Confirmatory Factor Analysis of the Trauma Symptom Checklist for Children in an At-Risk Sample of Youth
Abstract
Objective: The Trauma Symptom Checklist for Children (TSCC) is a widely used assessment tool in both community and clinical samples. Although the reliability and validity of the TSCC has been studied extensively, its factor structure has been understudied. Particularly, researchers have conceptualized the TSCC as a summative total scale score without evidence of its psychometric properties. Method: The current study examines the factor structure of the TSCC conceptualized as a total score using second-order confirmatory factor analysis (CFA) in a large sample of juvenile justice involved youth (N = 2,268). Results: Results indicate that a total score does not fit the data. However, an alternative model that included 2 factors that grouped together (a) anxiety, dissociation, and posttraumatic stress with (b) anger and depression fit the data well. Conclusions: Although the findings do not support the use of the TSCC as a total score, a 2-factor solution provides a more parsimonious conceptualization of the TSCC. Implications for both social work practice and research are discussed.
A large number of children and adolescents have experienced traumatic events both in their lifetimes and in their recent past (Bal, Van Oost, De Bourdeaudhij, & Crombez, 2003; Finkelhor, Turner, Shattuck, & Hamby, 2013; Singer, Anglin, Song, & Lunghofer, 1995). A recent study of a national sample of 4,503 children and adolescents found that more than 40% reported being a victim of physical assault in the past year (Finkelhor et al., 2013). In addition to exposure to interpersonal violence, 24.1% reported being a victim of a property offense in the past year (Finkelhor et al., 2013). In a national community sample of 3,004 youth, ages 12 to 17 years, 50.2% of youth reported other types of trauma exposure such as natural disasters and automobile accidents (Adams et al., 2013). Youth exposed to violent and nonviolent trauma may develop mental health symptoms associated with these traumatic events (Finkelhor, Ormrod, & Turner, 2007; Singer et al., 1995).
A number of assessment tools measuring symptoms of traumatic stress in children and adolescents have shown good psychometric properties (Hawkins & Radcliffe, 2006). In this study, we tested the factor structure of the Trauma Symptom Checklist for Children (TSCC), a measure of child and adolescent trauma symptoms widely used by researchers and in clinical settings (Elhai, Gray, Kashdan, & Franklin, 2005). Studies have conceptualized the TSCC as a six-factor solution originally reported in Briere (1996) on a normative sample comprised of three nonclinical samples. The TSCC has also been used in research as a total symptom scale (e.g. Singer et al., 1995). This study sought to expand on the current literature by testing the factor structure of the TSCC as a total scale score in a clinical sample of juvenile justice involved youth. In the following sections, we discuss the TSCC literature with particular attention paid to the psychometric properties of the scale.
Literature Review
A large number of youth who have significant exposure to trauma and who suffer from associated trauma symptoms are served by child welfare agencies and the juvenile justice system (Ko et al., 2008). Screening and assessment to identify youth affected by exposure to trauma is a key aspect of providing trauma-informed treatment (Kerig, 2013). Although a large number of trauma screening and assessment tools are available, the TSCC is among a few instruments specifically designed for and standardized on a large sample of children and adolescents (Hawkins & Radcliffe, 2006; Wolfpaw, Ford, Newman, Davis, & Briere, 2005). The TSCC consists of 54 items that measure symptoms associated with trauma that are written to be applicable and understandable to youth aged 8 to 16 years (Briere, 1996). A reduced 44-item TSCC-A, which omits the questions making up the sexual concerns subscale, is also available. The TSCC was designed as a self-report tool to assess the presence of symptoms rather than as a diagnostic interview, with elevated TSCC scores indicating that a full diagnostic interview might be warranted (Briere, 1996; Wolfpaw et al., 2005). Generally, the TSCC is regarded as a trauma assessment that can be completed quickly to provide reliable information about trauma symptomatology in children (Strand, Sarmiento, & Pasquale, 2005).
In developing the TSCC, Briere (1996) conducted an exploratory factor analysis (EFA) of the data that revealed six subscales including anger (ANG), anxiety (ANX), depression (DEP), dissociation (DIS), posttraumatic stress (PTS), and sexual concerns (SC). Data from a nonclinical sample of schoolchildren (n = 807) and a clinical sample of children receiving treatment for physical and sexual abuse (n = 91) in Sweden showed a similar six-factor solution for the TSCC translated into Swedish (Nilsson, Wadsby, & Svedin, 2008). However, results of the Swedish study showed some discrepancies in the factor structure as it pertains to items loading onto the original factors described by Briere (1996). Nilsson and colleagues (2008) attributed these discrepancies to the items in the TSCC and its constructs being interrelated. Given the interrelatedness of the items on the instrument, researchers have conceptualized the TSCC as both a six-factor solution and a total symptom score.
In a study of the effect of violence exposure on a community sample of children and adolescents, regression models were conducted for each of the five TSCC-A subscales (ANG, ANX, DEP, DIS, PTS) and a total score of the five factors (Singer et al., 1995). Other studies have also reported on summed scores of all TSCC subscales (i.e., Nilsson, Gustafsson, & Svedin, 2010; Sadowski & Friedrich, 2000) or selected TSCC subscales (i.e., Bal, De Bourdeaudhuij, Crombez, & Van Oost, 2005; Finkelhor et al., 2007; Turner, Finkelhor, & Ormrod, 2010). Generally, researchers have found high internal consistency (α = .94–.97) for the total trauma symptom scale (Nilsson et al., 2010; Sadowski & Friedrich, 2000; Singer et al., 1995). However, Finkelhor and colleagues (2007) cautioned against the use of the TSCC as a total trauma symptom score because the psychometric properties of such a total score remain untested. Use of the TSCC as a total symptom score can potentially affect the administration of the instrument by allowing social workers in practice settings to quickly sum all of the items rather than identifying the items that make up each of the six subscales. Scoring a measure that includes 54 items loading onto six factors can be time consuming, particularly in clinical settings where scores are often calculated by hand. A total symptom scale can also benefit researchers who are seeking a parsimonious solution to the measure’s factor structure.
