Comparison of psychiatric diagnoses from interview reports with those from best-estimate procedures

Abstract

Objective: The aim of this study was to compare psychiatric diagnoses based on interview information with those based on best-estimate procedures, to evaluate information used in such procedures, and to use 5-year follow-up data to determine whether the best-estimate diagnosis is an improvement over the interview-based diagnosis.

Method: Psychiatric diagnoses were based on interview reports from 373 probands and 2615 relatives participating in a high-risk family study of alcoholism. The diagnosis also included clinician ratings in a best-estimate procedure of this study.

Results: For most diagnoses, both sensitivity and specificity, using the best-estimate diagnosis (BED) as the gold standard, were excellent, in both relatives and probands. Substance abuse was an exception, with very low sensitivity, although specificity rates were excellent. For nonsubstance diagnoses, specificity was high, but sensitivity ranged from 59% to 84% across relatives and probands. In general, BED procedures led to higher prevalence estimates than those from the interview only. In the BED process, family history data were especially useful for conduct and antisocial personality disorders. Follow-up interview data supported the fact that BED procedures led to both enhancements of, as well as errors in, diagnosis.

Conclusions: Our data attest to the utility of family history information, particularly for antisocial personality disorder and conduct disorder, and indicate that, for the phenotype of substance-dependence disorder, an interview-based diagnosis alone is adequate in classifying individuals with a minimum of error. These results should be reassuring for research studies in which costs and resources required for best-estimate procedures are not affordable.