A short screening tool for borderline personality disorder (Short-Bord): Validated by Rasch analysis

• Main Authors: Nahathai Wongpakaran, Tinakon Wongpakaran, Pimolpun Kuntawong
• Format: Journal
• Published: 2019
• Subjects: Medicine; Psychology
• Online Access: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85070529723&origin=inward; http://cmuir.cmu.ac.th/jspui/handle/6653943832/66714
• Institution: Chiang Mai University

© 2019 Elsevier B.V. The study aimed to development a short screening scale for borderline personality disorder (Short-Bord), and to validate its psychometric properties using Rasch analysis. Ninety-eight outpatients undergoing psychotherapy were evaluated using a semistructured diagnostic interview for DSM-IV Personality disorders. Correlational analysis and Rasch analysis were used to identify the best-fitted items for the shorter scale. Rasch analysis identified three underfitted items. The best five items were selected for the Short-Bord using two analyses, resulting in two sets of Short-Bord which included item 1 (becoming frantic when someone left), item 2 (up- and- down relationships), item 3 (sudden change of sense of self), item 8 (self-harm or suicide), item 9 (self-mutilation), item 10 (sudden mood change) and item 11 (feeling empty inside). Each set of the five-item Short-Bord were tested against the original 15-item BPD scale. Results showed that both sets of the Short-Bord yielded minimally lower in area under curve (AUC = 0.95 and 0.96, respectively) compared with the total score of 15 items (AUC = 0.97), but none significantly differed (chi-square = 0.89-2.87, df 1, p >.05). Internal consistency for the set from Rasch analysis was slightly higher than correlation methods (Cronbach's alpha = 0.80, and 0.78, respectively). The Short-Bord presents promising tool to screen for borderline personality disorder. Its diagnostic validity was comparable to the total 15 items despite completing in a shorter time. The Short-Bord derived from Rasch analysis was, however, preferable as all items were shown to have unidimensional construct with good fit statistics and good internal consistency.