Research has shown that obesity appears to spread through social ties. However, the association between other characteristics of social networks and obesity is unclear. This study aimed to identify the association between social network characteristics and body mass index (BMI, kg/m2) in an elderly Korean population.
This cross-sectional study analyzed data from 657 Koreans (273 men, 384 women) aged 60 years or older who participated in the Korean Social Life, Health, and Aging Project. Network size is a count of the number of friends. Density of communication network is the number of connections in the social network reported as a fraction of the total links possible in the personal (ego-centric) network. Average frequency of communication (or meeting) measures how often network members communicate (or meet) each other. The association of each social network measure with BMI was investigated by multiple linear regression analysis.
After adjusting for potential confounders, the men with lower density (<0.71) and higher network size (4-6) had the higher BMI (β=1.089, p=0.037) compared to the men with higher density (>0.83) and lower size (1-2), but not in the women (p=0.393). The lowest tertile of communication frequency was associated with higher BMI in the women (β=0.885, p=0.049), but not in the men (p=0.140).
The study of the effects of social networks on health rose to the surface in the 1970s through pioneers such as Cassel, Cobb, and Berkman, who theorized or proved empirically that social networks could affect mortality [1-6]. After those mortality studies, several studies have reported that social networks are related to infectious diseases, such as sexually transmitted disease [7-9], tuberculosis , severe acute respiratory syndrome , and pneumonia . Additionally, evidence suggests that emotions and related behaviors such as depression , suicide , and happiness  are also associated with social networks. Social networks affect health through several mechanisms, including the provision of social support (both perceived and actual), social influence (e.g., norms, social control), social engagement, person-to-person contact (e.g., pathogen exposure, secondhand cigarette smoke), and access to resources (e.g., money, jobs, information) .
Obesity is a well-known risk factor for cardiovascular diseases . Adipose tissue has been recognized as an active endocrine organ, capable of releasing many cytokines that play parts in the pathogenesis of many obesity-related diseases . Social network assessment over 32 years in the Framingham Heart Study found that obesity tended to spread through intimate friends . It found that an individual chances of being obese increased by 57% if she or he had a friend who also be obese. Additionally, the evidence for influence of friends on body weight comes from other studies examining the relationships between the weight of individuals and their friends [19-23]. These studies tested the relationship between the body mass index (BMI, kg/m2) of individuals and their friends in order to address the peer effect.
It is known that obesity appears to spread through social ties. However, the association between other characteristics of social networks and obesity is unclear. Thus, we investigated whether social network characteristics such as network size, density of communication network, average frequency of communication, and average frequency of meeting were associated with BMI in an elderly Korean population.
The Korean Social Life, Health, and Aging Project (KSHAP) conducted a social network survey and health examination from November 2011 through July 2012 among 860 community-dwelling adults aged 60 or older and their spouses living in Yangsa-myeon, Ganghwa-gun, Incheon, Korea. The institutional review board of Yonsei University approved this study (YUIRB-2011-012-01) and informed consent was obtained from all of the participants. A total of 814 people responded to the social network survey (response rate 94.7%). Potentially confounding factors were assessed as well, including age, education, smoking status (never or ever smoker), alcohol drinking status (non-drinker or drinker), depression score (Center for Epidemiologic Studies-Depression Scale, CES-D), and self-reported comorbidity (hypertension, diabetes, stroke, and arthritis). Among them, 657 participants (80.7%) were examined for height and weight in a public health center or at home. BMI was calculated from measured weight and height.
In order to collect social network data, KSHAP adopted a model similar to that of the National Social Life, Health, and Aging Project (NSHAP) . On 'Social network card I' (Appendix 1), a respondent (ego) was asked questions about three types of alters: their spouse (roster A), a maximum of five people with whom the respondents discussed things that were important to them (roster B), and someone to whom the respondents felt especially close (roster C). On 'Social network card II' (Appendix 2), the respondent answered questions about the relationships among alters listed on the rosters, and this information was used to build up egocentric network variables. The difference between egocentric data and common survey questions is that egocentric questionnaires collect data on the characteristics of the persons named, that is, on the respondent's personal network. The additional network information provides some insight into personal network characteristics and their potential influence on behavior .
