Abstract
This article describes the development and validation of a brief screen for developmental trauma appraisals. The 7-item screen was administered to a non-clinical sample of 477 South African adolescents who had experienced interpersonal violence during childhood. In all, 216 participants (45%) met the study criteria for a diagnosis of posttraumatic stress disorder and 41 (9%) met the study criteria for a diagnosis of complex posttraumatic stress disorder. Exploratory factor analysis yielded one factor which had a high level of internal consistency (α = .91) and acceptable levels of construct and concurrent validity for both posttraumatic stress disorder and complex posttraumatic stress disorder outcomes. Mediation analyses indicated that the appraisal screen measures a construct which effectively mediates the relationship between traumatic exposure and the severity of posttraumatic outcomes across a broad range of maltreatment types.
A history of child maltreatment has been found to be associated with a variety of adverse behavioural and mental health outcomes, including posttraumatic stress disorder (PTSD; Erolin, Wieling, & Parra, 2014; Shenk, Putnam, Rausch, Peugh, & Noll, 2014). Available studies indicate that maltreated children are 5–10 times more likely than non-victimized controls to qualify for a diagnosis of PTSD (Scott, Smith, & Ellis, 2010), with exposure to multiple childhood traumas having been found to be associated with a greater risk of PTSD outcomes (Copeland, Keeler, Angold, & Costello, 2007).
It is, however, generally acknowledged that the consequences of child maltreatment are complex and variable (Játiva & Cerezo, 2014), with some victimized children going on to develop PTSD symptoms, while others do not (Collishaw et al., 2007; Widom, 1999). Attempts to identify factors associated with an increased risk of adverse mental health outcomes following child maltreatment have, to the large part, focused on abuse-related variables, with the severity of posttraumatic outcomes having been found to be associated with (a) the type, severity, and chronicity of traumatic exposure (English et al., 2005; Erolin et al., 2014; Jackson, Gabrielli, Fleming, Tunno, & Makanui, 2014); (b) the extent of poly-victimization (Copeland et al., 2007); (c) female gender (Udwin, Boyle, Yule, Bolton, & O’Ryan, 2000); and (d) non-supportive reactions to disclosure (Trickey, Siddaway, Meiser-Stedman, Serpell, & Field, 2012; Tyler, 2002).
In another line of research, the mediating role of trauma-related cognitions on child maltreatment outcomes has been explored in a number of studies (Calvete, 2014; Cromer & Smyth, 2010; Foa, Ehlers, Clark, Tolin, & Orsillo, 1999; Játiva & Cerezo, 2014; Shenk et al., 2014). Drawing on both general cognitive theories of stress and stress reactions (e.g., Lazarus & Folkman, 1984; Scherer, Klaus, Schorr, & Johnstone, 2001) and on more specific cognitive theories of reactions to traumatic exposure (e.g., Calvete, 2014; Ehlers & Clark, 2000; Janoff-Bulman, 1992; Spaccarelli, 1994), this line of research is predicated on the assumption that victim’s appraisals of traumatic events are likely to play a key mediating role in the development and maintenance of posttraumatic outcomes.
Specific trauma-related appraisals, which have been found to be associated with posttraumatic outcomes, include (a) negative trauma-related cognitions (Ehlers & Clark, 2000; Hembree & Foa, 2004), (b) negative cognitions about the self (Calvete, 2014; Cromer & Smyth, 2010; Foa et al., 1999; Játiva & Cerezo, 2014), (c) negative cognitions about others (Calvete, 2014), and (d) negative cognitions about the world (Cromer & Smyth, 2010; Foa et al., 1999; Shenk et al., 2014). Moreover, a number of studies have found that trauma-related appraisals predict an equivalent, or greater, proportion of the variance in PTSD outcomes than is predicted by trauma-related variables (e.g., Trickey et al., 2012).
