This study uses a pre-post design to evaluate the impact of vaping advertisement restrictions on the number of vaping advertisements surrounding secondary schools (n = 18) in London, Ontario. These eighteen schools represent all secondary schools in the two largest school boards in London (Thames Valley District School Board and London District Catholic School Board), for which students’ socio-demographic data were available.
Vape advertisement audits
To evaluate changes in vape advertising near schools, audits were conducted from December 2 to December 18, 2019 (pre-ban) and again from January 16 to February 6, 2020 (post-ban) within the area surrounding each school. Areas surrounding schools were defined by an 800-m Euclidean buffer around each school’s property boundary, which represents approximately a ten-minute walk for youth (Martin et al. 2019). Prior to the start of the audits, a database of potential vaping advertisement locations was created by identifying four types of sites where vaping advertisements could be displayed: (1) specialty shops (this included vape shops as well as cannabis shops, as sellers of vape products and devices); (2) convenience stores; (3) gas stations; and (4) billboards and transit shelters. Convenience stores and gas stations were identified from the Middlesex London Health Unit’s Public Health Inspector Database. Billboards and transit shelters were identified from the City of London’s Billboards dataset, available through their Open Data Portal. Online searches (including Google and Yellow Pages) were used to identify specialty vape and cannabis shops within the city. These data were uploaded to Collector for ArcGIS (ESRI, Redlands, CA, USA). Researchers used Collector on their smartphones, as it provided a convenient method to capture images of all vaping advertisements in the field and to log the number of advertisements at each location. Because online sources, such as Google, may have missed some specialty shops, a full street audit was completed of all commercial land use areas within 800 m of the schools.
Audits were conducted by trained researchers in teams of two, who surveyed the area within 800 m around each of the 18 schools to identify vaping advertisements. Once at a potential vape advertising location, the external advertisements of the property and building were examined to identify vaping advertisements. Advertisements were counted if they contained images of vaping products (e.g., devices, e-juice, flavour pods, brand names) or vaping-related phrases or words (e.g., vape, vaping, vaporizer, electronic cigarettes, e-cigarettes). The storefront signs of specialty shops were included. “No vaping” signs were excluded. Specialty retailer storefront signs are a potential location for vape promotion not specifically addressed under the Smoke-Free Ontario Act. Vaping advertisements that appeared more than once at the same retailer were counted individually.
No vaping advertisements were found on billboards or transit shelters either before or after the ban. Therefore, these location types were not included in further analysis.
The primary outcomes for this study were the number and density (advertisements/km2) of vaping advertisements near each school (within 800 m), pre- and post-ban. A weighted count was also derived for each school, based on the proportion of retailers with vaping advertisements from the total number of retailers (gas stations, convenience stores, and specialty shops) surrounding each school, to account for the fact that not all of the specified retailers sell vaping products (Giovenco et al. 2016a).
Vaping advertisements were also measured at 400 m to examine whether effects differed when potential advertising exposure was measured closer to schools. All measures were calculated using ArcMap version 10.7 (ESRI, Redlands, CA, USA).
School socio-demographic characteristics
Socio-demographic characteristics for each of the 18 schools were calculated by taking the proportion of students in the school who resided in a dissemination area (DA) classified as the most deprived quintile based on the Canadian Index of Multiple Deprivation (CIMD) for Ontario (CIMD-Ontario data are from Statistics Canada) (i.e., % of students per school who live in a deprived area). DAs are the smallest administrative area for which socio-demographic data are available from the Canadian Census; they typically contain 400–700 people (Statistics Canada 2019). Data on students’ DA of residence was obtained from an anonymized school bus eligibility database, which included residential postal codes for every student by school (but did not include any other information). The postal codes were used to link students to their respective DAs.
The CIMD was based on the 2006 Canadian Marginalization Index (CAN-Marg) (Matheson et al. 2012). The CIMD-Ontario is an index derived for each DA, specifically for Ontario, from the 2016 Census and is comprised of four dimensions of deprivation: (1) residential instability; (2) economic dependency; (3) ethno-cultural composition; and (4) situational vulnerability. The dimension of residential instability takes into consideration the tendency of inhabitants of a neighbourhood to change over time and includes five indicators (i.e., proportion of: dwellings that are apartment buildings; dwellings that are rented; persons living alone; population who moved in last 5 years; population who are single, divorced, separated, or widowed). Economic dependency considers reliance on the workforce or a dependence on sources of income other than employment; this dimension also includes five indicators (i.e., proportion of population who are: aged 65 and older; aged 15 and older who are not participating in the labour force; aged 0–14 and 65 and older divided by population aged 15–64; unemployed; receiving government transfer payments). Ethno-cultural composition considers the community make-up of immigrant populations and those who self-identify as a visible minority and includes four indicators (i.e., proportion of population who: are foreign-born; self-identify as visible minority; have no knowledge of English or French; are recent immigrants). Situational vulnerability takes into account factors such as educational attainment, housing conditions, and Indigenous identity; this dimension includes three indicators (i.e., proportion of: dwellings needing major repair; population who identify as Indigenous; population who are aged 25–64 without a high school diploma). Factor analysis was used to derive the CIMD-Ontario. A set of initial variables was selected for the factor analysis as established by the CAN-Marg and through expert consultation. Different indicator variables loaded within the four dimensions of deprivation to create the final index; some indicators were not included because they were not significantly correlated with any factors. (See Statistics Canada (2019) for more details.)
Commercial land use
The percent of land classified as commercial within the 800-m buffer surrounding each school was derived from City of London land use data. Commercial land use was considered as a co-variate because of the possibility that associations between school socio-demographic characteristics and number or density of vaping advertisements could be a function of municipal land use zoning and general commercial activity in the areas surrounding schools.
To examine changes in vape advertising surrounding secondary schools, paired t tests were conducted on the vaping advertisement measures for 800-m and 400-m buffers. Cohen’s d was used to determine the effect size for the paired t test, where 0.20 indicates a weak effect, 0.50 a moderate effect, and 0.80 a large effect (Cohen 1988). To address any non-normality in the change from pre- to post-ban, we also utilized Yuen’s robust paired test. Yuen’s method addresses issues that arise from violating the normality assumption (Fradette et al. 2003).
A series of Kendall’s tau (rΤ) tests were used to assess correlations between vaping advertisement measures and school socio-demographic characteristics, before and after the vaping advertisement ban. Kendall’s tau is an appropriate statistic for data with a small n (Field et al. 2012). As recommended by Cohen (1988), effect sizes for Kendall’s tau correlations were considered small, medium, and large at 0.10, 0.30, and 0.50, respectively. Partial correlations were also examined, adjusting for commercial land use density (Kim 2015), as commercial land use density may have a confounding effect on the relationship between the number of vaping advertisements surrounding schools and the deprivation dimensions.
A p value < 0.05 was considered statistically significant. All analyses were conducted in R version 3.5.2.