With several exceptions (i.e., Finkelhor et al., 2007; Nilsson et al., 2008; Singer et al., 1995), the TSCC has been primarily used in clinical settings, particularly to assess trauma for children and adolescents who have been victims of sexual abuse (Bal et al., 2004; Crouch, Smith, Ezzell, & Saunders, 1999), victims of assault (McCart et al., 2005), exposed to domestic violence (Carter, Kay, George, & King, 2003), and exposed to parental methamphetamine abuse (Ostler, Bahar, & Jesse, 2010). Although the TSCC has generally been applied to clinical samples, factor structures of the instrument were explored using nonclinical samples, and have yet to be confirmed in a clinical sample. The reliability and validity of the TSCC conceptualized as six subscales have been studied in both clinical (Fricker & Smith, 2001; Lanktree et al., 2008; Li et al., 2009; Sadowski & Friedrich, 2000) and nonclinical (Nilsson et al., 2008) samples.
Current Study
Although studies have conceptualized the TSCC to include a total symptom scale (e.g. Singer et al., 1995; Turner, Shattuck, Hamby, & Finkelhor, 2013), no study has examined the psychometric properties of the total symptom scale. The total symptom scale provides a more parsimonious way to capture trauma symptoms in children for researchers, and provides an easily interpretable measure of trauma symptoms for social workers. Further, even though several studies have examined the psychometric properties of the TSCC in clinical samples, including its reliability and validity (e.g. Lanktree et al., 2008; Li et al., 2009), the factor structure of the TSCC in clinical samples remains understudied.
To address these gaps in the literature, we proposed to test the psychometric properties of the scale as a total trauma score on a clinical sample of juvenile justice-involved youth by examining a series of confirmatory factor analysis (CFA) models. We hypothesized that the 54 items in the TSCC consisted of six first-order latent factors (ANG, ANX, DEP, DIS, PTS, and SC) and a higher-order latent factor (total trauma). Figure 1 presents a path model showing the 54 observed variables along with the associated latent factors. The TSCC manual (Briere, 1996) includes descriptions of the hypothesized first-order latent factors, factor loadings, and cross loadings as shown in Figure 1. In addition, we proposed to test the shortened TSCC-A, an alternative 44-item version of the TSCC without the SC subscale. The hypothesized factor structure of the shortened TSCC-A scale is a similar higher-order factor model with all first-order latent factors loading on one higher-order latent factor. Hypothesized model
Method
Sample
The sample consisted of juvenile justice-involved youth with behavioral health issues participating in the Behavioral Health Juvenile Justice initiative (BHJJ), a community-based diversion program in 11 urban and rural counties in Ohio. Data from the BHJJ sample were collected during an intake interview on youth enrolled in the program between 2006 and 2013. All youth enrolled in the program were given the option to participate in the evaluation of the program. For youth who consented to the data collection, de-identified data were then sent to researchers at a Midwestern U.S. university for evaluation. Entry into the BHJJ program was typically through the juvenile court of each participating county. To participate in BHJJ, youth had to have a history of juvenile justice involvement, at least one diagnosis meeting criteria in the Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM‐IV; American Psychiatric Association, 1994), and between the ages of 10 and 18 years old. In addition, the state identified additional optional eligibility criteria, including co-occurring substance abuse, a pattern of criminal behavior, exposure to trauma or domestic violence, and a history of multi-system involvement. Each participating county was able to decide which of these optional criteria they would use to determine program eligibility. These criteria suggested by the state were used by the juvenile court of each county as a guide to identify youth who were most likely to benefit from treatment in BHJJ. Descriptive statistics of the study sample are presented in Table 1. The total sample consisted of 2,268 youth who had completed the TSCC as part of their intake assessment. The sample contained more males (59.0%, n = 1,337) than females, and the average age of youth was 15.07 years (SD = 1.53). The majority of the sample consisted of youth who self-identified as Caucasian (52.6%, n = 1,192), followed by those who self-identified as African American (38.9%, n = 882) or multiracial (6.4%, n = 145). All study protocols were approved by the Institutional Review Board of a university in the U.S. Midwest.
| M | SD | Range | % | n | |
|---|---|---|---|---|---|
| Gender | |||||
| Male | 59.0 | 1337 | |||
| Female | 41.0 | 931 | |||
| Age | 15.07 | 1.53 | 2268 | ||
| Race | |||||
| White | 52.6 | 1192 | |||
| Non-White | 47.4 | 1074 | |||
| Trauma Symptom Checklist for Children Subscales | |||||
| Anger (ANG) | 8.06 | 6.01 | 0–27 | 2202 | |
| Anxiety (ANX) | 3.86 | 3.98 | 0–23 | 2201 | |
| Depression (DEP) | 5.12 | 4.80 | 0–27 | 2202 | |
| Dissociation (DIS) | 5.87 | 5.11 | 0–30 | 2193 | |
| Posttraumatic stress (PTS) | 6.04 | 5.57 | 0–27 | 2202 | |
| Sexual concerns (SC) | 3.33 | 3.47 | 0–24 | 2199 |
Measure
The TSCC is a 54-item measure of trauma symptoms in children and adolescents designed to measure psychological trauma in children who are between the ages of 8 and 16 years (Briere, 1996). In accordance with the recommendations made in the professional manual, researchers have also administered the TSCC to 17-year-old youth (i.e., Butcher, Kretschmar, Lin, Singer, & Flannery; 2014; Finkelhor et al., 2007; Fricker & Smith, 2001; Singer et al., 1995). The TSCC is normed on two age groups: youth ages 8 to 12 years, and youth 13 to 17 years old. The current sample was composed of 2,134 youth who were 13 to 17 years old (91.4%) and 134 youth who were 8 to 12 years old (5.9%). The TSCC consists of six subscales: ANG, ANX, DEP, DIS, PTS, and SC. Possible responses on the items that make up the TSCC range from never (coded 0) to almost all of the time (coded 3). Previous data from a large national study (N = 2,030) of a community sample of youth showed good reliability (α = .72−.83) on all TSCC subscales (Finkelhor et al., 2007). The subscales also exhibited good reliability (α = .77−.89) in a small clinical sample (N = 100) of sexually abused adolescents (Bal et al., 2004).
