Meeting Physical Activity Guidelines: The Role of Personal Networks Among Residents of Low-Income Communities

Meeting Physical Activity Guidelines: The Role of Personal Networks Among Residents of Low-Income Communities

RESEARCH ARTICLE Meeting Physical Activity Guidelines: The Role of Personal Networks Among Residents of Low-Income Communities Stephanie Child, PhD,1...

236KB Sizes 0 Downloads 106 Views


Meeting Physical Activity Guidelines: The Role of Personal Networks Among Residents of Low-Income Communities Stephanie Child, PhD,1 Andrew T. Kaczynski, PhD,2,3 Spencer Moore, PhD2,4 Introduction: Despite known benefits of regular physical activity (PA), residents of low-income communities have disproportionately high rates of physical inactivity. Mounting evidence suggests that social network characteristics may be associated with health behaviors, including PA. The purpose of the current study was to examine associations between egocentric network characteristics and meeting PA guidelines among residents of low-income and predominantly African-American communities. Methods: Data from the Greenville Healthy Neighborhoods Project (2014), a cross-sectional study, examined social network characteristics, including the PA behavior of social ties, and whether participants met PA guidelines (150 minutes per week of aerobic exercise). Respondent-driven sampling (non-random) was utilized to recruit participants (n¼430) within eight low-income communities. Logistic regression analyses, performed in 2016, included robust sandwich estimation to account for clustering (non-independence) of observations. Results: Participants were predominantly older (M¼54.4 years, SD¼15.1 years), African American (88.0%), and female (70.7%). More than one third of participants had an annual household income o$15,000 (41.6%) or reported meeting the current aerobic PA guidelines (45.8%). Controlling for sociodemographic characteristics, greater network extensity (based on the occupation of ego’s network ties; OR¼1.11, 95% CI¼1.03, 1.20, p¼0.02) and a higher percentage of physically active network members (OR¼1.97, 95% CI¼1.02, 3.82, p¼0.04) were associated with higher odds of meeting PA guidelines. Conclusions: Social network characteristics are associated with individual PA behavior among residents of low-income communities. Interventions to increase PA among low-income and predominantly African-American communities should leverage personal networks, including the implementation of walking groups or buddy systems. Am J Prev Med 2017;](suppl 1):]]]–]]]. & 2017 American Journal of Preventive Medicine. Published by Elsevier Inc. All rights reserved.



hysical activity (PA) is associated with improved health and well-being,1 including the prevention of chronic diseases such as obesity, diabetes, cardiovascular disease, and some cancers.2 Despite these benefits, less than half of Americans meet aerobic PA guidelines (150 minutes of moderate to vigorous PA per week).3,4 Moreover, PA is substantially lower among residents of underserved communities, with low-income and African-American (AA) populations least likely to meet PA guidelines.5,6

Prior studies have shown mixed results about which network characteristics are important for PA among From the 1Department of Sociology, University of California, Berkeley, Berkeley, California; 2Department of Health Promotion, Education, and Behavior, Arnold School of Public Health, University of South Carolina, Columbia, South Carolina; 3Prevention Research Center, University of South Carolina, Columbia, South Carolina; and 4School of Kinesiology and Health Studies, Queen’s University, Kingston, Ontario, Canada Address correspondence to: Stephanie Child, PhD, Department of Sociology, University of California, Berkeley, 2232 Piedmont Avenue, Office 209, Berkeley CA 94720. E-mail: [email protected]. 0749-3797/$36.00

& 2017 American Journal of Preventive Medicine. Published by Elsevier Inc. All rights reserved.

