Modelling Predictors of Homophily on Perceived Oral Health Status Among Social Network Ties in a Population of Public Housing Residents.

PURPOSE: Individual behaviours are often shared within social networks (homophily), suggesting network-level interventions hold promise for health promotion. Yet, little is known about oral health homophily. This study aimed to identify individual- and network-based predictors of oral health homophily among individual's (ego) social networks of public housing residents. METHODS: Respondents self-reported demographics, oral health status and associated risk behaviours (n = 277). They named social contacts (alters), reported on relationship attributes, demographics and behavioural characteristics (n = 889). Hypothesised predictors of oral health homophily included relationship attributes (e.g., contact frequency), respondent-level and shared characteristics. Oral health homophily was modelled using multilevel (hierarchical) logistic regression evaluating model attributes (AIC) to determine gains in explanatory power. RESULTS: Relationship strength, including high frequency of shared meals and contact, was associated with higher odds of oral health homophily (OR [95% CI]: 1.92 [1.05, 3.52] and 1.62 [1.00, 2.63], respectively). The best performing model included daily shared meals and contact, respondent age, smoking and oral health status. CONCLUSIONS: Oral health homophily is predicted by relationship strength and 'excellent/very good/good' oral health. Respondents with poorer oral health and a smoking history were less homophilous in oral health. Multilevel interventions targeting oral health outcomes may benefit from accounting for social relationships.