Introduction and objective:
Vaccine hesitancy is dynamic, with longitudinal evidence beyond two waves scarce. The aim of the study is to characterise, across a four-wave Polish panel (2021–2025), how stated vaccination intentions translate into actual vaccination, and to identify modifiable determinants of persistent non-vaccination versus delayed conversion.

Material and methods:
4,205 wave records from 2,559 unique Polish adults (Rise or Fall [ROF] panel, 2021/2022/2024/2025) were harmonised. The analytical cohort (n=859, ≥2 waves) was classified into five trajectory classes: Baseline vaccinated, Early convert, Late convert, Persistent unvax willing (PuW), and Persistent unvax hesitant (PuH; reference). Predictors were identified by Multinomial logistic regression and discrete-time hazard models.

Results:
By 2025, 83.4% of baseline-willing respondents were vaccinated vs. 38.9% of baseline-hesitant – a 44.5-percentage-point gap nearly identical to 2022 (45.7 pp). Persistent non-vaccination tracked higher fear of side-effects (hazard OR=0.63 per point; p<0.001) and conspiracy beliefs (OR=0.79; p<0.001); a Bachelor’s degree raised the conversion hazard 2.6-fold (p<0.001). Low fear emerged as a strong predictor of Baseline vaccinated (OR=6.0) and Early convert (OR=3.0). PuW had the highest mean HLS-Q12 (3.04; omnibus p=0.040), with a borderline multinomial association (OR=1.56; p=0.077); HLS-Q12 robustly predicted lower vaccination hazard (OR=0.58; p=0.001).

Conclusions:
Vaccine hesitancy in Poland is durable; the willing/hesitant behavioural gap persists. Post-rollout strategy should combine safety-centred dialogue targetting fear of side-effects with structural facilitators for the willing-but-unvaccinated subgroup.