Personal network characteristics and body mass index: the role of education among Black Americans

Personal network characteristics and body mass index: the role of education among Black Americans Abstract Background Personal (i.e. egocentric) network characteristics are associated with health outcomes, including overweight and obesity. Previous research suggests educational attainment may interact with network characteristics to buffer these relationships. Limited research has examined the personal network characteristics of Black Americans, who have increased risk of overweight and obesity. The purpose of the current study was to examine associations between network characteristics and body mass index (BMI), and whether educational attainment modified these associations among Black Americans. Methods In 2014, using respondent-driven sampling, we recruited 430 adult residents of eight low-income neighborhoods in Greenville, SC. Self-administered questionnaires assessed structural and compositional characteristics (i.e. size, density) of respondents’ personal networks, socio-demographic characteristics, and health-related behaviors and conditions. Multilevel regression models with robust sandwich estimation accounted for clustering within respondent chains. Results Among Black adults overall, network density—the number of connections among network members—was positively associated with BMI. Higher education moderated this relationship; among Black adults with a college degree, higher network density was inversely associated with BMI. Conclusions Our data suggest low educational attainment may reflect more homogenous and less resourceful networks. Multiple pathways are discussed for how education interacts with network density on BMI among Black Americans. education, employment and skills, obesity, relationships Personal (i.e. egocentric) networks are important for health. Through the provision of social support and resources, individuals (i.e. egos) with more social ties (i.e. alters) tend to report better health.1 Similarly, networks with greater diversity, and therefore unique resources,2 may promote healthy behaviors.3 Personal networks are also a source of social norms and influence which have the potential to promote health-related behaviors that are normative within the network, including risky behaviors such as smoking and drinking.4,5 Personal networks may also be related to other behavior-related outcomes such as overweight and obesity.6 Research on predominantly White and Latino populations underscores both structural (i.e. network size and density) and compositional (i.e. alter-level occupation) personal network characteristics associated with obesity-related behaviors (i.e. physical activity, diet) and body mass index (BMI).7–9 For example, having more friends and other network members who exercise regularly is associated with greater physical activity.7,9 Additionally, networks with greater resources, measured by alter-level occupational diversity, have been shown to have lower BMI and adiposity.8 This literature indicates there are multiple mechanisms by which personal networks influence overweight and obesity rates, including the provision of social support, peer influence and social control.1,10 For example, alters often provide support for healthy dietary practices and weight control behaviors.7,11 Conversely, alters may also facilitate undesirable behaviors.5,12 The current study examines structural (i.e. size, density) and compositional (alter-level education) network characteristics that may influence BMI in an all-Black sample. Additionally, because the provision of social control and resources may be geographically dependent, this study also asks whether the proximity of network ties is associated with BMI. Previous studies also indicate that ego-level characteristics, including educational attainment, modifies the association between personal networks and health. For example, Gorman and Sivaganesan13 found greater levels of social contact were associated with a higher probability of hypertension among those with less than a high school diploma, but a lower probability among those with a high school diploma, a college degree or higher. The data highlight negative health outcomes associated with greater social contact for low-educated individuals, potentially explained by network homophily, which refers to the principle that people tend to associate with others similar to themselves (e.g. demographically, socioeconomically).14 In this example, network homophily would suggest that individuals with higher exposure to health risks (e.g. potentially due to low educational attainment) are embedded within networks whose members also have higher health risks. Pearson and Geronimus15 find a protective pattern of network homophily. In their study, having co-ethnic social ties, a compositional network characteristic in which alters are the same ethnicity as the ego, was associated with better self-rated health among Jewish adults. Moreover, this association was stronger among persons with low educational attainment. Taken together, these studies offer contrasting, yet not entirely opposing narratives of how these factors may combine to affect health. Multiple pathways are plausible. For example, while greater educational attainment may protect against negative associations between network characteristics and health, it may also temper positive associations between network characteristics and health. Specifically, those with higher levels of education may see diminishing positive returns for health from their social networks, in part because highly educated populations are also likely to have personal resources associated with better health (i.e. higher income, stable employment, etc.) and may depend less on their networks for support. Conversely, less educated populations may rely on their networks more frequently, exposing them to greater influence from the network. Finally, despite recent interest in social networks and their role in obesity, few studies have examined personal network characteristics of Black Americans16,17 even given their increased risk of obesity, and further, substantial evidence to suggest that the formation and utilization of personal networks among Black Americans may be distinct from other demographic groups.17,18 For example, compared to Whites, Blacks are less likely to mobilize their social networks for personal gain, such as giving or receiving job referrals.19 This reduced capacity of Black social networks to connect individuals to economic opportunities (i.e. job referrals, recommendations for promotion, etc.) may be associated with diminished health outcomes, including overweight and obesity. The current study addresses several gaps in the existing literature. First, we examined the association between four personal network characteristics—core network size, density, educational attainment and geographic location—and BMI among Black adults in the US South, a population that is at high risk for obesity, and is relatively understudied within the social network and health literature. Second, to address previous mixed evidence in the potential downside of personal networks for health, we consider the extent to which ego-level educational attainment moderates associations between personal networks and BMI. Methods Data were collected throughout 2014 as part of the Greenville Healthy Neighborhoods Project (GHNP). The City of Greenville is located in upstate South Carolina, and is comprised of 62 252 residents, ~30% of whom are Black. Eight special emphasis neighborhoods were selected based on their economic adversity, and their partnership with the City of Greenville to identify resources within the community that could be utilized to enhance the health and well-being of its residents. Across the Special Emphasis neighborhoods, the percentage of Black residents ranged from 34 to 82% and median household income ranged from $15 550 to $19 316 USD (M= $17 802).20 The study employed respondent-driven sampling (RDS) methodology in order to engage hard-to-reach residents.21,22 Neighborhood association presidents served as a seed (recruiter) in each neighborhood and were asked to identify ten residents (of varying age, gender, occupation, etc.) in their neighborhood for the first wave of the sampling chain. These ten people were given a coupon from the president that served as their invitation and tracked how they entered the study (i.e. their recruitment chain). Upon survey completion, participants received a $10 gift card and were incentivized with a raffle to recruit three more individuals who lived in their neighborhood, and so on, for a total of four waves of participants. Identification numbers on study coupons were used to create sampling chains which informed the cluster variable for multilevel analysis. In total, 180 sampling chains were formed, with an average of 2.1 participants (range: 1–14) per chain. Participants completed the self-administered GHNP survey at their local community center or church. Eligibility for the survey included the ability to speak and comprehend English, being 18 years of age or older, and residing in one of the eight study neighborhoods. For the purposes of the current study, anyone who did not self-identity as Black was excluded from further analysis (n = 59) for an eligible sample size of 371. Measures The primary outcome variable was BMI. BMI was calculated using self-reported height and weight data and the standard formula for adults: BMI = [weight (lbs.)