A Reciprocal Relationship between Formal and Informal Social Engagement and Handgrip Strength of Older Adults in South Korea

Social engagement may protect the physical health of older adults. Social engagement has been found to be positively associated with physical health and mobility [1‒3] and negatively related to physical limitations and disability [4, 5] of older adults, although a minority of studies did not identify such a relationship [6]. Handgrip strength, an indicator of maximum voluntary muscle strength, is a convenient and reliable indicator of physical health status [7]. It has been widely adopted in clinical and epidemiological surveys of older adults due partly to convenient measurement with a handgrip dynamometer. It is also a diagnostic tool for sarcopenia, the clinical frailty of skeletal muscle mass and function loss [8, 9]. Studies showed that low handgrip strength is associated with the likelihood of falls, fractures, and physical disabilities among older adults [10, 11].

However, the literature lacks empirical studies identifying the relationship between social engagement and handgrip strength. Instead, the negative association between social isolation, the state opposite to social engagement, and handgrip strength has been examined [12, 13]. A recent study thus lamented that “the specific association between social relationship factors and handgrip strength in older adults is not yet well understood” [14]. This is partly because handgrip strength is ignored or counted as one of many indices of physical function.

Further, it is uncertain which type of social engagement is more strongly related to handgrip strength – between informal social engagement, which consists of interactions with family, friends, and neighbors, and formal social engagement, which denotes joining various organizations and participating in their activities. Therefore, the present study examines whether and how formal and informal social engagement and handgrip strength are associated among older adults.

The study draws on the conceptual model by Berkman et al. [15] in examining the relationship between social engagement and handgrip strength. Berkman et al. [15] proposed a model that demonstrates how social networks may influence health. The conceptual model suggests a sequential causal association that links macro-social conditions with psychological and physiological mechanisms that jointly structure the processes whereby social networks may impact health.

Specifically, the conceptual model by Berkman et al. [15] integrates sociostructural, cultural, and institutional conditions at the macro level, social network structure and characteristics at the meso level, and psychosocial mechanisms at the micro level that may, in concert, impact health through behavioral, psychological, and physiologic pathways. In this model, social engagement is identified as a micro-level psychosocial factor that may reinforce “meaningful social roles,” strengthen “bonding/interpersonal attachment,” and offer regular opportunities for “physical and cognitive exercise” [15]. These social engagement features may be positively associated with health by encouraging health behaviors (behavioral pathway), enhancing self-efficacy (psychological pathway), or regulating hypothalamic-pituitary-adrenal axis response, allostatic load, or immune system function against stress (physiologic pathway).

According to the model, social engagement through informal ties and formal organizations may be associated with handgrip strength over time because it provides a long-term active lifestyle [2] that may be positively related to the behavioral, psychological, and physiologic pathways. However, empirical evidence is scarce. A study by Chen et al. [16], using small-sized cross-sectional data, reported that Taiwanese older adults with fewer social activities (e.g., leisure activities, religious activities, gathering, or chatting with neighbors) were more likely to exhibit the prevalence of frailty; one of the frailty measures was handgrip strength. It is thus necessary to test the association between social engagement and handgrip strength in isolation from other health consequences, employing longitudinal data.

Between formal and informal social engagement, formal social engagement may help older adults retain a routinized schedule of various activities presumably related to handgrip strength in the long run. Nemoto et al. [17] found that older Japanese adults who consistently engaged with various social groups (a senior citizens’ club, a volunteer group, a sports group, or a hobby and culture group) over 2 years experienced smaller declines in physical activities than those who were consistently disengaged. Thus, formal social engagement in diverse associations and their activities may solidify regular and greater involvement in the social environment than informal engagement with familiar persons.

Identifying the relationship between social engagement and handgrip strength may give older adults a plausible reason to be socially active. That is, social engagement may help promote handgrip strength, “a powerful biomarker of aging” [18], which is systematically related to wide-ranging morbidities such as hypertension, cardiovascular diseases, or cancer [19‒21]. Thus, the positive association, if verified, may support the social causation perspective by showing that a social network upstream factor affects an overall physical health indicator [22].