Conceptual and methodological issues
The literature summarized above reflects two broad perspectives on the dynamics of PTSD outcomes: a stimulus-based perspective and a cognitive-mediational perspective (see Figure 1). Researchers operating from a stimulus-based perspective have been inclined to attribute variations in PTSD outcomes to variations in the intensity of exposure to potentially traumatic events (i.e., frequency, severity, chronicity, and/or the extent of exposure to poly-victimization; Jackson et al., 2014) and/or to trauma-related variables such as the type of traumatic exposure, the victim’s gender, the victim’s age, and/or the nature of the victim–perpetrator relationship. In Figure 1, this causal perspective is presented as a simple Stimulus–Response (S-R) model in terms of which trauma-related variables (the severity of abuse, victim’s gender, etc.) are assumed to play a primary role in determining PTSD outcomes.
Causal models of PTSD outcomes (S: Stimulus; M: Mediator; R: Response).
By way of contrast, the cognitive-mediation perspective, which, in recent years, has emerged as the standard paradigm in the field of psychology (Schwarzer & Taubert, 2002), would define posttraumatic outcomes as the product of a transaction between the individual and the environment in which trauma-related appraisals constitute the primary causal pathway between traumatic exposure and posttraumatic outcomes. In Figure 1, this perspective is presented using a simple mediation model (S-M-R), in terms of which the relationship between traumatic exposure (S) and traumatic outcomes (R) is assumed to be mediated (M) by victim appraisals of traumatic events.
However, in order to demonstrate that cognitive appraisals mediate the relationship between traumatic exposure and traumatic outcomes, a number of conditions need to be met. According to Baron and Kenny (1986), cognitive appraisals could be said to function as a mediator if (a) there are significant zero-order correlations between traumatic exposure and both traumatic outcomes (Path 1) and cognitive appraisals (Path 2), (b) there are significant zero-order correlations between cognitive appraisals and traumatic outcomes (Path 3), and (c) after controlling for Paths 2 and 3, the relationship between traumatic exposure and traumatic outcomes (Path 1) is either no longer significant (complete mediation) or there is a significant reduction in the significance of the relationship (partial mediation). The first two of these requirements have of course been consistently met in studies reviewed above, with there being some evidence to suggest that the direct relationship between traumatic exposure and traumatic outcomes may be mediated by cognitive appraisals (Calvete, 2014).
Thus, available findings would appear to provide support for a cognitive-mediation perspective on PTSD outcome. However, a number of authors (e.g., Holmbeck, 2002; Preacher & Hayes, 2004) have pointed out that changes (or non-changes) in the significance of Path 1, after controlling for Paths 2 and 3, may not reliably indicate the presence (or absence) of mediation effects, with there being an associated risk of researchers drawing incorrect conclusions regarding mediation. As such, Preacher and Hayes (2004) propose an alternate/additional strategy for determining the presence or absence of mediation effects (i.e., the Sobel test), which is designed to determine whether a mediator variable significantly carries the influence of an independent variable to a dependent measure.
In addition to statistical considerations, available attempts to validate a cognitive-mediation perspective have been hampered by definitional and measurement problems with respect to key variables. Operational definitions of traumatic exposure have, for example, tended to be restricted to direct forms of victimization occurring predominantly in the home (Calvete, 2014). As a result, little, if any, attention has been paid to children’s vicarious or ambient exposure to traumatic events or to children’ experiences of traumatic events occurring outside of the home.
Operational definitions of PTSD outcomes have also been somewhat narrowly defined in terms of the Diagnostic and Statistical Manual of Mental Disorders, 5th edition, criteria for PTSD (DSM-5; American Psychiatric Association, 2013), which fail to adequately address a constellation of symptoms (including affective and behavioural dysregulation, somatization, and dissociation) which have been found to be associated with a history of chronic exposure to interpersonal developmental trauma (D’Andrea, Ford, Stolbach, Spinazzola, & van der Kolk, 2012; Herman, 2009, 2012). Although various terms have been used to describe this constellation of symptoms – for example, disorders of extreme stress not otherwise specified (DESNOS: Pelcovitz et al., 1997), type II trauma (Terr, 1991), developmental trauma disorder (van der Kolk, 2005) – the term complex PTSD has been employed by a number of authors in the field and will be used in this article.