/height (in.)2] × 703 kg/m2 (in2/lbs). Previous research indicates that self-reported height and weight used to calculate BMI scores are valid for measurement of overweight and obesity in epidemiological studies.23 Raw BMI scores were maintained and used as a continuous variable for analyses. Participants were asked to report the highest level of education they had completed. Educational attainment was categorized as follows: less than high school, high school diploma/GED, some college/associate’s degree and college or graduate degree. Participants’ social network characteristics were assessed using four measures, of which the correlation between each measure was expectedly low (−0.10 to 0.27). First, the number of core ties was assessed using a name generator.24 Participants were asked to name up to three people (alters) with whom they had discussed important personal matters over the last 6 months. The number of core ties a person designates has been shown to approximate the number of close ties within the network and is representative of the level of an individual’s social integration.24 Consistent with a previous study, core ties were dichotomized, such that persons who named all three alters were coded ‘1’ (maximum discussant network), and those who named less than three alters were coded ‘0’ (partial discussant network). A name interpreter consisting of several follow-up questions asked for more details about alters listed in the aforementioned name generator. First, participants were asked whether each of the three alters knew one another. For example, participants indicated (yes/no) whether Person A knew Person B, whether Person B knew Person C, and whether Person A knew Person C. From this, network density was calculated by dividing the number of actual ties between alters by the number of potential ties between alters.2 These scores ranged from 0 to 1 and were treated as a continuous predictor. The name interpreter also included questions about alters’ educational attainment and residential location. From these, we were able to assess the average educational attainment of a participant’s network where ‘0’ = less than a high school diploma, ‘1’ = a high school diploma and ‘2’ = more than a high school diploma for each alter. The average educational attainment of the network was calculated by summing these values and dividing by the number of alters within the network. Scores ranged from 0 to 2 and were treated as a continuous variable. Participants reported on network proximity by indicating whether each of the three alters resided in their home, their neighborhood, within the City of Greenville, or outside of Greenville. Similar to previous research,10 the number of alters who resided in their home or neighborhood was summed, and ranged from 0 to 3. Demographic covariates of the ego included age (in years), gender (male or female), marital status (married/cohabiting or single/widowed/divorced), employment status (employed or unemployed/retired/disabled) and annual household income (five categories ranging from <$15 000 to more than $60 000). Analytic approach Multilevel linear regression analyses were used to examine the relationship between each of the four network characteristics and BMI separately, before including all four predictors in a single model. Additional models examined moderation between ego-level educational attainment and each of the four network characteristics on the BMI relation. A Wald F-test was used to determine the significance of this interaction effect in the final model. All models controlled for ego-level demographic characteristics. In line with previous health studies utilizing RDS,25 the current study employed multilevel regression with robust estimation to best account for the unknown clustering of observations. Originally, a three-level model was employed to account for additional clustering at the neighborhood level. However, it was observed that variance at the neighborhood level was insignificant for all models. Subsequently, all analyses were re-estimated using two-level hierarchical models, with individuals nested within sampling chains. A robust (sandwich) covariance estimator was employed to account for additional errors associated with the unknown clustering of observations within the sampling chains.26 Cases with missing data for BMI (n = 34), gender (n = 1) and education (n = 3) were excluded from the analyses for a final sample size of 333. Additional missing data were imputed using a multivariate imputation by chained equations procedure,27 which generates a specified set of values by drawing from estimated conditional distributions of each variable, given all available information. This method is most practical for the current analyses due to its ability to impute seamlessly across continuous, binary, categorical and nominal variables.27 A total of twenty imputations were used to calculate missing entries on participant’s income (n = 33), and network characteristics (n = 72). All analyses were performed in STATA software version 14.1. Results Participant demographics and network characteristics are presented in Table 1. Most participants in this all-Black sample were female (70.3%), older (M = 56.2 years, SD = 14.6 years), and overweight (mean BMI: 30.3, SD = 7.2), with a range in BMI from 16.3 to 54.1. More than half of the sample had a high school diploma (43.6%) or less (17.4%). The majority of participants listed three core network members (i.e. alters; 73.0%). Networks were moderately dense (M = 0.6, SD = 0.4, range: 0–1). Participants reported that, on average, 1.2 alters lived at home or in their neighborhood and that alters education was slightly higher than a high school diploma, on average (M = 1.4, SD = 0.5, range: 0–2). Table 1 Sample characteristics (n = 333) % or M (SD) Body mass index 30.3 (7.2) Network structure  Core ties (named all 3) 73.0  Density (0–1) 0.6 (0.4) Network composition  Ties live in home/neighborhood (0–3) 1.2 (0.7)  Average network education (0–2) 1.4 (0.5) Educational attainment  Less than high school 17.4  High school diploma 43.6  Some college 26.4  College/Grad. degree 12.6 Female 70.3 Age 56.2 (14.6) Married 23.1 Employed 33.5 Income  <$15 000 47.3  $15 000–$29 999 22.9  $30 000–$44 999 6.4  $45 000–$59 999 14.2  $60 000+ 9.2 % or M (SD) Body mass index 30.3 (7.2) Network structure  Core ties (named all 3) 73.0  Density (0–1) 0.6 (0.4) Network composition  Ties live in home/neighborhood (0–3) 1.2 (0.7)  Average network education (0–2) 1.4 (0.5) Educational attainment  Less than high school 17.4  High school diploma 43.6  Some college 26.4  College/Grad. degree 12.6 Female 70.3 Age 56.2 (14.6) Married 23.1 Employed 33.5 Income  <$15 000 47.3  $15 000–$29 999 22.9  $30 000–$44 999 6.4  $45 000–$59 999 14.2  $60 000+ 9.2 Data: Greenville Healthy Neighborhoods Project, 2014. View Large Table 1 Sample characteristics (n = 333) % or M (SD) Body mass index 30.3 (7.2) Network structure  Core ties (named all 3) 73.0  Density (0–1) 0.6 (0.4) Network composition  Ties live in home/neighborhood (0–3) 1.2 (0.7)  Average network education (0–2) 1.4 (0.5) Educational attainment  Less than high school 17.4  High school diploma 43.6  Some college 26.4  College/Grad. degree 12.6 Female 70.3 Age 56.2 (14.6) Married 23.1 Employed 33.5 Income  <$15 000 47.3  $15 000–$29 999 22.9  $30 000–$44 999 6.4  $45 000–$59 999 14.2  $60 000+ 9.2 % or M (SD) Body mass index 30.3 (7.2) Network structure  Core ties (named all 3) 73.0  Density (0–1) 0.6 (0.4) Network composition  Ties live in home/neighborhood (0–3) 1.2 (0.7)  Average network education (0–2) 1.4 (0.5) Educational attainment  Less than high school 17.4  High school diploma 43.6  Some college 26.4  College/Grad. degree 12.6 Female 70.3 Age 56.2 (14.6) Married 23.1 Employed 33.5 Income  <$15 000 47.3  $15 000–$29 999 22.9  $30 000–$44 999 6.4  $45 000–$59 999 14.2  $60 000+ 9.2 Data: Greenville Healthy Neighborhoods Project, 2014. View Large Multilevel regression estimates of associations between network characteristics and BMI are presented in Table 2. Model 1 accounts for all demographic covariates (i.e. age, gender, income, marital status, employment status), except educational attainment. Once accounting for ego-level education in Model 2, levels of network density were positively associated with BMI (b = 2.19, 95% CI: 0.14, 4.24, P < 0.05). Finally, Model 3 introduced the interaction term between network density and the ego’s education. This interaction was significant for those with a college degree (b = −7.80, 95% CI: −14.31, −1.28, P < 0.05), compared to those with less than a high school diploma. While overall network density was positively associated with BMI, this relationship was different for those with a college degree (Fig. 1). Specifically, greater network density was negatively associated with BMI among those with a college degree, but still positively associated with BMI for all other education groups. Table 2 Multilevel regression estimates of associations between network characteristics and BMI (n = 333) b (95% CI) Model 1 Model 2 Model 3 Network structure  Core ties (named all 3) 1.45 (−0.31, 3.22) 1.52 (−0.21, 3.25) 1.41 (−0.29, 3.12)  Density 1.97 (−0.25, 4.11) 2.19* (0.14, 4.24) 4.40* (0.59, 8.22) Network composition  Ties live in home/neighborhood −0.55 (−1.28, 0.19) −0.65 (−1.39, 0.11) −0.60 (−0.29, 3.12)  Average network education −0.87 (−2.30, 0.56) −1.11 (−2.57, 0.34) −1.13 (−2.57, 0.31) Education  Less than HS – –  High school 1.08 (−1.17, 3.33) 2.48 (−0.84, 5.81)  Some college 1.49 (−1.33, 4.32) 2.05 (−2.12, 6.21)  College/Grad. degree 0.19 (−3.44, 3.83) 5.23 (−0.64, 11.09) Interactions  Less than high school × density –  High school × density −2.49 (−7.12, 2.13)  Some college × density −1.20 (−6.52, 4.11)  College/Grad. degree × density −7.80* (−14.31, −1.28) Constant 26.82* (21.31, 32.32) 26.18* (20.69, 31.67) 25.24* (19.64, 30.84) Wald F 3.85* b (95% CI) Model 1 Model 2 Model 3 Network structure  Core ties (named all 3) 1.45 (−0.31, 3.22) 1.52 (−0.21, 3.25) 1.41 (−0.29, 3.12)  Density 1.97 (−0.25, 4.11) 2.19* (0.14, 4.24) 4.40* (0.59, 8.22) Network composition  Ties live in home/neighborhood −0.55 (−1.28, 0.19) −0.65 (−1.39, 0.11) −0.60 (−0.29, 3.12)  Average network education −0.87 (−2.30, 0.56) −1.11 (−2.57, 0.34) −1.13 (−2.57, 0.31) Education  Less than HS – –  High school 1.08 (−1.17, 3.33) 2.48 (−0.84, 5.81)  Some college 1.49 (−1.33, 4.32) 2.05 (−2.12, 6.21)  College/Grad. degree 0.19 (−3.44, 3.83) 5.23 (−0.64, 11.09) Interactions  Less than high school × density –  High school × density −2.49 (−7.12, 2.13)  Some college × density −1.20 (−6.52, 4.11)  College/Grad. degree × density −7.80* (−14.31, −1.28) Constant 26.82* (21.31, 32.32) 26.18* (20.69, 31.67) 25.24* (19.64, 30.84) Wald F 3.85* All models control for age, gender, marital status, employment status and annual household income. Data: Greenville Healthy Neighborhoods Project, 2014. *P < 0.05. View Large Table 2 Multilevel regression estimates of associations between network characteristics and BMI (n = 333) b (95% CI) Model 1 Model 2 Model 3 Network structure  Core ties (named all 3) 1.45 (−0.31, 3.22) 1.52 (−0.21, 3.25) 1.41 (−0.29, 3.12)  Density 1.97 (−0.25, 4.11) 2.19* (0.14, 4.24) 4.40* (0.59, 8.22) Network composition  Ties live in home/neighborhood −0.55 (−1.28, 0.19) −0.65 (−1.39, 0.11) −0.60 (−0.29, 3.12)  Average network education −0.87 (−2.30, 0.56) −1.11 (−2.57, 0.34) −1.13 (−2.57, 0.31) Education  Less than HS – –  High school 1.08 (−1.17, 3.33) 2.48 (−0.84, 5.81)  Some college 1.49 (−1.33, 4.32) 2.05 (−2.12, 6.21)  College/Grad. degree 0.19 (−3.44, 3.83) 5.23 (−0.64, 11.09) Interactions  Less than high school × density –  High school × density −2.49 (−7.12, 2.13)  Some college × density −1.20 (−6.52, 4.11)  College/Grad. degree × density −7.80* (−14.31, −1.28) Constant 26.82* (21.31, 32.32) 26.18* (20.69, 31.67) 25.24* (19.64, 30.84) Wald F 3.85* b (95% CI) Model 1 Model 2 Model 3 Network structure  Core ties (named all 3) 1.45 (−0.31, 3.22) 1.52 (−0.21, 3.25) 1.41 (−0.29, 3.12)  Density 1.97 (−0.25, 4.11) 2.19* (0.14, 4.24) 4.40* (0.59, 8.22) Network composition  Ties live in home/neighborhood −0.55 (−1.28, 0.19) −0.65 (−1.39, 0.11) −0.60 (−0.29, 3.12)  Average network education −0.87 (−2.30, 0.56) −1.11 (−2.57, 0.34) −1.13 (−2.57, 0.31) Education  Less than HS – –  High school 1.08 (−1.17, 3.33) 2.48 (−0.84, 5.81)  Some college 1.49 (−1.33, 4.32) 2.05 (−2.12, 6.21)  College/Grad. degree 0.19 (−3.44, 3.83) 5.23 (−0.64, 11.09) Interactions  Less than high school × density –  High school × density −2.49 (−7.12, 2.13)  Some college × density −1.20 (−6.52, 4.11)  College/Grad. degree × density −7.80* (−14.31, −1.28) Constant 26.82* (21.31, 32.32) 26.18* (20.69, 31.67) 25.24* (19.64, 30.84) Wald F 3.85* All models control for age, gender, marital status, employment status and annual household income. Data: Greenville Healthy Neighborhoods Project, 2014. *P < 0.05. View Large Fig. 1 View largeDownload slide Interaction of educational attainment and network density on BMI. (Data: Greenville Healthy Neighborhoods Project, 2014.) Fig. 1 View largeDownload slide Interaction of educational attainment and network density on BMI. (Data: Greenville Healthy Neighborhoods Project, 2014.) Discussion Main findings of this study Among a sample of Black residents of low-income neighborhoods in Greenville, SC, educational attainment moderated the association between network density and BMI, such that network density was inversely associated with BMI among participants with a college degree but was positively associated with BMI among participants with less than a college degree. What is already known on this topic A vast literature on social networks and health offers mixed evidence on the associations between networks and health, across not only multiple network characteristics (i.e. structure, composition, type), but across numerous health outcomes, including health behaviors, chronic disease and mortality. While early research indicated certain network characteristics, including size, were associated with better health outcomes,1,2 recent research has also highlighted potential downsides or negative consequences of networks.6,8 These studies also find that the patterning of associations between networks and health may be due to social circumstances,28 including the education level of network members.13,15 What this study adds Few studies have explored the personal network characteristics of Black Americans, despite previous evidence to indicate that social networks may differ among this population.17,19 Participants in the current study were predominantly female, low-income and older Black adults, representing an important and understudied sample in which to examine links between social networks and BMI. In contrast with previous work,29 network density was positively associated with BMI among participants overall. Dense networks, characterized by high levels of trust, are known to foster support and the sharing of available resources.30 However, dense networks are less likely to contain bridging ties, and may limit the ability to access new information or support outside of the network.31 The clustering and sharing of similar resources within dense networks may not affect college educated individuals, who are more likely to have personal resources and be less dependent on their networks for support. However, among networks with less educated individuals, the redundancy (or perhaps, lack) of information and resources among highly dense networks may restrict access to opportunities that could improve their well-being.17,31 Similar to prior work, we find that education moderates the association between network characteristics and BMI.13,15 The specific mechanisms that may help explain effect modification are less clear. Individuals with more education may be better positioned to develop personal networks that are beneficial to them. College educated individuals may also be more effective in leveraging support and resources available within their network, leading to a stronger network effect among the highly educated. Alternatively, individuals with less education may be more susceptible to network influences, including social control and peer pressure,32 than individuals with greater educational attainment.33 While the current study is not able to empirically test these pathways due to the cross-sectional nature of the study design and the measurement limitations, the identification of these pathways in future studies will lead to a clearer understanding of when, why, and how educational attainment interacts with network characteristics. Finally, the findings corroborate previous work which finds no association between respondent (ego-level) education and BMI among Blacks. While educational attainment is frequently found inversely associated with BMI among other populations,34 this finding has not been consistent among Black populations in the United States.35,36 The moderation effect found in the current study implies the benefits of educational attainment for BMI among Blacks may also depend on personal network structure. Future studies that examine changes in network structure over time may yield further insight into how education interacts with network characteristics, such as density, in association with BMI, and health. Limitations of this study These findings should be considered in light of several limitations. First, the relatively small sample size may limit the ability to detect statistically significant relationships within this study. However, little research has explored the social network characteristics of Blacks, and how these characteristics are associated with overweight and obesity among this population. As such, the data provide a unique, if preliminary, opportunity to examine such relationships. Second, the study is cross-sectional and does not allow us to infer causality from the findings. This is particularly pertinent to the discussion of whether networks influence weight-related behaviors and outcomes, or whether individuals select into certain types of networks.37,38 Thus, while networks may influence health behaviors that lead to overweight and obesity, it is also likely that overweight and obese adults form relationships based on these same, shared behaviors or status. Third, the self-reporting of height and weight may have reflected social desirability bias.39 Finally, the name generator used to collect information about network members limits the number of alters named to control for recall bias.40 Thus, information about the personal network is limited to three or fewer core members and may not be reflective of an individual’s larger social network. Despite these limitations, the current study contributes to the literature on personal networks of Blacks and their associations with BMI among this at-risk population. This study sheds light on personal network characteristics, particularly network density, associated with BMI among Black residents in the US South, and provides further support for the patterning of these associations by education. While the direct mechanisms for this patterning remains unclear, the contingency of these relationships on education has important implications for public health interventions and policies. For example, our findings suggest it may be important to tailor network-based interventions and policies based on characteristics of the ego within the network. Future studies examining associations between network characteristics and health among distinct population groups are warranted in order to leverage social networks for population health improvements and equity. Acknowledgements The authors would like to acknowledge funding support from the BlueCross BlueShield Foundation of South Carolina as well as the Office of the Vice President for Research at the University of South Carolina that made this work possible. The research presented in this paper is that of the authors and does not reflect the positions or views of either funding source. Funding No financial disclosures were reported by the authors of this article. References 1 Berkman LF , Glass T . Social integration, social networks, social support, and health. In: Kawachi I , Berkman L (eds) . Social Epidemiology . New York: Oxford University Press , 2000 , 137–173. 