, as a favor to her). The research evidence to date suggests strongly that it is the perception that one has good social support that reduces stress, rather than the actual receipt of support (Thoits, 2011). This makes sense if we think of stress as the perception that one has inadequate resources for the challenges one faces. Knowing one has support is a resource, like money in the bank. It acts as a resource, even if one does not need to spend it now. Berkman et al. (2000) pointed to health behaviors, such as smoking and exercise; psychological pathways, such as depression and self-efficacy; and physiological pathways, such as allostatic load and immune function, as examples of pathways through which social support affects health and mortality.

We measure perceived social support in NSHAP through a series of questions asked, separately, about a spouse/partner, members of one’s family, and one’s friends. How often can you open up to [CURRENT PARTNER] if you need to talk about your worries? How often can you rely on them for help if you have a problem? The fourth question on social support was added to this series in 2015/2016: How often do they really understand you? In addition, we assess social support from neighbors through two questions: How often do you and other people in this area do favors for each other? And, How often do you and other people in this area ask each other for advice about personal things? The response categories are 0 (hardly ever), 1 (some of the time), and 3 (often). Responses to the series of questions on social support and on social strain (below) are generally combined to produce a scale of social support or social strain and we follow that practice here (Table 1). We define high social support as reporting that you “often” can turn to friends, family, the spouse, and neighbors.

Social strain.

—We know quite a bit about the presence or absence of social ties but less about their quality. More research has been done, for example, on the impact of being married than on the quality of the marital relationship. Strains in social dyads are a source of chronic stress and appear more often in relationships that are obligatory, as in the parent–child or sibling relationships (Wong et al., 2019). As people have more ability to shed or avoid relationships with a negative component, such as conflict, criticism, or demands, they do so; as a result, their negative relationships become rarer. Divorce or relationship dissolution can rid people of a poor-quality marriage or romantic partnership (Kalmijn & Monden, 2006), one can avoid a sibling or in-law with whom one does not get along, and friends are generally retained only if they provide greater benefits than costs (Offer & Fischer, 2017). Recent research suggests that mild strain, such as nagging or criticizing, may be a benefit in some close relationships. Warner and Adams (2016) found that for disabled married men, increases in negative marital quality, as indexed by criticism, making too many demands, and getting on one’s nerves, reduced loneliness. These relatively mild negatives in the marriage seem to encourage men to persist in social activities that they might give up without the wife’s pushing. In a study that asks directly about difficult people in social networks, Offer and Fischer (2017) found that these people tend to be in close and obligatory social roles with the alter, particularly women relatives and aging parents. One could measure good social health by the lack of negative relationships or poor social health by their presence. Social strain—or strain in social relationships—is measured by asking the individual the extent to which they perceive various relationships to be negative or difficult in some way.

We measure social strain through a series of questions asked, separately, about one’s spouse/partner, members of one’s family, and one’s friends. These are as follows: How often does [CURRENT PARTNER] criticize you? How often do they make too many demands on you? How often do they get on your nerves? The response categories are 0 (hardly ever), 1 (some of the time), and 3 (often). As with social strain, we combine the answers to the three questions on social strain to create a scale, given in Table 1. We categorize individuals as experiencing high social strain if they report “often” to one or more of the questions related to friends, family, and the spouse. Note that the coding of strain scale differs from that for the support scale, as people tend to report high levels of support from all the categories of others, whereas they report generally low levels of strain.

Loneliness or perceived social isolation.

About one person in five is lonely at any given time, with about half of current feelings of loneliness due to situational factors such as a recent move, and about half being hereditary (Hawkley & Capitanio, 2015). Because loneliness is a feeling—the perception that one’s social relationships are lacking or inadequate—one can measure loneliness by asking people how they feel. The simplest measure consists of a single direct question, included as part of the Center for Epidemiological Studies—Depression scale: “I felt lonely,” asked about the last 2 weeks or the last month (Payne et al., 2014), with responses ranging from (never) to (often). Loneliness is also often measured using the University of California, Los Angeles (UCLA) Loneliness Scale, which contains 20 items about perceptions of their social relationships (Russell et al., 1980). All rounds of NSHAP include a short scale, also asked in the Health and Retirement Study, in which respondents are asked how often over the past 2 weeks they have experienced feelings such as “I lacked companionship,” “I felt left out,” or “I felt isolated from others” (Hughes et al., 2004). Response categories include 0 (never), 1 (hardly ever), 2 (some of the time), and 3 (often) (Payne et al., 2014). We suggest categorization as lonely those who answer “some of the time” on at least one question (Payne et al., 2014). Alternatively, measures could categorize individuals as “high loneliness” if they answer “often” on at least one question.

