Background: Vaccine hesitancy—defined by the World Health Organization as the delay in acceptance or refusal of vaccination despite availability—has emerged as one of the top ten threats to global health. Community-based educational campaigns are widely deployed to address hesitancy, but rigorous causal evidence of their effectiveness remains sparse. Observational evaluations are compromised by confounding from self-selection, access to care, income, and political ideology. 
Objective: This study demonstrates how two quasi-experimental methods—instrumental variable (IV) analysis and regression discontinuity (RD) design—can be applied to estimate the causal effect of a statewide community health worker (CHW) vaccine education program on childhood vaccination uptake in a setting where randomization was not feasible. 
Methods: We analyzed administrative vaccination records for 68,714 children aged 12–35 months across 412 communities in a Midwestern state (2016–2021). The CHW program was allocated using a composite community risk score; communities scoring above a threshold received the intervention. We exploited this assignment mechanism using a fuzzy RD design at the eligibility threshold and, independently, used distance to the nearest CHW training center as an instrument for program participation. Outcomes included completion of the combined 7-vaccine series by 24 months and the measles-mumps-rubella (MMR) first-dose rate by 15 months. 
Results: The RD estimate of the local average treatment effect (LATE) on 7-vaccine series completion was 8.6 percentage points (95% CI: 3.4–13.8, p = 0.001). The IV estimate was 7.1 percentage points (95% CI: 2.8–11.4, p = 0.001). For MMR first-dose by 15 months, the RD estimate was 6.2 pp (95% CI: 2.1–10.3) and the IV estimate was 5.1 pp (95% CI: 1.4–8.8). Both methods produced estimates that were concordant and substantially larger than naïve OLS estimates, consistent with attenuation from confounding bias. 
Conclusions: Community-based CHW vaccine education programs produce meaningful increases in childhood vaccination rates. The concordance of IV and RD estimates strengthens confidence in a causal interpretation. These methods should be more widely adopted in public health program evaluation to overcome the limitations of observational comparisons.