Social network characteristics are composed of network structure (network size, density, etc.) and network activation (frequency of nonvisual contact, frequency of face-to-face interaction, etc.) [16,26]. The following are the four social network measures we used.
The egocentric network was composed of rosters A and B. Network size was simply a count of the number of alters provided in response to questions on 'Social network card I.'
Density of Communication Network
KSHAP data can be used to construct a personal (ego-centric) network density variable that reflects the extent to which someone's closest contacts are connected to one another. Density of communication network (or density) is calculated as:
Node nj and nk represent the adjoined nodes connected to node ni and 'e' represents the number of adjoined nodes connected to ni. If node nj and nk are connected, d(nj, nk) becomes 1 and if not, it becomes 0. After summing and dividing by every possible number of connections between adjoined nodes, this becomes the density of the communication network. Figure 1 shows examples of various personal network densities.
Average Frequency of Communication and Meeting
Respondents were asked to rate how often they talked to each network member on an 8-point scale, ranging from 'everyday' to 'less than once per year'. The scores were coded by assigning the approximate number of times per year egos interacted with alters (e.g., 'everyday'=365; 'once a month'=12) and the scores were summed across all alters on roster A and roster B to obtain a measure of the overall volume of contact with network members. Then, the scores were summed and divided by the network size. The average frequency of communication (or "communication frequency") indicates how often network members communicated. The average frequency of meeting (or "meeting frequency") was likewise calculated to determine how often network members met.
All analyses were performed separately for the men and women. To gain information on the distribution of the continuous variables (age, depression score, BMI, network size, density of communication network, and average frequency of communication and meeting), we drew a histogram and calculated the minimum, lower quartile, median, upper quartile, maximum, and mean values. In order to solve the problem of non-normal distribution, we used two approaches before multiple linear regression analyses. One was the rank transformation approach and the other was the tertile method. The Spearman correlation test was used to examine the possibility of multicollinearity among social network measures. Network size and density were strongly and inversely correlated in the men (Spearman correlation coefficient (rs)=-0.98, p<0.001) and women (rs=-0.97, p<0.001). We created three groups of social network structure: 'higher density (>0.83) and lower size (1-2)', 'middle density (0.71-0.83) and size (3)', and 'lower density (<0.71) and higher size (4-6)' (Appendix 3). Before the regression analysis, unclassified individuals (3 men and 8 women) were removed. For univariate analysis, the trend test was used. For the continuous variables, p-values were calculated by using a contrast to test for a linear trend. For the categorical variables, the Cochran-Armitage test was used. Multiple linear regression analyses were performed to assess the linear relationship between each of the social network measures and BMI. In the matter of network activation, communication frequency was more significant than meeting frequency from the aforementioned analyses. Lastly, both the group of social network structure (density and size) and communication frequency were included in the multiple regression model. All statistical tests were performed with SAS version 9.2 (SAS Inc., Cary, NC, USA). All analyses were two-sided and p-values <0.05 were regarded as statistically significant.
The characteristics of the study population are summarized in Table 1. We assessed the linear trends between the social network measures and BMI (Appendices 4-6). The group of social network structure and communication frequency were significantly associated with BMI in both the men and women. However, meeting frequency was only significant in the women.
Multiple linear regression analyses were performed to assess the relationship between each social network measure and BMI (Table 2). As regards network structure, rank-transformed network size and density were only associated with the men's BMI in model 3, while the group of network structure was significantly associated with BMI among all of the models in the men. The men with lower density and higher size had a higher BMI (β=1.253, p=0.014) than the men with higher density and lower size, while the men with middle density and size also had a higher BMI (β=1.022, p=0.044) compared with the reference group. We did not observe a significant association between network structure and BMI in the women. As regards network activation, communication frequency was significantly associated with the women's BMI among all of the models. The association between rank-transformed meeting frequency and the women's BMI was significant, but the tertile of meeting frequency was only associated in model 3. We did not observe such an association between network activation and BMI in the men.