Social participation

The social participation dimension of social health is generally defined as attending organized groups or gatherings, getting together with friends or relatives, attending religious services, or doing volunteer work. Participation in an organized group might include attending meetings of clubs, exercise groups or bowling leagues, playing on a sports team, singing in a choir, being a member of a book club, or being active in a local political or community organization. Volunteering in a soup kitchen, as a docent in a museum, or at the information desk in a hospital all involve organized groups of people doing activities together. One could participate in social events by getting together with family, going out with friends, or attending a neighborhood potluck. Social participation through group or community activities creates weak links between people, may link participants to sources of support—or provide support to others. It offers opportunities to display and act on shared values and beliefs with others and to spend time with others in useful and/or pleasant activities. Social participation is linked to better sleep among older adults (Chen et al., 2016), to better cognitive function (Bowling et al., 2016; Kotwal et al., 2016), to lower or higher levels of depression, depending on the type of organization in which one participates (Croezen et al., 2015), to health behaviors (Lindström et al., 2001), and to the preservation of general competence.

Social participation is almost always measured by asking respondents whether they participate in various social activities and if so, how often they participate. All rounds of NSHAP include questions on five types of social participation: (a) attending meetings of organized groups, (b) attending religious services, (c) doing volunteer work, (d) getting together socially with friends or relatives and getting together with neighbors, and (e) getting together with neighbors. The respondent is asked, in the past 12 months how often he or she attended meetings of organized groups, such as a choir, a committee or board, a support group, a sports or exercise group, a hobby group or a professional society, about how often he or she attended religious services, how often he or she did volunteer work for religious, charitable, political health-related or other organizations, and how often he or she got together socially with friends or relatives. The response categories for all of these questions was 0 (never), 1 (less than once a year), 2 (about once or twice a year), 3 (several times a year), 4 (about once a month), 5 (every week), and 6 (several times a week) (Table 1). Each item can be examined individually to understand different dimensions of group social activities, which can be seen in the Table 2 below. Alternatively, items can be combined into a dichotomous categorization to identify individuals who are actively participating in at least one community activity, one social activity, or any type of activity, or a scale could be created with a count of the number of different types of social activities in which an individual participates (Litwin & Stoeckel, 2016). We describe separate scales of measuring social participation below.

Table 2.

