Background: In population surveys, the assessment of mania is commonly done by trained lay interviewers using structured diagnostic instruments: the validity of this approach has been questioned. We examined the criterion validity and prevalence of lifetime mania in a survey of Swedish twins conducted with interview methodology usually applied in psychiatric epidemiology.
Methods: 41 838 individuals in the Swedish Twin Registry were evaluated via a telephone interview that included the eight DSM-IV mania items, and these data were merged with inpatient hospitalization discharge diagnoses from two comprehensive national registries (the criterion). An algorithm with eight cut-points was used to diagnose lifetime mania, and compared by a receiver operator characteristic curve to the criterion. The algorithm requiring at least four positive items resembling a DSM-IV diagnosis.
Results: History of hospitalization for a psychiatric condition that included a manic episode was present for 0.7% of all living twins, and predicted non-response to the survey (OR = 0.5; 95% CI = 0.4-0.6). The incidence rate for first hospitalization was 2.1/10 000 year(-1). For > or =1 symptom (first cut-point), the prevalence, sensitivity and specificity were 3.6%, 39.0% and 96.6%; for > or = 4 symptoms (DSM-IV-like cut-point) 2.6%, 36.5% and 97.6%; and for eight symptoms 0.3%, 18.0% and 99.8%. Positive predictive values were, respectively, 5.5%, 7.0% and 29.8%.
Conclusions: The performance of the telephone screening for mania by lay interviewers in terms of positive predictive power was not satisfactory; despite a high specificity, the false positive rate was high. The low population prevalence of mania, non-response bias, criterion choice and inherent limitations of the interviewing method are among the explanations. Assessment of a lifetime manic episode based on lay interviewer screening may yield misleading data.