Am J Prev Med 2017;](suppl 1):]]]–]]] 1


Child et al / Am J Prev Med 2017;](suppl 1):]]]–]]]

residents of low-income communities. For example, structural characteristics of personal networks, such as the size (i.e., number of network members or “alters”), extensity (i.e., the number of different occupations held by alters), and density (i.e., number of connections between alters) of a network, as well as the strength of ties within a network, have been associated with PA. Network composition, which refers to information about the network alters, such as an alters’ health behavior, has also been associated with PA.7,8 Some studies suggest having fewer than three close ties is associated with lower PA.6,9 Conversely, having a greater number of close ties, particularly those who exercise regularly, may increase the odds of engaging in PA behaviors and meeting PA guidelines.10–13 Other research has found no association between network size and PA, and instead emphasized that the PA behavior of a person’s network may be more important than network size for PA.8 Still another study found that greater network extensity was associated with lower physical inactivity.14 Thus, although several network characteristics have been examined, including size, extensity, and composition, the literature remains inconclusive about which factors are consistently associated with higher levels of PA, especially among residents of low-income communities. Personal network characteristics are also associated with a variety of health behaviors and outcomes,6,15–17 which may better inform the current hypotheses around networks and PA. Recent studies suggest that having larger, more-extensive networks is associated with better health and higher levels of PA.6,11,12,14 Network size and extensity may positively impact health through several mechanisms, including greater access to resources, such as services and social support.18 Moreover, individuals who have network members that engage in PA are more likely to engage in PA,13 suggesting that social networks may impact health by influencing social norms, the standards by which behaviors, including PA, are deemed socially or culturally appropriate.15,16,19 Conversely, the density of a person’s social network (i.e., the number of connections among alters) may also influence health.20 Dense networks are more likely to be homophilous (e.g., exhibiting similar characteristics and behaviors) and may therefore exert greater social influence on health behaviors than less-dense networks.21,22 Even though dense networks are often associated with greater levels of social support, this support may also hamper behaviors, including PA, especially among networks or communities where PA is not the social norm. Although some studies have examined the role of network density on other health-related behaviors, including smoking and binge drinking,23,24 the authors are not aware of any studies that have examined network density in association with

meeting PA guidelines among residents of low-income communities. Given high rates of leisure time physical inactivity among residents of underserved communities, including low-income and AA populations,5 as well as previous research that suggests network characteristics are associated with PA,9,11–13 the current study investigated the association between personal network characteristics and meeting PA recommendations among residents of lowincome communities. Both structural (number of close ties, density of close ties, and network extensity) and compositional (percentage of physically active close ties) characteristics of a person’s social networks and their associations with meeting PA guidelines were assessed. The authors hypothesized that individuals with a greater number of close ties, a greater percentage of physically active close ties, and greater network extensity would be more likely to meet PA guidelines than those with fewer close ties, lower percentages of physically active close ties, and low network extensity. Consistent with previous literature on homophily among high-density networks and health,7 it was also hypothesized that, among residents of low-income communities, the greater the density of a person’s close network ties, the less likely they were to meet PA guidelines.

METHODS Data were collected in 2014 as part of the Greenville Healthy Neighborhoods Project (GHNP). The GHNP aimed to assess the social and physical environmental characteristics of low-income communities and their associations with residents’ health behaviors and outcomes. The GHNP occurred in eight “special emphasis” neighborhoods within the city of Greenville, South Carolina. Greenville is located in the upstate region of South Carolina and has a municipal population of 58,409 residents (within a county of 4450,000 people). The “special emphasis” designation represents a heightened effort on behalf of the city to partner with underserved communities to leverage existing resources and promote well-being among residents. Median household incomes for the eight study neighborhoods ranged from $15,550 to $19,316 (M¼$17,802).

Study Sample The GHNP, a cross-sectional study, employed a non-random respondent-driven sampling (RDS) technique to engage residents from each of the eight neighborhoods in a household survey.25,26 RDS was used to garner trust and support for the study among historically hard-to-engage groups.27,28 To begin, the presidents from each neighborhood association served as the “seeds” who recruited ten other residents from their neighborhood to complete the survey. After each of these ten participants completed the survey, they were asked to recruit three more individuals from their neighborhood to complete the survey. Participants recruited by the first wave were also asked to recruit three others, and so forth, for four waves of participants in total. All participants were

Child et al / Am J Prev Med 2017;](suppl 1):]]]–]]] given a $10 gift card for completing the survey, and were incentivized to recruit others with a raffle. For each coupon that was returned by a new participant, the recruiter was entered to win a $50 gift card to a local grocery store. Participants completed surveys at a community center or church in their neighborhood. Eligibility for the survey included the ability to speak and comprehend English, being aged Z18 years, non-institutionalized, and residing in one of the eight study neighborhoods. Although most participants were invited to participate in the survey through RDS, eligible residents who did not have coupons, but had been informed of the study through another study participant, were permitted to complete the survey.