Distributions of NSHAP’s Social Health Measures

	Total					Men					Women
Age in 2015a	50–59	60–69	70–79	80–95	Total	50–59	60–69	70–79	80–95	Total	50–59	60–69	70–79	80–95	Total
Married or partnered	78%	73%	72%	48%	73%	81%	82%	85%	71%	81%	75%	65%	61%	29%	65%
N	1,327	1,338	1,207	732	4,604	591	592	568	327	2,078	736	746	639	405	2,526
Very happily partnered	56%	58%	68%	72%	60%	57%	59%	72%	74%	62%	55%	58%	63%	69%	58%
N	1,017	1,017	914	375	3,323	477	499	497	244	1,717	540	518	417	131	1,606
Partnered sex last year	89%	76%	54%	45%	77%	91%	78%	58%	42%	78%	87%	73%	50%	49%	76%
N	1,130	1,055	883	389	3,457	513	511	476	239	1,739	617	544	407	150	1,718
Frequent sexual ideation	83%	73%	60%	41%	72%	96%	90%	82%	66%	90%	73%	56%	41%	22%	58%
N	1,282	1,281	1,128	665	4,356	566	571	532	293	1,962	716	710	596	372	2,394
Network size	80%	72%	71%	56%	74%	76%	68%	68%	60%	71%	84%	76%	73%	52%	77%
N	1,327	1,338	1,207	732	4,604	591	592	568	327	2,078	736	746	639	405	2,526
Network diversity	79%	69%	65%	51%	71%	75%	63%	61%	52%	67%	83%	73%	69%	50%	75%
N	1,327	1,338	1,207	732	4,604	591	592	568	327	2,078	736	746	639	405	2,526
Loneliness	51%	46%	42%	51%	48%	48%	44%	35%	46%	44%	53%	47%	48%	55%	50%
N	1,019	1,133	1,075	626	3,853	442	497	491	278	1,708	577	636	584	348	2,145
Strained by partner	16%	18%	18%	21%	18%	15%	18%	17%	21%	17%	17%	19%	20%	21%	19%
N	1,018	1,018	912	375	3,323	478	500	496	244	1,718	540	518	416	131	1,605
Strained by friends	5%	4%	2%	2%	4%	6%	4%	2%	2%	4%	5%	4%	3%	2%	4%
N	1,028	1,145	1,092	643	3,908	447	501	502	288	1,738	581	644	590	355	2,170
Strained by family	16%	14%	7%	7%	13%	13%	13%	8%	7%	12%	18%	14%	7%	7%	14%
N	1,031	1,154	1,095	644	3,924	448	504	504	288	1,744	583	650	591	356	2,180
Supported by partner	49%	48%	50%	51%	49%	52%	49%	54%	48%	51%	47%	46%	45%	57%	47%
N	1,018	1,017	914	375	3,324	478	500	497	244	1,719	540	517	417	131	1,605
Supported by friends	19%	20%	13%	8%	17%	11%	11%	9%	4%	10%	25%	28%	17%	11%	23%
N	1,029	1,146	1,091	640	3,906	447	501	502	287	1,737	582	645	589	353	2,169
Supported by family	23%	24%	29%	28%	25%	22%	17%	24%	20%	21%	24%	30%	33%	35%	29%
N	1,031	1,154	1,095	643	3,923	448	504	504	288	1,744	583	650	591	355	2,179
Supported by neighbors	61%	59%	67%	69%	62%	66%	59%	65%	71%	64%	57%	60%	69%	68%	61%
N	1,030	1,143	1,092	640	3,905	448	500	500	287	1,735	582	643	592	353	2,170
Religious services	40%	42%	54%	58%	45%	37%	38%	50%	57%	41%	42%	46%	58%	59%	48%
N	1,014	1,109	1,067	613	3,803	444	486	491	270	1,691	570	623	576	343	2,112
Volunteering	24%	30%	35%	33%	29%	21%	27%	32%	26%	26%	26%	31%	37%	38%	31%
N	1,006	1,104	1,053	603	3,766	441	484	488	266	1,679	565	620	565	337	2,087
Attending group meetings	39%	42%	49%	41%	42%	36%	37%	44%	36%	38%	42%	46%	54%	45%	46%
N	1,006	1,109	1,061	609	3,785	441	486	490	270	1,687	565	623	571	339	2,098
Socialize with friends family	77%	80%	82%	82%	79%	76%	75%	78%	78%	76%	78%	84%	86%	85%	82%
N	1,010	1,109	1,066	614	3,799	442	487	492	270	1,691	568	622	574	344	2,108
Socialize with neighbors	51%	53%	54%	54%	52%	52%	48%	52%	52%	51%	50%	58%	55%	56%	54%
N	1,023	1,142	1,088	634	3,887	445	498	500	283	1,726	578	644	588	351	2,161
Overall	1,327	1,338	1,207	732	4,604	591	592	568	327	2,078	736	746	639	405	2,526

Note: NSHAP = National Social Life, Health, and Aging Project.

^aNote this may be different from actual age at interview: aged 50–59 years in 2015 = born 1956–1965, aged 60–69 years in 2015 = born 1946–1955, aged 70–79 years in 2015 = born 1936–1945, and aged 80–95 years in 2015 = born 1920–1935.

Scales of community participation, socializing, and social disconnectedness

A number of scales of social health that capture important concepts have been developed by researchers using NSHAP. We refer the reader to publications describing the three-item loneliness scale (Hughes et al., 2004) and scales of social support and social strain (Kim & Waite 2014; Wong & Hsieh, 2017). We present two scales of social participation used in the work of Kotwal et al. (2016). These are a scale of community participation, which sums volunteering, attending meetings of organized groups, and attending religious services, and a scale of socializing, which adds getting together with friends and family and getting together with neighbors. We provide details on these two scales in Table 3. Table 2 presents the distribution of each of the individual items in these scales, and Table 3 presents the distribution of each of these scales by age and gender.

Table 3.

