Statistical Comparison Between Interview Questions and Rating Scales in Psychiatry.

Abstract

Background: Psychiatric evaluations consist of both qualitative questions and quantitative assessments, sometimes questioning the same issue. The present study attempts to investigate the statistical equivalency of several close-ended questions of a procedural psychiatric examination and rating scales addressing a similar problem.

Methods: The current retrospective analysis included 314 patients who made their first visit to a private psychiatry clinic. Subjects underwent a routine psychiatric examination, including close-ended questions and related clinical scales. Questions included sleep and sexual problems, problems in marriage, parent relationship problems, and childhood abuse. The related psychiatric scales were Jenkins Sleep Scale, Arizona Sexual Experience Scale, Dyadic Adjustment Scale, and Childhood Trauma Questionnaire, respectively. First, receiver operating curve analysis was conducted for each yes/no question and clinical scale. Then, area under curve sensitivity and specificity values were calculated. Multinomial logistic regression analysis was also performed to observe paired predictor variables.

Results: Among clinical questionnaires, the receiver operating curve model provided good area under curve values as prediction criteria for Dyadic Adjustment Scale (0.78; P < .001), Childhood Trauma Questionnaire (0.74; P < .001), Childhood Trauma Questionnaire-physical abuse (0.826; P < .001), Childhood Trauma Questionnaire-sexual abuse (0.828; P < .001), Arizona Sexual Experience Scale (0.796; P < .001), and Jenkins Sleep Scale (0.920; P < .001). Multinomial logistic regression models also revealed good correct classification values for Dyadic Adjustment Scale-Childhood Trauma Questionnaire (61%), Childhood Trauma Questionnaire-Physical abuse-Childhood Trauma Questionnaire-Sexual abuse (87.6%), and Arizona Sexual Experience Scale-Jenkins Sleep Scale (67%).

Conclusion: When the symptoms are investigated in general terms, the present study reveals that an experienced clinician could rely on clinical questions as much as the quantitative scales in both clinical and research domains.