Non-participation may bias the results of a psychiatric survey

Objective A major reason for limited validity of research is non-participation. Subjects with severe mental illness tend to cumulate in the group of non-participants, causing selection bias. We tested the hypothesis that severe psychosis is linked to nonparticipation in a field survey including magnetic resonance imaging (MRI) of the brain, psychiatric interviews and cognitive testing among subjects with psychosis. Furthermore, we wanted to explore other associative factors expected to affect non-participation. Methods Members of the Northern Finland 1966 Birth Cohort with a lifetime diagnosis of psychosis (N = 145) were invited to participate in the survey conducted in 1999–2001. Non-participation was determined by refusal or loss of contact. Data were gathered in earlier phases of the follow-up study or using register data. Results Ninety-one (63%) subjects attended the study. Compared to participants, non-participants were more often patients with schizophrenia and had more psychiatric hospitalisations, they had more positive psychosis symptoms during their illness course and they were more often on disability pension. Married subjects participated more often than those who were not married. Conclusion This study suggests that among subjects with psychosis, particularly those subjects who have the most severe course of illness are less willing to participate. This may lead to biased estimates when studying subjects with severe mental disorders.