Distributions of Social Health and Participation Scales

	Total					Men					Women
Age in 2015a	50–59	60–69	70–79	80–95	Total	50–59	60–69	70–79	80–95	Total	50–59	60–69	70–79	80–95	Total
0–5 Socializing Scale	3.10	3.17	3.26	3.29	3.17	3.10	3.03	3.19	3.15	3.09	3.11	3.30	3.32	3.39	3.24
N	1,002	1,099	1,055	598	3,754	438	483	488	263	1,672	564	616	567	335	2,082
0–6 Community Scale	1.66	1.84	2.18	2.04	1.85	1.49	1.63	1.92	1.88	1.65	1.80	2.02	2.40	2.17	2.01
N	1,001	1,099	1,045	595	3,740	438	482	485	265	1,670	563	617	560	330	2,070
Overall	1,327	1,338	1,207	732	4,604	591	592	568	327	2,078	736	746	639	405	2,526

^aNote this may be different from actual age at interview: aged 50–59 years in 2015 = born 1956–1965, aged 60–69 years in 2015 = born 1946–1955, aged 70–79 years in 2015 = born 1936–1945, and aged 80–95 years in 2015 = born 1920–1935.

Objective social isolation or social disconnectedness

The theories on social isolation and a large body of research both point to the mechanisms linking both perceived social isolation, which psychologists call loneliness, and objective social isolation to other domains of health. And these links appear to be strong. To summarize briefly, the socially isolated face higher risks than the well connected of poor sleep, unhealthy behaviors such as alcohol use and smoking, obesity, early cognitive decline and Alzheimer’s disease, poor mental health including depression, poor self-rated health, and early mortality (Hawkley & Capitanio, 2015; York Cornwell & Waite, 2009a). Objective social isolation is a source of stress, increases exposure to other stressors, and exacerbates their effects. The socially isolated are cut off from sources of instrumental, emotional, advisory, financial, or other support. They use up attention and executive function worrying about social threats and so have less of these resources for the rest of life. If they get sick, they are less likely than others to have people who will help them. This is a powerful case for the inclusion of assessments of both loneliness and objective social isolation in any composite or global measure of social health.

Because being objectively isolated means having relatively few people around one, a fairly vague concept, there are many ways to measure it. York Cornwell and Waite (2009a) created a “social disconnectedness scale” from eight items: (a) network size, (b) network range, (c) frequency of interaction with network members, (d) proportion of alters in the home, (e) number of friends, (f) socializing with family or friends, (g) attending group meetings, and (h) volunteering. Additional items that can be considered as part of overall social isolation assessments include whether an individual lives alone, being unmarried/unpartnered, or having infrequent contact with others, small social networks, or perceptions of low social support (Berkman et al., 2000; Ertel et al., 2008; House et al., 1988). McPherson et al. (2006) operationalized social isolation as not having a confidant: someone to talk to about matters that are important to one. We present here details on the York Cornwell and Waite social disconnectedness scale as originally defined, with Stata code available to recreate this measure using variables available in R3, 2015–2016 (Supplementary Appendix Table 1, top panel). We also suggest a modification of this scale for use with the 2015 measures, to maintain the underlying concept but with an improved Cronbach’s alpha. This scale adaptation is included in the lower panel of Supplementary Appendix Table 1.

Data and Analysis

We use data collected from the 4,777 respondents who comprise Round 3 (Cohort 1 Wave 3 [C1W3], and Cohort 2 Wave 1 [C2W1]) of the NSHAP study. These data consist of an in-person home interview and a supplemental Leave-Behind Questionnaire (LBQ) given to respondents to complete and return by mail. All surviving respondents from Rounds 1 and 2 were interviewed, along with a new cohort of respondents (N = 2,368) born 1948–1965 together with their spouses/partners. The new cohort received a different version of the LBQ from that given to the returning respondents of the original cohort.

We restrict our analyses of social health to the 4,607 respondents who meet our age criterion of the birth year between 1920 and 1965. Table 2 summarizes the distributions of these dichotomized measures across age and gender. All results reflect sample weights and nonresponse weight adjustments. Table 3 presents means on the scales of community participation and socializing by age and gender.

Results

We categorized respondents into birth cohorts, using the information on age at interview and date of interview. Birth year was used to create age categories in Table 2 as follows: aged 50–59 years (born in 1956–1965), aged 60–69 years (born 1946–1955), aged 70–79 years (born 1936–1945), and aged 80–95 years (born 1920–1935). Table 2 reports the weighted prevalence of each social health measure and comparisons by age group and gender. The majority of the sample reported being married or partnered (73%), very happily partnered (60%), having had sex in the last year (77%), frequent sexual ideation (72%), a large network size (74%), network diversity (71%), being supported by neighbors (62%), socializing with friends and family (79%), and socializing with neighbors (52%). In contrast, a minority of participants reported any loneliness (48%), being strained by their partner (18%), being strained by friends (4%), being strained by family (13%), being supported by friends (17%), being supported by family (25%), religious participation (45%), volunteering (29%), or attending group meetings (42%), or social strain overall (23%), by family or friends (13%), or by their partner (18%).

