Predicting Nonsuicidal Self-Injury: Complex Interactions Among Posttraumatic Symptoms and Dissociation

Background and Purpose: Nonsuicidal self-injury (NSSI) is a significant problem in clinical and nonclinical (adolescent) populations. Recently, models for understanding risk to engage in NSSI have determined linkages among avoidant behaviors/coping mechanisms, posttraumatic arousal and NSSI. The purpose of this study was to further this research by modeling avoidant posttraumatic stress symptoms (PTSS) as a moderator of NSSI, with intrusive PTSS as the focal variable in a two-way interaction, and a three-way interaction that added dissociation. It was hypothesized that as the gap between intrusive and avoidant PTSS widened, treatment-seeking adults sexually abused in childhood would be more likely to have engaged in NSSI. This presentation will present the findings of the study, and discuss further how to model, graph, interpret, and present interactions in logistic regression analysis. Methods: This was a cross-sectional study of 78 adult survivors of sexual abuse currently in treatment or entering outpatient treatment over a nine-month period. All measures were self-report. Variables in the analysis were intrusive and avoidant PTSS and attachment security. NSSI was captured using a measure developed for this study. Two different macros were used to help graph the interactions. MODPROBE (Hayes & Matthes, 2009), an SPSS macro, probed the two-way interactions, and an Excel macro ( Dawson & Richter, 2006) probed the three-way interaction. Results: Variables were entered into logistic regression analysis in three blocks, with all two-way interactions among dissociation and avoidant and intrusive PTSS entering the second block and the three-way interaction entering the third block. In block 1, all variables but avoidant PTSS were significant. In block 2, three interactions were significant (dissociation x avoidant PTSS; dissociation x intrusive PTSS; and avoidant x intrusive PTSS). As the gap between avoidant and intrusive PTSS increased, so did the probability of engaging in NSSS. In the three-way interaction, probability to engage in NSSI increased in three conditions: (a) with low avoidant and high intrusive PTSS, regardless of the level of dissociative symptoms; (b) high avoidant and low intrusive PTSS, but only under the condition of low dissociation; and (c) high avoidant/high intrusive PTSS and low avoidant/low intrusive PTSS, but only under the condition of high dissociation. Conclusions and Implications: The significant interactions supported the hypothesis for this study, buttressing a conceptualization of NSSI as an attempt to rectify an imbalance between the avoidant and approach PTSD systems. Because interactions were in logistic regression analysis rather than linear regression analysis, they were more complicated to draw and interpret. The interpretation benefited from the use of two different macros developed to probe graph two- and three-way interactions in logistic regression. Dawson, J. F., & Richter, A. W. (2006). Probing three-way interactions in moderated multiple regression: Development and application of a slope difference test. Journal of Applied Psychology, 91, 917-926. Hayes, A. F., & Matthes, J. (2009). Computational procedures for probing interactions in OLS and logistic regression: SPSS and SAS implementations. Behavior Research Methods, 41(3), 924-936.