نوجوانان عمدی خودآسیبی، استرس بین فردی و اثرات تعدیل خودتنظیمی: مطالعه طولی دو موجی

Abstract The predictive effects of peer victimization and harsh parenting on deliberate self-harm were examined. As derived from the experiential avoidance model, the study also tested whether these links were moderated by individual self-regulation approaches. Data were collected at two points in time from 880 junior high school students (mean age = 13.72) in Sweden. Analyses using structural equation modeling revealed that Peer Victimization was predictive of self-harm. Although Harsh Parenting was not predictive of self-harm, this link was moderated by adolescents' gender. No moderating effect of self-regulation was revealed. The study concludes that the high prevalence of deliberate self-harm recently found in community samples of adolescents cannot be prevented without attending to environmental psychosocial factors.

1. Introduction Young people who intentionally hurt their own bodies with sharp objects or any other physical means typically report that the act of self-harm brings relief from emotional distress (Klonsky, 2007). Self-harm has been associated with problems such as victimization by peers, parental emotional neglect, childhood sexual abuse, insecure attachment, anxiety, depression, low self-esteem, body dissatisfaction, poor school achievement, drug consumption, dissociative symptoms, and general psychopathology (Bjärehed and Lundh, 2008, Brodsky et al., 1995, Gratz et al., 2002, Hawton et al., 2006, Klonsky et al., 2003, Laukkanen et al., 2009, Ross and Heath, 2002, van der Kolk et al., 1991 and Zlotnick et al., 1997). Definitions of self-harm vary with the inclusion or exclusion of behaviors such as attempted suicide, self-poisoning and general self-destructive behavior. Based on theoretical arguments that both the antecedent causes and the expected outcome of self-harm differ fundamentally from those of attempted suicide (Favazza, 1998, Gratz, 2001, Messer and Fremouw, 2008 and Muehlenkamp, 2005), we define self-harm in this study as the deliberate, direct destruction or alteration of body tissue without conscious suicidal intent but resulting in injury severe enough for tissue damage (e.g., scarring) to occur (see Gratz, 2001). Using the above definition of self-harm, recent investigations have reported one-year estimates of the prevalence of self-harm among adolescents to be as high as 40% in Sweden (Bjärehed & Lundh, 2008) and 26% to 46% in the United States (Hilt et al., 2008, Lloyd-Richardson et al., 2007 and Yates et al., 2008). Because these reports regard socially and ethnically diverse community samples recruited through schools, such a high prevalence of self-harm indicates that it deserves even more attention than previously recognized.