The share of older adults who are married or partnered across the ages of our sample shows different patterns for men and women. Marriage/partnership is approximately equally prevalent among men from their 50s (81%) through their 70s (85%) and then falls to 75% among those aged 80 and older, whereas for women, the percent married/partnered falls steadily from their 50s (75%) through ages 80 and older (29%). For both men and women, the percent very happily partnered rises with age. Men and women are quite similar in percent of those with a partner who was sexually active, very high for those in their 50s falling to less than half for those 80 and older. A substantial majority of men in their 50s report thinking about sex frequently, compared to two thirds among those 80 and older. For women, the starting point is lower (73%) and the decline with age is more dramatic (22% of those 80+).

We focus on two characteristics of the social networks of older adults: network size and network diversity. Table 2 displays that 76% of men in their 50s have networks of three or more people. Among men 80 and older, 60% have networks this large. The decline with age is even sharper for women, from 84% with sizeable networks among those in their 50s to 52% among those 80 and older. We see a similar pattern for network diversity—having at least three different types of relationships among members of one’s network. Three quarters of men in their 50s have diverse networks compared to 52% of those aged 80 and older. For women, share with diverse networks falls from 85% among the youngest in NSHAP to 50% among the oldest. To the extent to which having a larger, more diverse network provides resources, these are less available to older than younger men and women in our sample.

Older adults exchange social support with others and may experience strain from their relationships. Table 2 presents the share who report any symptoms of loneliness. The share of those who are lonely is fairly stable across age groups for both men and women, although levels are slightly higher for women than men at all ages. Social strain from friends and family is quite low generally and tends to fall with age. Strain from a partner is a bit higher overall (17% men and 19% women) but shows slight increases with age rather than the decreases we see for strain from others.

About half of older men and just a bit less than half of older women report very high levels of support from their spouse/partner across all ages. However, few men or women report very high levels of support from friends (10% men, 23% women) or family (21% men, 29% women). Recall that our measures count frequency of support across all three dimensions—talking, help, and understanding. Apparently, only spouses often provide all three types. Friends and family either provide one or two types of support or do not do so consistently. Note that the NSHAP measure of support from neighbors, which comes from a series of questions on neighborhood social cohesion (York Cornwell & Cagney, 2014), includes items measuring perceptions of how often neighbors do favors and or can be asked for advice. About 60% of both men and women in our sample say that people in their neighborhood do these things sometimes or often.

NSHAP’s measures of social participation include attending religious services, volunteering, and attending meetings of organized groups. Here we present the share of NSHAP respondents who engage in each activity once a month or more versus less frequently. A sizeable minority of both men and women regularly attend religious services (41% men, 48% women), with attendance rising with age for both (Cornwell et al., 2008). A smaller share volunteer regularly (26% men, 31% women) but with different age patterns for men and women. Older men show about the same level of volunteering across the ages we examine but older women show a consistent rise in volunteering with age, from 26% of those in their 50s to 38% of those 80 and older. The share attending group meetings is fairly similar across ages for both men and women. The scale of community participation, described above and in Table 3, ranges from 0, for those who do not participate in any community activity—religious services, volunteering, or meetings of organized groups—at least once or twice a year, to 6 for those who participate in all three activities more than once a month. The mean of this scale—1.7 for men and 2.0 for women—suggests that NSHAP respondents participate in one community activity sometimes but not regularly or several activities occasionally—either gets you the score we observe. Community participation rises with age for both men and women but not dramatically. If we view social participation at older ages with a wide lens, we see clearly that the social lives of older adults revolve around activities with friends and family that do not diminish in frequency with age, combined with friendly relationships with neighbors that involve visiting and chatting. Fewer older adults organize their social lives around the church, or volunteering, or their book club or bowling league, although religious participation does tend to be higher at older ages than at the younger ones we study.

Of all the types of social participation we examine, older adults are most likely to socialize with friends and family once a month or more than to do any of the other social things on our list. Three quarters of men and 82% of women say they get together with family and friends this frequently, with slight increases with age for both genders. And about half say they socialize with neighbors “often or sometimes” (51% men, 54% women). The scale of socializing adds together getting together with friends or relatives with visiting with neighbors. The scale ranges from 0 to 5, with 0 indicating people who never or rarely do either of these and 5 indicating those who do both “often” or more than once a month. We saw above that getting together with friends and relatives at least monthly is the most popular type of social participation and chatting or visiting with neighbors is also very common. The mean of the socializing scale reflects this (3.1 men, 3.2 women).