Finally, attempts to validate the cognitive-mediation model have been hampered by an absence of validated instruments which have been designed to address the full range of trauma-related appraisals which have been found to be associated with PTSD outcomes. Thus, for example, instruments such as the Posttraumatic Cognitions Inventory (PTCI; Foa et al., 1999) focus on negative appraisals of the self, others, and the world, while failing to adequately address appraisals relating to traumatic experiences themselves. Similarly, the Young Schema Questionnaire-3 (Young, 2006) focuses on negative appraisals relating to the self and to relationships with others while failing to adequately address negative appraisals of either the world in general or trauma-related experiences.
In an attempt to address some of the issues raised above, this study describes the development and preliminary validation of a screen for developmental trauma appraisals which was designed to assess the range of appraisals which have been found, or theorized, to be associated with posttraumatic outcomes. In order to ensure that key constructs were adequately captured in the study, (a) child maltreatment experiences were surveyed using an inventory which was specifically designed to provide a comprehensive measure of individual’s exposure to interpersonal violence, and (b) PTSD symptoms were assessed using validated measures of both PTSD and complex PTSD. In addition to exploring standard psychometric characteristics of the new measure, a mediational analysis was conducted in order to assess the conceptual validity of the appraisal measure.
Method
Participants
Data for the study were obtained from a convenience sample of students attending an English medium high school located in Durban (KwaZulu-Natal, South Africa) in 2011. Research questionnaires were distributed to 550 consenting participants, and 531 usable responses (96.5%) were returned. The study sample comprised 477 participants (89.8% of usable responses) who reported that they had been exposed to one or more incidents of interpersonal violence during childhood. Participants were predominantly male (66.9%) and Black African (86.9%) with a mean age of 15.7 years (range: 12–20 years; standard deviation [SD)]= 1.48). Participants were in Grades 8–12, with there being approximately 95 students per grade level.
An analysis of school registration records indicated that participants did not differ significantly from nonparticipating registered students in terms of age, race, or gender.
Instruments
In addition to completing the appraisal measure which was developed and validated in this study, participants completed the following measures.
Developmental Trauma Inventory
Developed by Collings, Valjee, and Penning (2013), the Developmental Trauma Inventory (DTI) is a 33-item screen for interpersonal childhood trauma developed specifically for the South African context, which assesses traumatic exposure across eight domains: emotional abuse, community assault, domestic assault, witnessing domestic violence, witnessing community violence, indecent assault, rape, domestic non-accidental injury, and domestic neglect. The authors report adequate levels of internal consistency for DTI subscales and evidence of concurrent criterion-related validity, in the sense that DTI subscale scores are significantly correlated with scores on clinical measures of both PTSD and complex PTSD. In the present sample, Cronbach’s alphas for DTI subscales were as follows: emotional abuse, α = .79; community assault, α = .77; domestic assault, α = .71; witnessing domestic violence, α = .70; witnessing community violence, α = .73; indecent assault, α = .71; rape, α = .72; domestic non-accidental injury, α = .71; and domestic neglect, α = .71.
PTCI
Developed by Foa et al. (1999), the PTCI is a 31-item measure of negative trauma-related appraisals which assesses negative appraisals in three domains: self-blame, negative cognitions about the self, and negative cognitions about others and the world. Items are scored on a 7-point scale ranging from 1 (totally disagree) to 7 (totally agree), with the authors reporting adequate levels of internal consistency and test–retest reliability for the scales (Foa et al., 1999). In the present sample, Cronbach’s alphas for the PTCI were as follows: total score, α = .94; self-blame, α = .90; negative cognitions about the self, α = .84; and negative cognitions about others and the world, α = 87.
Davidson Trauma Scale
Developed by Davidson (2003), the Davidson Trauma Scale (DTS) is a 17-item scale designed to assess for the severity and presence of PTSD symptoms. Each item is scored on a 5-point scale ranging from 0 (none) to 4 (every day) for frequency and from 0 (not at all distressing) to 4 (extremely distressing) for severity. The author reports adequate levels of test–retest reliability for the scale, with construct validity being demonstrated by high correlations between DTS scores and scores on other measures of PTSD (Davidson, 2003). In this study, Cronbach’s alphas for DTS scales were as follows re-experiencing, α = .89; avoidance, α = .88; hyper-arousal, α = .90; and full scale, α = .95.
