A Neural Network Approach to Comparing AMPD and Object Relations Theory for Personality Disorder Assessment.

Abstract

While prior research has explored the relationship between Object Relations Theory (ORT) and the Alternative Model for Personality Disorders (AMPD), comprehensive comparisons across diverse clinical populations and methodologies remain limited. This study investigated the predictive accuracy of AMPD and ORT in identifying personality psychopathology using neural network models within a mixed sample of 639 participants (229 non-clinical undergraduates, 410 psychiatric inpatients). Data were collected using Persian translations of the Level of Personality Functioning Scale-Self-Report (LPFS-SR), the Personality Inventory for DSM-5 (PID-5) (AMPD measures), and the Structured Interview of Personality Organization-Revised (STIPO-R) (ORT measure). Results indicated significant differences in all subscales of both models between clinical and non-clinical groups. Notably, the borderline personality disorder group showed elevated scores on specific STIPO-R subscales and all AMPD constructs except empathy. Neural network models achieved over 65% accuracy in predicting group membership, with AMPD slightly surpassing ORT (66%+ vs. 65%+). Receiver Operating Characteristic (ROC) analysis demonstrated high sensitivity for both models, with Area Under the Curve (AUC) values ranging from 0.79 to 0.94. These findings underscore the significant utility of both AMPD and ORT for the assessment, early identification, and diagnosis of personality disorders.