Measures The authors assessed PA using the short-form International Physical Activity Questionnaire, which elicited self-reported time and frequency of vigorous and moderate activity over the previous 7 days.29 From these data, adherence to the 2008 Physical Activity Guidelines for Americans aerobic activity component (Z150 minutes of moderate to vigorous PA per week) was determined and treated as a dichotomous outcome (meeting or not meeting recommendations). Network extensity was evaluated using a position generator.30 The position generator included a list of 12 occupations that covered a spectrum of prestige scores (i.e., accountant, high school teacher, janitor). These served as indicators of a range of accessible social capital (resources).30,31 Participants indicated whether or not they knew someone on a first-name basis who held each of these occupations. From these items, Lin32 maintains that three separate dimensions of network social capital can be assessed: highest accessed prestige score (reach), number of different accessed occupations (diversity or extensity), and the difference between the highest and lowest accessed prestige score (range). In line with previous research underscoring the salience of network extensity specifically, for both health outcomes and PA,14,17 network extensity was similarly assessed based on the number of occupations to which a participant had access (range, 0–12). A name generator/interpreter instrument was used to gather information on the characteristics of participants’ core network members.33 The name generator first asked participants to name up to three individuals (i.e., alters) with whom they had discussed important personal matters within the last 6 months. Next, they reported whether or not their alters exercised or walked regularly. From this information, the number of close ties was calculated (i.e., the number of named people; range, 0–3) as well as the percentage of close ties who were physically active. Additionally, participants were asked to indicate whether each of their alters knew the other named alters. If the participant reported that Alter A knew Alter B, it was assumed that Alter B knew Alter A as well. The density of close ties was calculated by dividing the number of reported ties among alters by the number of potential ties among alters.7 For example, in a network of three alters (A, B, and C), if A and B knew each other but neither knew C (or vice versa), the network density was calculated as 0.33. Network density ranged from 0.0 to 1.0, where higher numbers indicated greater density. Sociodemographic measures were also asked of each participant, including age (years) and gender. Race was categorized as AA, white, or other for the current analysis. Educational attainment was assessed as the highest completed degree, and categorized as ] 2017


high school diploma or less, some college (including an associate’s degree), and bachelor’s degree or higher. Marital status was categorized as never married (single), married (cohabiting), or other (divorced, separated, widowed). Household income was categorized as o$15,000 (low), $15,000–$44,999 (middle), or Z$45,000 (high). Employment status was categorized as employed (full or part time), unemployed or on disability, retired, or other (e.g., student).

Statistical Analysis A total of 430 participants completed the GHNP survey. Case-wise deletion was used for observations with missing PA data (n¼85). Additionally, observations with outlier values for the self-reported PA variable (reported PA 410 hours per day) were dropped from the analysis (n¼3) for a final sample size of 342. Logistic regression examined associations between meeting PA guidelines and network extensity, density of close ties, number of close ties, and percentage of physically active ties, while controlling for individual socioeconomic and sociodemographic characteristics. To account for the clustered nature of the RDS design (i.e., observations within recruitment chains), a clustered robust sandwich estimator was used. This allows for a “working covariance matrix” during the estimation step under circumstances when the correlation structure among observations is unknown.34 This method has been used previously in regression-based analyses of RDS data.23,24 Analyses were conducted in 2016 using STATA, version 13.1, with the vce (cluster clusterid) command for the clustered robust sandwich estimator. This study was approved by the IRB at the University of South Carolina.

