The co-evolution of social networks and health-related behaviors in university students: A two-year longitudinal...

Health-related behaviors adopted during emerging adulthood have important implications for long-term health trajectories. While homophily has been widely documented for substance use, less is known about whether behavioral similarity arises from contagion or selection across a broader range of health-related behaviors. Distinguishing between these mechanisms has direct implications for intervention design. We conducted a four-wave longitudinal sociocentric cohort study of undergraduate freshmen at two universities in Guangzhou, China (October 2022-July 2024). Eligible participants from five schools completed online surveys at baseline and three follow-up waves. Measures included physical activity, fruit and vegetable consumption, sleep quality, smoking, drinking, sugar-sweetened beverage (SSB) consumption, and video gaming. Friendship networks were assessed using roster-based nominations. Stochastic Actor-Oriented Models were used to examine the co-evolution of networks and behaviors and to estimate the relative contributions of contagion and selection processes. Parameter estimates were pooled across schools using meta-analytic procedures. The analytic sample included 609 students at baseline (mean age = 18.7 years; 65.2% female). Significant contagion effects were observed for physical activity, fruit consumption, sleep quality, smoking, drinking, and video gaming. In contrast, homophily in SSB consumption was primarily explained by selection processes. Overall, contagion accounted for a larger share of observed homophily than selection, with stronger effects for health-risk behaviors than for health-promoting behaviors. Mechanisms generating peer similarity vary by behavior. These findings suggest that mechanism-informed interventions, network diffusion strategies for influence-sensitive behaviors and structural or individual-level approaches for selection-driven behaviors, may help inform intervention design during the transition to university, although this study did not test network-based interventions.