Diagnostic accuracy of DSM-5 borderline personality disorder criteria: Toward an optimized criteria set

•DSM-5 Personality criteria for borderline personality disorder (BPD) exhibit different levels of accuracy in diagnosing BPD•BPD criteria of abandonment fears, unstable relationships, identity disturbance, impulsivity, affective instability, and chronic emptiness demonstrated good accuracy•A score of 3 or more positive criteria yielded an excellent balance of sensitivity (SN=.90) and specificity (SP=.99)•The optimized set evidenced substantially lower heterogeneity (42 combinations) than the traditional polythetic model (256 combinations). The polythetic system used by the Diagnostic and Statistical Manual for Mental Disorders (DSM-5) for diagnosing borderline personality disorders (BPD) is far from optimal; however, accumulated research and clinical data are strong enough to warrant ongoing utilization. This study examined diagnostic efficiency of the nine DSM-IV BPD criteria, then explored the feasibility of an optimized criteria set in classifying BPD. Adults (N=1,623) completed the Structured Clinical Interviews for DSM-IV Axis II Disorders resulting in a BPD group (n=352) and an inpatient psychiatric control group (PC) with no personality disorders (n=1,271). Receiver operator characteristics and diagnostic efficiency statistics were calculated to ascertain the relative diagnostic efficiency of each DSM-5 BPD criterion in classifying BPD cases. Affective instability (Criterion 6) evidenced the strongest capacity to differentiate the groups (AUC = .84, SE = .01, p < .0001). Abandonment fears (Criterion 1), unstable relationships (Criterion 2), identity disturbance (Criterion 3), impulsivity (Criterion 4), and chronic emptiness (Criterion 7) yielded good-to-moderate discrimination (AUC range = .75-.79). A composite index of these six criteria yielded excellent accuracy (AUC = .98, SE = .002, p < .0001), sensitivity (SN=.99), and specificity (SP=.90). The current findings add to evidence that affective instability is a useful gate criterion for screening, and the optimized criteria set evidences equivalent accuracy to the original 9 criteria, with a substantial reduction in estimated heterogeneity (from 256 combinations with the original set to 42 combinations with the optimized set).