RESULTS Table 1 presents sample sociodemographic characteristics. Participants were predominantly older (M¼54.4 years, SD¼15.1 years), AA (88.0%), and female (70.7%). More than half of the sample had a high school diploma or less and 440% had an annual household income o$15,000. Nearly half of participants (45.9%) reported meeting the current PA guidelines. The average density of close ties was 0.7, indicating that most core networks were moderately dense (i.e., two of the three reported ties knew one another; Table 1). The number of close ties reported on average by participants was 2.8, out of a maximum of three. A fifth of the sample (21.6%) reported that none of their close ties walked or exercised on a regular basis, whereas one quarter reported that all of their close ties walked or exercised on a regular basis (26.9%). Network extensity averaged 4.35, indicating that participants reported knowing at least one person on a first name basis from approximately four of 12 listed occupations (Table 1). Table 2 presents the results of the logistic regression model. Controlling for sociodemographic characteristics, greater network extensity (OR¼1.11, 95% CI¼1.03, 1.20, p¼0.02) and having a higher percentage of ties that walked or exercised (OR¼1.97, 95% CI¼1.02, 3.82, p¼0.04) were significantly related to greater odds of


Child et al / Am J Prev Med 2017;](suppl 1):]]]–]]]

Table 1. Sample Characteristics (n¼342) Characteristic Age, years, M (SD) Female Race Black White Other Educational attainment High school or less Some college College or advanced degree Employment status Employed Retired Unemployed/disability Other (i.e., student) Marital status Single Separated/divorced/widowed Married/cohabiting Household income o$15,000 $15,000–$44,999 Z$45,000 Meets physical activity guidelines Density of close ties, range 0–1, M (SD) Number of close ties, range 0–3, M (SD) Physically active close ties (proportion of alters) No physically active ties 1/3 physically active ties 1/2 physically active tiesa 2/3 physically active ties All ties are physically active Network extensity, range 0–12, M (SD)

Table 2. Logistic Regression of Social Network Characteristics and Meeting PA Guidelines, n¼342 Data 54.4 (15.13) 70.7 88.0 11.1 0.9 53.5 26.4 20.1 37.2 28.0 27.4 7.4 37.0 36.5 26.5 41.6 30.5 27.9 45.9 0.7 (0.41) 2.8 (0.55) 21.6 26.3 1.3 24.9 26.9 4.35 (3.40)

Source: Greenville Healthy Neighborhoods Project, Greenville, SC, 2014. Note: Values are percentages unless otherwise noted. a These were a small number of participants who named only two alters (instead of three), one of whom was reported to be physically active.

meeting PA guidelines. Thus, for every additional known occupation, the odds of meeting PA guidelines increased by 11%. For every additional close tie who walked or exercised within the network, the odds of meeting PA guidelines nearly doubled. Network density and the number of close ties were not significantly associated with meeting PA recommendations.

DISCUSSION The current study examined the role of social network characteristics on meeting PA guidelines among a sample

Network characteristics

OR (95% CI)

Network extensity Density of close ties Number of close ties Physically active close ties

1.11 (1.03, 1.20) 0.61 (0.28, 1.36) 1.32 (0.75, 2.34) 1.97 (1.02, 3.82)

Source: Greenville Healthy Neighborhoods Project, Greenville, SC, 2014. Note: Boldface indicates statistical significance (po0.05). Model controls for the following covariates: age, gender, race, educational attainment, employment status, marital status, and annual household income. PA, physical activity.

of predominantly AA residents of low-income communities. This study contributes to the growing literature around the role of social networks and PA in several ways. First, it examined the association between meeting PA guidelines and the behavioral characteristics of alters. This is based on previous work that examined the role of peer influence among social ties.13 Other mechanisms, including social control (i.e., the extent to which social relationships facilitate or regulate health behaviors), may explain why having active ties is associated with meeting PA guidelines.35,36 For example, social control has been positively associated with PA behaviors among older adults.35 Second, the results support limited sections of previous literature around network extensity and PA,17 in that greater network extensity was associated with meeting PA guidelines. The effect of greater network extensity on PA may occur through access to broader resources,21 which may include resources for PA (e.g., information about ways/places to be physically active). Similarly, network extensity could promote health by exposing less active individuals to diverse resources or to people whose PA behaviors may differ from their own, which may encourage more PA. By contrast, less extensive networks tend to be redundant in both their resources and behaviors,7,37 and among those who are less active, this redundancy may enable continued physical inactivity. Additionally, greater network extensity is an indicator of increased social integration, which is associated with both health behaviors and outcomes,21 and may also make an individual more susceptible to social influence, or social control, over health behaviors.36,38 In contrast with previous work that has found associations between the number of close ties reported by an individual and PA,6,9,11 the authors found no relationship between number of close ties and meeting PA guidelines. This could be due to differences in PA measurement. Specifically, past studies examined other

