Non-response to population surveys is a common problem in epidemiological and public health research. Systematic non-response threatens the validity of results. Researchers rarely evaluate the magnitude of systematic non-response because of limited access to population data. This study explores how well morbidity and mortality in postal survey respondents aged 65 years and older represented that of the target population.
The 2010 Stockholm Public Health Survey and the Swedish Population Register were linked to the Cause of Death Register and the National Patient Register in Sweden. Differences were analysed between the response group and the corresponding population in mortality, hospital admission, days spent in hospital and number of diagnoses. Finally, data were weighted for non-response to see if this improved generalizability.
Non-response increased with age, and this increase was more pronounced among women than men. Respondents were marginally less often admitted to hospital, hospitalized fewer days and had slightly fewer diagnoses than the population, in particular after age 80. Significantly fewer women died in the response group than in the population as a whole. In terms of mortality among men and in terms of hospitalizations for most age groups, the respondents represented the population fairly well. Non-response weighting adjustment did not improve generalizability.
Postal questionnaires are likely to capture morbidity (hospitalization) among women and men aged 65-80 years old and mortality among men, while morbidity after age 80 and mortality in women are likely to be underestimated.