The health status of nonparticipants in a population-based health study: the Hordaland Health Study.

The authors aimed to examine whether nonparticipation in a population-based health study was associated with poorer health status; to determine whether specific health problems were overrepresented among nonparticipants; and to explore potential consequences of participation bias on associations between exposures and outcomes. They used data from the Hordaland Health Study (HUSK), conducted in western Norway in 1997-1999. Of 29,400 persons invited, 63.1% participated in the study. Information from HUSK was linked with the Norwegian national registry of disability pensions (DPs), including information about DP diagnosis. The risk of DP receipt was almost twice as high among nonparticipants as participants (relative risk = 1.88, 95% confidence interval: 1.81, 1.95). The association was strongest for DPs received for mental disorders, with a 3-fold increased risk for nonparticipation. Substance abuse, psychotic disorders, and personality disorders were especially overrepresented among nonparticipants. The authors simulated the impact of nonparticipation on associations between exposures and outcomes by excluding HUSK participants with higher symptoms of common mental disorders (exposure) and examining the impact on DP (outcome). This selective exclusion modestly reduced associations between common mental disorders and DP. The authors conclude that nonparticipants have poorer health, but this is disorder-dependent. Participation bias is probably a greater threat to the validity of prevalence studies than to studies of associations between exposures and outcomes.