Article
Victimization, Posttraumatic Stress Disorder Symptomatology, and Later Nonsuicidal Self-Harm in a Birth Cohort
Journal of Interpersonal Violence 26(18) 3667–3681 © The Author(s) 2011 Reprints and permission: http://www. sagepub.com/journalsPermissions.nav DOI: 10.1177/0886260511403757 http://jiv.sagepub.com
Shyamala Nada-Raja1 and Keren Skegg2
Abstract This longitudinal population-based study examined pathways to nonsuicidal self-harm (NSSH) in relation to childhood sexual abuse (CSA), assault victimization in early adulthood, posttraumatic stress disorder symptomatology (PTSD), and other mental disorders. At age 21, 476 men and 455 women completed interviews on assault victimization, PTSD, and other mental disorders. At age 26, they completed independent interviews on self-harm and childhood sexual abuse (CSA). Multivariate logistic regression analyses were conducted to determine predictors for NSSH at age 26. For men, anxiety and depressive disorders at age 21 were the only significant predictors of NSSH at age 26. For women, victimization, PTSD, and other anxiety disorders at age 21 all significantly predicted NSSH. CSA predicted later NSSH only indirectly, by increasing the risk of anxiety disorders among men and of assault victimization among women. In conclusion, pathways to nonsuicidal self-harm differed by sex. For women there were direct links with assault victimization 1
Injury Prevention Research Unit, Department of Preventive and Social Medicine, University of Otago, Dunedin, New Zealand 2 Department of Psychological Medicine, University of Otago, Dunedin, New Zealand Corresponding Author: Shyamala Nada-Raja, Senior Research Fellow, Injury Prevention Research Unit, University of Otago, PO Box 56, Dunedin, New Zealand Email:
[email protected]
3668
Journal of Interpersonal Violence 26(18)
and PTSD in early adulthood, whereas for men only internalizing disorders predicted future NSSH. Keywords self-injurious behavior, posttraumatic stress disorder, child sexual abuse, attempted suicide, crime victims
Victimization and suicidality are often related to each other and are strongly associated with mental disorder (Belik, Cox, Stein, Asmundson, & Sareen, 2007; Klonsky & Moyer, 2008; Seedat, Stein, & Forde, 2005; Ullman, 2004). Recent reviews on childhood sexual abuse (CSA, a major type of victimization) in relation to self-harm that includes both attempted suicide and nonsuicidal self-injury (NSSI) or nonsuicidal self-harm (NSSH) have emphasized that any observed associations are likely to be largely indirect and to include common psychosocial risk factors (Klonsky & Moyer, 2008; Ullman, 2004). There are also associations between sexual assault and revictimization in adult women and attempted suicide or NSSH (Gladstone et al., 2004; Noll, Horowitz, Bonanno, Trickett, & Putnam, 2003, Ullman, 2004; Ullman & Najdowski, 2009) and between physical assault or intimate partner violence and suicidality (Seedat et al., 2005; Simon, Anderson, Thompson, Crosby, & Sacks, 2002). The above studies of NSSI or NSSH in relation to prior victimization have generally been based on college rather than community samples. However, community-based samples of longitudinal studies have not always found a positive and direct association between CSA and later suicidality (e.g., Fergusson, Woodward, & Horwood, 2000), thus highlighting a need for longitudinal community studies to clarify the relationships that might exist between a range of psychosocial factors that predict suicidality following CSA (Ullman et al., 2009). Depression and revictimization are important factors in the pathways that have been suggested as connecting victimization and suicidality (Gladstone et al., 2004). Posttraumatic stress disorder (PTSD) is another important mediator that might link violence and self-harm (Kessler, 2000; Weierich & Nock, 2008). Current evidence is inconclusive as to whether PTSD predicts self-harm independently of other anxiety disorders and depression (Krysinska & Lester, 2010). Studies in this area therefore need to adjust for the role of internalizing disorders. (Breslau, Peterson, Poisson, Schultz, & Lucia, 2004). Women are more likely than men to report CSA and to experience depression and anxiety (Nelson et al., 2002). In a study of depressed women,
Nada-Raja and Skegg
3669
revictimization was much more frequent among women with a history of CSA than among comparison women with no CSA history, and sexual revictimization was correlated with PTSD symptoms (Noll et al., 2003). PTSD, CSA, and internalizing disorders are all risk factors for self-harm (Oquendo et al., 2003). Weierich and Nock (2008) highlighted a need for longitudinal research that could test the relative timing of CSA, PTSD symptoms, and self-harm. In U.S. studies, risk of PTSD was much greater after exposure to assault than to any other type of trauma (Kessler, 2000). In a community study about a variety of traumatic events, respondents overselected assault and sudden unexpected death of someone close as the most stressful events they had experienced (Breslau et al., 2004).A high proportion of this group also met criteria for PTSD. Several questions remain to be answered about assault, PTSD, and subsequent self-harm, and the possible role of CSA. These include (a) whether assault sensitizes or predisposes victims to later self-harm directly or whether this occurs via experiencing PTSD symptomatology; (b) whether, given comorbidity between PTSD and other mental disorder, mental disorder per se is the overriding factor that predicts self-harm among assault victims; (c) whether the association between CSA and self-harm is direct or mediated by revictimization and other mental disorder; and finally (d) whether the pathways differ for men and women. The aim of the present study was to examine these questions in relation to nonsuicidal self-harm (NSSH) in a birth cohort of New Zealand men and women separately. It was hypothesized that assault victimization resulting in significant PTSD symptomatology at age 21, in addition to other anxiety, depressive, and substance dependence disorders, would predict NSSH at age 26, and that CSA would also independently predict NSSH.