2 Valente TW . Social Networks and Health: Models, Methods, and Applications . New York: Oxford University Press , 2010 . Google Scholar CrossRef Search ADS 3 Ball K , Jeffery RW , Abbott G et al. . Is healthy behavior contagious: associations of social norms with physical activity and healthy eating . Int J Behav Nutr Phys Act 2010 ; 7 : 86 . doi:10.1186/1479-5868-7-86 . Google Scholar CrossRef Search ADS PubMed 4 Mercken L , Steglich C , Sinclair P et al. . A longitudinal social network analysis of peer influence, peer selection, and smoking behavior among adolescents in British schools . Health Psychol 2012 ; 31 ( 4 ): 450 – 9 . doi:10.1037/a0026876 . Google Scholar CrossRef Search ADS PubMed 5 Child S , Stewart S , Moore S . Perceived control moderates the relationship between social capital and binge drinking: longitudinal findings from the Montreal Neighborhood Networks and Health Aging (MoNNET-HA) panel . Ann Epidemiol 2017 ; 27 ( 2 ): 128 – 34 . doi:10.1016/j.annepidem.2016.11.010 . Google Scholar CrossRef Search ADS PubMed 6 Powell K , Wilcox J , Clonan A et al. . The role of social networks in the development of overweight and obesity among adults: a scoping review . BMC Public Health 2015 ; 15 : 996 . doi:10.1186/s12889-015-2314-0 . Google Scholar CrossRef Search ADS PubMed 7 Tamers SL , Okechukwu C , Allen J et al. . Are social relationships a healthy influence on obesogenic behaviors among racially/ethnically diverse and socio-economically disadvantaged residents? Prev Med 2013 ; 56 ( 1 ): 70 – 4 . doi:10.1016/j.ypmed.2012.11.012 . Google Scholar CrossRef Search ADS PubMed 8 Moore S , Daniel M , Paquet C et al. . Association of individual network social capital with abdominal adiposity, overweight and obesity . J Public Health 2009 ; 31 ( 1 ): 175 – 83 . doi:10.1093/pubmed/fdn104 . Google Scholar CrossRef Search ADS 9 Child S , Kaczynski AT , Moore S . Meeting physical activity guidelines: the role of personal networks among residents of low-income communities . Am J Prev Med 2017 ; 53 ( 3 ): 385 – 91 . Google Scholar CrossRef Search ADS PubMed 10 Moore S , Bockenholt U , Daniel M et al. . Social capital and core network ties: a validation study of individual-level social capital measures and their association with extra- and intra-neighborhood ties, and self-rated health . Health Place 2011 ; 17 ( 2 ): 536 – 44 . doi:10.1016/j.healthplace.2010.12.010 . Google Scholar CrossRef Search ADS PubMed 11 Andersson MA , Christakis NA . Desire for weight loss, weight-related social contact, and body mass outcomes . Obesity (Silver Spring) 2016 ; 24 ( 7 ): 1434 – 7 . doi:10.1002/oby.21512 . Google Scholar CrossRef Search ADS PubMed 12 Borgatti SP , Foster PC . The network paradigm in organizational research: a review and typology . J Manag 2003 ; 29 ( 6 ): 991 – 1013 . doi:10.1016/S0149-2063_03_00087-4 . 13 Gorman BK , Sivaganesan A . The role of social support and integration for understanding socioeconomic disparities in self-rated health and hypertension . Soc Sci Med 2007 ; 65 ( 5 ): 958 – 75 . doi:10.1016/j.socscimed.2007.04.017 . Google Scholar CrossRef Search ADS PubMed 14 McPherson M , Smith-Lovin L , Cook J . Birds of a feather: homophily in social networks . Annu Rev Sociol 2001 ; 27 ( 1 ): 415 – 44 . doi:10.1146/annurev.soc.27.1.415 . Google Scholar CrossRef Search ADS 15 Pearson JA , Geronimus AT. Race/Ethnicity . Socioeconomic characteristics, coethnic social ties, and health: evidence from the national jewish population survey . Am J Public Health 2011 ; 101 ( 7 ): 1314 – 21 . doi:10.2105/AJPH.2009.190462 . Google Scholar CrossRef Search ADS PubMed 16 Lochner KA , Kawachi I , Brennan RT et al. . Social capital and neighborhood mortality rates in Chicago . Soc Sci Med 2003 ; 56 ( 8 ): 1797 – 1805 . doi:10.1016/S0277-9536(02)00177-6 . Google Scholar CrossRef Search ADS PubMed 17 Wilson WJ . The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy , 2nd edn . Chicago: University of Chicago Press , 2012 . 18 Holtgrave DR , Crosby R . Is social capital a protective factor against obesity and diabetes? Findings from an exploratory study . Ann Epidemiol 2006 ; 16 ( 5 ): 406 – 8 . doi:10.1016/j.annepidem.2005.04.017 . Google Scholar CrossRef Search ADS PubMed 19 Smith SS . ‘Don’t put my name on it’: social capital activation and job‐finding assistance among the Black Urban poor . Am J Sociol 2005 ; 111 ( 1 ): 1 – 57 . doi:10.1086/428814 . Google Scholar CrossRef Search ADS 20 U. S. Census Bureau . 2010–2014 American Community Survey 5-Year Estimates. http://factfinder.census.gov/faces/tableservices/jsf/pages/productview.xhtml?fpt=table. Published 2014. (15 July 2016 , date last accessed). 21 Frost SDW , Brouwer KC , Cruz MAF et al. . Respondent-driven sampling of injection drug users in two U.S.–Mexico Border Cities: recruitment dynamics and impact on estimates of HIV and syphilis prevalence . J Urban Health 2006 ; 83 ( 1 ): 83 – 97 . doi:10.1007/s11524-006-9104-z . Google Scholar CrossRef Search ADS 22 Salganik MJ , Heckathorn DD . Sampling and estimation in hidden populations using respondent-driven sampling . Sociol Methodol 2004 ; 34 ( 1 ): 193 – 240 . doi:10.1111/j.0081-1750.2004.00152.x . Google Scholar CrossRef Search ADS 23 Spencer EA , Appleby PN , Davey GK et al. . Validity of self-reported height and weight in 4808 EPIC–Oxford participants . Public Health Nutr 2002 ; 5 ( 4 ): 561 – 5 . doi:10.1079/PHN2001322 . Google Scholar CrossRef Search ADS PubMed 24 McPherson M , Smith-Lovin L , Brashears ME . Social isolation in America: changes in core discussion networks over two decades . Am Sociol Rev 2006 ; 71 ( 3 ): 353 – 75 . doi:10.1177/000312240607100301 . Google Scholar CrossRef Search ADS 25 Rhodes SD , McCoy TP . Condom use among immigrant latino sexual minorities: multilevel analysis after respondent-driven sampling . AIDS Educ Prev 2015 ; 27 ( 1 ): 27 – 43 . doi:10.1521/aeap.2015.27.1.27 . Google Scholar CrossRef Search ADS PubMed 26 Kauermann G , Carroll RJ The Sandwich Variance estimator: efficiency properties and coverage probability of confidence intervals . https://epub.ub.uni-muenchen.de/1579/. Published 2000 (13 November 2015 , date last accessed). 27 White IR , Royston P , Wood AM . Multiple imputation using chained equations: Issues and guidance for practice . Stat Med 2011 ; 30 ( 4 ): 377 – 99 . doi:10.1002/sim.4067 . Google Scholar CrossRef Search ADS PubMed 28 Uphoff EP , Pickett KE , Cabieses B et al. . A systematic review of the relationships between social capital and socioeconomic inequalities in health: a contribution to understanding the psychosocial pathway of health inequalities . Int J Equity Health 2013 ; 12 : 54 . doi:10.1186/1475-9276-12-54 . Google Scholar CrossRef Search ADS PubMed 29 Lee WJ , Youm Y , Rhee Y et al. . Social network characteristics and body mass index in an elderly Korean population, social network characteristics and body mass index in an elderly Korean population . J Prev Med Public Health 2013 ; 46 ( 6 ): 336 – 45 . doi:2013.46.6.336 . Google Scholar CrossRef Search ADS PubMed 30 Portes A . Downsides of social capital . Proc Natl Acad Sci 2014 ; 111 ( 52 ): 18407 – 08 . doi:10.1073/pnas.1421888112 . Google Scholar CrossRef Search ADS PubMed 31 Granovetter M . The strength of weak ties: a network theory revisited . Sociol Theory 1983 ; 1 : 201 – 33 . doi:10.2307/202051 . Google Scholar CrossRef Search ADS 32 Stronks K , van de Mheen HD , Looman CWN et al. . Cultural, material, and psychosocial correlates of the socioeconomic gradient in smoking behavior among adults . Prev Med 1997 ; 26 ( 5 ): 754 – 66 . doi:10.1006/pmed.1997.0174 . Google Scholar CrossRef Search ADS PubMed 33 Rowley TJ . Moving beyond Dyadic ties: a network theory of stakeholder influences . Acad Manage Rev 1997 ; 22 ( 4 ): 887 – 910 . doi:10.2307/259248 . 34 Burke GL , Bild DE , Hilner JE et al. . Differences in weight gain in relation to race, gender, age and education in young adults: The CARDIA study . Ethn Health 1996 ; 1 ( 4 ): 327 –- 35 . doi:10.1080/13557858.1996.9961802 Google Scholar CrossRef Search ADS PubMed 35 Walsemann KM , Ailshire JA , Bell BA et al. . Body mass index trajectories from adolescence to midlife: differential effects of parental and respondent education by race/ethnicity and gender . Ethn Health 2012 ; 17 ( 4 ): 337 – 62 . doi:10.1080/13557858.2011.635374 . Google Scholar CrossRef Search ADS PubMed 36 Mujahid MS , Diez Roux AV , Borrell LN et al. . Cross-sectional and longitudinal associations of BMI with socioeconomic characteristics . Obes Res 2005 ; 13 ( 8 ): 1412 – 21 . doi:10.1038/oby.2005.171 . Google Scholar CrossRef Search ADS PubMed 37 Schaefer DR , Simpkins SD . Using social network analysis to clarify the role of obesity in selection of adolescent friends . Am J Public Health 2014 ; 104 ( 7 ): 1223 – 9 . doi:10.2105/AJPH.2013.301768 . Google Scholar CrossRef Search ADS PubMed 38 Steglich C , Snijders TAB , Pearson M . Dynamic networks and behavior: separating selection from influence . Sociol Methodol 2010 ; 40 ( 1 ): 329 – 93 . doi:10.1111/j.1467-9531.2010.01225.x . Google Scholar CrossRef Search ADS 39 Ambwani S , Chmielewski JF . Weighing the evidence: social desirability, eating disorder symptomatology, and accuracy of self-reported body weight among men and women . Sex Roles 2013 ; 68 ( 7–8 ): 474 – 83 . doi:10.1007/s11199-012-0244-1 . Google Scholar CrossRef Search ADS 40 Henry AD , Lubell M , McCoy M . Survey-based measurement of public management and policy networks . J Policy Anal Manage 2012 ; 31 ( 2 ): 432 – 52 . doi:10.1002/pam.21623 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Public Health Oxford University Press

Personal network characteristics and body mass index: the role of education among Black Americans

Loading next page...