3. Results Intercorrelations and standard deviations of the study variables in the baseline model are displayed in Table 2. All parameters in the results are presented as standardized estimates. Mean values of deliberate self-harm were not significantly different across genders, either at T1, t(863) = 1.91, p = .056, or T2, t(861) = .95, p = .340. The proportion of adolescents who reported one or more acts of deliberate self-harm during the previous 6 months was 34% at T1 and 36% at T2. Table 2. Intercorrelations and standard deviations (SD) of study variables in baseline model (N = 880). Variable SD 1 2 3 4 5 6 7 8 9 1. T1 Deliberate self harm .68 2. T2 Deliberate self harm .79 .45 3. T1 Personal harassment .73 .28 .29 4. T2 Personal harassment .77 .17 .34 .39 5. T1 Victim of bullying .78 .29 .26 .60 .29 6. T2 Victim of bullying .78 .18 .30 .29 .53 .32 7. T1 Angry outbursts .79 .21 .14 .17 .14 .17 .12 8. T2 Angry outbursts .82 .20 .30 .24 .28 .21 .20 .56 9. T1 Coldness-rejection .68 .20 .17 .12 .14 .15 .20 .58 .40 10. T2 Coldness-rejection .68 .20 .39 .20 .25 .20 .26 .37 .64 .43 Note. T1 and T2 = Time 1 and Time 2 of the data collection (approximately one year apart). Study variables were centered at the grand mean. All coefficients are significant at p < .01. Table options 3.1. Measurement-model properties To verify that Harsh Parenting and Peer Victimization can be treated as two separate latent constructs, we tested a measurement model. The model included all four concurrent observed measures for T1 and T2, with personal harassment and victim of bullying loading on the same latent variable (labeled Peer Victimization) and angry outbursts and coldness-rejection loading on the other latent variable (labeled Harsh Parenting). The model also included two-way paths between the concurrent latent constructs and one-way paths from T1 to T2 for each latent construct. This model yielded a reasonable fit to the data, χ2(16, N = 880) = 105.63, CFI = .97; RMSEA = .08 (90% CI = .07–.09). Factor loadings for T1 were .81 (personal harassment) and .74 (victim of bullying) for Peer Victimization, and .85 (angry outbursts) and .68 (coldness-rejection) for Harsh Parenting. Factor loadings for T2 were .79 (personal harassment) and .66 (victim of bullying) for Peer Victimization and .90 (angry outbursts), and .71 (coldness-rejection) for Harsh Parenting. To test whether each observed measure operated similarly at T1 and T2, we evaluated the factorial invariance across time for each latent construct separately. For this purpose, we designed measurement models that included a temporal path between the T1 and T2 latent construct—each of which were built up by the concurrent observed measures. As a first step in these analyses, each pair of the observed measures was constrained to be equal across T1 and T2. Changes in model fit were, thereafter, evaluated as one pair at a time was released to allow free estimation. None of these modifications resulted in significant change in model fit, with Td(Δdf = 1) ranging from .27 (p = .604) to .68 (p = .411). 3.2. Effects of interpersonal stress on deliberate self-harm To examine the first aim of this study, the predictive effect of environmental stress on deliberate self-harm, a two-wave panel model controlling for both stability over time and cross-sectional intercorrelations of the included constructs was tested (see Fig. 1a). The hypothesized effect of environmental stressors on deliberate self-harm in this model is represented by the parameters of T1 Peer Victimization predicting T2 deliberate self-harm and T1 Harsh Parenting predicting T2 deliberate self-harm. No other paths or constraints were added to this model. The estimation of this hypothesized structural model yielded an acceptable fit to the data, χ2(26, N = 880) = 108.28, CFI = .97, RMSEA = .06 (90% CI = .05–.07). Estimates for the conceptual links are displayed in Fig. 2. As the figure shows, T1 Peer Victimization was predictive of T2 deliberate self-harm (depicted by the continuous diagonal line; β = .18, p = .006), whereas T1 Harsh Parenting was not (depicted by the dashed diagonal line; β = .03, p = .539). Structural model of relationships among Deliberate Self-harm, Peer ... Fig. 2. Structural model of relationships among Deliberate Self-harm, Peer Victimization, and Harsh Parenting across Time 1 and Time 2 (N = 880). Dashed line indicates that the relationship was not significant at p < .05. Figure options Gender differences were examined by using gender to perform a multiple-group model where boys (n = 445) and girls (n = 435 girls) had separate path values, χ2(60, N = 880) = 212.06, CFI = .95, RMSEA = .08 (90% CI = .06–.09). Tests for gender differences in terms of the predictive effect of T1 Peer Victimization and T1 Harsh Parenting on T2 deliberate self-harm were performed by constraining each path across groups, one at a time. A significant change in model fit was revealed only for the predictive effect of T1 Harsh Parenting, Td(Δdf = 1) = 6.59, p = .010, suggesting that gender moderates the effect of Harsh Parenting on deliberate self-harm. Overall, Harsh Parenting was a greater risk factor for girls (β = .13, p = .061) than for boys (β = −.07, p = .253), although neither effect was statistically significant. 3.3. Moderating effects of self-regulation The second aim of this study was to examine the possible moderating effects of individual self-regulation approaches. For this purpose, we designed a series of models that extended the baseline model by first adding the self-regulation concepts, one at a time, each with its temporal and concurrent paths. Then, a direct path from each T1 self-regulation concept to T2 deliberate self-harm was added. Finally, an interaction term of the T1 self-regulation concept and each of the concepts of interpersonal stress, first T1 Peer Victimization and then T1 Harsh Parenting, was included in the model. An overview of these models with the interaction terms included appears in Fig. 1b and c. A significant path from the interaction term to T2 deliberate self-harm would indicate a significant moderating effect. No other paths or constraints were added to these models. Analyses for each self-regulation concept were performed separately. All three extended baseline models with the respective self-regulation concept added, including both concurrent and temporal paths, had acceptable fit to the data, χ2(51, N = 880) < 183, CFI > .95, RMSEA < .06. For all instances, adding the interaction terms to these extended baseline models slightly decreased both AIC and BIC values, indicating that model fit did not change unfavorably. The results show that the interaction term based on T1 integrative emotion regulation did not load significantly on T2 deliberate self-harm, either in combination with T1 Harsh Parenting (β = .09, p = .078) or T1 Peer Victimization (β = .00, p = .996). Also, when based on T1 emotion dysregulation the interaction term did not affect T2 deliberate self-harm significantly, either when combined with T1 Harsh Parenting (β = .05, p = .383) or T1 Peer Victimization (β = .08, p = .347). The path from the interaction term to T2 deliberate self-harm was also nonsignificant when based on impulsivity, both in combination with T1 Harsh Parenting (β = −.17, p = .109) or T1 Peer Victimization (β = .01, p = .893).