Structured Interview for Disorders of Extreme Stress–Self-Report
Developed by Pelcovitz et al. (1997), the Structured Interview for Disorders of Extreme Stress–Self-Report (SIDES-SR) is a 45-item self-administered questionnaire designed to measure symptoms of complex developmental trauma. Subscales of the SIDES-SR comprise three clinical syndromes (affective and behavioural dysregulation, somatization, and dissociation) as well as three measures of altered perceptions (perceptions of self, others, and meaning). Validation studies on the SIDES-SR indicate that the instrument has adequate levels of internal consistency for both total and subscale scores (Luxenberg, Spinazzola, & van der Kolk, 2001). In the present sample, Cronbach’s alphas for the SIDES-SR were as follows: full scale, α = .82; affective and behavioural dysregulation, α = .73; somatization, α = .70; dissociation, .71; alterations in self-perceptions, α = .76; alterations in perceptions of others, α = 68; and alterations in systems of meaning, α = .71.
Procedure
Identifying scale items
In the initial phase of the research, a large pool of probes for trauma-related cognitions was sourced from the available research and clinical literature on interpersonal developmental trauma (Cromer & Smyth, 2010; Foa et al., 1999; Herman, 1992; Trickey et al., 2012). Thereafter, an item analysis was conducted in order to ensure that items were mutually exclusive and unambiguously phrased. Given that the study aimed to develop a brief screen for trauma-related cognitions, items were then further screened for redundancy and salience, with this process reducing the item pool to seven items (see Table 1).
Exploratory facture structure of items on the DTI appraisal screen (N = 475).
DTI: Developmental Trauma Inventory.
Pretesting
The items in Table 1 were formatted into a scale, with participants being requested to indicate ‘To what extent did you experience each of the following feelings and emotions because of what happened?’ For each item, participants were asked to complete a 5-point Likert scale ranging from 0 (not at all) to 4 (extremely). The scale was then pretested on a sample of 50 adolescents (approximately 10 from each grade) from the target school. Apart from indicating a need to change the wording of two items (in order to enhance comprehension), pretesting did not indicate the need for any further modifications/adaptations.
The validation study
In the final phase of the research, the appraisal scale was incorporated into a questionnaire which contained (a) standard demographic questions (age, grade, race, and gender), (b) the DTI appraisal scale which was developed in this study, (c) the DTI, (d) the PTCI, (e) the DTS, and (f) the SIDES-SR. Research questionnaires were administered to groups of consenting participants during Life Orientation classes.
Ethical considerations
Prior to questionnaire administration, ethical clearance for the research was obtained from the Humanities and Social Science Research Ethics Committee at the University of KwaZulu-Natal. Parental consent as well as participant assent was obtained from all participants, with participants being assured of confidentiality and anonymity and with offers of free counselling support (from a guidance teacher and/or a university clinic) being made to all participants.
Data analysis
Scores on the DTI appraisal scale were subjected to factor analysis using IBM SPSS 21, with factors being retained if they met the criteria of an eigenvalue of more than 1 and a factor loading of at least .40. Coefficient alpha was used to assess the internal consistency of scale items, with the construct validity of the scale being explored with respect to correlations between DTI appraisal scale scores and scores on the PTCI and scores on SIDES-SR subscales which assess trauma-related cognitions (i.e., alterations in self-perceptions, perceptions of others, systems of meaning).
The concurrent validity of the DTI appraisal scale was assessed using discriminant analyses in which scores on the appraisal measure were entered as independent variables and the presence of PTSD and complex PTSD was entered as criterion measures. Following Davidson (2003), the presence of PTSD was operationally defined as a minimum frequency score of 2 (2–3 times a week) and a minimum severity score of 2 (somewhat upsetting) for at least one re-experiencing symptom, three avoidance symptoms, and two hyper-arousal symptoms as assessed using the DTS. Complex PTSD was defined as being present if participants scored above the clinical cut-off on all six scales of the SIDES-SR.