Method Sample The participants belong to a cohort of 1,037 born between April 1, 1972 and March 31, 1973, enrolled in the Dunedin Multidisciplinary Health and Development Study (DMHDS; Silva & Stanton, 1996). Data on their health, behavior, and background were available at birth and thereafter at regular intervals including at ages 21 and 26. The longitudinal study has a high followup rate of participants with more than 90% of the original cohort having been assessed at most phases of the study. At age 26 assessment, 980 study
3670
Journal of Interpersonal Violence 26(18)
participants gave written informed consent to take part in an interview on selfharm. More than 90% of the cohort self-identified as New Zealand European. The entire socioeconomic spectrum is covered adequately. The Otago Ethics Committee approved the study. Complete data were available for 449 women and 467 men for all measures described below.
Measures Nonsuicidal Self-Harm (NSSH). Participants were asked about a range of self-harm behaviors they had experienced in the past year at age 26 and about suicidal intent (Nada-Raja, Skegg, Langley, Morrison, & Sowerby, 2004). If no suicidal intent was reported for any of the behaviors the study member had engaged in (81 men and 54 women), they were classified in the NSSH category. Questions were framed in the context of dealing with mental or psychological pain, emotions, or stress rather than using the term “suicide” initially. Participants viewed a list of 16 specific self-harm behaviors and reported on any “other” similar behaviors, followed by questions on the number of episodes that involved suicidal intent for each method used in the past year (Nada-Raja et al., 2004). The behaviors comprised all methods specified in the International Classification of Diseases (ICD-9) E-codes 950-958 for suicide and purposely self-inflicted injury. In addition, other self-harmful behaviors were enquired about deliberately hitting oneself or putting one’s fist through a wall (i.e., self-hitting), denying oneself a necessity such as food to punish oneself, exercising excessively to deliberately hurt oneself, self-biting, or other bodily harmful behaviors (Nada-Raja et al., 2004). Most NSSH behaviors involved self-hitting, overdosing, or self-cutting.
Potential Predictors
Assault victimization at age 21 years. A total of 944 participants answered questions about any form of assault they had experienced in the previous year (Martin et al., 1998). Assault victimization was defined as the report of physical or sexual victimization by another person, involving anyone from a partner to a complete stranger, including attempted or threatened assault experienced in the past year (238 men and 144 women). Over half (55%) of physical assault incidents reported by women were perpetrated by a partner, in contrast to 12% of physical assault incidents against men being perpetrated by a partner (Langley, Martin, & Nada-Raja, 1997). Of those who had experienced physical assault, 12% of the men and 7% of the women had received some medical treatment, including being hospitalized.
Nada-Raja and Skegg
3671
Psychological distress and impairment following assault. Participants who reported a physical or sexual assault were asked to describe characteristics of their principal assault event, one that they remembered in particular had affected them the most. This procedure is considered an acceptable “shortcut”, rather than enquiring about every episode of trauma, and is considered a conservative approach to estimating risk of PTSD (Breslau et al., 2004). Then followed a set of 17 questions from the Diagnostic Interview Schedule (DIS) on Posttraumatic Stress Disorder for the principal assault event (Feehan, Nada-Raja, Martin, & Langley, 2001). Included were questions about persistent avoidance of stimuli, increased arousal, diminished responsiveness to the outside world, and symptoms of psychological reexperience of the event. The responses were coded as 0 = none, 1 = somewhat, 2 = yes, in accordance with modifications made to the DIS (Feehan, McGee, Nada Raja, & Williams, 1994). A PTSD diagnosis was not made because of interview time constraints that prevented a reliable examination of the duration for each symptom. Instead a total score was produced by summing individual responses. Cronbach’s coefficient alpha for the total symptom scale was .82 (Feehan et al., 2001). To gauge the degree of distress for PTSD symptoms, participants rated how much each symptom interfered with their life or daily activities on an impairment scale of 1 = very little to 5 = very much, and significant impairment was indicated by ratings of 2 or above on the scale (Feehan et al., 2001). Mean scores, differences between the sexes, and assaulted and nonassaulted groups have been published elsewhere (Feehan et al., 2001). Mental disorders at age 21. Disorders were diagnosed according to DSM-III-R criteria as per a modified version of the DIS referred to above. Diagnoses included anxiety other than PTSD, depression, and substance dependence (alcohol, cannabis) disorders experienced in the previous year (Newman et al., 1996). Childhood sexual abuse. At age 26, a computer-administered interview was used to elicit retrospective information from participants on their sexual behavior and experiences of sexual abuse or unwanted sexual contact by age 16 years. Childhood sexual abuse (CSA), reported by 18% of women and 7% of men, comprised having one’s genitals touched by another, being forced to touch another’s genitals, and experiencing forced or attempted sexual intercourse.
