Social engagement pattern, health behaviors and subjective well-being of older adults: an international perspective using WHO-SAGE survey data

This study used nationally representative data from the World Health Organization’s Study on Global Ageing and Adult Health (WHO SAGE) Wave 1, which was carried out in six LMICs (China, Russia, India, Ghana, South Africa and Mexico) from 2007 to 2010. Although SAGE is a longitudinal study and a baseline survey (SAGE Wave 0) was created during 2002–2004, only Wave 1 data were used in this study because China and South Africa did not follow up Wave 0 respondents for Wave 1. China, Russia, India, Ghana and South Africa all used stratified multi-stage clustered sampling strategy, and Mexico included extra supplementary and replacement samples in Wave 1 to account for losses to follow up since Wave 0 [15]. SAGE was approved by the Ethics Review Committee of the World Health Organization.

The 9 social engagement questions asked of participants in all six countries are listed in Appendix 1. These questions asked about the frequency of several social activities in the previous 12 months, which had been used before and reported elsewhere [16]. The response options were rated on a 5-point Likert scale ranging from 1 (never) to 5 (daily). Scores from these questions were summed and divided by 9 to derive a mean social engagement score for India, Ghana, South Africa and Mexico. Due to the low prevalence of religious practices in China and Russia, only 8 items (excluding the religious service question) were used for these two countries. For China and Russia, the summed score from the 8 questions was divided by 8, the total number of questions used in these two countries. Social engagement was used as a continuous variable in this study, with higher scores representing higher levels of social engagement. The Cronbach’s alpha internal consistency coefficients for these groups of social engagement items ranged from 0.63 to 0.81 across six countries. (See more details in Appendix 1).

Table 2 shows the mean score of social engagement across six countries stratified by sociodemographic variables. Ghana had the highest total mean score, while Mexico had the lowest. Males, those who are currently married or cohabiting, those living in rural areas, those younger and better educated were more likely to have higher social engagement scores. Those with higher income levels also reported higher social engagement, except for Ghana, where the middle-income quintile group had the highest level of social engagement, and South Africa, where the second-income quintile group had the highest mean score. People who were still working had a higher mean score than those who were not working in all six countries.

The logistic regression results are presented in Appendix 2 (unadjusted model) and Table 3 (adjusted model). Appendix 2 presents the univariate logistic regression results examining the relationship between social engagement and six unhealthy lifestyle behaviors as well as subjective well-being across six countries. The results changed slightly after adjusting for sex, age group, marital status, education level, place of residence, income quintiles and working status (see Table 3). In China, higher level of social engagement was associated with higher risk of smoking, while in Ghana and South Africa, an inverse relationship was observed. For India, the relationship was no longer significant after controlling for the sociodemographic characteristics.

Social engagement was not associated with harmful drinking in any of these six countries. And only in Ghana, high social engagement had a positive relationship with low fruit and vegetable intake, while an inverse relationship was found in China, Russia, and India.

Social engagement was found to have an inverse relationship with physical inactivity across all six countries. It was also found that social engagement was inversely associated with prolonged sitting time in Ghana and South Africa, while no significant relationship in China, Russia, India, and Mexico. Social engagement was inversely associated with the risk of unhealthy sleep duration in China and Ghana only.

There was strong evidence for an association between high levels of social engagement and poor self-rated health and low quality of life in all six countries, as well as perceived depression (not significant in India, Ghana and South Africa) even after adjusting for the other sociodemographic variables.

Our study showed that, in China, higher levels of social engagement were associated with higher risk of smoking, while an inverse association was found in Ghana and South Africa. The results from China are contrary to those found in other studies. Results from the longitudinal British Household Panel Survey (BHPS), showed that active social participation was positively associated with smoking cessation [11]. In a US Study, Samuel et al. also reported that emotional social support and neighborhood social cohesion were generally linked to lower smoking rates [7]. However, in the same paper, the authors pointed out that social support and social cohesion may be associated with higher smoking rate in groups with high rates of smoking, which is in line with previous research [25‐27] and the current study. Past research has shown that the greater the social network size, the more likely a person is to smoke, especially in Asian cultures where collectivism is valued [28]. China consumes about 40% of the world’s cigarettes, and the prevalence of smoking remained high in men (54.0% prevalence for current-smoking) [29]. In many Asian countries [30], smoking with others is seen as a way to foster relationships between family members, peers, and business associates [31]. Tobacco can also be exchanged as “social currency” for social opportunities and benefits, which permeates every aspect of family life and wider social interactions [32, 33]. As there were no widespread anti-smoking public health program implemented in these countries, people view smoking in social activities as acceptable or even desirable, rather than a behavior to be reprimanded or punished [33]. This might explain the positive association between social engagement and smoking in China.

Social engagement was not found to be associated with alcohol consumption across the six LMICs, which is different from that reported in some HICs, where highest levels of alcohol consumption were noted [34]. Research in the United States [35] and Australia [14] found that social drinking was part of the social fabric among older peoples and retirees. Many participants in HICs emphasized the social nature of their alcohol use, as something to be enjoyed with others. During social activities, people may be less aware of how much they were drinking, and thus consume more than they would in non-social situations [14].