Finally, a mediational analysis was conducted in order to explore the conceptual validity of the DTI appraisal measure. The mediational analysis was conducted using procedures suggested by Baron and Kenny (1986), with the Sobel test being used to determine whether the mediator variable (appraisals) significantly carried the influence of the independent measure (traumatic exposure) to the dependent measure (PTSD outcomes). Consistent with the general rule of thumb for mediational analyses (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002), a sample size of at least 100 per form of child maltreatment was regarded as being necessary in order to detect the medium effect sizes which have been noted in the available literature (Calvete, 2014; Cromer & Smyth, 2010; Foa et al., 1999; Játiva & Cerezo, 2014).
Results
Preliminary analyses
Preliminary analyses indicated that seven DTI scales yielded sample sizes of at least 100, which was deemed to be the minimum necessary to detect expected medium effect sizes in mediational analyses (MacKinnon et al., 2002): witnessing community violence (n = 334, 70.0%), domestic assault (n = 253, 53.0%), community assault (n = 213, 44.7%), indecent assault (n = 198, 41.5%), witnessing domestic violence (n = 165, 34.6%), emotional abuse (n = 126, 26.4%), and domestic veglect (n = 101, 21.2%). Only 24 participants (5.0%) reported that they had experienced rape or attempted rape during childhood, with 28 participants (5.9%) reporting that they had experienced one or more incidents of non-accidental domestic injury during childhood.
Thus, while all nine maltreatment domains of the DTI were considered in the psychometric evaluation of the DTI appraisal scale, incidents of rape and non-accidental domestic injury were excluded from mediational analyses due to small sample sizes.
Factor analysis and internal consistency
The factor analysis of items on the appraisal scale yielded a single factor solution, with all seven items demonstrating high (>.70) loadings on this factor (Table 1). Cronbach’s alpha for the seven items was .91.
Construct validity
Scores for the DTI appraisal scale were significantly correlated with total and subscale scores for the PTCI and with symptom severity scores for subscales of the SIDES-SR which measure trauma-related cognitions (see Table 2). Taken together, these findings provide support for the convergent construct validity of the DTI appraisal screen.
Convergent construct validity of the DTI appraisal screen.
DTI: Developmental Trauma Inventory; PTCI: Posttraumatic Cognitions Inventory; SIDES-SR: Structured Interview for Disorders of Extreme Stress–Self-Report.
N = 475.
p < .001.
Concurrent validity
A total of 216 participants (45.3%) met the study criteria for a diagnosis of PTSD and 41 (8.6%) met the study criteria for a diagnosis of complex PTSD. Scores on the DTI appraisal screen significantly discriminated between participants who (a) met the criteria for a diagnosis of PTSD (M = 14.13) from those who did not meet the criteria for a PTSD diagnosis (M = 8.46), λ = .863, χ2(1) = 69.55, p = .000, or (b) met the criteria for a diagnosis of complex PTSD (M = 18.05) from those who did not qualify for a complex PTSD diagnosis (M = 11.10), λ = .937, χ2(1) = 30.78, p = .000.
With respect to classification accuracy, an exploration of all possible cut-off points on the DTI appraisal screen indicated that a median split (<12 versus 12+) provided the highest positive predictive value for both PTSD (.67) and complex PTSD (.61). Estimates of sensitivity (i.e., the proportion of participants who qualified for a diagnosis who fell above the median split) were .68 for PTSD and .85 for complex PTSD, with estimates of specificity (i.e., the proportion of participants who did not qualify for a diagnosis who fell below the median split) being .67 for PTSD and .59 for complex PTSD.
Consistent with findings from previous studies, mean DTI appraisal scores for females were significantly higher than mean scores for males, with this trend being evident across all forms of developmental trauma examined in the study (Table 3).
Gender differences in trauma appraisals.
DTI: Developmental Trauma Inventory.
p < .01; ***p < .001.
