Pairs of students at advanced stages of graduate training in clinical psychology reviewed each interview to achieve consensus about whether a given symptom was present. Consensus teams consisted of different individuals than those who had conducted the diagnostic interview, and each team was blind to subject identity. Different teams were assigned different family members. Computer algorithms based on DSM-III-R requirements were subsequently used to assign lifetime diagnoses. For the twins, these were best-estimate diagnoses combining information from the twin and parent (Kosten & Rounsaville, 1992; Leckman, Sholomskas, Thompson, Belanger, & Weissman, 1982). For the younger cohort information was aggregated across assessments. Because adolescents in this study were not through the risk period for most disorders assessed, we used a probable certainty level to assign offspring diagnoses, meaning that all diagnostic criteria but one had to be met. Several hundred interviews were reviewed by independent, blind teams to assess reliability. Kappa coefficients were all greater than .74, with coefficients for SUDs greater than .90. Every 25th case is evaluated in an ongoing way to insure continued reliability and protect against drift.