 
/lp/ou_press/personal-network-characteristics-and-body-mass-index-the-role-of-0Eb0cJTag2
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1741-3842
eISSN
1741-3850
D.O.I.
10.1093/pubmed/fdy020
Publisher site
See Article on Publisher Site

Abstract

Abstract Background Personal (i.e. egocentric) network characteristics are associated with health outcomes, including overweight and obesity. Previous research suggests educational attainment may interact with network characteristics to buffer these relationships. Limited research has examined the personal network characteristics of Black Americans, who have increased risk of overweight and obesity. The purpose of the current study was to examine associations between network characteristics and body mass index (BMI), and whether educational attainment modified these associations among Black Americans. Methods In 2014, using respondent-driven sampling, we recruited 430 adult residents of eight low-income neighborhoods in Greenville, SC. Self-administered questionnaires assessed structural and compositional characteristics (i.e. size, density) of respondents’ personal networks, socio-demographic characteristics, and health-related behaviors and conditions. Multilevel regression models with robust sandwich estimation accounted for clustering within respondent chains. Results Among Black adults overall, network density—the number of connections among network members—was positively associated with BMI. Higher education moderated this relationship; among Black adults with a college degree, higher network density was inversely associated with BMI. Conclusions Our data suggest low educational attainment may reflect more homogenous and less resourceful networks. Multiple pathways are discussed for how education interacts with network density on BMI among Black Americans. education, employment and skills, obesity, relationships Personal (i.e. egocentric) networks are important for health. Through the provision of social support and resources, individuals (i.e. egos) with more social ties (i.e. alters) tend to report better health.1 Similarly, networks with greater diversity, and therefore unique resources,2 may promote healthy behaviors.3 Personal networks are also a source of social norms and influence which have the potential to promote health-related behaviors that are normative within the network, including risky behaviors such as smoking and drinking.4,5 Personal networks may also be related to other behavior-related outcomes such as overweight and obesity.6 Research on predominantly White and Latino populations underscores both structural (i.e. network size and density) and compositional (i.e. alter-level occupation) personal network characteristics associated with obesity-related behaviors (i.e. physical activity, diet) and body mass index (BMI).7–9 For example, having more friends and other network members who exercise regularly is associated with greater physical activity.7,9 Additionally, networks with greater resources, measured by alter-level occupational diversity, have been shown to have lower BMI and adiposity.8 This literature indicates there are multiple mechanisms by which personal networks influence overweight and obesity rates, including the provision of social support, peer influence and social control.1,10 For example, alters often provide support for healthy dietary practices and weight control behaviors.7,11 Conversely, alters may also facilitate undesirable behaviors.5,12 The current study examines structural (i.e. size, density) and compositional (alter-level education) network characteristics that may influence BMI in an all-Black sample. Additionally, because the provision of social control and resources may be geographically dependent, this study also asks whether the proximity of network ties is associated with BMI. Previous studies also indicate that ego-level characteristics, including educational attainment, modifies the association between personal networks and health. For example, Gorman and Sivaganesan13 found greater levels of social contact were associated with a higher probability of hypertension among those with less than a high school diploma, but a lower probability among those with a high school diploma, a college degree or higher. The data highlight negative health outcomes associated with greater social contact for low-educated individuals, potentially explained by network homophily, which refers to the principle that people tend to associate with others similar to themselves (e.g. demographically, socioeconomically).14 In this example, network homophily would suggest that individuals with higher exposure to health risks (e.g. potentially due to low educational attainment) are embedded within networks whose members also have higher health risks. Pearson and Geronimus15 find a protective pattern of network homophily. In their study, having co-ethnic social ties, a compositional network characteristic in which alters are the same ethnicity as the ego, was associated with better self-rated health among Jewish adults. Moreover, this association was stronger among persons with low educational attainment. Taken together, these studies offer contrasting, yet not entirely opposing narratives of how these factors may combine to affect health. Multiple pathways are plausible. For example, while greater educational attainment may protect against negative associations between network characteristics and health, it may also temper positive associations between network characteristics and health. Specifically, those with higher levels of education may see diminishing positive returns for health from their social networks, in part because highly educated populations are also likely to have personal resources associated with better health (i.e. higher income, stable employment, etc.) and may depend less on their networks for support. Conversely, less educated populations may rely on their networks more frequently, exposing them to greater influence from the network. Finally, despite recent interest in social networks and their role in obesity, few studies have examined personal network characteristics of Black Americans16,17 even given their increased risk of obesity, and further, substantial evidence to suggest that the formation and utilization of personal networks among Black Americans may be distinct from other demographic groups.17,18 For example, compared to Whites, Blacks are less likely to mobilize their social networks for personal gain, such as giving or receiving job referrals.19 This reduced capacity of Black social networks to connect individuals to economic opportunities (i.e. job referrals, recommendations for promotion, etc.) may be associated with diminished health outcomes, including overweight and obesity. The current study addresses several gaps in the existing literature. First, we examined the association between four personal network characteristics—core network size, density, educational attainment and geographic location—and BMI among Black adults in the US South, a population that is at high risk for obesity, and is relatively understudied within the social network and health literature. Second, to address previous mixed evidence in the potential downside of personal networks for health, we consider the extent to which ego-level educational attainment moderates associations between personal networks and BMI. Methods Data were collected throughout 2014 as part of the Greenville Healthy Neighborhoods Project (GHNP). The City of Greenville is located in upstate South Carolina, and is comprised of 62 252 residents, ~30% of whom are Black. Eight special emphasis neighborhoods were selected based on their economic adversity, and their partnership with the City of Greenville to identify resources within the community that could be utilized to enhance the health and well-being of its residents. Across the Special Emphasis neighborhoods, the percentage of Black residents ranged from 34 to 82% and median household income ranged from $15 550 to $19 316 USD (M= $17 802).20 The study employed respondent-driven sampling (RDS) methodology in order to engage hard-to-reach residents.21,22 Neighborhood association presidents served as a seed (recruiter) in each neighborhood and were asked to identify ten residents (of varying age, gender, occupation, etc.) in their neighborhood for the first wave of the sampling chain. These ten people were given a coupon from the president that served as their invitation and tracked how they entered the study (i.e. their recruitment chain). Upon survey completion, participants received a $10 gift card and were incentivized with a raffle to recruit three more individuals who lived in their neighborhood, and so on, for a total of four waves of participants. Identification numbers on study coupons were used to create sampling chains which informed the cluster variable for multilevel analysis. In total, 180 sampling chains were formed, with an average of 2.1 participants (range: 1–14) per chain. Participants completed the self-administered GHNP survey at their local community center or church. Eligibility for the survey included the ability to speak and comprehend English, being 18 years of age or older, and residing in one of the eight study neighborhoods. For the purposes of the current study, anyone who did not self-identity as Black was excluded from further analysis (n = 59) for an eligible sample size of 371. Measures The primary outcome variable was BMI. BMI was calculated using self-reported height and weight data and the standard formula for adults: BMI = [weight (lbs.)