Predicting COVID-19 Vaccination Uptake Using a Small and Interpretable Set of Judgment and Demographic Variables: A Population Survey.

BACKGROUND: Despite COVID-19 vaccine mandates, many chose to forgo vaccination - raising questions about the psychology underlying how judgment affects choice. Research shows that judgments about reward and aversion are important for vaccination choice, however no studies have integrated such cognitive science with machine learning to predict COVID-19 vaccine uptake.

OBJECTIVE: This study aimed to determine the predictive power of a small, but interpretable set of judgment variables using three machine learning algorithms to predict COVID-19 vaccine uptake, and then interpret what profile of judgement variables was important for this prediction.

METHODS: We surveyed 3,476 adults across the United States in December 2021. Participants answered demographic, COVID-19 vaccine uptake (i.e., whether participants were fully vaccinated at the time of the survey), and COVID-19 precaution questions. Participants also completed a picture rating task using images from the International Affective Picture System. Images were rated on a Likert-like scale to calibrate the degree of liking and disliking. Ratings were computationally modeled using relative preference theory to produce a set of graphs for each participant (minimum R2>0.8). Fifteen judgment features were extracted from these graphs - two of which were analogous to risk and loss aversion from behavioral economics. These judgment variables, along with demographics, were compared between those who were fully vaccinated and those who were not. Three machine learning approaches [random forests, balanced random forests (BRF), and logistic regression] were used to test how well judgment, demographic, and COVID-19 precaution variables predicted vaccine uptake. Mediation and moderation were then implemented to assess statistical mechanisms underlying successful prediction.

RESULTS: Age, income, marital status, employment, ethnicity, education level, and sex all differed by vaccine uptake (Wilcoxon rank sum, chi-square P<.001). The majority of judgment variables also differed by vaccine uptake (Wilcoxon rank sum P<.05). Similar AUROC was achieved by the three machine learning frameworks, although random forests and logistic regression produced specificities between 30-38% (versus 74.2% for BRF) indicating lower performance predicting unvaccinated participants. BRF achieved high precision (87.8%) and AUROC (79.0%) with moderate-high accuracy (70.8%), and balanced recall (69.6%) and specificity (74.2%). It should be noted that for BRF the negative predictive value was below 50% despite good specificity. From BRF and random forests, 63-75% of the feature importance came from the 15 judgment variables. Further, age, income, and education level mediated relationships between judgment variables and vaccine uptake.

CONCLUSIONS: Findings demonstrate the underlying importance of judgment variables for vaccine choices and uptake, suggesting that vaccine education and messaging might target varying judgment profiles to improve uptake. These methods could also be used to aide vaccine roll outs and healthcare infrastructure by providing location-specific details (e.g., areas that may experience low vaccination and high hospitalization).