Child et al / Am J Prev Med 2017;](suppl 1):]]]–]]]

aspects of PA, including number of daily steps taken, physical inactivity, and total amount of moderate to vigorous PA. However, two studies that examined network characteristics among a Latino population found that larger networks were associated with meeting PA guidelines.8,12 Future examinations of AAs’ social networks may yield further insight into these discrepancies. Finally, this study advances the literature around social networks and PA by extending this research to an understudied population. Although several studies have examined the role of social network characteristics on PA,8,12,13 including within adolescent populations,39,40 few studies have assessed these relationships among AAs or low-income adult populations.41 Given that these populations are less likely to engage in regular PA,4 better understanding the role of social networks on meeting PA guidelines provides a potential strategy for increasing PA among these groups. These results have implications for policies and interventions aimed at increasing PA among residents of low-income and predominantly AA neighborhoods. Health promotion programs may aim to increase opportunities for insufficiently active individuals to meet and socialize with others who are regularly physically active as a way to increase exposure and generate support for PA. For example, implementation of walking groups, or a “buddy system,” may be one way to leverage social networks for PA.7 Additionally, structural interventions that enhance communal environments, such as neighborhood parks, would provide opportunities for both PA and social interaction.42 Finally, shifting cultural and social norms around PA, including regular discussion of PA among religious and community-based organizations,43 may underscore the ability of physically active individuals to serve as a resource within their respective networks.

Limitations This work is not without limitations. First, because these data are cross-sectional, it is not possible to assess whether network characteristics influence PA patterns or rather physically active individuals tend to form relationships around this shared behavior. Future studies that replicate this work using longitudinal data would further strengthen these findings. Another limitation includes the self-reporting of PA behavior. Although the International Physical Activity Questionnaire measure has been shown to provide valid and reliable estimates of PA,29 self-reported PA is subject to inherent bias and is often overestimated.44 However, even with these limitations of self-reported PA, its low participant burden, cost effectiveness, and ease of administration make it a frequent choice of measurement, allowing for ] 2017


comparison across other studies. Additionally, the study sample, which was non-randomly selected, and comprised mostly older AA women, prevents generalizability to other populations or residents of all low-income communities. Finally, although consistent with previous literature examining the role of close ties and PA,9,11 the name generator and interpreter used in the current study assessed only three network members. A name generator is often limited to three to five alters to reduce recall error and bias across participants, but these measures may not necessarily provide comprehensive information about an individual’s broader social network. Despite these limitations, this study had several strengths, including being one of few to examine the behavior of close ties as a predictor of meeting PA guidelines. Furthermore, rarely have such studies examined this relationship among residents of low-income communities, the majority of whom were AA. It is important to examine the social network characteristics of residents of low-income and AA communities because the networks of underserved groups tend to have fewer overall resources and may operate in distinct ways to affect health. As such, a better understanding of the social network characteristics of these populations could yield opportunities to improve health among these groups.

CONCLUSIONS This study is one of the first to examine the role of social networks and the behaviors of social ties on meeting PA recommendations among a predominantly low-income and AA sample. Results from this study resound with previous findings to indicate that social relationships and close social ties in particular play an important role in meeting PA guidelines among residents of low-income communities. Additionally, this work implies that interventions aimed at increasing PA rates among insufficiently active populations should leverage social networks as a resource for meeting PA guidelines, including strategies related to shifting social norms around PA, as well as the implementation of walking groups or buddy systems to foster support for PA.

ACKNOWLEDGMENTS 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. No financial disclosures were reported by the authors of this paper.


