To estimate the immunity of the Dutch population against vaccine-preventable diseases, a
population-based serum bank was established. Since a multi-tiered approach to enrol eligible
individuals was used, both the overall non-response selection and the effect, on this selection,
of including additional participants and of excluding a subgroup of non-participants (i.e. those
without questionnaire data) could be studied. For some characteristics associated with non-participation, an association with seroprevalence of vaccine-preventable diseases is likely (e.g.
age, gender). For other characteristics (e.g. marital status, receipt of reminder, degree of
urbanization) the association with immune status is unclear but probably small. If the
distribution in the population, or information on all participants and non-participants, of the
characteristic is available, then the effect on the seroprevalence can be estimated. However,
investigators have to be aware that studying only a subgroup of non-participants might lead to
a biased insight into non-participation selection. Furthermore, merely including additional
participants might not always reduce this bias.