Assessing the Validity of Confirmed Ad Recall Measures for Public Health Communication Campaign Evaluation

ABSTRACT

Evidence strongly suggests that public health communication campaigns can succeed in changing health-related cognitions and behaviors. For many evaluation studies, however, inferences of campaign effects are only valid to the extent that measures of campaign exposure are themselves valid. This study compares the validity of “aided” and “confirmed” ad recall measures in the context of a statewide tobacco countermarketing campaign using data from the Florida Anti-Tobacco Media Evaluation (FAME) surveys. Both aided and confirmed ad recall measures exhibited positive associations with cumulative gross ratings points (GRPs), a measure of the relative availability of specific countermarketing ads on broadcast television. In addition, both recall measures were significant predictors of campaign-targeted beliefs. Confirmed ad recall, however, was not a significantly better predictor of cumulative GRPs or campaign-targeted beliefs than aided ad recall, and the magnitude of association between both recall measures and targeted beliefs was quite small. These findings raise questions about the marginal utility of confirmed ad recall measures, compared with aided ad recall, in public health communication campaign evaluations. Nevertheless, results do provide evidence that both aided and confirmed ad recall measures are valid measures of campaign exposure.

Notes

* = p < 0.05;

** = p < 0.01;

*** = p < 0.001.

¹Preliminary analyses revealed that linear estimates were the most parsimonious approximation for the observed relationship between both recall measures and cumulative GRPs.

²Bivariate analyses between potential confounders and the main variables of study interest revealed that some relationships were not adequately captured with a linear approximation. Therefore, variable transformations among potential confounders were used where necessary, including a quadratic transformation for age, a cubic transformation for average daily television viewing, indicator variables for each wave (secular trends), and indicator variables for the number of national “truth” ads confirmed.

Note. Multivariate models also included indicators for each survey wave (wave 1 omitted; coefficients not shown). Preliminary models also included controls for age (quadratic transformation), school enrollment, and hours of daily television viewing (cubic transformation), but these variables were not statistically significant in the full model and were thus dropped.

*Denotes coefficient significantly different from zero, p < 0.05;

** = p < 0.01;

*** = p < 0.001.