Response trends and nonresponse bias in a mail survey of oral and facial pain.

While mail surveys offer a low-cost method for collecting health data, they have been unpopular because of concerns about low response rates and nonresponse bias. This paper examines the response to a mail survey concerning oral and facial pain and uses regression techniques to assess the degree and direction of nonresponse bias on estimated prevalence rates. The analysis shows that early and late responders to the survey differed in terms of sociodemographic variables and responses to items concerning pain. The regression analysis suggests that the rather high prevalence rates revealed by the survey are not a product of bias induced by nonresponse and deserve further investigation.