The group of social network structure and the communication frequency as an index of network activation were further analyzed for the association with BMI in the men and women (Table 3). After adjusting for age, education, smoking, alcohol drinking, depression score, comorbidity (hypertension, diabetes, stroke, and arthritis), and communication frequency, the men with lower density (<0.71) and higher size (4-6) had a higher BMI (β=1.089, p=0.037) than the men with higher density (>0.83) and lower size (1-2), but this was not the case in the women (β=0.391, p=0.393). In contrast, after adjusting for potential confounders and social network structure, the lower tertile of communication frequency was associated with a higher BMI in the women (β=0.885, p=0.049), but not in the men (β=0.702, p=0.140). In the women, less education was prominently related with a higher BMI among potential confounders.
The present study examined the association between social network measures and BMI in an elderly Korean population. The men with a sparse and large communication network had a higher BMI compared with the men embedded in dense and small communication networks. The women in networks with a low frequency of communication had a higher BMI than the women in networks with a high frequency of communication.
Social contacts can promote, discourage, and sanction attitudes and behaviors [27,28]. Social contacts may influence participation in organized sports , dieting , and food choices . A study has linked unhealthy weight-control behaviors among adolescent girls to the dieting behaviors of their peers . We found that the men's BMI was more strongly associated with network structure, that is, network size and density, than communication frequency as an index of network activation. KSHAP data from a restricted rural area had a simpler network structure than the NSHAP data because the NSHAP study population was a nationally representative sample . Almost the entire KSHAP study population existed in three of the nine possible density and size groups. Therefore, it was difficult to discriminate statistically the effect of density from the association between social network structure and BMI, because network size and density were strongly correlated. We postulate a possible explanation for the association between network structure and the men's BMI. Dense personal networks provide reinforcement for prevailing norms and practices and can provide protection from outside sources of influence or risk . The greater the density, the more likely a network is to be considered a cohesive community, a source of social support, and an effective transmitter . By contrast, the women's BMI was more strongly associated with network activation such as communication frequency than network structure. In the women, education was prominently related with BMI among potential confounders, such that more advanced education correlated with lower BMI. The well-educated have a healthier lifestyle . They are more likely to exercise, less likely to smoke, and more likely to drink moderately rather than abstain or drink heavily. It could be inferred that individual attributes such as education level may be most important in determining women's BMI. A previous study explained possible reasons for gender differences in factors associated with BMI . Social contact among friends may be qualitatively different for the men and women in relation to how they affect health behavior. Although not statistically significant in interaction terms (social network measures by gender), β-coefficient estimates that are markedly different from each other suggest the possibility of true interaction between gender and social network characteristics .
Previous studies have been mainly interested in the peer effect on obesity. There have been few reports about the association between other characteristics of social networks and obesity. Our findings were not definite, but might illuminate the novel association between social network structure or activation and BMI.
This study has several limitations. First, because of the cross-sectional design, this study could not establish a temporal relationship between BMI and social network characteristics. Second, the survey data were open to measurement error. In particular, the reported relationships among alters might differ from actual relationships. It is important to determine not only the association between an ego's health behavior and an actual social network, but the perceived social network as well. Third, we assumed an inverse linear relationship between BMI and health. However, an increased risk of death has been observed to be associated with a low BMI in other studies [37,38]. Inadequate or incomplete control for confounding or reverse-causation bias could in part explain the increased risk . A low BMI can be an indicator of certain other chronic medical conditions . We adjusted for comorbidities, such as hypertension, diabetes, stroke, and arthritis, in order to overcome this unintended effect. This study population was dwelling in the community, not institutions, and they completely responded to the long survey (mean response time: 70 minutes). It can thus be concluded that they were not unhealthy enough to induce a reverse-causation bias. Fourth, the questionnaires did not measure diet and physical activity. These factors are determinants for obesity, and taking them into account may be necessary to identify the behavioral process (indirect effect) of social networks on BMI . We could not distinguish behavioral processes from psychological processes in this study. Fifth, external validity could be limited because all of the subjects in the study population resided in a single rural community. Accordingly, further longitudinal studies in urban and rural areas are needed.