Statistical Analyses The chi-square statistic was used to examine associations between predictors and outcome variables. Logistic regression analyses were conducted to
3672
Journal of Interpersonal Violence 26(18)
examine potential predictors as described above for self-harm as per odds ratios (OR) and 95% confidence intervals (CI). Population attributable risk percentages (PAR) were calculated for each of the significant predictors identified in the final multivariate models to determine the proportion of selfharm that could be prevented if any one of the identified predictors were totally eliminated (Miettinen, 1974).
Results Assault victimization at age 21 and later NSSH. For men, there was no significant association between assault victimization at age 21 and NSSH at age 26. Women who reported that they had been assaulted at age 21 were, however, significantly more likely than other women to report NSSH at age 26 (χ2 = 10.3, p = .0013). Of the 144 women who reported assault victimization at age 21, 19% reported past-year NSSH at age 26 compared with 8% of the 295 women who did not report assault victimization. We examined three specific characteristics of assault in relation to later NSSH for women and men separately. Women who reported a sexual assault (23% vs. 10%, χ2 = 10.3, p = .0024), whose assailant was a partner (22% vs. 10%, χ2 = 6.3, p = .0122), or had received medical treatment for their assault (23% vs. 11%, χ2 = 4.3, p = .0390) were significantly more likely than women who did not report these characteristics to also report later NSSH. The results were not significant for men. PTSD symptomatology following assault victimization. Two dummy variables, one for victimization and PTSD symptomatology (hereafter referred to as PTSD) and the other for victimization without PTSD were created and included in logistic regression analyses to determine whether PTSD uniquely contributed to the associations described above. The reference group was no victimization. Odds ratios (OR) and 95% confidence intervals (CI) are summarized for each of the explanatory variables in Table 1, for 916 participants with complete data on all measures of interest. As shown in Table 1, there were no significant associations for men between victimization and PTSD at age 21 and NSSH at age 26. Women who were victimized at age 21, however, irrespective of experiencing PTSD, were at significantly higher risk of NSSH at age 26 than nonvictimized women. Role of mental disorders. Table 2 summarizes relations between victimization with or without PTSD and mental disorders at age 21. Victimization and PTSD were significantly associated with other anxiety disorders and with depressive disorders. A significantly higher number of victims with PTSD
3673
Nada-Raja and Skegg
Table 1. The Associations Between Past-Year Assault Victimization With and Without Associated PTSD Reported at Age 21 Years and Past-Year NSSH Behavior Reported at Age 26 Years Men (n = 467)
Women (n = 449)
OR
95% CI
OR
95% CI
No victimization Victimization, but no PTSD symptomatology Victimization and PTSD symptomatology
1.0 0.9
[0.6, 1.6]
1.0 2.1
[1.1, 4.2]
2.2
[0.9, 5.8]
3.9
[1.8, 8.6]
Note: NSSH refers to methods for suicide and purposely self-inflicted injury behaviors specified by ICD-9 E-codes 950-958, and “other” behaviors, for example, self-battery, self-biting, engaged in during the previous year in the absence of any self-reported suicidal intent. OR = odds ratio; CI = confidence interval.
compared with those in the remaining two groups also met criteria for a substance dependence disorder (alcohol or cannabis) at age 21. As victimization was significantly associated with mental disorder, it was of interest to determine if mental disorder mediated the associations between victimization and later NSSH. The results from multivariate logistic regression analyses, adjusting for three categories of disorder, are summarized in Table 3. For men, initial multivariate logistic regression analyses showed that anxiety disorder other than PTSD at age 21 (OR = 2.1) was a significant risk factor for NSSH at age 26. When the model excluded substance dependence disorders, depressive disorder (OR = 2.1, CI = [1.0, 4.1], p = .0373) was a significant risk factor, and the odds ratio was slightly increased for other anxiety disorder (OR = 2.2, CI = [1.1, 4.2]). For women, even after adjusting for mental disorders, victimization and PTSD remained significant risk factors for later NSSH (ORs reduced from 3.9 [univariate analysis] to 2.6). Anxiety disorder (other than PTSD) was also an independent and significant risk factor. When substance dependence disorder (not a significant risk factor) was removed from the multivariate regression model, victimization without PTSD (OR = 2.0, CI = [1.0, 4.1]) was a significant risk factor for later NSSH. Victimization at age 21 was also significantly associated with anxiety disorders at the same age (OR = 4.0, CI = [2.1, 7.6]).
3674 n 26 74 23 61 33 13
Men Women Men Women Men
Women
Any anxiety disorder Any depressive disorder Any substance dependence disorder 4
11 24 10 20 14
%
No Assault Victimization
13
24 25 20 24 60
n
14
12 27 10 26 29
%
Assault Victimization but No PTSD
12
10 25 10 25 14
n
27
43 56 43 56 61
%
Assault Victimization and PTSD
30.3
20.3 19.9 24.8 27.6 34.3
Chi-Square With 2 Degrees of Freedom
