Background:
School-based prevention activities continue to be an important tobacco control resource, however there is little guidance for school-based tobacco control programming within Ontario. The objective of this study is to identify real-world changes in school-based tobacco control programs or policies in the COMPASS study and examine of those interventions (natural experiments) had any impact on the school-level prevalence of smoking susceptibility and current smoking over time.

Methods:
This paper uses longitudinal school-level smoking behaviour data from Year 1 (Y1: 2012–13) and Year 2 (Y2: 2013–14) of the COMPASS study. Changes to school-level tobacco control programs and policies were measured using the COMPASS School Programs and Policies Questionnaire and knowledge broker follow-up interviews. Quasi-experimental tests of proportion and difference-in-difference models were used to evaluate the impact of the interventions identified between Y1 and Y2 on school-level prevalence of smoking susceptibility among never smokers and current smoking.

Results:
Between Y1 and Y2, 17 schools reported a change in their tobacco control programming or policies. In four of the intervention schools, the increase in the within-school prevalence of susceptible never smokers between Y1 and Y2 was significantly greater than the natural change observed in the control schools. In five of the intervention schools, the decrease in the within-school prevalence of current smokers between Y1 and Y2 was significantly greater than the natural change observed in the control schools. Only two of the new interventions evaluated (both focused on policies of progressive punishment for students caught smoking on school property), were associated with significant desirable changes in both smoking susceptibility and current smoking between Y1 and Y2.

Discussion:
Interventions specific to effective and enforced tobacco control were the most common and consistently had the desired impact on the school-level prevalence of smoking susceptibility and current smoking. Due to the variation in the types of interventions implemented and their effectiveness, additional evaluation evidence is necessary to determine the most successful activities and contexts among individual students. The results presented here highlight which of these real-world promising interventions should be further evaluated using the longitudinal individual-level data in COMPASS over time.