Our study suggests that social network structure (network size and density) and activation (communication frequency and meeting frequency) are associated with obesity among the elderly. Lower density of communication network and higher network size may be associated with higher BMI in men, while lower frequency of communication may be associated with higher BMI in women.
National Research Foundation of KoreaNRF-2011-330-B00137
This study was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF-2011-330-B00137).
CONFLICT OF INTEREST
The authors have no conflicts of interest with the material presented in this paper.
1. Cassel J. The contribution of the social environment to host resistance: the Fourth Wade Hampton Frost Lecture. Am J Epidemiol 1976;104(2):107-123 782233.
2. Cobb S. Presidential Address-1976. Social support as a moderator of life stress. Psychosom Med 1976;38(5):300-314 981490.
3. Berkman LF, Syme SL. Social networks, host resistance, and mortality: a nine-year follow-up study of Alameda County residents. Am J Epidemiol 1979;109(2):186-204 425958.
4. Blazer DG. Social support and mortality in an elderly community population. Am J Epidemiol 1982;115(5):684-694 7081200.
5. House JS, Robbins C, Metzner HL. The association of social relationships and activities with mortality: prospective evidence from the Tecumseh Community Health Study. Am J Epidemiol 1982;116(1):123-140 7102648.
7. Helleringer S, Kohler HP. Sexual network structure and the spread of HIV in Africa: evidence from Likoma Island, Malawi. AIDS 2007;21(17):2323-2332 18090281.
8. Laumann EO, Youm Y. Racial/ethnic group differences in the prevalence of sexually transmitted diseases in the United States: a network explanation. Sex Transm Dis 1999;26(5):250-261 10333277.
9. Liljeros F, Edling CR, Amaral LA, Stanley HE, Aberg Y. The web of human sexual contacts. Nature 2001;411(6840):907-908 11418846.
10. Klovdahl AS, Graviss EA, Yaganehdoost A, Ross MW, Wanger A, Adams GJ, et al. Networks and tuberculosis: an undetected community outbreak involving public places. Soc Sci Med 2001;52(5):681-694 11218173.
11. Mettikolla P, Calander N, Luchowski R, Gryczynski I, Gryczynski Z, Zhao J, et al. Cross-bridge kinetics in myofibrils containing familial hypertrophic cardiomyopathy R58Q mutation in the regulatory light chain of myosin. J Theor Biol 2011;284(1):71-81 21723297.
12. Ancel Meyers L, Newman ME, Martin M, Schrag S. Applying network theory to epidemics: control measures for Mycoplasma pneumoniae outbreaks. Emerg Infect Dis 2003;9: 204-210 12603991.
13. Rosenquist JN, Fowler JH, Christakis NA. Social network determinants of depression. Mol Psychiatry 2011;16(3):273-281 20231839.
14. Bearman PS, Moody J. Suicide and friendships among American adolescents. Am J Public Health 2004;94(1):89-95 14713704.
15. Fowler JH, Christakis NA. Dynamic spread of happiness in a large social network: longitudinal analysis over 20 years in the Framingham Heart Study. BMJ 2008;337: a2338. 19056788.
16. Berkman LF, Kawachi I. Social epidemiology. New York: Oxford University Press; 2000. p. 145-149.
17. Van Gaal LF, Mertens IL, De Block CE. Mechanisms linking obesity with cardiovascular disease. Nature 2006;444(7121):875-880 17167476.
18. Christakis NA, Fowler JH. The spread of obesity in a large social network over 32 years. N Engl J Med 2007;357(4):370-379 17652652.
19. Fowler JH, Christakis NA. Estimating peer effects on health in social networks: a response to Cohen-Cole and Fletcher; and Trogdon, Nonnemaker, and Pais. J Health Econ 2008;27(5):1400-1405 18692263.