Data Analysis
The purpose of the current study was to examine the factor structure of the TSCC and TSCC-A conceptualized as a total symptom scale by presenting a series of second-order CFA models and their fit to the data. In addition to the hypothesized model presented in Figure 1, two alternative models are proposed as comparison models (see Table 2). An EFA on the normative sample revealed overlapping items on the ANX, PTS, and DIS subscales whereas the three remaining subscales did not include any overlapping items (Briere, 1996). One alternative model (Model 2) was used to test the conceptualization of the TSCC as consisting of two higher-order factors that included the three overlapping subscales as one higher-order factor (ANX, PTS, and DIS) and the remaining subscales as a separate higher-order factor. For the full TSCC measure, a third model was tested in which the five factors included in the TSCC-A were conceptualized as two higher-order factors and the SC subscale was isolated as a first-order latent factor. For both the TSCC and the shortened TSCC-A, Model 1 was hypothesized to consist of one higher-order factor and nested in Model 2, a solution consisting of two higher-order factors. Model 3 and Model 2 for the full TSCC measure both had two higher-order factors, and therefore, were not nested models.
| First-order latent factors | Higher-order latent variable(s) | |
|---|---|---|
| TSCC | ||
| Model 1 | ANG, ANX, DEP, DIS, PTS, SC | Total Scale Score |
| Model 2 | ANG, ANX, DEP, DIS, PTS, SC | ANX, DIS, PTS – Factor 1 |
| ANG, DEP, SC – Factor 2 | ||
| Model 3 | ANG, ANX, DEP, DIS, PTS, SC | ANX, DIS, PTS – Factor 1 |
| ANG, DEP – Factor 2 | ||
| TSCC-A | ||
| Model 1 | ANG, ANX, DEP, DIS, PTS | Total Scale Score |
| Model 2 | ANG, ANX, DEP, DIS, PTS | ANX, DIS, PTS – Factor 1 |
| ANG, DEP – Factor 2 |
All CFA models were estimated using Mplus version 7.11 (Muthén & Muthén, 1998–2012). Given that TSCC items were on a 4-point categorical scale, the means and variance adjusted weighted least squares (WLSMV) estimation method was used. Previous studies have shown the WLSMV estimator to be effective in fitting CFA models with categorical and non-normal data (Beauducel & Herzberg, 2006). Modification indices were examined to identify potential model misspecification and models were respecified when both theoretically and statistically appropriate using an iterative process.
Several commonly used goodness-of-fit indices were considered when evaluating the models presented in the following sections. Indices reported include a χ2 test, comparative fit index (CFI), Tucker-Lewis Index (TLI), and the root mean square error of approximation (RMSEA). The first of these indices, the χ2 test, is a universally used index to assess CFA models. A significantly large χ2 statistic indicates that the hypothesized model adequately fitting the data is an unlikely event, and therefore, the model should be rejected (Byrne, 2012). However, the χ2 statistic is sensitive to sample size, and larger sample sizes tend to result in an inflated χ2 statistic. Further, model complexity also leads to an inflated χ2 statistic. An attempt to circumvent the problems inherent in the application of the χ2 test is the relative χ2 statistic calculated as the χ2/df. Although there is no consensus, particularly given its sensitivity to sample size, a χ2/df of less than 5.0 can be considered an adequate fit (Marsh & Hocevar, 1985). In addition to χ2/df less than 5.0, we considered a CFI value above .95, a TLI value above .95, and a RMSEA value less than .06 to indicate good model fit (Hu & Bentler, 1999).
To compare nested models, we used the Mplus Difftest procedure, which is designed to test model differences in nonnormally distributed data (Muthén & Muthén, 1998–2012). A significant χ2 statistic indicates that the additional restriction worsens model fit. For the models described in this study, the total scale score model (Model 1) is nested within the models with two higher-order factors (Models 2 and 3). For the full TSCC, Models 2 and 3 are not nested models, and difference testing cannot be conducted. Ultimately, we were interested in determining whether statistical evidence existed to justify restricting the TSCC to a total scale score.
Results
For each of the 54 items included in the TSCC, less than 2% were missing. Little’s test for MCAR was violated, χ2 (9378) = 10753.93, p < .001. However, given the low percentage of missing data on each of the observed variables, no further corrections for missing data were applied. By default, the WLSMV estimator in the Mplus program handles missing data using pairwise present data (Muthén & Muthén, 1998–2012). The hypothesized total scale score showed good internal consistency for both the full TSCC (α = .96) and the shortened TSCC-A (α = .96) measures. The subscales also displayed adequate internal consistency ranging between α = .73 (SC) and α = .90 (ANG). For the current sample, means and standard deviations for the six subscales are reported in Table 1.
TSCC Model Fit
Baseline models were estimated separately on the sample for both the TSCC and the shortened TSCC-A measure. Table 3 presents fit indices for the hypothesized and alternative TSCC models. Fit indices demonstrated that Model 1 approached an acceptable fit to the data, χ2 (1368) = 9449.31, χ2/df = 6.91, CFI = .92, TLI = .92, RMSEA = .05. Fit indices for Model 2, χ2 (1367) = 9463.76, χ2/df = 6.92, CFI = .92, TLI = .92, RMSEA = .05, and Model 3, χ2 (1366) = 9530.54, χ2/df = 6.98, CFI = .92, TLI = .92, RMSEA = .05, also demonstrated that the models approached an acceptable fit to the data.
