gms | German Medical Science

Children under the age of five and elderly people with a weak immune system are especially susceptible to pneumococcus infections and different vaccines exist for children and adults. In Germany there is no general recommendation to vaccinate young children against streptococcus pneumoniae. As this is contrary e.g. to the practice in the USA, we analysed the consequences of different vaccination programs using a stochastic age-structured compartment model. Special attention was paid to possible effects on herd immunity.

The compartment model describes the spread of the infection in a population with eight age classes (0-2, 3-4, 5-12, 13-20, 21-39, 40-64, >65). The equations track the changes in the numbers of newborns with and without maternal antibodies (B, M), of susceptible (S), infected (I) and vaccinated (V) individuals, also accounting for failure of immunization by the vaccine (S_V, I_V). The stochastic individual-based model presented her was derived from a deterministic version [Ref. 1] which in turn was inspired by an approach of Halloran et al. [Ref. 2] for varicella. The time step used in the simulations was half a year and each person could undergo at most one transition within this period. Initial values for the comparments and transition rates came from expert knowledge and the literature, in particular [Ref. 3] and the Statistical Yearbooks. There are, for instance, 380 cases of pneumococcus-meningitis in Germany each year, in which 140 occur under the age of two years.

After validating the model with data from invasive pneumococcus diseases (e.g. pneumonia, sinusitis and meningitis) in the USA it was used to simulate the dynamics of pneumococcus-meningitis cases in Germany.

Our results indicate that vaccinating children under 2 years of age with realistic coverage rates could lead to a 75% reduction of disease cases within 30 years. The accuracy of these results is indicated by 95% confidence intervals from stochastic simulations.

The stochastic model described above lead to a faster initial decrease in the number of infected people but levelled off much earlier and at higher numbers than the comparable deterministic model. Further time series data are necessary in order to refine and validate the models.

We thank L. Fahrmeir and L. Held for discussions.