Child et al / Am J Prev Med 2017;](1):]]]–]]]

REFERENCES 1. WHO. Global Health Risks: Mortality and Burden of Disease Attributable to Selected Major Risks. Geneva: WHO; 2009. 2. Lee IM, Shiroma EJ, Lobelo F, Puska P, Blair SN, Katzmarzyk PT. Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy. Lancet. 2012;380(9838):219–229. 61031-9. 3. Haskell W, Lee IM, Pate R, et al. Physical activity and public health: updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Circulation. 2007;116(9):1081–1093. 107.185649. 4. CDC. Leisure-time physical activity: early release of selected estimates based on data from the National Health Interview Survey. 2014. www. Published 2015. Accessed February 8, 2016. 5. Tucker JM, Welk GJ, Beyler NK. Physical activity in U.S. adults: compliance with the Physical Activity Guidelines for Americans. Am J Prev Med. 2011;40(4):454–461. 2010.12.016. 6. Shelton RC, McNeill LH, Puleo E, Wolin KY, Emmons KM, Bennett GG. The association between social factors and physical activity among low-income adults living in public housing. Am J Public Health. 2011;101(11):2102–2110. 7. Valente TW. Social Networks and Health: Models, Methods, and Applications. Oxford, UK: Oxford University Press; 2010. https://doi. org/10.1093/acprof:oso/9780195301014.001.0001. 8. Marquez B, Gonzalez P, Gallo L, Ji M. Latino civic group participation, social networks, and physical activity. Am J Health Behav. 2016;40 (4):437–445. 9. Willey JZ, Paik MC, Sacco R, Elkind MSV, Boden-Albala B. Social determinants of physical inactivity in the Northern Manhattan Study (NOMAS). J Community Health. 2010;35(6):602–608. 10.1007/s10900-010-9249-2. 10. Booth ML, Owen N, Bauman A, Clavisi O, Leslie E. SocialCognitive and perceived environment influences associated with physical activity in older Australians. Prev Med. 2000;31(1):15–22. 10.1006/pmed.2000.0661. 11. 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):70–74. 12. Marquez B, Elder JP, Arredondo EM, Madanat H, Ji M, Ayala GX. Social network characteristics associated with health promoting behaviors among Latinos. Health Psychol. 2014;33(6):544–553. https: // 13. Firestone MJ, Yi SS, Bartley KF, Eisenhower DL. Perceptions and the role of group exercise among New York City adults, 2010–2011: an examination of interpersonal factors and leisure-time physical activity. Prev Med. 2015;72:50–55. 14. Legh-Jones H, Moore S. Network social capital, social participation, and physical inactivity in an urban adult population. Soc Sci Med. 2012;74 (9):1362–1367. 15. Christakis NA, Fowler JH. The collective dynamics of smoking in a large social network. N Engl J Med. 2008;358(21):2249–2258. https: // 16. 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. https://doi. org/10.1056/NEJMsa066082. 17. Moore S, Bockenholt U, Daniel M, Frohlich K, Kestens Y, Richard L. Social capital and core network ties: a validation study of individuallevel social capital measures and their association with extra- and





22. 23.









32. 33. 34.