It is also crucial to test if a reverse relationship between handgrip strength and social engagement exists. If confirmed, it may indicate that a health selection mechanism enables some older adults with greater handgrip strength to increase their social engagement [23]. However, whether older adults’ social engagement and handgrip strength are mutually associated over time has not been studied. The present study thus examines whether formal or informal social engagement is reciprocally associated with handgrip strength. The reciprocal relationship between social engagement and handgrip strength may validate both social causation and health selection mechanisms in a balanced manner. To test this reciprocal association, the study uses a nationally representative longitudinal survey dataset of older adults from the Republic of Korea (Korea hereafter).

The study used the seven waves of data from the Korean Longitudinal Study of Aging (KLoSA) survey administered biannually from 2006 to 2018. The nationally representative sample of adults aged 45 years or older was drawn in 2006 using multistage stratified probability sampling. The KLoSA was modeled after representative panel studies of older adults, such as the Health and Retirement Study (HRS) in the USA [24] and the English Longitudinal Study of Aging (ELSA) in the UK [25].

The KLoSA recruited 10,254 respondents in the first wave, with a response rate of 75.4 percent [26]. Its sample retention rate was 78 percent at the seventh wave. The survey added 920 new respondents to the fifth wave to compensate for sample attrition, but this study excluded them from the analysis. The study also dropped 2,327 respondents who passed away between the first and seventh waves. According to random-effects logistic regression analysis, respondents who were unmarried older men with low income and education suffering from functional difficulties and depression had a higher likelihood of exiting the sample due to death. The number of respondents and sample attrition across waves are summarized in Fig. 1. Further data details, such as the research design, sampling, and specific protocols, can be found on the KLoSA website [27].

Handgrip strength was measured using the Tanita dynamometer. The respondents were instructed to squeeze the dynamometer’s handle with their best effort while sitting with an elbow bent at ninety degrees and their forearm and wrist in a neutral position [28]. They conducted the test twice with each hand. The interviewers were instructed not to measure if respondents were sick or refused to take the test. The final handgrip strength score was the average of the four measurements in kilograms. Given that the multivariable analysis method the study adopted is better fitted to a continuous outcome measure, the study did not apply a cutoff point to convert the handgrip strength scale into a dummy variable [19, 29].

Additionally, the study employed a parallel measure of the isometric relative handgrip strength for sensitivity analyses. The alternative variable was calculated by dividing the averaged handgrip strength score by the body mass index [30].

Two indicators measured formal social engagement are as follows:

Number of association memberships: this was the summated number of memberships a respondent held in the following seven types of organizations (range: 0–7): (1) religious groups, (2) social clubs, (3) leisure/culture/sports groups, (4) alumni associations, hometown alliance, clan gatherings, (5) volunteer groups, (6) political parties, NGOs, interest groups, and (7) other organizations.

Frequency of organizational activities: this indicator measured how often a respondent participated in seven associational activities. It ranged from 1 = almost never to 10 = almost every day. We assigned the average frequency of up to seven activities to each respondent.

Given that the membership and the frequency of activity in the third type of association (i.e., leisure/culture/sports groups) may be directly related to physical activities, we conducted supplementary analyses that excluded the type. Next, informal social engagement was measured with the following indicator:

Frequency of contact with familiar persons: this indicator denoted how frequently a respondent met friends, relatives, or neighbors. It ranges from 1 = almost never to 10 = almost every day.

We controlled for demographic features [4, 6, 30, 31], socioeconomic characteristics [1, 4], health risk behaviors [15, 32, 33], comorbidities [33, 34], Activities of Daily Living [35, 36], and depression [36, 37] because they were reported to be associated with older adults’ social engagement and handgrip strength in the literature. Among these, the time-varying controls are as follows:

Medical comorbidities: this measure counted the number of medical diagnoses such as hypertension, diabetes, cancer, lung disease, liver disease, heart disease, cerebrovascular disease, and arthritis. The possible range of the summated scale was from 0 to 8; the higher the number, the greater the comorbidities.

