Compensating for non-response in a study estimating the incidence of mental disorders in long-term sickness absence by a two-phased design

Aims: The study compensates for the non-response that was observed in a previous study that estimated the frequencies of mental disorders in long-term sickness absence (LSA) (more than eight weeks of continuous sickness absence). In this study, the frequency of any mental disorder was estimated at 48% by a two-phase design and weighted logistic regression. The total non-response rate was 53.6%. This motivated the present study to compensate for non-response by applying adjustment of the weights and by multiple imputation of missing data in the estimation of the frequencies of mental disorders. Methods: The study took place in a Danish population of 120,000 inhabitants. During one year, all 2,414 incident individuals on LSA were identified. By a two-phase design 1,121 individuals were screened in Phase 1. In Phase 2, which was a subsample of Phase 1, 337 individuals participated in a psychiatric diagnostic examination applying Present State Examination as gold standard. The weighted analyses were based on scores of the screening instrument SCL-8AD compiled of SCL-8, SCL-ANX4, and SCL-DEP6 from the Common Mental Disorders Screening Questionnaire. In the present study, the variables sex, age, municipality, and social transfer income variables were used for the adjustment of weights in weighted analyses and in the imputation models. Results: The frequencies were: any mental disorder 46%-49%, depression 31%-36%, anxiety 13%-15%, and somatoform disorder 8%-9%. Conclusions: Irrespective of whether compensation for non-response was applied, the frequencies of mental disorders were similar. The variables used for the compensation were of problematic value.