Early Indicators of COVID-19 Infection Prevention Behaviors: Machine Learning Identifies Personal and Country-Level Factors

Abstract

The Coronavirus is highly infectious and potentially deadly. In the absence of a cure or a vaccine, the infection prevention behaviors recommended by the World Health Organization constitute the only measure that is presently available to combat the pandemic. The unprecedented impact of this pandemic calls for swift identification of factors most important for predicting infection prevention behavior. In this paper, we used a machine learning approach to assess the relative importance of potential indicators of personal infection prevention behavior in a global psychological survey we conducted between March-May 2020 (N = 56,072 across 28 countries). The survey data were enriched with society-level variables relevant to the pandemic. Results indicated that the two most important indicators of self-reported infection prevention behavior were individual-level injunctive norms—beliefs that people in the community should engage in social distancing and self-isolation, followed by endorsement of restrictive containment measures (e.g., mandatory vaccination). Society-level factors (e.g., national healthcare infrastructure, confirmed infections) also emerged as important indicators. Social attitudes and norms were more important than personal factors considered most important by theories of health behavior. The model accounted for 52above the performance of psychological models of health behavior. These results suggest that individuals are intuitively aware that this pandemic constitutes a social dilemma situation, where their own infection risk is partly dependent on the behaviors of others. If everybody engaged in infection prevention behavior, the virus could be defeated even without a vaccine.