"Everything had stopped, no meeting, no gathering": Social interactions during the COVID-19 pandemic in the Central African Republic, the Democratic Republic of Congo, and Bangladesh.

Understanding the spread of COVID-19 in humanitarian and fragile settings is challenging for many reasons, including the lack of data on social dynamics and preventive behaviors during an epidemic. We investigate social interactions in three such settings - Democratic Republic of the Congo (DRC), Central African Republic (CAR), Cox's Bazar (CXB), Bangladesh - and how they changed during the first year of the pandemic. This comparative mixed-methods study uses a representative household survey and focus group discussions or key informant interviews in each site. Descriptive weighted analysis of survey responses was conducted; multivariate logistic regression identified factors associated with changes in social interactions. Thematic analysis was conducted on qualitative data. Nearly all participants had social interactions the day before the survey, although the average number of daily interactions was low. Interactions primarily occurred indoors, at home and without masks. We saw a discrepancy between knowledge about and practice of preventive behaviors. Most respondents reported interacting less often (77.3% CXB, 86.7% CAR, 58.8% DRC) and having shorter meetings (80% CXB, 77.8% CAR, and 47.8% DRC). Reluctance towards the COVID-19 vaccine was a risk factor for non-compliant behaviors in CAR (OR increased frequency = 3.51, 95%CI = 1.41-8.75; OR increased duration = 2.47, 95%CI = 1.15-5.29) and DRC (OR increased duration = 3.06, 95%CI = 1.71-5.49), likely pointing to distrust towards institutional policies. Respondents from IDP communities in DRC were less likely to reduce the frequency of interaction, likely because living conditions did not facilitate physical distancing. Increased knowledge in CXB was associated with compliant behavior (for 1pt-increase: OR increased frequency = 0.47, 95%CI = 0.32-0.68; OR increased duration = 0.46, 95%CI = 0.31-0.69). Understanding social dynamics is fundamental to predict infectious disease spread, particularly in humanitarian settings. More evidence is needed to understand behaviors influencing disease dynamics and drivers of behaviors, including trust in authorities, social, and economic factors. Peace, community engagement, and reduction of misinformation remain critical for epidemic responses in humanitarian settings.