Predicting social network homophily on common chronic disease risk behaviors among public housing residents

Attributes and behaviors tend to cluster (homophily) among connected individuals (social networks). Accordingly, the design of effective interventions to address chronic disease risk behaviors at the individual level has proven challenging. To effectively move behavior-change interventions upstream, beyond the individual, an understanding of behavior clustering within a social context is required. This ego-centric/participant reported social network study aimed to identify individual-level (gender, behavior) and relational (closeness) factors that predict homophily on the consumption frequency of both sugar-sweetened beverages (SSB) and sugar-sweetened foods (SSF) among residents of public housing developments in Boston, MA. Egos/participants (n = 272) named alters/social contacts (n = 889) with whom they discuss important matters, share meals, and interact within their housing development. Egos reported sociodemographics, relationship attributes and health behaviors, including SSB and SSF consumption for themselves and alters. Data were collected between March 2019–2020. Multilevel regression models evaluated homophily on SSB and SSF. The best fitting model predicting homophily on SSB included daily contact (OR 1.99, 95% CI:1.33, 2.98), education homophily (OR 1.68, 95% CI:1.15, 2.46), and SSF homophily (OR 1.79, 95% CI:1.21, 2.66). The best fitting model predicting homophily on SSF included daily contact (OR 1.72, 95% CI:1.11, 2.68), education homophily (OR 1.75, 95% CI:1.15, 2.67), individual SSF consumption level (OR 0.55, 95% CI:0.29, 1.07), and SSB homophily (OR 1.89, 95% CI:1.23, 2.89). Shared common chronic disease risk behaviors within social networks can be predicted by network attributes, holding promise for multilevel approaches to behavior change.