A proposed social disconnectedness scale is shown in Supplementary Appendix Table 1. The upper panel shows a scale estimated with the eight items in the original scale (York Cornwell & Waite, 2009a) using values for these variables in 2015. The combination of the eight items has reasonable internal consistency reliability with a Cronbach’s alpha of 0.64. The social connectedness scale was standardized to have a mean of 0 and SD of 1. This measure is included here to allow data users to construct an identical social disconnectedness scale in each wave of the survey, using the same variables. We note, however, that the Cronbach’s alpha for the original scale created from NSHAP R1 (C1W1 in 2005–2006) was 0.73. This scale estimated with values of the variables in 2015 has worse scale properties, with a Cronbach’s alpha of 0.64. For this reason, we created a modified version of the scale in R3 (2015–2016) using somewhat different coding and measures, which we include in Supplementary Appendix Table 1, lower panel. We recommend this for cross-sectional analyses of social disconnectedness using the 2015 round of data.

Discussion

The NSHAP was designed to cast light on the complex, dynamic, and powerful processes that link the social world to other dimensions of health. The resulting three rounds of NSHAP data over 10 years provide researchers with the opportunity to develop a comprehensive understanding of older adults’ social health across numerous dimensions. This article provides a description of commonly used social measures that can be constructed from all rounds and waves of NSHAP and a rationale for categorization schemes that are easily interpretable for a broad audience of investigators. We also provide the Stata code to construct these measures to help facilitate their wide application. Simple analyses of these categorized measures reveal differences by age and gender that deserve closer attention.

If we return to the conceptual framing underlying the study, particularly the work of Berkman et al. (2000), we see the mechanisms through which various dimensions of social health affect and can be affected by processes at other levels, ranging from large-scale public policy to individual-level gene expression. The various facets of social health are intertwined and integrated with each other and with the processes affecting biological and psychological health. Thus, it is valuable and often essential to have measures of, say, intimate relationships in the same survey as measures of social networks for a study on a health condition such as erectile dysfunction (Cornwell & Laumann, 2011) or as used in the York Cornwell and Waite (2009b) paper linking different social processes of social isolation and loneliness to mental and physical health. Rich, granular, extensive measures of the social world let researchers explore the various levels, from macro to meso to micro, as Berkman et al. argue, and the ways that they might work through each other to produce various outcomes.

NSHAP promotes examining social health in the context of cognition, physical well-being, functional health, emotional health, and health behaviors, which is helping to unravel the complex processes tying these health domains together. This is increasingly relevant to clinical researchers, especially with the coronavirus disease 2019 pandemic which has highlighted how social health affects multiple other facets of health. Recently, the National Academy of Medicine published a report on the importance of clinical research and education on the social determinants of health (including community and social contexts). Separately the National Academy of Sciences included a report on the health effects of loneliness and social isolation. These reports emphasize the need for the consistent use of validated social measures to improve the interpretation of studies for medical professionals and place future work in the context of a growing body of research. By providing clear, consistent, and validated social constructs in the NSHAP, we can promote future clinical research while bridging interdisciplinary research principles.

NSHAP has more information about the positives of social ties than about the negatives, although measures of social strain and elder mistreatment are a step in the right direction (see Wong et al., this volume). NSHAP measures of social strain, discussed here, reflect fairly low levels of everyday conflict, and other than questions on elder mistreatment, we do not attempt to capture more serious conflicts. More can be done. Plans to add questions on discrimination in housing, employment, and everyday interactions in the next round are part of this effort.

Perhaps more important, NSHAP and other surveys must think carefully—if surveys can think—about how to capture rare but cataclysmic events at all the levels at which they occur. Researchers everywhere are dealing with the coronavirus disease 2019 pandemic and its consequences for social health as part of the broad effects of this worldwide event. Clearly, knowing what happened to older adults and those around them will be essential to modeling changes in health and well-being now and over the next decades.

Funding

This paper was published as part of a supplement supported by funding for the National Social Life, Health, and Aging Project from the National Institute on Aging, National Institutes of Health (R01AG043538, R01AG048511). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Aging, National Institutes of Health, or NORC at the University of Chicago.

Conflict of Interest

None declared.