Activities of Daily Living (ADL): the seven-item ADL scale developed by the Korean Geriatrics Society measured whether a respondent experienced limitations in performing basic daily self-care activities such as dressing, washing face and hair, and brushing teeth, bathing, self-feeding, getting out of bed and walking across a room, toileting alone, and controlling urination and defecation [38]. The scale reliability coefficient ranged between 0.91 and 0.97 over the seven waves. The summated variable ranged between 0 and 7.

Depressive symptoms: this was a ten-item short form of the Center for Epidemiologic Studies Depression Scale (CES-D-10) [39]. The questions probed whether a respondent experienced depression in the past week: I felt depressed, I felt that everything I did was an effort, my sleep was restless, or I could not get going. Each had four response categories from 0 = rarely or none of the time (less than 1 day) to 3 = most or all of the time (5–7 days). The scale reliability coefficient ranged between 0.79 and 0.86 over the seven waves. We constructed a summated depression scale ranging from 0 to 30; the higher the score, the greater the number of depressive symptoms.

Smoking: it was a dichotomous variable where 1 = current smoker and 0 = nonsmoker.

Drinking: it was a dichotomous variable where 1 = current drinker and 0 = nondrinker.

Exercise: it was a dichotomous variable where 1 = exercise regularly at least once a week and 0 = no regular exercise.

Household income: this was the total household income for the last year. The mean household income was KRW 20,570,000 (approximately USD 14,500) at wave 1 and KRW 27,250,000 (approximately USD 19,000) at wave 7. The study used logged total household income.

Working: it was a dichotomous variable where 1 = currently working and 0 = not working.

Urban: it was a dichotomous variable where 1 = urban residence and 0 = rural residence.

Marital status: it was a dichotomous variable where 1 = married and 0 = unmarried.

The time-invariant controls were as follows:

Age at baseline: it ranged from 45 to 90 in the first wave.

Female: it was a dichotomous variable where 1 = female and 0 = male.

Education: it measured the level of education: 1 = primary school or less, 2 = middle school, 3 = high school, and 4 = university or above.

Religion: it comprised four dichotomous dummies – Protestant, Catholic, Buddhist, and other religions. They were compared against nonreligious respondents.

We employed a fixed-effects dynamic panel model using structural equation modeling [40, 41]. Because the model used Moral-Benito’s [42] maximum likelihood estimator, it is called ML-SEM. The model estimates the influence of cross-lagged time-varying predictors on an outcome measure while controlling for the lagged effect of the outcome on itself at later time points. Cross-lagged means the effect of a past independent variable on the future realization of the dependent variable. The self-lagged effect of a variable on itself at a later time is called autoregression because a current state depends on its past state. ML-SEM is an advanced version of the usual time-lagged regression model.

A critical difference between ML-SEM and traditional cross-lagged SEMs is that ML-SEM accounts for unmeasured time-invariant individual traits by a latent construct that absorbs unmeasured unit-specific fixed effects (α_i and η_i in the equations below). In sum, ML-SEM uses within-individual changes across time to estimate the relationship between time-varying predictors and an outcome while accounting for all stable individual characteristics [43].

Next, δ and γ are row vectors of coefficients, wit are time-varying control variables, while z_i are time-invariant control variables, and α_i and η_i are unit-specific fixed effects comprising all unobserved time-invariant confounders. In Fig. 2, alpha, a fixed-effects term, is correlated with the time-varying exogenous variable at all time points (“soceng1–soceng7”) and the time-varying endogenous variable at the baseline (“hgs1”) while predicting the time-varying endogenous variable at later time points (“hgs2–hgs7”). Therefore, ML-SEM produces unbiased estimates after accounting for unobserved heterogeneity between respondents. Lastly, εit and υit are random errors.

