OBJECTIVES

: Age-structured models are widely used to simulate the impact of vaccination on health and economic outcomes. The population in such models is split into groups including one or several years of age. A simple simulation was developed alongside a complex dynamic model to demonstrate that the use of multi-year groups may bias the results for vaccinated individuals.

METHODS

: The simulation tracked a hypothetical population split into 19 age groups. The first ten groups included one year of age; the length of other groups increased progressively. Ageing occurred annually. In multi-year groups, the rate of ageing (transition to next group) was the inverse of the group length. Life expectancy was set to 75 years. Different ages of vaccination were tested and the average life expectancy was estimated for the vaccinated population.

RESULTS

: With the vaccination of the first cohorts, the population within multi-year groups was no longer uniformly distributed across the single age years; part of the vaccinees aged already after one year. Their faster transition between the groups resulted in the underestimation of their life expectancy. With vaccination at birth the average life expectancy in vaccinated individuals was 73.2 years. The life expectancy decreased to 72.5 years when increasing the vaccination age to 4 years, and 71.3 years with vaccination at 9 years (before considering any protective effect of the vaccine).

CONCLUSIONS

: The use of multi-year age groups is often dictated by computational intensity of age-structured models. The present study demonstrates that it results in faster ageing of vaccinated individuals. Apart from the demonstrated impact on the life expectancy, other biases may arise if model parameters are age-dependent. Single-year age groups are the recommended option. When not feasible, it is advised to estimate the potential biases caused by the use of multi-year groups and introduce correction factors.