Factor Analysis of the Borderline Personality Inventory Using Item Response Theory

The current study used a sample of 251 undergraduate females &om Old Dominion University to assess the item properties of the Borderline Personality Inventory (BPI). The BPI is a true/false self-report instrument used to measure borderline personality disorder. The BPI was created using classical test theory techniques, and traditional methods were used to assess the scale's properties (Leichsenring, 1999). Due to the dichotomous nature of the BPI, traditional classical test theory techniques are not ideal. The current study used Item Response Theory (IRT) to characterize the scale and item properties. The study was conducted using an online survey that also included other scales for psychometric comparison: the Borderline Syndrome Index, the Curious Experiences Scale, the International Personality Item Pool version of the NEO-PI, the Defense Mechanisms Scale, the Revised Borderline Personality Inventory, and the Revised Borderline Syndrome Index. Results from a traditional exploratory factor analysis indicated that the items did not load reliably, and did not replicate the findings in Leichsenring (1999). Further analysis indicated that four factors best fit the data: Anxiety, Impulsivity, Impaired Reality Testing, and Relationship Dysfunction. IRT provided alpha (discrimination) and beta (difficulty) values for each item, as well as theta (true score) values for each participant for each scale. Item and scale properties were estimated. Concurrent and discriminant validity was established by confirming appropriate correlations between the BPI and the psychometric comparison scales. In conclusion, only five items should be deleted from the factors, and 21 items should be deleted in total from the scale. New items should be constructed to reflect a wider range of beta values. New items should also be created to measure borderline traits that are not measured with the current items.