War, Belief, and the Needle: An Agent-Based Analysis of Vaccine Hesitancy in Ukraine

Background Vaccine uptake in conflict zones is hindered by disrupted supply chains, instability, and eroded trust, leaving millions unvaccinated. Traditional models overlook behavioral and social dynamics shaping vaccine attitudes. This study integrates the 5C model into a spatially explicit agent-based model, using survey data from conflict-affected populations in Ukraine to simulate evolving beliefs and design targeted interventions to strengthen equity and trust. Methods Data from a survey conducted in Ukraine (n = 2,526) were analyzed utilizing the 5C mode. This encompasses confidence, complacency, constraints, calculation, and collective responsibility. Respondents were categorized into four profiles: Deliberators, Acceptors, Indifferent, and Barriered. These clusters informed the development of a spatially explicit agent-based model (ABM), wherein agents were geolocation, access, and exposure variables related to war. The ABM simulated 40 evolutionary steps of beliefs under Low, Moderate, and High Access conditions. Cluster validity was assessed through silhouette width and the Calinski-Harabasz (CH) indices, while principal component analysis (PCA) was used for dimensionality reduction. Results Survey-based estimates revealed striking spatial heterogeneity in baseline vaccination coverage, forming the foundation for agent-based simulations. The ABM projected notable increases in acceptance, with resistant regions converging toward higher uptake under improved accessibility. Behavioral clustering identified four groups. Acceptors, Deliberators, Indifferent, and Barriered, each exhibiting distinct acceptance trajectories. Simulations showed that access strongly determined adoption speed, while peer influence accelerated convergence. High access conditions produced near-universal uptake, whereas low access was sustained by resistant groups. Conflict exposure and access jointly shaped tolerance shifts, explaining 75% of the variance in belief states. Validation metrics confirmed robust clustering, highlighting the model’s capacity to capture dynamic, context-specific drivers of vaccine acceptance in fragile settings. Conclusion The most compelling determinant of vaccine acceptance was equitable access, with peer influence and exposure to conflict facilitating transitions toward acceptance. Efforts to expedite adoption in insecure regions should focus on addressing structural inequalities and promoting social diffusion.