Interviewer effects on epidemiologic diagnoses of posttraumatic stress disorder.

In an epidemiologic study of 641 interviewed subjects in the Australian Vietnam Veterans Health Study, three diagnoses of Vietnam combat-related posttraumatic stress disorder (PTSD) were obtained: lifetime prevalence using a variant of the Diagnostic Interview Schedule and lifetime and current (1-month) PTSD prevalence using the Standardized Clincical Interview for the Diagnostic and Statistical Manual of Mental Disorders, Third Revision. Prevalence estimates using the Standardized Clinical Interview varied according to interviewer characteristics (female vs. male, clinician vs. nonclinician) but not for the Diagnostic Interview Schedule. The authors use a simple variant of logistic regression to distill estimates of two informative parameters characterizing interviewers' judgments: severity threshold (related to the individual interviewer's criterion of "caseness") and reliability (related to degree of classification error of the individual interviewers). Examination of these estimates shows that female clinicians adopted lower severity thresholds for diagnosis of PTSD than other interviewers and hence had higher prevalence estimates while being relatively reliable in their judgments. Examination also shows that nonclinician interviewers can perform at least as reliably as clinicians. The Diagnostic Interview Schedule measure of PTSD was not moderated by these interviewer aspects. This use of threshold and reliability parameters is offered for routine use in epidemiologic field studies to examine potential interviewer effects.