Towards a predictive model of COVID-19 vaccine hesitancy among American adults.

Designing effective public health campaigns to combat COVID-19 vaccine hesitancy requires an understanding of i) who the vaccine hesitant population is, and ii) the determinants of said population's hesitancy. While researchers have identified a number of variables associated with COVID-19 vaccine hesitancy that could inform such campaigns, little is known about the cumulative or relative predictive power of these factors. In this article, we employ a machine learning model to analyze online survey data collected from 3353 respondents. The model incorporates an array of variables that have been shown to impact vaccine hesitancy, allowing us to i) test how well we can predict vaccine hesitancy, and ii) compare the relative predictive impact of each covariate. The model allows us to correctly classify individuals that are vaccine acceptant with 97% accuracy, and those that are vaccine hesitant with 72% accuracy. Trust in and knowledge about vaccines is, by far, the strongest predictor of vaccination choice. While our results demonstrate that public health campaigns designed to increase vaccination rates must find a way to increase public trust in COVID-19 vaccines, our results cannot speak to the malleability of such beliefs, nor how to enhance trust.

The belief that politics drive scientific research & its impact on COVID-19 risk assessment.

We use survey data collected from 12,037 US respondents to examine the extent to which the American public believes that political motives drive the manner in which scientific research is conducted and assess the impact that such beliefs have on COVID-19 risk assessments. We find that this is a commonly held belief and that it is negatively associated with risk assessments. Public distrust in scientists could complicate efforts to combat COVID-19, given that risk assessments are strongly associated with one's propensity to adopt preventative health measures.