Bias due to non-participation and heterogenous sub-groups in population surveys.

A random sample of men in the age-group 30-39 years from the general population in Göteborg, Sweden, has been investigated with respect to socioeconomic factors and risk factors for coronary artery disease. The total sample could well be characterized with socioeconomic variables obtained from public registers. All the individuals of the sample were invited to an examination which 68% attended (participants). It was found that those not attending the examination (non-participants) greatly differed from the participants. The non-participants were more often unmarried, and had lower annual incomes and more sickness benefit days. There were more foreigners and more individuals registered for intemperance among the non-participants than the participants. Among the participants the foreigners reported lower physical activity and had higher serum cholesterol than the participating Swedes and individuals registered for intemperance stated a higher tobacco consumption and had higher systolic and diastolic blood pressures than those not registered. This highlights that consideration of factors discriminating participants and non-participants is important for proper estimation of population parameters. The same is true for comparisons between cases and controls recruited from cross-sectional population surveys.