Background: Observational data on the reduction in hospitalisations after rotavirus vaccine introduction in Belgium suggest that vaccine impact plateaued at an unexpectedly high residual hospitalisation rate. The objective of this analysis was to identify factors that influence real-world vaccine impact. Methods: Data were collected on hospitalisations in children aged < 5 years with rotavirus disease from 11 hospitals since 2005 (the RotaBIS study). The universal rotavirus vaccination campaign started late in 2006. A mathematical model simulated rotavirus hospitalisations in different age groups using vaccine efficacy and herd effect, influenced by vaccine coverage, vaccine waning, and secondary infection sources. The model used optimisation analysis to fit the simulated curve to the observed data, applying Solver add-in software. It also simulated an 'ideal' vaccine introduction maximising hospitalisation reduction (maximum coverage, maximum herd effect, no waning), and compared this with the best-fit simulated curve. Modifying model input values identified factors with the largest impact on hospitalisations. Results: Compared with the 'ideal' simulation, observed data showed a slower decline in hospitalisations and levelled off after three years at a higher residual hospitalisation rate. The slower initial decline was explained by the herd effect in unvaccinated children. The higher residual hospitalisation rate was explained by starting the vaccine programme in November, near the rotavirus seasonal peak. This resulted in low accumulated vaccine coverage during the first rotavirus disease peak season, with the consequential appearance of secondary infection sources. This in turn reduced the herd effect, resulting in a diminished net impact. Conclusions: Our results indicate that countries wishing to maximise the impact of rotavirus vaccination should start vaccinating well ahead of the rotavirus seasonal disease peak. This maximises herd effect during the first year leading to rapid and high reduction in ...