The latent structure of self-harm

The underlying structure of self-harm behaviors is not well-understood; for example, whether suicidality and nonsuicidal self-injury (NSSI) lie on a single dimension or two separate dimensions is unknown. We used confirmatory factor analyses to examine the factor structure of self-harm items in a clinical/community sample (N = 641). Of three alternative factor structures (one-factor, correlated-factors, bifactor), the bifactor model fit best. The general factor, representing overlap between suicidality and NSSI, captured the majority of model variance and was the strongest predictor of psychosocial correlates. The NSSI-specific factor captured a moderate amount of variance and correlated uniquely with both antagonistic traits and obsessive- compulsive tendencies; this factor was named NSSI. The suicidality-specific factor explained little model variance and was weakly associated with external criteria; this factor was named low attraction to life. Results are interpreted as preliminary evidence for the utility of bifactor modeling in understanding the latent structure of self-harm.