Previous research has not compared the psychometric properties of diagnostic interviews of community samples and clinically referred subjects within a single study. As part of a multisite cooperative agreement study funded by the National Institute of Mental Health, 97 families with clinically referred children and 278 families identified through community sampling procedures participated in a test-retest study of version 2.1 of the Diagnostic Interview Schedule for Children (DISC 2.1).
The DISC was separately administered to children and parents, and diagnoses were derived from computer algorithms keyed to DSM-III-R criteria. Three sets of diagnoses were obtained, based on parent information only (DISC-P), child information only (DISC-C), and information from either or both (DISC-PC).
Test-retest reliabilities of the DISC-PC ranged from moderate to substantial for diagnoses in the clinical sample. Test-retest κ coefficients were higher for the clinical sample than for the community sample. The DISC-PC algorithm generally had higher reliabilities than the algorithms that relied on single informants. Unreliability was primarily due to diagnostic attenuation at time 2. Attenuation was greatest among child informants and less severe cases and in the community sample.
Test-retest reliability findings were consistent with or superior to those reported in previous studies. Results support the usefulness of the DISC in further clinical and epidemiologic research; however, closely spaced or repeated DISC interviews may result in significant diagnostic attenuation on retest. Further studies of the test-retest attenuation phenomena are needed, including careful examination of the child, family, and illness characteristics of diagnostic stability.