We used SPSS package version 23 (SPSS Inc., Chicago IL, USA) to analyse the data. At the three-month follow-up, and with a response rate of 54.1%, we tested for patterns in the missing data [19 ,20 (link)]. To this end, we used the Chi-square test to compare the characteristics of the participants and non-participants at the three-month follow-up (Table 1 ). Because we detected no statistically significant differences between the groups, any patterns in the missing data were considered Missing Completely At Random (MCAR) (i.e., individuals with missing data at the follow-up were randomly scattered throughout the sample). We therefore conducted the analysis using available cases without fear of bias in the findings [21 ].
To assess the effect of variables such as the baseline knowledge score, time after viewing the campaign, viewing the campaign and socio-demographic variables on the mean score for post-campaign knowledge we conducted a generalised estimating equation (GEE) analysis. While the GEE analysis revealed significant interaction between the two main factors of time after viewing the campaign and group (viewing the campaign or not), we computed the estimated mean difference in post-campaign knowledge scores between the participants who had viewed the campaign and those who had not separately at each measurement point [22 (link)].
To assess the effect of variables such as the baseline knowledge score, time after viewing the campaign, viewing the campaign and socio-demographic variables on the mean score for post-campaign knowledge we conducted a generalised estimating equation (GEE) analysis. While the GEE analysis revealed significant interaction between the two main factors of time after viewing the campaign and group (viewing the campaign or not), we computed the estimated mean difference in post-campaign knowledge scores between the participants who had viewed the campaign and those who had not separately at each measurement point [22 (link)].