JMIR Publications

ABSTRACT

Background:

The COVID-19 pandemic has demonstrated the need for digital health interventions, such as those led by chatbots, to surpass physical, economic, and temporal barriers and deliver support to those in need. During this emergency context however, typical facilitators of participant recruitment (e.g., clinicians, healthcare organizations) are engaged with high priority pandemic response tasks and are limited in their ability to facilitate participant recruitment. To have maximum impact therefore, chatbot developers must understand the audience for such tools, and leverage social media to recruit participants independently.

Objective:

The current paper examines how the demographic and psychographic profile users seeking support during the pandemic influences their use of digital health interventions and which health areas they seek support.

Methods:

Utilizing Facebook data from the chatbot-led digital health intervention “Elena+: Care for COVID-19”, we report app downloads, user engagement metrics, and comments posted from all campaigns conducted in U.K., Ireland, U.S.A, Spain, Colombia, and Mexico between April 2020 and December 2021, including a quasi-experiment conducted in the U.K. and Ireland that ran eight advertisements targeting different health areas. To analyze Facebook comments, we utilized sentiment analysis. Using survey data collected in January 2022, we examined the role of participant health status across seven health areas in driving intention to use digital health interventions and link behavior (i.e., selecting to view a link to a digital health intervention). Data were analyzed using a structural equation model.

Results:

The Facebook advertising campaigns reached 1 381 897 individuals, resulting in 4’994 app downloads (63.1% female, 69.3% aged between 35-54) with similar patterns across other engagement metrics. From 164 comments posted on advertisements, 61.6% (n=101) were in Spanish. Of the 364 words classified in the sentiment analysis, prominent themes were fear (n=66), trust (n=62), and sadness (n=61). The quasi-experiment in the U.K. and Ireland resulted in 726 downloads: 39.9% of all downloads related to a mental health topic, with loneliness (17.1% each) and physical activity (16.5%) the single most popular topics. Results of the structural equation model showed mediation (i.e., that the relationship between link behavior and COVID-19 risk perceptions is mediated by intention to use) with indirect effects (P=.019, β=.348), and also moderated mediation (i.e., that the relationship between link behavior and mental resources is fully mediated by intention to use, but that this depends on the participants level of institutional trust) with conditional indirect effects (p=.017, β=.005).

Conclusions:

Findings outline demand for chatbot-led digital health interventions during the COVID-19 pandemic, and show both demographic (females, middle-aged) and psychographic (institutional trust, COVID-19 risk perceptions) factors influence participant recruitment. Results indicate the particular importance of providing mental health support during the COVID-19 pandemic.