In our study, low levels of social engagement were associated with low fruit and vegetable consumption in China, Russia and India, but an inverse association was found in Ghana. This is a novel finding to our knowledge. A healthy diet is important to health and well-being at all stages in life; however, the determinants of dietary health change with age. Older adults are prone to developing an unhealthy dietary pattern for many reasons, including reduced mobility and/or fewer financial resources to spend on food [6, 7]. Moreover, socially isolated older adults are at a greater risk of poor dietary behavior because of the lack of social support [36]. Jones et al. found that getting out of the house and being active were effective in stimulating appetite [37], and eating with others has been shown to increase food intake by 60% in healthy older adults aged 65 and over [38]. Holmes also suggested that those who usually eat alone at home may substitute a cooked or hot meal with convenient food that can be easily accessed and prepared, such as sandwiches [38]. There was no evidence in the literature that explained why low social engagement was inversely associated with low fruit and vegetable intake in Ghana. One possible reason might be that when eating out, people usually have a high intake of meat products and beverage instead of fruit and vegetable [39, 40]. To fully understand this question, more culturally specific information about social eating behavior and the cultural context of food in Ghana is required.

Our study showed that social engagement was inversely related to physical inactivity, prolonged sitting time and unhealthy sleep duration. These results are similar across the six countries and consistent with previous research. In the study by Samuel et al. [7], social support and neighborhood social cohesion were associated with achieving the recommended level of physical activity. Hiroyuki and colleagues [8] had similar findings among a Japanese sample, where men and women reporting higher social participation were less likely to be physically inactive. The possible mechanism could be that social engagement provides older people with more opportunities and social reasons to go outside to join physical activities [8].

According to ecological models [41], the physical and social environment in which people live are important determinants of sedentary behavior. In our study, we found that social engagement was inversely associated with prolonged sitting time in Ghana and South Africa, while no relationship was found in China, Russia, India and Mexico. Hiroyuki’s study [8] showed that higher levels of social participation was associated with less sedentary time, such as television watching, sitting around and listening or talking while sitting. But Van Holle et al. [10] found no evidence for the association between social isolation and talking with neighbors, with sedentary time. One possible explanation for the inconsistent results across countries might be the unmeasured domains of sedentary behaviors, for example mentally-active sedentary behavior (e.g. reading books). It is important to distinguish between different types of sedentary behaviors instead of considering them as one entity because social participation pattern may vary across domains of sedentary activities. Additionally, these results may also be partly influenced by age and working status, as well as whether the sitting time was self-reported or objectively measured, which requires further study.

Our study showed that people with higher levels of social engagement were less likely to be outside the optimal sleep range. This might be because a lack of social contacts in older people’s lives may result in the flattening of their circadian rhythm and reduced needs for sleep in the evening [42]. Thida et al. [43] found that having lower neighborhood social capital was associated with insufficient sleep among Japanese adults, particularly in the men. Tarja and colleagues [44] found that people with high levels of social support were more likely to have adequate duration of sleep. Such findings from observational studies have been confirmed by a cluster randomized trial by Joachim and colleagues among nursing home residents, where social activity sessions, including parlor games and group discussions, improved subjective sleep quality of the participants at both clinical and statistical significance levels [45].

In our study, higher level of social engagement was consistently associated with less perceived depression, better self-rated health and higher quality of life. A Japanese longitudinal study showed that social engagement improves older people’s mental health, including depressive symptoms and psychological distress. Results from the 4^th National Household Health Survey [46] also found that social contacts were positively associated with quality of life among Chinese older adults in urban areas. Snorri [47] observed that social network size and contact frequency were positively and independently related to future subjective well-being in English adults aged 50 and older. Findings from the National Social Life, Health, and Aging Project [48] among older Americans confirmed that networks with a wider range of social ties were related to better well-being, independent of demographic and health characteristics.

Social engagement could increase people’s social networks, which leads to attachment, esteem, social approval, belongingness, social identity and increasing access to social support [49]. In particular, social network could provide access to functional support and assistance from family members, neighbors and friends. This type of support is important and have been shown to be related to older adults’ improved sense of control, enhanced quality of life and wellbeing [47].

Overall, this international study found high level of social engagement as a consistent correlate of health in some LMICs. Regarding subjective well-being, social engagement appears to be protective of perceived depression, poor self-rated health and low quality of life. However, for lifestyle risk behaviors, the associations varied by the outcome and the country, possibly as a result of the different ways people socialize. It is also important to note that as a cross-sectional study, the relationships between social engagement, lifestyle behaviors and subjective well-being could be bidirectional, that is either social engagement promotes healthier behaviors and mental health or healthy behaviors and subjective well-being lead to higher levels of social engagement. Additionally, across countries, social engagement was associated with almost all outcomes in Ghana and China, but less so in other countries. This suggests that the relationship between social engagement and lifestyle behavior is complicated and culture-specific. Future research should focus on examining various cultural elements when measuring both social engagement and health behaviors.

Strengths of this study include comparable data from six LMICs with nationally representative samples. This paper is among the first to examine a broad range of lifestyle behaviors and subjective well-being, in relation to social engagement with a large sample size across LMICs. However, some limitations should be noted. First, a cross-sectional design limits causal inferences. Second, self-reported measures are subject to reporting bias. Third, some unknown confounders, such as BMI and existing chronic disease, were not collected and therefore could not be adjusted for. Finally, it is arguable that there might be differences in the interpretation of social engagement questions across countries because of the inherent differences in cultural and social norms. Thus, these societal aspects are worthy of further exploration to help explain our findings.