By the end of 2021, approximately 15% of U.S. adults remained unvaccinated against COVID-19, and vaccination initiation rates had stagnated. We used unsupervised machine learning (K-means clustering) to identify clusters of unvaccinated respondents based on Behavioral and Social Drivers (BeSD) of COVID-19 vaccination and compared these clusters to vaccinated participants to better understand social/behavioral factors of non-vaccination. The National Immunization Survey Adult COVID Module collects data on U.S. adults from September 26-December 31,2021 (n = 187,756). Among all participants, 51.6% were male, with a mean age of 61 years, and the majority were non-Hispanic White (62.2%), followed by Hispanic (17.2%), Black (11.9%), and others (8.7%). K-means clustering procedure was used to classify unvaccinated participants into three clusters based on 9 survey BeSD items, including items assessing COVID-19 risk perception, social norms, vaccine confidence, and practical issues. Among unvaccinated adults (N = 23,397), 3 clusters were identified: the "Reachable" (23%), "Less reachable" (27%), and the "Least reachable" (50%). The least reachable cluster reported the lowest concern about COVID-19, mask-wearing behavior, perceived vaccine confidence, and were more likely to be male, non-Hispanic White, with no health conditions, from rural counties, have previously had COVID-19, and have not received a COVID-19 vaccine recommendation from a healthcare provider. This study identified, described, and compared the characteristics of the three unvaccinated subgroups. Public health practitioners, healthcare providers and community leaders can use these characteristics to better tailor messaging for each sub-population. Our findings may also help inform decisionmakers exploring possible policy interventions.
Keywords: COVID-19 vaccines; Cluster analysis; Health communication; Health policy; SARS-CoV-2; Vaccine hesitancy.
Published by Elsevier Inc.