An evaluation of DSM-III-R and ICD-10 benzodiazepine dependence criteria using Rasch modelling.

AIMS To evaluate the homogeneity of the elements of the Substance Dependence Syndrome (SDS) as applied to benzodiazepines (BZDs) by Rasch modelling. MEASUREMENTS The Rasch scaling model was applied to data obtained by administering the SCAN (Schedules for Clinical Assessments in Neuropsychiatry) substance dependence sections. Subsequently, Rasch-homogeneous sets of DSM-III-R and ICD-10 BZD dependence criteria were assessed for subject and item discriminability. To support their construct validity a theoretical rationale was formulated based on the Rasch scale values. PARTICIPANTS A heterogeneous sample of 599 outpatient BZD users. FINDINGS Only particular subsets of the DSM-III-R and ICD-10 BZD dependence criteria met the requirements for Rasch-homogeneity, which appears to be due to medical aspects of BZD use. The subject and item discriminability results were sufficiently good. CONCLUSIONS The DSM-III-R and ICD-10 BZD dependence constructs may need to be redefined. The use of a BZD dependence severity model based on a Rasch-homogeneous scale appears to have greater clinical value than a dichotomous diagnostic model based on an arbitrary cut-off point. We recommend Rasch modelling to investigate the homogeneity of the elements of the SDS across other psychoactive substances.