An alternative sampling approach to the study of diabetes prevalence.

Sampling methods hitherto used in epidemiological surveys have generally faced two major sorts of potential bias: a latent period between listing and testing of listed individuals which can cause a misrepresentation of the population under study, compared with the population basis; population migration and death during the study period, which may result in a discrepancy between the population selected for study and the part of the population examined. A Central Population Register, found in all Scandinavian countries, permits an evaluation of the size of these problems of bias and makes it possible to use an alternative sampling method based on test samples successively drawn from the actual dynamic population. The feasibility of this new approach was tested in an epidemiological survey of diabetes mellitus among individuals in the age group 60-74 years living in a Danish municipality (Fredericia). It is concluded that this alternative dynamic method produces a more representative study population with less bias, thanks to the smaller number of deaths and migration among non-responders during the study, than do the methods previously used. Furthermore, the length of the study period, the size of the test samples, as well as the intervals between selecting samples may be adjusted to the capacity available, thus reducing the resources required. By using a correction in the computer program for selecting test sample size, the sex and age profile of the whole study population may be chosen in accordance with the incidence of the disease studied. In the Scandinavian countries a unique possibility exists to carry out prevalence studies on all kinds of diseases by the method described and evaluated here.