Vaccine Refusal and Hesitancy: An Agent-Based Modeling Approach

Background: Parents regularly cite pediatric care providers as an important source of vaccine information and counseling. Providers may be more or less tolerant of requests for alternative vaccine schedules or declined vaccines. Parental and provider preferences interact in a dynamic way to produce patterns of vaccine refusal within pediatric practices. These patterns have important implications for disease outbreak risk, as was seen in San Diego in 2008 when a pediatric medical office was the site of measles transmission.

Objectives: To construct an agent-based model of parent preferences for vaccination and provider tolerance for alternative schedules that simulates the resulting spatial and social clustering of unvaccinated children within certain medical practices.

Methods: The model is populated by agents (parents) who are assigned characteristics including vaccination preferences, network homophily, and provider preferences. In each tick of the model, agents make decisions about socializing with other agents, choosing a provider, and adhering to the vaccine schedule. Providers make decisions about whether to retain or dismiss non-vaccinating parents from their practice. Decision rules are based on the agent’s characteristics as well the characteristics of nearby agents and providers. The model is programmed in Netlogo, an agent-based modeling platform.

Results: Model output reveals the distribution of unvaccinated children in neighborhoods and physician practices over time and under varying scenarios of agent characteristic distributions and decision rules. For example, higher rates of physician tolerance for non-vaccination results in higher population-level exposure to unvaccinated children, but lower exposure for unvaccinated children.

Conclusions: This modeling approach can identify triggers or tipping points that result in particularly risky clusters of non-vaccination within certain groups.  These clusters have implications both for the risk of disease outbreak and policy and program interventions to preserve herd immunity. The model can also be used to test competing policy interventions related to immunization coverage.