/height (in.)2] × 703 kg/m2 (in2/lbs). Previous research indicates that self-reported height and weight used to calculate BMI scores are valid for measurement of overweight and obesity in epidemiological studies.23 Raw BMI scores were maintained and used as a continuous variable for analyses. Participants were asked to report the highest level of education they had completed. Educational attainment was categorized as follows: less than high school, high school diploma/GED, some college/associate’s degree and college or graduate degree. Participants’ social network characteristics were assessed using four measures, of which the correlation between each measure was expectedly low (−0.10 to 0.27). First, the number of core ties was assessed using a name generator.24 Participants were asked to name up to three people (alters) with whom they had discussed important personal matters over the last 6 months. The number of core ties a person designates has been shown to approximate the number of close ties within the network and is representative of the level of an individual’s social integration.24 Consistent with a previous study, core ties were dichotomized, such that persons who named all three alters were coded ‘1’ (maximum discussant network), and those who named less than three alters were coded ‘0’ (partial discussant network). A name interpreter consisting of several follow-up questions asked for more details about alters listed in the aforementioned name generator. First, participants were asked whether each of the three alters knew one another. For example, participants indicated (yes/no) whether Person A knew Person B, whether Person B knew Person C, and whether Person A knew Person C. From this, network density was calculated by dividing the number of actual ties between alters by the number of potential ties between alters.2 These scores ranged from 0 to 1 and were treated as a continuous predictor. The name interpreter also included questions about alters’ educational attainment and residential location. From these, we were able to assess the average educational attainment of a participant’s network where ‘0’ = less than a high school diploma, ‘1’ = a high school diploma and ‘2’ = more than a high school diploma for each alter. The average educational attainment of the network was calculated by summing these values and dividing by the number of alters within the network. Scores ranged from 0 to 2 and were treated as a continuous variable. Participants reported on network proximity by indicating whether each of the three alters resided in their home, their neighborhood, within the City of Greenville, or outside of Greenville. Similar to previous research,10 the number of alters who resided in their home or neighborhood was summed, and ranged from 0 to 3. Demographic covariates of the ego included age (in years), gender (male or female), marital status (married/cohabiting or single/widowed/divorced), employment status (employed or unemployed/retired/disabled) and annual household income (five categories ranging from <$15 000 to more than $60 000). Analytic approach Multilevel linear regression analyses were used to examine the relationship between each of the four network characteristics and BMI separately, before including all four predictors in a single model. Additional models examined moderation between ego-level educational attainment and each of the four network characteristics on the BMI relation. A Wald F-test was used to determine the significance of this interaction effect in the final model. All models controlled for ego-level demographic characteristics. In line with previous health studies utilizing RDS,25 the current study employed multilevel regression with robust estimation to best account for the unknown clustering of observations. Originally, a three-level model was employed to account for additional clustering at the neighborhood level. However, it was observed that variance at the neighborhood level was insignificant for all models. Subsequently, all analyses were re-estimated using two-level hierarchical models, with individuals nested within sampling chains. A robust (sandwich) covariance estimator was employed to account for additional errors associated with the unknown clustering of observations within the sampling chains.26 Cases with missing data for BMI (n = 34), gender (n = 1) and education (n = 3) were excluded from the analyses for a final sample size of 333. Additional missing data were imputed using a multivariate imputation by chained equations procedure,27 which generates a specified set of values by drawing from estimated conditional distributions of each variable, given all available information. This method is most practical for the current analyses due to its ability to impute seamlessly across continuous, binary, categorical and nominal variables.27 A total of twenty imputations were used to calculate missing entries on participant’s income (n = 33), and network characteristics (n = 72). All analyses were performed in STATA software version 14.1. Results Participant demographics and network characteristics are presented in Table 1. Most participants in this all-Black sample were female (70.3%), older (M = 56.2 years, SD = 14.6 years), and overweight (mean BMI: 30.3, SD = 7.2), with a range in BMI from 16.3 to 54.1. More than half of the sample had a high school diploma (43.6%) or less (17.4%). The majority of participants listed three core network members (i.e. alters; 73.0%). Networks were moderately dense (M = 0.6, SD = 0.4, range: 0–1). Participants reported that, on average, 1.2 alters lived at home or in their neighborhood and that alters education was slightly higher than a high school diploma, on average (M = 1.4, SD = 0.5, range: 0–2). Table 1 Sample characteristics (n = 333) % or M (SD) Body mass index 30.3 (7.2) Network structure  Core ties (named all 3) 73.0  Density (0–1) 0.6 (0.4) Network composition  Ties live in home/neighborhood (0–3) 1.2 (0.7)  Average network education (0–2) 1.4 (0.5) Educational attainment  Less than high school 17.4  High school diploma 43.6  Some college 26.4  College/Grad. degree 12.6 Female 70.3 Age 56.2 (14.6) Married 23.1 Employed 33.5 Income  <$15 000 47.3  $15 000–$29 999 22.9  $30 000–$44 999 6.4  $45 000–$59 999 14.2  $60 000+ 9.2 % or M (SD) Body mass index 30.3 (7.2) Network structure  Core ties (named all 3) 73.0  Density (0–1) 0.6 (0.4) Network composition  Ties live in home/neighborhood (0–3) 1.2 (0.7)  Average network education (0–2) 1.4 (0.5) Educational attainment  Less than high school 17.4  High school diploma 43.6  Some college 26.4  College/Grad. degree 12.6 Female 70.3 Age 56.2 (14.6) Married 23.1 Employed 33.5 Income  <$15 000 47.3  $15 000–$29 999 22.9  $30 000–$44 999 6.4  $45 000–$59 999 14.2  $60 000+ 9.2 Data: Greenville Healthy Neighborhoods Project, 2014. View Large Table 1 Sample characteristics (n = 333) % or M (SD) Body mass index 30.3 (7.2) Network structure  Core ties (named all 3) 73.0  Density (0–1) 0.6 (0.4) Network composition  Ties live in home/neighborhood (0–3) 1.2 (0.7)  Average network education (0–2) 1.4 (0.5) Educational attainment  Less than high school 17.4  High school diploma 43.6  Some college 26.4  College/Grad. degree 12.6 Female 70.3 Age 56.2 (14.6) Married 23.1 Employed 33.5 Income  <$15 000 47.3  $15 000–$29 999 22.9  $30 000–$44 999 6.4  $45 000–$59 999 14.2  $60 000+ 9.2 % or M (SD) Body mass index 30.3 (7.2) Network structure  Core ties (named all 3) 73.0  Density (0–1) 0.6 (0.4) Network composition  Ties live in home/neighborhood (0–3) 1.2 (0.7)  Average network education (0–2) 1.4 (0.5) Educational attainment  Less than high school 17.4  High school diploma 43.6  Some college 26.4  College/Grad. degree 12.6 Female 70.3 Age 56.2 (14.6) Married 23.1 Employed 33.5 Income  <$15 000 47.3  $15 000–$29 999 22.9  $30 000–$44 999 6.4  $45 000–$59 999 14.2  $60 000+ 9.2 Data: Greenville Healthy Neighborhoods Project, 2014. View Large Multilevel regression estimates of associations between network characteristics and BMI are presented in Table 2. Model 1 accounts for all demographic covariates (i.e. age, gender, income, marital status, employment status), except educational attainment. Once accounting for ego-level education in Model 2, levels of network density were positively associated with BMI (b = 2.19, 95% CI: 0.14, 4.24, P < 0.05). Finally, Model 3 introduced the interaction term between network density and the ego’s education. This interaction was significant for those with a college degree (b = −7.80, 95% CI: −14.31, −1.28, P < 0.05), compared to those with less than a high school diploma. While overall network density was positively associated with BMI, this relationship was different for those with a college degree (Fig. 1). Specifically, greater network density was negatively associated with BMI among those with a college degree, but still positively associated with BMI for all other education groups. Table 2 Multilevel regression estimates of associations between network characteristics and BMI (n = 333) b (95% CI) Model 1 Model 2 Model 3 Network structure  Core ties (named all 3) 1.45 (−0.31, 3.22) 1.52 (−0.21, 3.25) 1.41 (−0.29, 3.12)  Density 1.97 (−0.25, 4.11) 2.19* (0.14, 4.24) 4.40* (0.59, 8.22) Network composition  Ties live in home/neighborhood −0.55 (−1.28, 0.19) −0.65 (−1.39, 0.11) −0.60 (−0.29, 3.12)  Average network education −0.87 (−2.30, 0.56) −1.11 (−2.57, 0.34) −1.13 (−2.57, 0.31) Education  Less than HS – –  High school 1.08 (−1.17, 3.33) 2.48 (−0.84, 5.81)  Some college 1.49 (−1.33, 4.32) 2.05 (−2.12, 6.21)  College/Grad. degree 0.19 (−3.44, 3.83) 5.23 (−0.64, 11.09) Interactions  Less than high school × density –  High school × density −2.49 (−7.12, 2.13)  Some college × density −1.20 (−6.52, 4.11)  College/Grad. degree × density −7.80* (−14.31, −1.28) Constant 26.82* (21.31, 32.32) 26.18* (20.69, 31.67) 25.24* (19.64, 30.84) Wald F 3.85* b (95% CI) Model 1 Model 2 Model 3 Network structure  Core ties (named all 3) 1.45 (−0.31, 3.22) 1.52 (−0.21, 3.25) 1.41 (−0.29, 3.12)  Density 1.97 (−0.25, 4.11) 2.19* (0.14, 4.24) 4.40* (0.59, 8.22) Network composition  Ties live in home/neighborhood −0.55 (−1.28, 0.19) −0.65 (−1.39, 0.11) −0.60 (−0.29, 3.12)  Average