Problems and considerations in the valid assessment of personality disorders

This article reviews evidence for the reliability and diagnostic concordance of structured-interview and self-report questionnaire methods for the diagnosis of personality disorders. The findings of nine studies that compared two or more axis II diagnostic instruments administered to the same groups of subjects are summarized. Across the eight studies with sufficient data, a summary of the overall diagnostic agreement between any two instruments yielded a low reliability (median kappa = 0.25) for making individual personality disorder diagnoses. Diagnostic concordance was lower between self-report questionnaire and interview methods than between interview methods. Comparing dimensional scores of different methods did not appreciably improve the level of agreement. The author concludes that current methods for making personality disorder diagnoses have high reliability but yield diagnoses that are not significantly comparable across methods beyond chance, which is not scientifically acceptable. Sources for the disagreement include variance due to different raters, interview occasions, data sources (self-report versus observer report), information bases obtained, and instrument sensitivity to state effects (e.g., mood). Serious problems in assessment validity may also arise from the yes/no format, which, despite probes for confirmatory examples, may fail to distinguish adequately between sporadic occurrences and longstanding patterns. Efforts should be made to improve and demonstrate the validity of axis II diagnostic methods. One route to increasing validity is to improve the clinical interview, because personality patterns are best revealed by the recurring patterns one finds when taking a systematic history.