| χ2 | df | χ2/df | CFI | TLI | RMSEA [CI] | |
|---|---|---|---|---|---|---|
| Model 1 | 9449.31 | 1368 | 6.91 | .92 | .92 | .05 [.050, .052] |
| Model 1 - Adjusted | 6977.82 | 1364 | 5.11 | .95 | .94 | .04 [.042, .044] |
| Model 2 | 9463.76 | 1367 | 6.92 | .92 | .92 | .05 [.050, .052] |
| Model 2 - Adjusted | 6995.74 | 1363 | 5.13 | .95 | .94 | .04 [.042, .044] |
| Model 3 | 9530.54 | 1366 | 6.98 | .92 | .92 | .05 [.050, .052] |
| Model 3 - Adjusted | 6997.41 | 1362 | 5.14 | .95 | .94 | .04 [.042, .044] |
| χ2 difference tests | ||||||
| Model 1–2 | 1.51 | 1 | ||||
| Model 1–3 | 21.40* | 2 |
Modification indices showed four additional parameters as possible reasons for misspecification for all three models. Covariances between two items on the DEP subscale and two items on the SC subscale were specified. Two items on the SC subscale were also identified as possibly cross loading on additional subscales. The item that relates to mistrusting someone based on perceived sexual interest cross loaded on the ANX subscale and the item that relates to a desire to express dirty words cross loaded on the ANG subscale. Modifications were made individually in separate models and improvements on model fit indices were observed at each step. Final adjusted models demonstrated excellent model fit for Model 1, χ2 (1364) = 6977.82, χ2/df = 5.11, CFI = .95, TLI = .94, RMSEA = .04, Model 2, χ2 (1363) = 6995.74, χ2/df = 5.13, CFI = .95, TLI = .94, RMSEA = .04, and Model 3, χ2 (1362) = 6997.41, χ2/df = 5.14, CFI = .95, TLI = .94, RMSEA = .04. Although both the χ2 and the relative χ2 indices suggest poor model fit, incremental fit indices CFI, TLI, and the RMSEA suggest excellent model fit. The current sample is large and complex, contributing to inflated χ2 statistics for the three models.
Model comparisons indicated that the additional restriction posed in Model 1 over Model 2 did not significantly worsen model fit, χ2 (1) = 1.51, p = .21. However, when the SC subscale was isolated in Model 3 as a first-order latent factor and compared to a total score (Model 1), the additional restriction significantly worsened model fit χ2 (2) = 21.40, p < .001. We therefore propose Model 3 as the best fitting model.
TSCC-A Model Fit
Fit indices for two competing models of the shortened TSCC-A measure are presented in Table 4. The hypothesized model (Model 1) exhibited adequate fit to the data, χ2 (894) = 5494.00, χ2/df = 6.14, CFI = .95, TLI = .95, RMSEA = .05. Similarly to the full TSCC instrument, modification indices suggested specifying a covariance between two items in the DEP subscale that relate to thinking about self-harm and suicidal ideation. Specifying this covariance improved model fit indices, χ2 (893) = 4765.94, χ2/df = 5.34, CFI = .96, TLI = .96, RMSEA = .04. Although fit indices for Model 2 also exhibited an adequate fit, χ2 (893) = 5499.81, χ2/df = 6.16, CFI = .95, TLI = .95, RMSEA = .05, specifying the additional covariance improved model fit, χ2 (892) = 4767.01, χ2/df = 5.34, CFI = .96, TLI = .96, RMSEA = .04. A comparison of the two models revealed that the additional restriction to a one second-order factor model resulted in a significantly worse fit to the data, χ2 (1) = 11.80, p < .01. We therefore propose Model 2 over Model 1 as a better fit to the data.
| χ2 | df | χ2/df | CFI | TLI | RMSEA [CI] | |
|---|---|---|---|---|---|---|
| Model 1 | 5494.00 | 894 | 6.14 | .95 | .95 | .05 [.046, .049] |
| Model 1 - Adjusted | 4765.94 | 893 | 5.34 | .96 | .96 | .04 [.043, .045] |
| Model 2 | 5499.81 | 893 | 6.16 | .95 | .95 | .05 [.046, .049] |
| Model 2 - Adjusted | 4767.01 | 892 | 5.34 | .96 | .96 | .04 [.043, .045] |
| χ2 difference tests | ||||||
| Model 1–2 | 11.80* | 1 |
Factor Loadings
All factor loadings for the best fitting models for the full TSCC instrument and the TSCC-A are presented in Table 5. As conceptualized and presented in the TSCC manual, the TSCC includes three cross loading factors (Briere, 1996). Although factor loadings of Items 24 and 25 were positively and significantly associated with the ANX subscale for the full TSCC scale, the same items were not significantly associated with the PTS subscale. For the shortened TSCC-A scale, Item 24 (being fearful of men) and Item 25 (being fearful of women) were significantly associated with the ANX subscale, but only Item 24 was significantly and positively associated with the PTS subscale. Both items exhibited weak factor loadings on the PTS subscale while loading strongly on the ANX subscale. As a result of model respecification, two cross loadings were found on the full TSCC instrument for the SC subscale where Item 4 (desire to express dirty words) was significantly and positively associated with the ANG subscale, and Item 34 (mistrusting people because of perceived sexual interest) was significantly and positively associated with the PTS subscale. Both items exhibited weak factor loadings on the SC subscale.