network education −0.87 (−2.30, 0.56) −1.11 (−2.57, 0.34) −1.13 (−2.57, 0.31) Education  Less than HS – –  High school 1.08 (−1.17, 3.33) 2.48 (−0.84, 5.81)  Some college 1.49 (−1.33, 4.32) 2.05 (−2.12, 6.21)  College/Grad. degree 0.19 (−3.44, 3.83) 5.23 (−0.64, 11.09) Interactions  Less than high school × density –  High school × density −2.49 (−7.12, 2.13)  Some college × density −1.20 (−6.52, 4.11)  College/Grad. degree × density −7.80* (−14.31, −1.28) Constant 26.82* (21.31, 32.32) 26.18* (20.69, 31.67) 25.24* (19.64, 30.84) Wald F 3.85* All models control for age, gender, marital status, employment status and annual household income. Data: Greenville Healthy Neighborhoods Project, 2014. *P < 0.05. View Large Table 2 Multilevel regression estimates of associations between network characteristics and BMI (n = 333) b (95% CI) Model 1 Model 2 Model 3 Network structure  Core ties (named all 3) 1.45 (−0.31, 3.22) 1.52 (−0.21, 3.25) 1.41 (−0.29, 3.12)  Density 1.97 (−0.25, 4.11) 2.19* (0.14, 4.24) 4.40* (0.59, 8.22) Network composition  Ties live in home/neighborhood −0.55 (−1.28, 0.19) −0.65 (−1.39, 0.11) −0.60 (−0.29, 3.12)  Average network education −0.87 (−2.30, 0.56) −1.11 (−2.57, 0.34) −1.13 (−2.57, 0.31) Education  Less than HS – –  High school 1.08 (−1.17, 3.33) 2.48 (−0.84, 5.81)  Some college 1.49 (−1.33, 4.32) 2.05 (−2.12, 6.21)  College/Grad. degree 0.19 (−3.44, 3.83) 5.23 (−0.64, 11.09) Interactions  Less than high school × density –  High school × density −2.49 (−7.12, 2.13)  Some college × density −1.20 (−6.52, 4.11)  College/Grad. degree × density −7.80* (−14.31, −1.28) Constant 26.82* (21.31, 32.32) 26.18* (20.69, 31.67) 25.24* (19.64, 30.84) Wald F 3.85* b (95% CI) Model 1 Model 2 Model 3 Network structure  Core ties (named all 3) 1.45 (−0.31, 3.22) 1.52 (−0.21, 3.25) 1.41 (−0.29, 3.12)  Density 1.97 (−0.25, 4.11) 2.19* (0.14, 4.24) 4.40* (0.59, 8.22) Network composition  Ties live in home/neighborhood −0.55 (−1.28, 0.19) −0.65 (−1.39, 0.11) −0.60 (−0.29, 3.12)  Average network education −0.87 (−2.30, 0.56) −1.11 (−2.57, 0.34) −1.13 (−2.57, 0.31) Education  Less than HS – –  High school 1.08 (−1.17, 3.33) 2.48 (−0.84, 5.81)  Some college 1.49 (−1.33, 4.32) 2.05 (−2.12, 6.21)  College/Grad. degree 0.19 (−3.44, 3.83) 5.23 (−0.64, 11.09) Interactions  Less than high school × density –  High school × density −2.49 (−7.12, 2.13)  Some college × density −1.20 (−6.52, 4.11)  College/Grad. degree × density −7.80* (−14.31, −1.28) Constant 26.82* (21.31, 32.32) 26.18* (20.69, 31.67) 25.24* (19.64, 30.84) Wald F 3.85* All models control for age, gender, marital status, employment status and annual household income. Data: Greenville Healthy Neighborhoods Project, 2014. *P < 0.05. View Large Fig. 1 View largeDownload slide Interaction of educational attainment and network density on BMI. (Data: Greenville Healthy Neighborhoods Project, 2014.) Fig. 1 View largeDownload slide Interaction of educational attainment and network density on BMI. (Data: Greenville Healthy Neighborhoods Project, 2014.) Discussion Main findings of this study Among a sample of Black residents of low-income neighborhoods in Greenville, SC, educational attainment moderated the association between network density and BMI, such that network density was inversely associated with BMI among participants with a college degree but was positively associated with BMI among participants with less than a college degree. What is already known on this topic A vast literature on social networks and health offers mixed evidence on the associations between networks and health, across not only multiple network characteristics (i.e. structure, composition, type), but across numerous health outcomes, including health behaviors, chronic disease and mortality. While early research indicated certain network characteristics, including size, were associated with better health outcomes,1,2 recent research has also highlighted potential downsides or negative consequences of networks.6,8 These studies also find that the patterning of associations between networks and health may be due to social circumstances,28 including the education level of network members.13,15 What this study adds Few studies have explored the personal network characteristics of Black Americans, despite previous evidence to indicate that social networks may differ among this population.17,19 Participants in the current study were predominantly female, low-income and older Black adults, representing an important and understudied sample in which to examine links between social networks and BMI. In contrast with previous work,29 network density was positively associated with BMI among participants overall. Dense networks, characterized by high levels of trust, are known to foster support and the sharing of available resources.30 However, dense networks are less likely to contain bridging ties, and may limit the ability to access new information or support outside of the network.31 The clustering and sharing of similar resources within dense networks may not affect college educated individuals, who are more likely to have personal resources and be less dependent on their networks for support. However, among networks with less educated individuals, the redundancy (or perhaps, lack) of information and resources among highly dense networks may restrict access to opportunities that could improve their well-being.17,31 Similar to prior work, we find that education moderates the association between network characteristics and BMI.13,15 The specific mechanisms that may help explain effect modification are less clear. Individuals with more education may be better positioned to develop personal networks that are beneficial to them. College educated individuals may also be more effective in leveraging support and resources available within their network, leading to a stronger network effect among the highly educated. Alternatively, individuals with less education may be more susceptible to network influences, including social control and peer pressure,32 than individuals with greater educational attainment.33 While the current study is not able to empirically test these pathways due to the cross-sectional nature of the study design and the measurement limitations, the identification of these pathways in future studies will lead to a clearer understanding of when, why, and how educational attainment interacts with network characteristics. Finally, the findings corroborate previous work which finds no association between respondent (ego-level) education and BMI among Blacks. While educational attainment is frequently found inversely associated with BMI among other populations,34 this finding has not been consistent among Black populations in the United States.35,36 The moderation effect found in the current study implies the benefits of educational attainment for BMI among Blacks may also depend on personal network structure. Future studies that examine changes in network structure over time may yield further insight into how education interacts with network characteristics, such as density, in association with BMI, and health. Limitations of this study These findings should be considered in light of several limitations. First, the relatively small sample size may limit the ability to detect statistically significant relationships within this study. However, little research has explored the social network characteristics of Blacks, and how these characteristics are associated with overweight and obesity among this population. As such, the data provide a unique, if preliminary, opportunity to examine such relationships. Second, the study is cross-sectional and does not allow us to infer causality from the findings. This is particularly pertinent to the discussion of whether networks influence weight-related behaviors and outcomes, or whether individuals select into certain types of networks.37,38 Thus, while networks may influence health behaviors that lead to overweight and obesity, it is also likely that overweight and obese adults form relationships based on these same, shared behaviors or status. Third, the self-reporting of height and weight may have reflected social desirability bias.39 Finally, the name generator used to collect information about network members limits the number of alters named to control for recall bias.40 Thus, information about the personal network is limited to three or fewer core members and may not be reflective of an individual’s larger social network. Despite these limitations, the current study contributes to the literature on personal networks of Blacks and their associations with BMI among this at-risk population. This study sheds light on personal network characteristics, particularly network density, associated with BMI among Black residents in the US South, and provides further support for the patterning of these associations by education. While the direct mechanisms for this patterning remains unclear, the contingency of these relationships on education has important implications for public health interventions and policies. For example, our findings suggest it may be important to tailor network-based interventions and policies based on characteristics of the ego within the network. Future studies examining associations between network characteristics and health among distinct population groups are warranted in order to leverage social networks for population health improvements and equity. Acknowledgements The authors would like to acknowledge funding support from the BlueCross BlueShield Foundation of South Carolina as well as the Office of the Vice President for Research at the University of South Carolina that made this work possible. The research presented in this paper is that of the authors and does not reflect the positions or views of either funding source. Funding No financial disclosures were reported by the authors of this article. References 1 Berkman LF , Glass T . Social integration, social networks, social support, and health. In: Kawachi I , Berkman L (eds) . Social Epidemiology . New York: Oxford University Press , 2000 , 137–173. 