AI-driven imputation of a synthetic personality severity index from the NEO-FFI

Personality disorder severity (PDS) is a novel measure of personality in the ICD-11 with potential prognostic value. Due to its recent introduction, however, no information on it exists in most clinical cohorts. Here, we explored the value of its replacement in statistical models by an AI-based synthetic score based on NEO-FFI data, a widely used normative assessment of personality. In linear models with sex, age, traumas in childhood, and occupational status, the synthetic PDS scores reproduced coefficients of the observed scores closely. Unlike the linear model, the neural synthetic scores did not reproduce a spurious association between sex and personality functioning, suggesting that the neural model captured clinically meaningful variance while suppressing confounding demographic signals. Null hypotheses in synthetic models were rejected with only a modest loss of power. These findings suggest that clinically relevant information is latent within normative measures, despite their distinct theoretical origins. The approach enables retrospective severity estimation in datasets that include NEO-FFI items but lack direct assessments of personality functioning, thereby embedding a dimensional, transdiagnostic framework for personality pathology into large health databases without additional data collection and supporting immediate, cross-cohort research on risk and prognosis.

AI-driven imputation of a synthetic personality severity index from the NEO-FFI