| Items | Factor | TSCC | TSCC-A | Items | Factor | TSCC | TSCC-A |
|---|---|---|---|---|---|---|---|
| Item 2 | ANX | .76* (.01) | .76* (.01) | Item 11a | PTS | .24* (.04) | .23* (.04) |
| Item 15 | ANX | .79* (.01) | .80* (.01) | Item 12 | PTS | .83* (.01) | .83* (.01) |
| Item 24a | ANX | .62* (.06) | .40* (.06) | Item 35 | PTS | .83* (.01) | .83* (.01) |
| Item 25a | ANX | .69* (.09) | .47* (.10) | Item 43 | PTS | .86* (.01) | .86* (.01) |
| Item 32 | ANX | .79* (.01) | .79* (.01) | Item 51 | PTS | .73* (.01) | .73* (.01) |
| Item 33 | ANX | .83* (.01) | .84* (.01) | Item 24a | PTS | .06 (.06) | .27* (.06) |
| Item 39 | ANX | .54* (.03) | .55* (.03) | Item 25a | PTS | −.13 (.10) | .08 (.10) |
| Item 41 | ANX | .75* (.01) | .76* (.01) | Item 5 | DIS | .56* (.02) | .55* (.02) |
| Item 50 | ANX | .69* (.02) | .68* (.02) | Item 18 | DIS | .70* (.02) | .70* (.02) |
| Item 34 | ANXb | .57* (.03) | Item 29 | DIS | .78* (.01) | .78* (.01) | |
| Item 7 | DEP | .78* (.01) | .78* (.01) | Item 30 | DIS | .69* (.01) | .69* (.01) |
| Item 9 | DEP | .83* (.01) | .83* (.01) | Item 31 | DIS | .78* (.02) | .78* (.02) |
| Item 14 | DEP | .69* (.01) | .70* (.01) | Item 38 | DIS | .69* (.02) | .69* (.02) |
| Item 20 | DEP | .74* (.02) | .74* (.02) | Item 45 | DIS | .74* (.01) | .74* (.01) |
| Item 26 | DEP | .58* (.02) | .56* (.03) | Item 48 | DIS | .76* (.01) | .76* (.01) |
| Item 27 | DEP | .82* (.01) | .81* (.01) | Item 53 | DIS | .61* (.02) | .61* (.02) |
| Item 28 | DEP | .76* (.01) | .76* (.01) | Item 11a | DIS | .53* (.04) | .53* (.04) |
| Item 42 | DEP | .78* (.01) | .77* (.01) | Item 4b | SC | .19* (.03) | |
| Item 52 | DEP | .71* (.02) | .71* (.02) | Item 8 | SC | .66* (.04) | |
| Item 6 | ANG | .70* (.01) | .70* (.01) | Item 17 | SC | .57* (.03) | |
| Item 13 | ANG | .82* (.01) | .82* (.01) | Item 22 | SC | .71* (.04) | |
| Item 16 | ANG | .78* (.01) | .78* (.01) | Item 23 | SC | .77* (.03) | |
| Item 19 | ANG | .82* (.01) | .82* (.01) | Item 34b | SC | .11* (.03) | |
| Item 21 | ANG | .71* (.02) | .70* (.02) | Item 40 | SC | .89* (.4) | |
| Item 36 | ANG | .52* (.02) | .51* (.02) | Item 44 | SC | .83* (.02) | |
| Item 37 | ANG | .80* (.01) | .80* (.01) | Item 47 | SC | .76* (.02) | |
| Item 46 | ANG | .84* (.01) | .84* (.01) | Item 54 | SC | .77* (.04) | |
| Item 49 | ANG | .82* (.01) | .82* (.01) | ANX | Factor 1 | .94* (.01) | .95* (.01) |
| Item 4 | ANGb | .50* (.02) | DIS | Factor 1 | .94* (.01) | .93* (.01) | |
| Item 1 | PTS | .62* (.02) | .63* (.02) | PTS | Factor 1 | .88* (.01) | .88* (.01) |
| Item 3 | PTS | .79* (.01) | .79* (.01) | ANG | Factor 2 | .76* (.01) | .77* (.01) |
| Item 10 | PTS | .80* (.01) | .80* (.01) | DEP | Factor 2 | .94* (.01) | .95* (.01) |
Discussion
The current study examined the psychometric properties of the TSCC and the shortened TSCC-A as a total symptom scale in a sample of juvenile justice-involved youth with behavioral health issues. Although previous studies have operationalized the measure to include a total symptom scale (e.g., Singer et al., 1995), the psychometric properties of such a scale remain understudied. The current study proposed a series of CFA models to test whether there is empirical evidence for the use of the TSCC as a total score. Results indicated that restricting the TSCC to one higher order factor worsened model fit. This finding indicates that the data do not support the hypothesis that the TSCC is comprised of a total scale score in addition to the six subscales. The best fitting model showed two higher-order factors in addition to the six subscales. The first higher-order factor consisted of the ANX, DIS, and PTS subscales that had cross-loading items originally found in an EFA of the standardization samples described by Briere (1996). The remaining two subscales, ANG and DEP loaded onto a second higher order factor, whereas SC was isolated as a first-order factor. This finding was consistent across both the TSCC and the shortened TSCC-A instruments. A potential reason for the ANG and DEP subscales being retained as a second order factor is that the DEP subscale includes items that address thoughts of inflicting self-harm and suicidal ideation. Elevated levels of anger in concert with depression might indicate that the youth is at risk for non-suicidal self-harm (Hawton, Kingsbury, Steinhardt, James, & Fagg, 1999; Laye-Gindhu & Sconert-Reichl, 2005) and suicidal ideation (Boergers, Spirito, & Donaldson, 1998; Flannery, Singer, & Wester, 2001; Kerr & Schneider, 2008; Shain, 2007). First-order CFA models could not be compared with these second-order models because first-order models typically tend to exhibit better fit.
Further, the data showed that the SC subscale seems to be independent of the remaining subscales in the TSCC. Both the factor loadings and the reliability analysis seem to indicate a relative misfit and independence of the SC subscale from the other subscales. In testing the full version of the instrument, we found that the best fitting model kept the SC subscale from loading onto a higher-order factor. Further, two of the items included in the SC subscale cross loaded with other subscales. The first cross-loading item involved expressing dirty language, which might be construed by youth to involve profanity rather than language of a sexual nature. This interpretation might have contributed to the item cross-loading with the ANG subscale. The second cross-loading item involved a mistrust of people, which might be indicative of general avoidance, a concept measured by the PTS subscale (Briere, 1996). Although the reliability of the SC subscale in the current sample was within the acceptable range (α = .73), it nevertheless exhibited the lowest levels of reliability among the TSCC subscales. Previous studies have found the SC subscale to exhibit the lowest alphas among the TSCC subscales (Crouch et al., 1999; Kaslow & Thompson, 2008).
