Modelling the effect of conjugate vaccines in pneumococcal disease: Cohort or population models?

Abstract

Cohort and population models estimate vaccine impact on disease events, and yield different estimates in countries with different demographic compositions. We compared administration of the new 10-valent pneumococcal Haemophilus influenzae–protein D conjugate vaccine (PHiD-CV) with no vaccination in two countries, the United Kingdom (UK) and Mexico, using two modelling strategies: a cohort model and a population model. The cohort model followed a birth cohort over a lifetime, beginning 10 years after initiation of the vaccine program, when vaccine efficacy steady state had been reached. The population model examined the country-specific population over 1 year, also beginning 10 years after initiation of the vaccine program. Both models included the same age-specific disease rates of meningitis, bacteraemia, pneumonia, and otitis media. The output variables were the numbers of specific events, with and without indirect vaccine effects. Without indirect effects, the cohort and population models produced similar results for both countries (deviation of <20% difference per output measure for most outcomes). The difference between the model types was much greater when indirect vaccine effects were included, especially in Mexico (up to 120% difference). Cohort and population modelling methods produce different results depending on the disease, the intervention, the demographic composition, and the time horizon evaluated. Results from the two model types provide different information about the impact of interventions on events: accumulated vaccine benefit for an individual in a cohort model; vaccine benefit for a whole population at a specific time point in a population model.