Blindness and reliability in lifetime psychiatric diagnosis.

Confidence in the assignment of lifetime psychiatric diagnosis is of great importance to genetic studies of psychiatric illness. To establish the credibility of a lifetime psychiatric history obtained via a structured interview, two paradigms were constructed to estimate reproducibility of the interview recording process. The first paradigm, simultaneous coding, was used to test comparability of four interviewers independently coding an interview form. Low variance/high reliability was demonstrated. The second paradigm, test-retest, provided for each subject to be interviewed twice, with a mean interim time of 6.7 months (SEM = .39). This paradigm demonstrated high reproducibility of psychiatric diagnosis over time. The overall k value for measurement of diagnostic agreement was .79. Only the diagnostic category of minor depression seemed to evade reliability. It was shown across both paradigms that an interviewer need not be blind (naive to previously held diagnosis) to obtain an unbiased interview. However, it is still recommended that the diagnosis of each interview should be determined by an independent diagnostician.