Modelling the epidemiological impact of rotavirus vaccination in Germany--a Bayesian approach

Abstract

Background: Rotavirus (RV) infection is the primary cause of severe gastroenteritis in children aged <5 years in Germany and worldwide. In 2013 the German Standing Committee on Vaccination (STIKO) developed a national recommendation for routine RV-immunization of infants. To support informed decision-making we predicted the epidemiological impact of routine RV-vaccination in Germany using statistical modelling.

Methods: We developed a population-based model for the dynamic transmission of RV-infection in a vaccination setting. Using data from the communicable disease reporting system and survey records on the vaccination coverage from the eastern federal states, where the vaccine was widely used before recommended at national level, we first estimated RV vaccine effectiveness (VE) within a Bayesian framework utilizing adaptive Markov Chain Monte Carlo inference. The calibrated model was then used to compute the predictive distribution of RV-incidence after achieving high vaccination coverage with the introduction of routine vaccination.

Results: Our model estimated that RV-vaccination provides high protection against symptomatic RV-infection (VE=96%; 95% credibility interval (CI): 91-99%) that remains at its maximum level for three years (95% CI: 1.43-5.80 years) and is fully waned after twelve years. At population level, routine vaccination at 90% coverage is predicted to reduce symptomatic RV-incidence among children aged <5 years by 84% (95% prediction interval (PI): 71-90%) including a 2.5% decrease due to herd protection. Ten years after vaccine introduction an increase in RV incidences of 12% (95% PI: -16 to 85%) among persons aged 5-59 years and 14% (95% PI: -6 to 109%) within the age-group >60 years was predicted.

Conclusion: Routine infant RV-vaccination is predicted to considerably reduce RV-incidence in Germany among children <5 years. Our work generated estimates of RV VE in the field and predicted the population-level impact, while adequately addressing the role of model and prediction uncertainty when making statements about the future.

Keywords: Bayesian inference; Epidemiological impact; Rotavirus; Routine vaccination; Transmission modelling; Vaccine effectiveness.