Menu Close

Interview: Rapid testing of COVID-19 messaging

June 1, 2021

Predoctoral Trainee Sophie Bartels (Health Behavior) is the lead author on a new research study on rapid testing of COVID-19 messaging published in Public Health Reports.Sophia Bartels

The paper describes the development of an interdisciplinary rapid message testing model to quickly create, test, and share messages with public health officials for use in health campaigns and policy briefings.

We spoke to Sophie about this research and the findings.

Tell us a little bit about the 4-step model that you developed and tested. 

Our 4-step rapid message testing model was developed during the height of the COVID-19 pandemic in the United States to respond to the North Carolina Department of Health and Human Services (NCDHHS) need for evidence-based messages about social distancing that they could use in their health campaigns and policy briefings.

Our interdisciplinary team from across UNC developed and executed the 4-step model which involved 1) message creation 2) survey development 3) survey administration (using Amazon Mechanical Turk (MTurk)), and 4) data analysis and presentation to NCDHHS. These steps were executed 4 times, each over a 7-day period, with findings from each week’s survey and new mandates from NCDHHS informing the subsequent week’s survey.

What were the most effective messaging strategies? 

Overall, we found that messages that centered around protecting oneself or others from COVID-19 were more effective at encouraging desire to social distance than norms or fear-based messages. The most effective messaging strategies were pairing social distancing behaviors with motivations for social distancing.

For example, when we added “protect your grandmother, your neighbor with cancer, and your best friend with asthma” (the motivation) to the social distancing behavior, “stay 6 feet apart from others when out in public,” respondents rated the message substantially higher in terms of desire to social distance than the message about the behavior alone.

How did you test messaging strategies aimed at different populations? 

Through our message testing, we aimed to create messages that would resonate with all North Carolina residents, including a few populations that NCDHHS determined as “key populations” including rural residents, populations that are disproportionately impacted by COVID-19, and people who were perceived as less likely to practice social distancing behaviors.

To create messages that might resonate with these populations, we integrated key theoretical principles for behavior change (e.g., attitudes, risk perceptions) into the messages that we tested, and disaggregated weekly findings and reports by these key populations to determine how well our messages were scoring with different groups.

How is this applicable to other public health emergencies? 

This model is applicable to other public health emergencies because it can be used to reduce the time that it takes for evidence-based public health messages to reach policymakers in rapidly changing public health emergency scenarios.

This model responds to one of the major barriers to evidence-based policymaking and emergency response: the amount of time that it takes for new evidence to be disseminated and used.

Our model both allows evidence-based messages to quickly (within 7 days) reach policymakers and public health departments and to be responsive to rapidly changing healthcare department needs or conditions during a public health emergency.

What do you plan to do next?

Next, I would like to test this model with different types of behaviors, such as vaccine uptake, to further refine it to ensure its utility beyond social distancing behaviors. Additionally, I would like to explore ways to reach a more diverse sample as part of this message testing model, given that MTurk participants are not representative of the North Carolina population and tend to overrepresent young, white individuals.

Integrating additional data collection techniques, like participatory research methods, at key points in the model could help ensure better representation of marginalized groups in our sample.