Contagious Misinformation Trial

The change in the belief in misinformation will be measured through Yes/No questions in the baseline and follow-up surveys. Using logistic regression models, the prevalence of the belief in misinformation about 2 aspects of an infectious disease compared to the control group will be analysed using an intention-to-treat and per-protocol analysis

To test whether the interventions might inadvertently have promoted the belief in misinformation, the same yes/no questions as the primary outcome will be used in ordinal logistic regression models to analyse if the there has been a backfire effect in the intervention groups, compared to the control group.

The Yes/No questions of the primary outcomes will be tested in an as-treated analysis. The investigators will only include participants who can correctly recall the basic storylines of the audio messages.

Using an open question, asking the respondent to name up to 3 preventive methods, a score will be created. For every correct answer, the participants gets a point, and one point is subtracted for every wrong answer; leading to a potential score of -3 to +3. Ordinal logistic regression models will be fitted to compare the scores of the intervention groups with the control group.

The question asking whether the participant has discussed the content of the audio messages will be used for this analysis, together with the question about how often the participant discussed health issues with family/friends. Logistic regression models will be fitted to test whether the interventions have influenced health-related discussions with family or friends, as compared to the control group.

To understand if the intervention works outside of WhatsApp, 60 additional people will be recruited who do not have WhatsApp. Participants will instead be called and listen to the audio dramas on the phone; 30 will listen to the audio drama of intervention group 1 and 30 will listen to the audio drama of intervention group 2. The two primary outcomes will be analysed similar to the primary outcome analysis (ITT and per protocol) among the 2 groups and compared to the control group, as well as to their equivalent group of respondents with WhatsApp.

Participants will answer 3 questions about their perceived self-efficacy on 3 specific preventive behaviours. Answers are on a 5-item scale: from not at all true to exactly true. Ordinal logistic regression models will be specified to test whether the interventions had an influence on people's self-efficacy about three specific preventive behaviours, compared to the control group.

Risk perception about the infectious disease will be measured with a question that asks how likely it is that the participants gets the disease in the next year. A question which asks what kind of actions the participants has undertaken, or is planning to undertake, to prevent infection with the disease - will be used to assess if and what kind of actions are taken. Several analyses will be carried out to test whether the intervention influenced risk perception and preventive methods. Furthermore, analyses will be carried out to determine if the a change in risk perception influenced preventive methods.

In the follow-up survey, there is a question asking if the participant feels like he/she learned from the audio messages (yes/no question). A Chi-Square analysis will be done to determine whether those who feel like they learned also learned objectively, using the two primary outcomes.