OBJECTVES Pre-launch economic models are constructed to simulate long-term changes in costs and effects.  Typically Markov cohort models are used, whereas the input often available to parameterize the models is obtained from cross-sectional, annual, population data. The question is how to make the link and reconcile results from long-term cohort models with annual observed population data? An illustration is given with modelled and observed hospitalisations due to rotavirus related acute gastroenteritis. METHODS The spread of hospitalisations of children up to the age of 5 years, observed over a one -year period follows a normal distribution (seasonality of the infection) with a peak around February March each year. The assessment is done by 1-year age-groups (0 to 1y; 1 to 2y; 2 to 3y; 3 to 4y; 4 to 5y). The parameters of this normal distribution are used to construct an overall modelled population density curve with the same annual spread. Within this construction the weekly spread of hospitalisations by age follows the density curve of the cohort model with an age-specific Weibull distribution. To compare the model results with the observation we analyse the age-group spread of hospitalisations but also the results following the introduction of a specific intervention such as vaccination. RESULTS Pre-vaccination, the fit of the age-related spread of hospitalisations modelled using the population model to the observed data was compelling (regression-scale model fit <0.05).  Post-vaccination the modelled and observed reduction in hospitalisations matched, however in the unvaccinated older children the model predicted a lower reduction than observed which could be explained by a herd protection effect in the observed population (indirect vaccine benefit).  Herd protection was not captured in the static model. CONCLUSIONS It is possible to make the link between cohort models and observed population data provided the underlying model characteristics reflect reality.