The utility of a TSCC total symptom scale is apparent for researchers and social work practitioners alike. For researchers using the TSCC in a regression model, the total scale can be a parsimonious alternative to presenting separate models for each of the six subscales. In addition, for social work practitioners, scoring the TSCC, particularly by hand, can be a difficult and time-consuming task. However, assessing trauma in children can be complex as repeated exposure to trauma can manifest itself in a wide range of symptoms (Briere & Spinazzola, 2005). Other measures of trauma in children such as the UCLA PTSD Reaction Index (Steinberg, Brymer, Decker, & Pynoos, 2004) and the Child Sexual Behavior Inventory (Friedrich, 1997) focus on a specific area of trauma symptoms. In contrast, the TSCC was developed to measure a broad array of trauma symptomatology (Briere, 1996). Although some overlap exists among the items and subscales, the results and findings from this current study indicate that the subscales provide important information that is lost when the subscales are combined to form a total score. We therefore recommend that a total score should not be used in the absence of subscale scores, both in research and clinical practice. Considering the factors retained by the CFA models presented in this study, we further recommend that clinicians pay close attention to the combination of the ANG and DEP subscales because elevated scores can suggest a potential for self-injurious behaviors. Elevated scores on this new factor might contribute information beyond the two items that directly ask whether the youth has thought about self-harm and suicide.
Limitations
A number of limitations of the current study are of note. The data are limited to a specific sample of juvenile justice-involved youth with behavioral health issues. Youth in this sample were participating in a juvenile justice diversion program designed to provide treatment for youth with behavioral health issues. Results here may not be generalizable to community samples. However, the TSCC has been administered in a variety of clinical samples of youth with a history of trauma exposure (e.g., Bal et al., 2004; Carter et al., 2003; McCart et al., 2005). Studies of juvenile justice-involved youth have found high rates of trauma exposure, with one study finding that more than 90% of juvenile detainees reported experiencing one or more traumas, with an average of more than 14 separate incidents (Abram et al., 2004). Results from the current sample may be applicable to other clinical samples of youth with a history of trauma exposure.
These findings are also limited by the small percentage of younger children in the sample. The normative sample was split into two groups of younger (ages 8 to 12 years) and older (ages 13 to 16 years) youth (Briere, 1996). Norms for each of the six subscales were established by age group and gender. In general, juvenile justice samples are older than 12 years, particularly because well over 90% of juvenile court cases in the United States involve children older than 13 years (Puzzanchera & Robson, 2014). Consistent with these numbers, children in the younger age group (age 8 to 12 years) made up 5.9% (n = 134) of the current sample.
One final limitation in this study is the method of data collection. For the current sample, the administration of the TSCC took place as part of an intake interview conducted by an intake professional, and data were intended to be used to inform treatment planning. In contrast to a self-administered and anonymous survey, having a case worker or other professional present can potentially affect a young person’s responses, particularly for survey instruments that contain sensitive questions (Tourangeau & Yan, 2007). Questions regarding sexuality might be particularly sensitive, and youth might be more likely to underreport these issues. Therefore, scores on the SC subscale might have been most affected by the presence of an interviewer. However, such interviews are common in clinical settings, and likely reflects the way in which the TSCC is used by social workers.
Conclusion
In applying the TSCC in a practice environment, a total symptom score would allow social workers to quickly and efficiently score the instrument and to interpret and communicate results to clinical teams. The TSCC is difficult to score by hand because items that make up each subscale are interspersed throughout the measure. Further, interpreting and communicating the results of six different subscales can be challenging. However, the data presented in this article do not support the use of the TSCC as a total symptom score. The factor structure indicates that the subscales provide information that is independent of one another, and can provide valuable information that can be used to inform treatment decisions.
Although a number of studies have reported on and used the TSCC as a total score, no study has examined the psychometric properties of a total scale score. This study attempts to fill this gap by examining a series of higher order CFA models. Data presented here provided no statistical evidence for the use of the TSCC as a total scale. The TSCC was developed to capture the breadth of symptomatology experienced by children exposed to trauma. The current study reinforces the original use and conceptualization of the TSCC as made up of six subscales. Although the complexity of scoring the TSCC as six separate subscales poses a challenge to using the scale efficiently in a clinical setting, the use of a total scale score can potentially limit the utility of the scale in assessing trauma symptoms in children. In addition to considering each of the subscales, we further recommend the careful examination of the combination of the ANG and DEP subscales. Future research should further examine the factor structure of the TSCC in a community sample of youth. An additional finding from this study is that there are cross-loading items from the SC subscale. Additional samples should be collected to confirm the factor structure of the SC subscale.
This research was supported by funding through grants from the Ohio Department of Youth Services (4AS3190; Jeff M. Kretschmar) and the Ohio Department of Mental Health and Addiction Services (4AS3190; Jeff M. Kretschmar).
Notes
Fredrick Butcher, is a research associate with the Begun Center for Violence Prevention Research and Education at Case Western Reserve University’s Jack, Joseph, and Morton Mandel School of Applied Social Sciences.
Jeff M. Kretschmar is a research assistant professor at the Mandel School of Applied Social Sciences Case Western Reserve University and senior research associate with the Begun Center for Violence Prevention Research and Education.
Mark I. Singer is the Leonard W. Mayo Professor of Family and Child Welfare at the Mandel School of Applied Social Sciences, Case Western Reserve University. He is currently the deputy director of the Semi J. and Ruth W. Begun Center for Violence Prevention Research and Education and co-director of the Center on Substance Abuse and Mental Illness.
Daniel Flannery is the Semi J. and Ruth Begun Professor and director of the Begun Center for Violence Prevention Research and Education at the Jack, Joseph, and Morton Mandel School of Applied Social Sciences at Case Western Reserve University.