intra-neighborhood ties, and self-rated health. Health Place. 2011; 17(2):536–544. Kawachi I, Berkman LF. Social cohesion, social capital, and health. In: Kawachi I, Berkman LF, eds. Social Epidemiology. Oxford, UK: Oxford University Press; 2000:174–190. 9780195377903.003.0008. Ball K, Jeffery RW, Abbott G, McNaughton SA, Crawford D. Is healthy behavior contagious: associations of social norms with physical activity and healthy eating. Int J Behav Nutr Phys Act. 2010;7:86. https://doi. org/10.1186/1479-5868-7-86. Haines VA, Beggs JJ, Hurlbert JS. Contextualizing health outcomes: do effects of network structure differ for women and men? Sex Roles. 2008;59(3–4):164–175. Berkman LF, Glass T. Social integration, social networks, social support, and health. In: Kawachi I, Berkman LF, eds. Social Epidemiology. Oxford, UK: Oxford University Press; 2000:137–173. Haines VA, Hurlbert JS. Network range and health. J Health Soc Behav. 1992;33(3):254–266. Mundt MP. The impact of peer social networks on adolescent alcohol use initiation. Acad Pediatr. 2011;11(5):414–421. j.acap.2011.05.005. McGloin JM, Sullivan CJ, Thomas KJ. Peer influence and context: the interdependence of friendship groups, schoolmates and network density in predicting substance use. J Youth Adolesc. 2014;43(9): 1436–1452. 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. Villanti A, German D, Sifakis F, Flynn C, Holtgrave D. Smoking, HIV status, and HIV risk behaviors in a respondent-driven sample of injection drug users in Baltimore, Maryland: the BeSure Study. AIDS Educ Prev. 2012;24(2):132–147. Johnston LG, Whitehead S, Simic-Lawson M, Kendall C. Formative research to optimize respondent-driven sampling surveys among hardto-reach populations in HIV behavioral and biological surveillance: lessons learned from four case studies. AIDS Care. 2010;22(6):784–792. Heckathorn DD. Respondent-driven sampling: a new approach to the study of hidden populations. Soc Probl. 1997;44(2):174–199. https: // Craig CL, Marshall AL, Sjöström M, et al. International physical activity questionnaire: 12-country reliability and validity. Med Sci Sports Exerc. 2003;35(8):1381–1395. 0000078924.61453.FB. Van der Gaag M, Snijders T, Flap H. Position generator measures and their relationship to other Social Capital measures. In: Lin N, Erickson B, eds. Social Capital: An International Research Program. Oxford, UK: Oxford University Press, 2008:27–49. oso/9780199234387.003.0011. Nakao K, Treas J. Updating occupational prestige and socioeconomic scores: how the new measures measure up. Sociol Methodol. 1994;24: 1–72. Lin N. Social Capital: A Theory of Social Structure and Action. Cambridge, UK: Cambridge University Press; 2002. Marsden PV. Core discussion networks of Americans. Am Sociol Rev. 1987;52(1):122–131. Kauermann G, Carroll RJ. The Sandwich variance estimator: efficiency properties and coverage probability of confidence intervals. https:// Published 2000. Accessed November 13, 2015. Newsom JT, Shaw BA, August KJ, Strath SJ. Physical activity–related social control and social support in older adults: cognitive and emotional pathways to physical activity. J Health Psychol. In press. Online July 28, 2016.

Child et al / Am J Prev Med 2017;](suppl 1):]]]–]]] 36. Umberson D. Family status and health behaviors: social control as a dimension of social integration. J Health Soc Behav. 1987;28(3): 306–319. 37. McPherson M, Smith-Lovin L, Cook JM. Birds of a feather: homophily in social networks. Annu Rev Sociol. 2001;27:415–444. 10.1146/annurev.soc.27.1.415. 38. Reifman A, Watson WK, McCourt A. Social networks and college drinking: probing processes of social influence and selection. Pers Soc Psychol Bull. 2006;32(6):820–832. 67206286219. 39. Gesell SB, Tesdahl E, Ruchman E. The distribution of physical activity in an after-school friendship network. Pediatrics. 2012;129(6): 1064–1071. 40. de la Haye K, Robins G, Mohr P, Wilson C. How physical activity shapes, and is shaped by, adolescent friendships. Soc Sci Med. 2011;73 (5):719–728.

] 2017


41. Andersen L, Gustat J, Becker AB. The relationship between the social environment and lifestyle-related physical activity in a low-income African American inner-city southern neighborhood. J Community Health. 2015;40(5):967–974. 42. Christian H, Giles-Corti B, Knuiman M, Timperio A, Foster S. The influence of the built environment, social environment and health behaviors on body mass index. Results from RESIDE. Prev Med. 2011;53(1–2):57–60. 43. Baruth M, Wilcox S, Laken M, Bopp M, Saunders R. Implementation of a faith-based physical activity intervention: insights from church health directors. J Community Health. 2008;33(5):304. 10.1007/s10900-008-9098-4. 44. Ainsworth BE, Caspersen CJ, Matthews CE, Mâsse LC, Baranowski T, Zhu W. Recommendations to improve the accuracy of estimates of physical activity derived from self report. J Phys Act Health. 2012; 9(suppl 1):S76–S84.