Use of a structured diagnostic interview to identify depressive episodes in an epidemiologic study: a posteriori internal validation

Background: During these last years, many structured and standardized diagn ostic interviews have been developed in order to identify psychiatric disor ders in a standardized way. These tools enable a systematic investigation o f these disorders according to international classifications. Their main dr awback is to be long. To assess the care of depression, we used a shorter a nd more simple tool: the Mini International Neuropsychiatric Interview (MIN I) to identify depressive subjects. Method: The study was conducted in the Gazel cohort from the French Nationa l Electricity and Gas Company. A stratified sample of 2394 civil servants s elected in order to overrepresent depressive subjects was asked to answer t o the MINI interview through a phone interview. Art epidemiological and sta tistical analysis was performed to rest the MINI internal validity: prevale nce of depressive disorders using different threshold of diagnosis (number of symptoms required to identify someone as depressive), frequency of diffe rent symptoms, variability between investigators and potential biases. Results: Respondents to the phone interview (1108 civil servants) had more often presented depression markers for the last 5 years. Prevalence of depr essive episodes changed little when we varied the threshold of diagnosis an d did not stress any threshold problem. The variability between investigato rs was important, bur the estimation of prevalence remained stable when we excluded extreme rates of prevalence. The choice of a classification system affected the prevalence estimation Using the Diagnostic and Statistical Ma nual of Mental Disorders (DSM IV) from the American Psychiatric Association , the prevalence of depressive episodes was lower and closer to the estimat ions shown in the literature than using the International Classification of Disease (ICD 10). Moreover, the stratification assigned very unbalanced we ights to the stratification strata. By excluding depressive episodes observ ed in the stratum "control" (no depression "marker" from 1989 to 1994 in th e database), the prevalence was very lower, whatever the classification was . Finally, factors which appeared linked to care of depression with the ICD definition remained the same when the DSM diagnosis definition was used, a nd relative risks were quite similar. Conclusion: The MINI appears to be a short and simple tool, suited to the e pidemiological studies. This analysis does not highlight any failure in the internal consistency of the MINI. The remaining question Is what the MINI really measures, particularly comparing to a psychiatrist's diagnosis.