Meta-analyses of agreement between diagnoses made from clinical evaluations and standardized diagnostic interviews

Standardized diagnostic interviews (SDIs) have become de facto gold standards for clinical research. However, because clinical practitioners seldom use SDIs, it is essential to determine how well SDIs agree with clinical diagnoses. In meta-analyses of 38 articles published from 1995 to 2006 (N = 15,967 probands), mean kappas (z-transformed) between diagnoses from clinical evaluations versus SDIs were 0.27 for a broad category of all disorders, 0.29 for externalizing disorders, and 0.28 for internalizing disorders. Kappas for specific disorders ranged from 0.19 for generalized anxiety disorder to 0.86 for anorexia nervosa (median = 0.48). For diagnostic clusters (e.g. psychotic disorders), kappas ranged from 0.14 for affective disorders (including bipolar) to 0.70 for eating disorders (median = 0.43). Kappas were significantly higher for outpatients than inpatients and for children than adults. However, these effects were not significant in meta-regressions.

Diagnostic agreement between SDIs and clinical evaluations varied widely by disorder and was low to moderate for most disorders. Thus, findings from SDIs may not fully apply to diagnoses based on clinical evaluations of the sort used in the published studies. Rather than implying that SDIs or clinical evaluations are inferior, characteristics of both may limit agreement and generalizability from SDI findings to clinical practice.