20. Halliday TJ, Kwak S. Weight gain in adolescents and their peers. Econ Hum Biol 2009;7(2):181-190 19497795.
21. Renna F, Grafova IB, Thakur N. The effect of friends on adolescent body weight. Econ Hum Biol 2008;6(3):377-387 18672412.
22. Trogdon JG, Nonnemaker J, Pais J. Peer effects in adolescent overweight. J Health Econ 2008;27(5):1388-1399 18565605.
23. Valente TW, Fujimoto K, Chou CP, Spruijt-Metz D. Adolescent affiliations and adiposity: a social network analysis of friendships and obesity. J Adolesc Health 2009;45(2):202-204 19628148.
24. Cornwell B, Schumm LP, Laumann EO, Graber J. Social Networks in the NSHAP Study: rationale, measurement, and preliminary findings. J Gerontol B Psychol Sci Soc Sci 2009;64(Suppl 1):i47-i55 19502574.
25. Valente TW. Social networks and health: models, methods, and applications. New York: Oxford University Press; 2010. p. 44-45 p. 69-70.
26. Uchino BN. Social support and health: a review of physiological processes potentially underlying links to disease outcomes. J Behav Med 2006;29(4):377-387 16758315.
27. Bahr DB, Browning RC, Wyatt HR, Hill JO. Exploiting social networks to mitigate the obesity epidemic. Obesity (Silver Spring) 2009;17(4):723-728 19148124.
28. Schlundt DG, Hill JO, Sbrocco T, Pope-Cordle J, Kasser T. Obesity: a biogenetic or biobehavioral problem. Int J Obes 1990;14(9):815-828 2228413.
29. Kohl HW 3rd, Hobbs KE. Development of physical activity behaviors among children and adolescents. Pediatrics 1998;101(3 Pt 2):549-554 12224661.
30. Haines J, Neumark-Sztainer D. Prevention of obesity and eating disorders: a consideration of shared risk factors. Health Educ Res 2006;21(6):770-782 16963727.
31. Cullen KW, Klesges LM, Sherwood NE, Baranowski T, Beech B, Pratt C, et al. Measurement characteristics of diet-related psychosocial questionnaires among African-American parents and their 8- to 10-year-old daughters: results from the Girls' health Enrichment Multi-site Studies. Prev Med 2004;38: S34-S42 15072857.
32. Eisenberg ME, Neumark-Sztainer D, Story M, Perry C. The role of social norms and friends' influences on unhealthy weight-control behaviors among adolescent girls. Soc Sci Med 2005;60(6):1165-1173 15626514.
33. Kadushin C. Understanding social networks: theories, concepts, and findings. New York: Oxford University Press; 2012. p. 29-30.
35. Oliveira AJ, Rostila M, de Leon AP, Lopes CS. The influence of social relationships on obesity: sex differences in a longitudinal study. Obesity (Silver Spring) 2013;21(8):1540-1547 23818388.
36. Szklo M, Nieto FJ. Epidemiology: beyond the basics. Sudbury: Jones and Bartlett Publishers; 2007. p. 211-212.
37. Zheng W, McLerran DF, Rolland B, Zhang X, Inoue M, Matsuo K, et al. Association between body-mass index and risk of death in more than 1 million Asians. N Engl J Med 2011;364(8):719-729 21345101.
38. Jee SH, Sull JW, Park J, Lee SY, Ohrr H, Guallar E, et al. Body-mass index and mortality in Korean men and women. N Engl J Med 2006;355(8):779-787 16926276.
39. Manson JE, Bassuk SS, Hu FB, Stampfer MJ, Colditz GA, Willett WC. Estimating the number of deaths due to obesity: can the divergent findings be reconciled? J Womens Health (Larchmt) 2007;16(2):168-176 17388733.
Network size and density of communication network
Participant characteristics according to social network structure
Participant characteristics according to average frequency of communication
Participant characteristics according to average frequency of meeting
Model 1 adjusted for age and education; Model 2 added smoking, alcohol drinking, and depression score; and Model 3 added hypertension, diabetes, stroke, and arthritis.