Conceptual validity
From Table 4, it is evident that for each form of traumatic exposure considered in the study, there were significant zero-order correlations between all three variables in the cognitive-mediation model specified in Figure 1. Subsequent regression analyses (Table 5) indicated that for each type of maltreatment considered in the analysis, (a) the significant S-R association (Path 1) was no longer significant after controlling for Paths 2 and 3, and (b) results of Sobel tests were significant for all forms of maltreatment considered in the study. Taken together, these findings indicate significant mediation effects for each of the types of child maltreatment considered in the analysis.
Two-tailed zero-order correlations between hypothesized causal pathways.
DTI: Developmental Trauma Inventory; PTSD: posttraumatic stress disorder; SIDES-SR: Structured Interview for Disorders of Extreme Stress–Self-Report.
Pathways 1, 2, and 3 correspond to the three numbered pathways presented in Figure 1.
p < .05; **p < .01; ***p < .001.
Test of mediation effects.
DTI: Developmental Trauma Inventory; SIDES-SR: Structured Interview for Disorders of Extreme Stress–Self-Report.
Discussion
The present findings indicate that the DTI appraisal scale constitutes a reliable and valid measure of trauma-related appraisals in relation to both PTSD and complex PTSD outcomes. Moreover, and consistent with the predictions of the cognitive-mediation model of PTSD outcomes, the present findings indicate that items on the DTI appraisal scale measure a construct which effectively mediates the relationship between traumatic exposure and the severity of posttraumatic outcomes.
Taken together, these findings are consistent with, and extend, the results of previous studies which have provided evidence in support of a cognitive-mediation model of posttraumatic reactions to child maltreatment (Calvete, 2014), with the extension to previous studies lying in the present finding that trauma-related appraisals have significant mediation effects on posttraumatic outcomes across a broad range of types of maltreatment and for both PTSD and complex PTSD outcomes.
Although the present findings are consistent with a transactional conceptualization of developmental trauma outcomes (e.g., Spaccarelli, 1994), a major limitation of the study relates to the use of a cross-sectional design which does not permit strong causal inferences. We cannot, therefore, confidently infer that cognitions changed as a result of traumatic exposure or that negative traumatic appraisals influence the severity of posttraumatic outcomes. As such, further research, employing prospective research designs, is indicated in order to more clearly establish the direction of causal links suggested by the present research. A further limitation of the study relates to the fact that data were obtained from a relatively small, convenience sample of school-going adolescents attending a high school in an urban area. As such, further research is indicated in order to assess the extent to which the present findings can be generalized to other samples (e.g., general population samples, clinical samples, and samples drawn from other socio-cultural contexts).
Conclusion
Although the present findings provide preliminary support for a cognitive-mediation model of posttraumatic outcomes, further research is indicated in order to assess the mediating effects of cognitive appraisals on other posttraumatic outcomes (e.g., depression, anxiety disorders, and posttraumatic re-enactment behaviours) which have been found to be associated with traumatic exposure (cf. Penning & Collings, 2014; Ward, Flisher, Zissis, & Lombard, 2001). Furthermore, given that many cases of child maltreatment involve exposure to multiple forms of victimization (Dong et al., 2004) – including poly-victimization (Finkelhor, Ormrod, & Turner, 2007), chronic victimization (Courtois & Ford, 2009), and/or continuous victimization (Eagle & Kaminer, 2013) – further research would appear to be indicated in order to more systematically explore the mediating effects of trauma-related appraisals on posttraumatic outcomes in the aftermath of (or in the context of) multiple forms of exposure to child maltreatment.
At a broader level, the study findings are congruent with an emerging body of literature which suggests that posttraumatic reactions are most usefully understood from a transactional perspective, in terms of which trauma-related appraisals are assumed to play a key mediating role in the development and maintenance of posttraumatic outcomes (Calvete, 2014; Ehlers & Clark, 2000; Janoff-Bulman, 1992; Spaccarelli, 1994. Although additional research is clearly indicated in order to further explore the heuristic value of such a transactional perspective, the present findings suggest that the DTI appraisal screen constitutes a reliable and valid instrument which could usefully be employed in such research efforts.
Footnotes
Funding
This work is based upon research supported by the South African National Research Foundation (NRF). Any opinion, findings and conclusions or recommendations expressed in this material are those of the authors, and therefore, the NRF does not accept any liability in regard thereto.