Correspondence regarding this article should be sent to Fredrick Butcher, Ph.D., Mandel School of Applied Social Sciences, Case Western Reserve University, 11402 Bellflower Rd., Cleveland, OH 44106, or via e-mail to [email protected]
References
Abram, K. M., Teplin, L. A., Charles, D. R., Longworth, S. L., McClelland, G. M., & Dulcan, M. K. (2004). Posttraumatic stress disorder and trauma in youth in juvenile detention. Archives of General Psychiatry, 61, 403–410. http://dx.doi.org/10.1001/archpsyc.61.4.403 Adams, Z. W., McCart, M. R., Zajac, K., Danielson, C. K., Sawyer, G. K., Saunders, B. E., & Kilpatrick, D. G. (2013). Psychiatric problems and trauma exposure in nondetained delinquent and nondelinquent adolescents. Journal of Clinical Child & Adolescent Psychology, 42, 323–331. American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed., text rev.). Washington, DC: Author. Bal, S., De Bourdeaudhuij, I., Crombez, G., & Van Oost, P. (2004). Differences in trauma symptoms and family functioning in intra- and extrafamilial sexually abused adolescents. Journal of Interpersonal Violence, 19, 108–123. http://dx.doi.org/10.1177/0886260503259053 Bal, S., De Bourdeaudhuij, I., Crombez, G., & Van Oost, P. (2005). Predictors of trauma symptomatology in sexually abused adolescents: A 6-month follow-up study. Journal of Interpersonal Violence, 20, 1390–1405. http://dx.doi.org/10.1177/0886260505278720 Bal, S., Van Oost, P., De Bourdeaudhuij, I., & Crombez, G. (2003). Avoidant coping as a mediator between self-reported sexual abuse and stress-related symptoms in adolescents. Child Abuse & Neglect, 27, 883–897. http://dx.doi.org/10.1016/S0145-2134(03)00137-6 Beauducel, A., & Herzberg, P. Y. (2006). On the performance of maximum likelihood versus means and variance adjusted weighted least squares estimation in CFA. Structural Equation Modeling, 13, 186–203. http://dx.doi.org/10.1207/s15328007sem1302_2 Boergers, J., Spirito, A., & Donaldson, D. (1998). Reasons for adolescent suicide attempts: Associations with psychological functioning. Journal of the American Academy of Child & Adolescent Psychiatry, 37, 1287–1293. http://dx.doi.org/10.1097/00004583-199812000-00012 Briere, J. (1992). Trauma Symptom Checklist for Young Children (TSCYC) professional manual. Odessa, FL: Psychological Assessment Resources. Briere, J. (1996). Trauma Symptoms Checklist for Children (TSCC): Professional manual. Odessa, FL: Psychological Assessment Resources. Briere, J., & Spinazzola, J. (2005). Phenomenology and psychological assessment of complex posttraumatic states. Journal of Traumatic Stress, 18, 401–412. http://dx.doi.org/10.1002/jts.20048 Butcher, F., Kretschmar, J. M., Lin, Y., Flannery, D. J., & Singer, M. I. (2014). Analysis of the validity scales in the Trauma Symptom Checklist for Children, Research on Social Work Practice. 24(6), 695–704. http://dx.doi.org/10.1177/1049731513516803 Byrne, B. M. (2012). Structural equation modeling with Mplus: Basic concepts, applications, and programming. New York, NY: Routledge. Carter, L., Kay, S. J., George, J. L., & King, P. (2003). Treating children exposed to domestic violence. Journal of Emotional Abuse, 3, 183–202. http://dx.doi.org/10.1300/J135v03n03_02 Crouch, J. L., Smith, D. W., Ezzell, C. E., & Saunders, B. E. (1999). Measuring reactions to sexual trauma among children: Comparing the Children’s Impact of Events Scale and the Trauma Symptom Checklist for Children, Child Maltreatment, 4, 255–263. http://dx.doi.org/10.1177/1077559599004003006 Elhai, J. D., Gray, M. J., Kashdan, T. B., & Franklin, C. L. (2005). Which instruments are most commonly used to assess traumatic event exposure and posttraumatic effects? A survey of traumatic stress professionals. Journal of Traumatic Stress, 18, 541–545. http://dx.doi.org/10.1002/jts.20062 Finkelhor, D., Turner, H. A., Shattuck, A., & Hamby, S. L. (2013). Violence, crime, and abuse exposure in a national sample of children and youth: An update. JAMA Pediatrics, 167, 614–621. http://dx.doi.org/10.1001/jamapediatrics.2013.42 Finkelhor, D., Ormrod, R. K., & Turner, H. A. (2007). Poly-victimization: A neglected component in child victimization. Child Abuse & Neglect, 31, 7–26. http://dx.doi.org/10.1016/j.chiabu.2006.06.008 Flannery, D. J., Singer, M. I., & Wester, K. (2001). Violence exposure, psychological trauma, and suicide risk in a community sample of dangerously violent adolescents. Journal of the American Academy of Child and Adolescent Psychiatry, 40, 435–442. http://dx.doi.org/10.1097/00004583-200104000-00012 Friedrich, W. N. (1997). Child Sexual Behavior Inventory: Professional manual. Odessa, FL: Psychological Assessment Resources. Fricker, A. E., & Smith, D. W. (2001). Trauma specific versus generic measurement of distress and the validity of self-reported symptoms in sexually abused children. Journal of Child Sexual Abuse, 10, 51–66. http://dx.doi.org/10.1300/J070v10n04_04 Hawkins, S. S., & Radcliffe, J. (2006). Current measures of PTSD for children and adolescents. Journal of Pediatric Psychology, 31, 420–430. http://dx.doi.org/10.1093/jpepsy/jsj039 Hawton, K., Kingsbury, S., Steinhardt, K., James, A., & Fagg, J. (1999). Repetition of deliberate self-harm by adolescents: The role of psychological factors. Journal of Adolescence, 22, 