Outcomes

When a 1 min interval contained any of the five types of tobacco occurrences, it was classified as having tobacco content. The number of total tobacco occurrences in the sample was calculated by summing all five tobacco occurrence types across all minutes. The hourly frequency is the number of occurrences divided by the total hours of programming and is presented by occurrence type.

Reach is the population exposure to tobacco imagery—a measure of the frequency of tobacco impressions seen by the audience aged 4–24 years. Reach was calculated by the type of occurrence, multiplying the number of tobacco occurrences in each 1 min interval by the estimated average number of people aged 4–24 years watching at that moment. Live audience viewing figures were collected and provided by Kantar IBOPE Media Chile from their peoples’ meter study (https://www.kantaribopemedia.cl/). Their sample represents the non-institutionalised people of 4+ years of age living for longer than 6 months in households with a TV of the most populated municipalities of Chile without consideration of their nationality. It represents a total of 7 702 676 individuals or approximately 40% of the Chilean population. The sample excludes households living in extreme poverty.

Programming airtime was summed across all 15 TV channels, along with estimated hours of exposure to the 4–24 years demographic and persons of all ages. These metrics are also reported by channel and programme characteristics. The total number of tobacco occurrences and impressions received by the population of 4-year to 24-year olds during the observation period is reported, along with the average number of impressions viewed per hour, overall and by channel and programme characteristics.

Using multivariable weighted logistic regression, we assessed the odds of exposure to tobacco impressions among 4-year to 24-year olds as a function of channel and programme characteristics. The dependent variable was whether a 1 min interval contained a tobacco occurrence. Independent variables included the type of channel the interval was broadcast on (broadcast or pay), interval programme genre, whether the programme in that interval was produced in Chile and whether the interval was broadcast during the minor’s protection schedule. Type of channel and country of production were collinear because 99.9% of pay channel productions were non-Chilean, so the type of channel was dropped from the analysis. We ran four models; the four dependent variables were: any tobacco occurrence, explicit tobacco use, smoke-free violation and tobacco brand appearance. The regression used robust variance estimates to account for the clustering of tobacco occurrences within programme units, defined as self-contained content or narrative with a start and an end during the same day. The analysis was weighted to account for 4–24 viewership, so that the ORs reflect odds of exposure among 4-year to 24-year olds. ORs were considered statistically significant when the p value was <0.05. The analysis was conducted using STATA V.13.