Public skepticism toward childhood vaccination remains a major challenge for infectious disease control. During the COVID-19 pandemic, support for pediatric vaccination varied across social groups, yet most evidence relies on average treatment effects. We re-analyze a survey experiment from Italy and the UK (N = 6,354) that tested two message frames—risk to children vs. societal benefit (herd immunity)—to assess how they shift attitudes toward childhood COVID-19 vaccination. Using a Bayesian finite mixture model, we identify three latent response profiles, skeptical, indecisive, and supportive, and estimate how treatments affect the probability of class membership across subgroups. While non-parents, especially women, respond to messaging with sharp increases in pro-vaccine sentiment, parents tend to shift from skepticism toward more moderate positions. Our approach highlights how messaging alters the shape, not just the mean, of attitudinal distributions, with implications for precision communication in public health.