Representativity of a postal public health questionnaire survey in Sweden, with special reference to ethnic differences in participation

Aim: Non-participation in health surveys is a common phenomenon. When differences between participants and non-participants are considerable, the external validity of the sample survey may decrease and false conclusions might be drawn about the health status of the population. For this reason, the authors aimed to investigate the representativity of a postal questionnaire survey performed in the county of Scania, Sweden, in 1999-2000. The survey, which was based on an 18- to 80-year-old population sample, had a 58% response rate (n = 13604). Methods: For some variables, the information obtained using the questionnaire was compared with information obtained from a population register that covers all the population in the county (for the 18- to 80-year-old group, n = 850476). The population register includes, among other data, information on age, gender, educational level, country of birth, and healthcare expenditure. Results: Men, individuals with a low level of education, and immigrants were under-represented in the survey. However, except for immigrants, the under-representation was not large. Among immigrants, particularly those born in former Yugoslavia, the Arabic-speaking countries, and Poland were very significantly under-represented in the study. By contrast, immigrants born in other Nordic countries had responded to almost the same extent as respondents born in Sweden. The survey sample had about the same healthcare utilization costs as did the general population. Conclusions: In summary, the "Health Survey for Scania, 2000'' seems largely representative of the total Scanian population. A major concern, however, is the under-representation of the immigrant population.