The ML-SEM uses model fit indices such as Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) and the root mean square error of approximation (RMSEA) and standardized root mean squared residual (SRMR) [45]. The cutoff criteria for goodness of fit between the proposed model and the observed data are 0.95 or higher for CFI and TLI, 0.06 or lower for RMSEA, and 0.08 or lower for SRMR [46]. Additionally, the study employed a full information maximum likelihood (FIML) estimator along with the ML-SEM to account for nonresponse bias and missing data.

Table 1 describes the sample characteristics. The mean handgrip strength score was 26.1 in 2006 and decreased by approximately 1.5 over 12 years. In addition, the relative handgrip strength also declined slightly from 1.13 to 1.07. The mean number of memberships was 1.15 at baseline and reduced to 0.96 over the same period. The average respondent participated in organizational activities bimonthly and met with familiar persons twice a month.

Regarding time-varying controls, respondents experienced deteriorating health over the study period: mean scores for medical comorbidities, limitations in Activities of Daily Living, and depressive symptoms increased. Regarding health behaviors, smokers in the sample decreased from 19 percent to 9 percent, while drinkers increased from 5 percent to 19 percent. Those who regularly exercised dropped from 41 percent to 33 percent. The portion of the married diminished from 83 percent to 74 percent. Household income increased over time, although the proportion of working older adults dropped from 44 percent to 34 percent. The percentage of urban residents reduced slightly from 79 to 74 percent.

Time-invariant controls at baseline indicate that, on average, the respondents were 59 years old, and 58 percent were women. The average respondents had completed middle school. About 56 percent of respondents had a religion, with Buddhists (26 percent) being the largest religious affiliation, followed by Protestants (20 percent) and Catholics (9 percent).

Table 2 shows the ML-SEM standardized coefficients of social engagement and other covariates that predicted handgrip strength over time. Models 1–3 employed the three indicators of formal and informal social engagement one at a time. Models 4–5 simultaneously examined the effects of formal and informal social engagement.

The main results are as follows: first, the lagged effect of handgrip strength, measured in previous waves, was the strongest time-varying predictor of current handgrip strength across models. Second, the two indicators of formal social engagement (i.e., number of association memberships [0.01*] and frequency of organizational activities [0.02***]) had positive and significant associations with handgrip strength when considered separately in models 1 and 2. They retained their significance when informal engagement (i.e., contact frequency with familiar persons) was considered together in models 4 and 5. Third, models 1–5 showed that formal and informal engagement indicators had significant contemporaneous associations with handgrip strength, except for the frequency of organizational activities. In particular, contact frequency with familiar persons formed a significant proximal association with handgrip strength, although it lacked a cross-lagged effect. These models confirm that formal social engagement had a significant cross-lagged relationship with handgrip strength over time, accounting for other confounders.

Time-varying controls of medical comorbidities and depressive symptoms were associated with weaker handgrip strength. Among time-invariant controls, younger age, male, and higher educational level were positively related to handgrip strength. Regarding religious affiliation, Protestants had lower handgrip strength than those without religion.

Model fit indices were satisfactory across models, with the CFI and TLI close to 1 and the RMSEA and SRMR close to 0. The ML-SEMs explained 69 % of the handgrip strength variance.

Next, Table 3 examined the effect of handgrip strength on the three indicators of social engagement. In line with what Table 2 showed, the standardized estimates of the autoregressive impact of social engagement on itself were strongest, ranging from 0.22*** to 0.28***. The cross-lagged effect of handgrip strength was significant for both formal engagement variables (i.e., number of association memberships [0.03*] and frequency of organizational activities [0.04**]). However, handgrip strength failed to make a positive association with informal engagement over time. In contrast, handgrip strength was proximally associated with both formal and informal social engagement. These results confirm that formal social engagement and handgrip strength form a reciprocal association.

Figure 3 presents the reciprocal relationship between formal social engagement and handgrip strength using unstandardized coefficients, accounting for all other covariates in Tables 2 and 3. The maximum number of association memberships or the highest frequency of organizational activities enabled an older adult to keep handgrip strength about 0.5–0.9 kg higher than that for those with no formal engagement. On the other hand, handgrip strength also helped increase the number of association memberships and the frequency of organizational activities somewhat over 2 years.

