Long-term test-retest reliability of personality disorder diagnoses in opiate dependent patients

This investigation reports the two year test-retest reliability of DSM-III-R personality disorder (PD) diagnoses in a sample of 219 patients with opiate dependence-admitted to methadone treatment. Different MA/PhD interviewers at each assessment used a semistructured diagnostic interview for PD, the Structured interview for DSM-III-R Personality Disorders (SIDP-R), to make their diagnoses. The reliability of any PD diagnosis versus no PD-was fair (kappa = .51). The reliability for any specific PD (weighted kappa = .31) was poor. Antisocial (kappa = .45) and sadistic (kappa = .42), were the only specific PDs for which at least fair reliability was achieved. At the cluster level, only Cluster B had fair reliability (kappa = .47). The intraclass correlation coefficients between number of criteria for the specific PDs at the two evaluation points were consistently higher (range .22 to .62.) than were the corresponding kappas for categorical diagnoses. In that the base rates for most of the PDs were low and agreement of no diagnosis tended to be high, percent exact agreement for the specific PDs typically exceeded 90%. Increasing the base rate by lowering the diagnostic threshold, or examining more severe cases by raising the diagnostic threshold, did not consistently effect reliability. Reasons for the low kappa coefficients and the implications for PD research are discussed.