2 Valente TW . Social Networks and Health: Models, Methods, and Applications . New York: Oxford University Press , 2010 . Google Scholar CrossRef Search ADS 3 Ball K , Jeffery RW , Abbott G et al. . Is healthy behavior contagious: associations of social norms with physical activity and healthy eating . Int J Behav Nutr Phys Act 2010 ; 7 : 86 . doi:10.1186/1479-5868-7-86 . Google Scholar CrossRef Search ADS PubMed 4 Mercken L , Steglich C , Sinclair P et al. . A longitudinal social network analysis of peer influence, peer selection, and smoking behavior among adolescents in British schools . Health Psychol 2012 ; 31 ( 4 ): 450 – 9 . doi:10.1037/a0026876 . Google Scholar CrossRef Search ADS PubMed 5 Child S , Stewart S , Moore S . Perceived control moderates the relationship between social capital and binge drinking: longitudinal findings from the Montreal Neighborhood Networks and Health Aging (MoNNET-HA) panel . Ann Epidemiol 2017 ; 27 ( 2 ): 128 – 34 . doi:10.1016/j.annepidem.2016.11.010 . Google Scholar CrossRef Search ADS PubMed 6 Powell K , Wilcox J , Clonan A et al. . The role of social networks in the development of overweight and obesity among adults: a scoping review . BMC Public Health 2015 ; 15 : 996 . doi:10.1186/s12889-015-2314-0 . Google Scholar CrossRef Search ADS PubMed 7 Tamers SL , Okechukwu C , Allen J et al. . Are social relationships a healthy influence on obesogenic behaviors among racially/ethnically diverse and socio-economically disadvantaged residents? Prev Med 2013 ; 56 ( 1 ): 70 – 4 . doi:10.1016/j.ypmed.2012.11.012 . Google Scholar CrossRef Search ADS PubMed 8 Moore S , Daniel M , Paquet C et al. . Association of individual network social capital with abdominal adiposity, overweight and obesity . J Public Health 2009 ; 31 ( 1 ): 175 – 83 . doi:10.1093/pubmed/fdn104 . Google Scholar CrossRef Search ADS 9 Child S , Kaczynski AT , Moore S . Meeting physical activity guidelines: the role of personal networks among residents of low-income communities . Am J Prev Med 2017 ; 53 ( 3 ): 385 – 91 . Google Scholar CrossRef Search ADS PubMed 10 Moore S , Bockenholt U , Daniel M et al. . Social capital and core network ties: a validation study of individual-level social capital measures and their association with extra- and intra-neighborhood ties, and self-rated health . Health Place 2011 ; 17 ( 2 ): 536 – 44 . doi:10.1016/j.healthplace.2010.12.010 . Google Scholar CrossRef Search ADS PubMed 11 Andersson MA , Christakis NA . Desire for weight loss, weight-related social contact, and body mass outcomes . Obesity (Silver Spring) 2016 ; 24 ( 7 ): 1434 – 7 . doi:10.1002/oby.21512 . Google Scholar CrossRef Search ADS PubMed 12 Borgatti SP , Foster PC . The network paradigm in organizational research: a review and typology . J Manag 2003 ; 29 ( 6 ): 991 – 1013 . doi:10.1016/S0149-2063_03_00087-4 . 13 Gorman BK , Sivaganesan A . The role of social support and integration for understanding socioeconomic disparities in self-rated health and hypertension . Soc Sci Med 2007 ; 65 ( 5 ): 958 – 75 . doi:10.1016/j.socscimed.2007.04.017 . Google Scholar CrossRef Search ADS PubMed 14 McPherson M , Smith-Lovin L , Cook J . Birds of a feather: homophily in social networks . Annu Rev Sociol 2001 ; 27 ( 1 ): 415 – 44 . doi:10.1146/annurev.soc.27.1.415 . Google Scholar CrossRef Search ADS 15 Pearson JA , Geronimus AT. Race/Ethnicity . Socioeconomic characteristics, coethnic social ties, and health: evidence from the national jewish population survey . Am J Public Health 2011 ; 101 ( 7 ): 1314 – 21 . doi:10.2105/AJPH.2009.190462 . Google Scholar CrossRef Search ADS PubMed 16 Lochner KA , Kawachi I , Brennan RT et al. . Social capital and neighborhood mortality rates in Chicago . Soc Sci Med 2003 ; 56 ( 8 ): 1797 – 1805 . doi:10.1016/S0277-9536(02)00177-6 . Google Scholar CrossRef Search ADS PubMed 17 Wilson WJ . The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy , 2nd edn . Chicago: University of Chicago Press , 2012 . 18 Holtgrave DR , Crosby R . Is social capital a protective factor against obesity and diabetes? Findings from an exploratory study . Ann Epidemiol 2006 ; 16 ( 5 ): 406 – 8 . doi:10.1016/j.annepidem.2005.04.017 . Google Scholar CrossRef Search ADS PubMed 19 Smith SS . ‘Don’t put my name on it’: social capital activation and job‐finding assistance among the Black Urban poor . Am J Sociol 2005 ; 111 ( 1 ): 1 – 57 . doi:10.1086/428814 . Google Scholar CrossRef Search ADS 20 U. S. Census Bureau . 2010–2014 American Community Survey 5-Year Estimates. http://factfinder.census.gov/faces/tableservices/jsf/pages/productview.xhtml?fpt=table. Published 2014. (15 July 2016 , date last accessed). 21 Frost SDW , Brouwer KC , Cruz MAF et al. . Respondent-driven sampling of injection drug users in two U.S.–Mexico Border Cities: recruitment dynamics and impact on estimates of HIV and syphilis prevalence . J Urban Health 2006 ; 83 ( 1 ): 83 – 97 . doi:10.1007/s11524-006-9104-z . Google Scholar CrossRef Search ADS 22 Salganik MJ , Heckathorn DD . Sampling and estimation in hidden populations using respondent-driven sampling . Sociol Methodol 2004 ; 34 ( 1 ): 193 – 240 . doi:10.1111/j.0081-1750.2004.00152.x . Google Scholar CrossRef Search ADS 23 Spencer EA , Appleby PN , Davey GK et al. . Validity of self-reported height and weight in 4808 EPIC–Oxford participants . Public Health Nutr 2002 ; 5 ( 4 ): 561 – 5 . doi:10.1079/PHN2001322 . Google Scholar CrossRef Search ADS PubMed 24 McPherson M , Smith-Lovin L , Brashears ME . Social isolation in America: changes in core discussion networks over two decades . Am Sociol Rev 2006 ; 71 ( 3 ): 353 – 75 . doi:10.1177/000312240607100301 . Google Scholar CrossRef Search ADS 25 Rhodes SD , McCoy TP . Condom use among immigrant latino sexual minorities: multilevel analysis after respondent-driven sampling . AIDS Educ Prev 2015 ; 27 ( 1 ): 27 – 43 . doi:10.1521/aeap.2015.27.1.27 . Google Scholar CrossRef Search ADS PubMed 26 Kauermann G , Carroll RJ The Sandwich Variance estimator: efficiency properties and coverage probability of confidence intervals . https://epub.ub.uni-muenchen.de/1579/. Published 2000 (13 November 2015 , date last accessed). 27 White IR , Royston P , Wood AM . Multiple imputation using chained equations: Issues and guidance for practice . Stat Med 2011 ; 30 ( 4 ): 377 – 99 . doi:10.1002/sim.4067 . Google Scholar CrossRef Search ADS PubMed 28 Uphoff EP , Pickett KE , Cabieses B et al. . A systematic review of the relationships between social capital and socioeconomic inequalities in health: a contribution to understanding the psychosocial pathway of health inequalities . Int J Equity Health 2013 ; 12 : 54 . doi:10.1186/1475-9276-12-54 . Google Scholar CrossRef Search ADS PubMed 29 Lee WJ , Youm Y , Rhee Y et al. . Social network characteristics and body mass index in an elderly Korean population, social network characteristics and body mass index in an elderly Korean population . J Prev Med Public Health 2013 ; 46 ( 6 ): 336 – 45 . doi:2013.46.6.336 . Google Scholar CrossRef Search ADS PubMed 30 Portes A . Downsides of social capital . Proc Natl Acad Sci 2014 ; 111 ( 52 ): 18407 – 08 . doi:10.1073/pnas.1421888112 . Google Scholar CrossRef Search ADS PubMed 31 Granovetter M . The strength of weak ties: a network theory revisited . Sociol Theory 1983 ; 1 : 201 – 33 . doi:10.2307/202051 . Google Scholar CrossRef Search ADS 32 Stronks K , van de Mheen HD , Looman CWN et al. . Cultural, material, and psychosocial correlates of the socioeconomic gradient in smoking behavior among adults . Prev Med 1997 ; 26 ( 5 ): 754 – 66 . doi:10.1006/pmed.1997.0174 . Google Scholar CrossRef Search ADS PubMed 33 Rowley TJ . Moving beyond Dyadic ties: a network theory of stakeholder influences . Acad Manage Rev 1997 ; 22 ( 4 ): 887 – 910 . doi:10.2307/259248 . 34 Burke GL , Bild DE , Hilner JE et al. . Differences in weight gain in relation to race, gender, age and education in young adults: The CARDIA study . Ethn Health 1996 ; 1 ( 4 ): 327 –- 35 . doi:10.1080/13557858.1996.9961802 Google Scholar CrossRef Search ADS PubMed 35 Walsemann KM , Ailshire JA , Bell BA et al. . Body mass index trajectories from adolescence to midlife: differential effects of parental and respondent education by race/ethnicity and gender . Ethn Health 2012 ; 17 ( 4 ): 337 – 62 . doi:10.1080/13557858.2011.635374 . Google Scholar CrossRef Search ADS PubMed 36 Mujahid MS , Diez Roux AV , Borrell LN et al. . Cross-sectional and longitudinal associations of BMI with socioeconomic characteristics . Obes Res 2005 ; 13 ( 8 ): 1412 – 21 . doi:10.1038/oby.2005.171 . Google Scholar CrossRef Search ADS PubMed 37 Schaefer DR , Simpkins SD . Using social network analysis to clarify the role of obesity in selection of adolescent friends . Am J Public Health 2014 ; 104 ( 7 ): 1223 – 9 . doi:10.2105/AJPH.2013.301768 . Google Scholar CrossRef Search ADS PubMed 38 Steglich C , Snijders TAB , Pearson M . Dynamic networks and behavior: separating selection from influence . Sociol Methodol 2010 ; 40 ( 1 ): 329 – 93 . doi:10.1111/j.1467-9531.2010.01225.x . Google Scholar CrossRef Search ADS 39 Ambwani S , Chmielewski JF . Weighing the evidence: social desirability, eating disorder symptomatology, and accuracy of self-reported body weight among men and women . Sex Roles 2013 ; 68 ( 7–8 ): 474 – 83 . doi:10.1007/s11199-012-0244-1 . Google Scholar CrossRef Search ADS 40 Henry AD , Lubell M , McCoy M . Survey-based measurement of public management and policy networks . J Policy Anal Manage 2012 ; 31 ( 2 ): 432 – 52 . doi:10.1002/pam.21623 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Journal

Journal of Public HealthOxford University Press

Published: Feb 13, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off