Concerning time-varying controls, limitations in Activities of Daily Living reduced the frequency of organizational activities and contact with familiar persons. Medical comorbidities diminished only contact frequency with familiar persons. The married had a lower frequency of organizational activities and contact with familiar persons. This result indicates that those who lacked a marital tie were more active in organizational and interpersonal engagement.

Time-invariant controls indicate that being a woman was significantly related to all three social engagement variables. Education was positively associated with the number of association memberships and the frequency of organizational activities. Being younger was related to a higher number of association memberships. Having a mainstream religious affiliation was associated with social engagement. In particular, being a Protestant was consistently related to all three social engagement indicators.

The models with formal engagement as the outcome (models 1 and 2) showed a better model fit than the model that predicted informal engagement (model 3). The models explained about 40 percent of the variance in social engagement variables. Overall, ML-SEMs reported better model fit to the data when they predicted handgrip strength compared to when social engagement was the outcome.

The reciprocal associations between formal social engagement and handgrip strength remained consistent when a sensitivity analysis employed relative handgrip strength that considered the respondent’s body weight and height (online suppl. Tables 1s, 2s; for all online suppl. material, see https://doi.org/10.1159/000540344). Membership and activities in leisure/culture/sports groups may have inflated the effect of formal social engagement on handgrip strength because these groups were naturally conducive to physical activity. Thus, we dropped this associational type and activities from the formal engagement variables and reestimated the ML-SEMs (online suppl. Tables 3s, 4s). Despite the change, the reciprocal associations between formal engagement and handgrip strength remained significant. Therefore, the supplementary analysis suggests that various types of formal social engagement, not limited to a particular type of sports activity, were positively associated with handgrip strength.

Handgrip strength is an indicator of the overall health status of older adults. In particular, it is a measure of sarcopenia, a condition denoting skeletal muscle mass loss. The condition is associated with lower quality of life, increased functional disability, physical frailty, obesity, resistance to insulin, type 2 diabetes, dyslipidemia, hypertension, cognitive decline, and higher all-cause mortality [10, 19, 47, 48]. Even with normal body weight, older adults may lose lean body mass but gain body fat due to a dormant lifestyle, which results in sarcopenic obesity [49]. The study identified that formal social engagement helped older adults maintain handgrip strength over time. This may be because formal organizations offer an active lifestyle to older adults, facilitating their regular and routinized involvement in social environments [17, 50]. Inversely, handgrip strength also increased the degree of formal social engagement.

An intervention study found that prefrail and frail older adults assigned to the home-based physical training and nutrition group increased their handgrip strength by 2.4 kg on average over 12 weeks [51]. The physical training program executed with buddy volunteers included mini squats in front of a chair, chest presses against elastic resistance, or reverse butterfly, and shoulder presses against elastic resistance, all targeted to increase handgrip strength. Its nutritional intervention also aimed to grow handgrip strength by increasing the intake of fluid, protein, and energy. Compared to such goal-oriented interventions in short periods [52], formal social engagement may provide noninvasive preservation of muscle strength by routinizing various physical and leisure activities and healthy diets in the daily lives of older adults. The cross-lagged association between formal social engagement and handgrip strength remained robust even when we omitted the membership and participation in leisure/culture/sports groups from the measures of formal engagement. Therefore, the influence of formal engagement may largely hinge on the active lifestyle it establishes for older adults based on diverse types of associations rather than on a particular type of group or activity that is believed to be closely related to handgrip strength.

Apart from and beyond the potential association between health behaviors – exercise, nonsmoking, and nondrinking – and handgrip strength, formal social engagement may be related to handgrip strength. A socially engaged lifestyle offers older adults ample opportunities to assume meaningful social roles, cultivate a sense of attachment, and remain physically and cognitively intact [15]. These social engagement benefits stemming from a sustained active lifestyle may pave a physiologic pathway in the long run (e.g., balancing hypothalamic-pituitary-adrenal axis response, allostatic load, immune system function, or cardiovascular reactivity) to handgrip strength. Treading this pathway to handgrip strength may not be what older adults had in mind when regularly engaging with various associations and their activities. Rather, the particular health benefit may often be an unintended consequence of formal engagement.