369–378. http://dx.doi.org/10.1006/jado.19990228 Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. Kaslow, N. J., & Thompson, M. P. (2008). Associations of child maltreatment and intimate partner violence with psychological adjustment among low SES, African American children. Child Abuse & Neglect, 32, 888–896. http://dx.doi.org/10.1016/j.chiabu.2007.09.012 Kerig, P. K. (2013). Trauma-informed assessment and intervention. Los Angeles, CA & Durham, NC: National Center for Child Trauma Stress. Kerr, M. A., & Schneider, B. H. (2008). Anger expression in children and adolescents: A review of the empirical literature. Clinical Psychology Review, 28, 559–577. http://dx.doi.org/10.1016/j.cpr.2007.08.001 Ko, S. J., Ford, J. D., Kassam-Adams, N., Berkowitz, S. J., Wilson, C., Wong, M., … Layne, C. M. (2008). Creating trauma-informed systems: Child welfare, education, first responders, health care, juvenile justice. Professional Psychology: Research and Practice, 39, 396–404. http://dx.doi.org/10.1037/0735-7028.39.4.396 Lanktree, C. B., Gilbert, A. M., Briere, J., Taylor, N., Chen, K., Maida, C. A., & Saltzman, W. R. (2008). Multi-informant assessment of maltreated children: Convergent and discriminant validity of the TSCC and TSCYC. Child Abuse & Neglect, 32, 621–625. http://dx.doi.org/10.1016/j.chiabu.2007.10.003 Laye-Gindhu, A., & Schonert-Reichl, K. A. (2005). Nonsuicidal self-harm among community adolescents: Understanding the “whats” and “whys” of self-harm. Journal of Youth and Adolescence, 34, 447–457. http://dx.doi.org/10.1007/s10964-005-7262-z Li, X., Fang, X., Stanton, B., Zhao, G., Lin, X., Zhao, J., … Chen, X. (2009). Psychometric evaluation of the Trauma Symptoms Checklist for Children (TSCC) among children affected by HIV/AIDS in China. AIDS Care, 21, 261–270. http://dx.doi.org/10.1080/09540120802195119 Marsh, H. W., & Hocevar, D. (1985). Application of confirmatory factor analysis to the study of self-concept: First and higher order factor models and their invariance across groups. Psychological Bulletin, 97, 562–582. McCart, M. R., Davies, W. H., Harris, R., Wincek, J., Calhoun, A. D., & Melzer-Lange, M. D. (2005). Assessment of trauma symptoms among adolescent assault victims. Journal of Adolescent Health, 36, 70.e7–70.e13. http://dx.doi.org/10.1016/j.jadohealth.2004.03.004 Muthén, L. K., & Muthén, B. O. (1998–2012). Mplus user's guide (7th ed.). Los Angeles, CA: Muthén & Muthén. Nilsson, D., Gustafsson, P. E., & Svedin, C. G. (2010). Self-reported potentially traumatic life events and symptoms of post-traumatic stress and dissociation. Nordic Journal of Psychiatry, 64, 19–26. http://dx.doi.org/10.3109/08039480903264846 Nilsson, D., Wadsby, M., & Svedin, C. G. (2008). The psychometric properties of the Trauma Symptom Checklist for Children (TSCC) in a sample of Swedish children. Child Abuse & Neglect, 32, 627–636. http://dx.doi.org/10.1016/j.chiabu.2007.09.009 Ostler, T., Bahar, O., & Jessee, A. (2010). Mentalization in children exposed to parental methamphetamine abuse: Relations to children's mental health and behavioral outcomes. Attachment & Human Development, 12, 193–207. http://dx.doi.org/10.1080/14616731003759666 Puzzanchera, C., & Robson, C. (2014). Delinquency cases in juvenile court, 2010 (NCJ Publication No. 243041). Washington DC; Office of Juvenile Justice and Delinquency Prevention. Retrieved from http://www.ojjdp.gov/pubs/243041.pdf Sadowski, C. M., & Friedrich, W. N. (2000). Psychometric properties of the Trauma Symptom Checklist for Children (TSCC) with psychiatrically hospitalized adolescents. Child Maltreatment, 5, 364–372. http://dx.doi.org/10.1177/1077559500005004008 Shain, B. N. (2007). Suicide and suicide attempts in adolescents. Pediatrics, 120, 669–676. http://dx.doi.org/10.1542/peds.2007-1908 .Singer, M. I., Anglin, T. M., Song, L. Y., & Lunghofer, L. (1995). Adolescents' exposure to violence and associated symptoms of psychological trauma. Journal of the American Medical Association, 273, 477–482. http://dx.doi.org/10.1001/jama.1995.03520300051036 .Steinberg, A. M., Brymer, M., Decker, K., & Pynoos, R. S. (2004). The UCLA PTSD Reaction Index. Current Psychiatry Reports, 6, 96–100. Strand, V. C., Sarmiento, T. L., & Pasquale, L. E. (2005). Assessment and screening tools for trauma in children and adolescents: A review. Trauma, Violence, and Abuse, 6(1), 55–78. http://dx.doi.org/10.1177/1524838004272559 Tourangeau, R., & Yan, T. (2007). Sensitive questions in surveys. Psychological Bulletin, 133, 859–883. http://dx.doi.org/10.1037/0033-2909.133.5.859 .Turner, H. A., Finkelhor, D., & Ormrod, R. (2010). Poly-victimization in a national sample of children and youth. American Journal of Preventive Medicine, 38, 323–330. http://dx.doi.org/10.1016/j.amepre.2009.11.012 Turner, H. A., Shattuck, A., Hamby, S., & Finkelhor, D. (2013). Community disorder, victimization exposure, and mental health in a national sample of youth. Journal of Health and Social Behavior, 54, 258–275. http://dx.doi.org/10.1177/0022146513479384 Wolfpaw, J. M., Ford, J. L., Newman E., Davis, J. L., & Briere, J. (2005). Trauma Symptom Checklist for Children. In T. Grisso, G. Vincent, & D. Seagrave (Eds.), Mental health screening and assessment in juvenile justice (pp. 152–165). New York, NY: Guilford Press.
