JMIR Publications

ABSTRACT

Background:

Despite COVID-19 vaccine mandates, many chose to forgo vaccination – raising questions about the psychology underlying these choices. Research shows that risk behaviors are important for vaccination choice, however no studies have integrated 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 dataset to predict COVID-19 vaccination status.

Methods:

We surveyed 3,476 adults across the United States in December 2021. Participants responded to questions pertaining to demographics, COVID-19 vaccine uptake (i.e., whether participants were fully vaccinated at the time of the survey), and COVID-19 precaution behaviors. Participants were also asked to compete a 48-item picture rating task. Ratings from this task were computationally modeled using relative preference theory to produce 15 features that quantitatively define biases in judgment. These judgment variables, along with demographics, were compared between those who were fully vaccinated and those who were not, i.e., vaccination status (Wilcoxon rank sum test α=0.05). A balanced random forest classifier was used to test how well judgment and demographic variables predicted vaccination status. Mediation and moderation were used to assess the statistical mechanisms underlying prediction.

Results:

Our small variable set predicted vaccine uptake with high precision (87.8%) and moderate-high accuracy (70.8%) and ROC AUC (79.0%). Collectively, 63% of the feature importance resulted from the 15 judgment variables. Furthermore, age, income, and education level statistically mediated relationships between judgment and vaccine uptake proposing a mechanism underlying prediction.

Conclusions:

Findings demonstrate the importance of biases in judgment for vaccine uptake, suggesting that vaccine education and messaging might target varying judgment profiles to improve vaccine use.