The present study found that informal social engagement failed to form a significant relationship with handgrip strength in either direction. It is plausible that informal engagement – social contacts with familiar others – falls short of creating a critical impact on handgrip strength due to its irregular nature with low intensity. In part, this may imply that older adults should go beyond a small circle of familiar people around them and get involved in an organized lifestyle to realize the benefit of social engagement.

The study contributes to the literature because the reciprocal relationship between handgrip strength and social engagement has been underexplored. Zhao et al. [53] found that handgrip strength was associated with social engagement among older Chinese adults. Nevertheless, they could not check the reciprocal relationship because they used cross-sectional data. Additionally, their measure of social engagement did not differentiate formal and informal engagement. Luo et al. [37] identified a reciprocal association between handgrip strength and depression but did not examine a plausible relationship between handgrip strength and social engagement.

In sum, a reciprocal relationship existed between formal social engagement and handgrip strength of older Korean adults, which supports both social causation and health selection perspectives [54]. The social causation perspective presumes that a socially active lifestyle helps preserve handgrip strength, an overall health indicator. On the other hand, the health selection perspective deduces that handgrip strength is associated with a higher level of social engagement.

The study found no significant relationship between informal social engagement and handgrip strength. However, this may be due to the weak measure of informal engagement. A single variable of contact frequency with familiar people may be inadequate to consider other important aspects of informal engagement, such as network size, psychological closeness with social ties, or the provision of social support [55].

The study, relying on the conceptual model by Berkman et al. [15], suggested that formal social engagement is likely to be associated with handgrip strength via a physiologic downstream pathway. However, the study could not test the mediatory path due to the lack of relevant data. Thus, future studies may examine the mediatory association between social engagement, physiological biomarkers, and health outcomes, including handgrip strength.

Among the health behavior measures, physical exercise lacked specificity by asking whether a respondent exercised regularly at least once a week. Thus, the measure was insufficient to capture longitudinal changes in handgrip strength. This finding aligns with a recent meta-analysis review, reporting that only five out of twenty-four physical exercise trials administered on community-dwelling older adults marginally increased handgrip strength compared to control groups [52]. The study concluded that only a particular type of task-specific training may be positively associated with handgrip strength. In other words, general physical exercise may not easily affect handgrip strength.

Also, it is uncertain if the reciprocal association between formal social engagement and handgrip strength applies to other countries because the present study used data from a single country. Thus, future studies are warranted to explore whether the reciprocal relationship exists in different national and regional contexts.

Handgrip strength is a convenient and effective indicator of overall physical health status. The present study demonstrated that a socially active lifestyle helps preserve handgrip strength.

Specifically, formal social engagement in and through diverse associations was positively associated with handgrip strength. This is mainly because formal engagement with various organizations allows older adults to maintain an active lifestyle, encouraging them to interact regularly with their social environment [56]. This active lifestyle embedded in various formal associations may increase social and physical activities that, in turn, may help retain handgrip strength. On the other hand, older adults with greater handgrip strength tend to reinforce their formal social engagement even further, establishing a reciprocal relationship between formal social engagement and handgrip strength.

The Institutional Review Board of Statistics Korea reviewed and approved the Korean Longitudinal Study of Aging (KLoSA) survey questionnaire (Approval No. 336,052). Written informed consent was obtained from all of the survey respondents.

The authors have no conflicts of interest to declare.

This work was not supported by any research grant.

Joonmo Son conceived the research questions and study design, interpreted the statistical results, wrote the first draft of the manuscript, and revised the manuscript. Pildoo Sung managed the longitudinal dataset, conducted the statistical analysis, wrote the data and measures sections, and critically reviewed the manuscript.

The secondary data employed by this empirical study are publicly available at https://survey.keis.or.kr/eng/klosa/klosa01.jsp.