Bimodal Contact Reductions and Social Homophily during COVID-19

Background: The COVID-19 pandemic disrupted social life and forced people to reconsider how, where, and with whom to spend time. These decisions are deeply personal and their intricacies are still poorly understood. Methods: To understand how people make such decisions, we conducted an online survey in summer 2023, collecting self-reported absolute contact numbers for four time points: 2019, 03/2020, summer 2021, and 01/2023. We analyzed the resulting contact data, focusing on the quantification of heterogeneities in reductions. Results: Analysis of the survey data revealed that the COVID-19 pandemic triggered substantial reductions in both the work and the leisure context. Mean reductions gradually decreased as time progressed, but by 01/2023 contact numbers remained below pre-pandemic levels. We found contact behavior to demonstrate heterogeneity in three different aspects. First, the distribution of contact reductions followed a bimodal pattern, with a distinct peak at either extreme: A large fraction of the survey participants initially strongly reduced their contacts, a smaller group maintained nearly normal contact levels, and the remainder of participants reduced their contacts intermediately. Consistent with the decrease of mean contact reductions, the relative sizes of these behavioral groups shifted over time, with participants relaxing their reductions incrementally. Second, we found risk perception to be an indicator for the strength of contact reductions: Risk-averse participants reduced their leisure contacts significantly more than risk-tolerant participants, resulting in a trend of both fewer and later COVID-19 infections. Neither age, gender, nor having a COVID-19- relevant comorbidity significantly influenced self-reported contact reductions. Third, the survey results provide evidence that social homophily persisted during the COVID-19 pandemic, revealing a correlation between participants' and their closest contacts' number of contacts during the COVID-19 pandemic. Risk-averse participants hereby especially preferred to maintain contact with equally careful individuals. Conclusions: Our study emphasizes the time-dependency and heterogeneity of contact reductions. On the one hand, our findings can easily be integrated into epidemiological models, improving their accuracy and predictive power. On the other hand, the results may guide the design of effective public health interventions, and help to predict and understand their effectiveness.