Mediation analysis investigating the mechanisms of two school-based smoking prevention interventions in adolescents from Northern Ireland and Bogotá

Background: Reviews have highlighted a lack of evidence on how successful intervention strategies for adolescent smoking prevention can be effectively adapted for low-middle income countries (LMICs). The MECHANISMS study compared behavioral mechanisms between two school-based smoking prevention programs for adolescents in Northern Ireland (NI; a high-income setting) and Bogotá, Colombia (middle-income). ASSIST works via peer education and diffusion. Dead Cool uses conventional classroom pedagogy. Both interventions were previously trialed in the UK and were culturally adapted for Bogotá. We investigated whether changes in smoking/vaping outcomes differed by intervention or setting. Mediation analyses were conducted to test the hypothesized intervention mechanisms.

Methods: Full school year groups in 12 secondary schools participated during one semester (n=1344, target age 12-13 years). Outcomes included willingness to pay (WTP) to support anti-smoking norms, self-report and objectively measured smoking behavior, smoking intentions, susceptibility, knowledge, and attitudes. Mediators included injunctive and descriptive smoking/vaping norms, selfefficacy to resist tobacco, perceived risks and benefits, perceived behavioral control, and exposure to advertising. Structural equation models examined intervention and setting effects on outcomes. Multiple mediator models were conducted using the product-of-coefficients approach.

Results: Most significant intervention effects showed more anti-smoking outcomes for Dead Cool compared to ASSIST, although exposure to advertising was lower for ASSIST (media: unstandardized coefficient [b]=-1.19, p<0.0001, shops: b=-0.14, p=0.04). ASSIST peer supporters also improved their knowledge (b=0.31, p=0.03), self-report smoking behavior (b=0.10, p=0.07), and perceived addiction risks (b=0.31, p=0.01) compared to Dead Cool. Most significant setting effects showed more anti-smoking outcomes for NI versus Bogotá. However, WTP to support anti-smoking norms (b=0.48, p=0.01), self-report descriptive norms (b=0.23, p=0.01), and exposure to advertising in shops (b=-0.34, p<0.0001) were more anti-smoking in Bogotá. Several significant mediators showed suppressive mediating effects, suggesting there were important but unmeasured mediators.

Conclusions: Our results suggest school-based programs may be an appealing target for adolescent smoking prevention in LMICs and support using social norms strategies. Future research should consider additional mediating constructs for adolescent smoking prevention in LMICs (integrating intrapersonal, social, physical environmental, cultural, and political factors), investigate how to optimize the communication channels in peer education and diffusion programs, and provide testable mechanisms.