Factor structure and reliability of the Childhood Trauma Questionnaire in a Canadian undergraduate student sample
Introduction
The purpose of the present study was to examine the factor structure and reliability of the Childhood Trauma Questionnaire (CTQ; Bernstein et al., 1994; Bernstein & Fink, 1993) in a sample of undergraduate students. The CTQ is a promising measure of diverse childhood experiences, but little is known about its psychometric properties in nonreferred student groups. A second and related purpose was to assess the prevalence of childhood abuse and neglect in this student sample which was drawn from a distinct region of North America. Information on local and regional prevalence can contribute to research on the sociocultural factors involved in different types of childhood maltreatment.
Despite increased attention to child abuse and neglect in recent years, large numbers of children continue to be victimized by some form of maltreatment (Pilkington & Kremer, 1995). Centrally important to research on the prevalence, causes, and consequences of childhood maltreatment is the need for reliable and valid assessments of diverse childhood experiences in both clinical and nonclinical populations. The CTQ is an easily administered, retrospective, self-report questionnaire that assesses the extent of maltreatment as an interaction between the severity and frequency with which a variety of experiences occurred. In many studies a limited range of childhood experiences are assessed with most research focussing on sexual and physical abuse (e.g., MacMillan et al., 1997), even though maltreatment types are not mutually exclusive and multiple types frequently co-occur in the same family (Briere & Runtz, 1990; Rind, Tromovitch, & Bauserman, 1998). The CTQ allows for a more ecologically valid approach to studying the separate and combined effects of different types of abuse and neglect.
Instructions and test items on the CTQ largely are phrased in terms of concrete, objective events and behaviors. Terms such as “trauma,” “abuse,” and “neglect” generally are avoided because these have a subjective, evaluative, and stigmatizing quality that can arouse defensiveness (Bernstein et al., 1994). Such a concrete and objective format also maximizes accuracy of recall. Research suggests that recall for childhood experience is most accurate for concrete events rather than the subjective experience of events (Brewin, Andrews, & Gotlib, 1993). As well, the researcher-defined questionnaire format used in the CTQ may be more sensitive than self-defined formats, which ask respondents whether they were abused (using that term), to less severe experiences. Silvern, Waelde, Baughan, Karyl, and Kaersvang (2000), for example, reported that, among college students, significantly less abuse was reported on self- versus researcher-defined question formats. Thus, the CTQ may be more appropriately used in research with nonclinical samples.
The CTQ demonstrated good reliability and validity in an initial evaluation study of 286 drug- or alcohol-dependent adult patients (Bernstein et al., 1994). Principal components analyses using the complete response matrix of 70 items yielded four factor scales of emotional/physical abuse, sexual abuse, physical neglect, and emotional neglect. Findings included internal consistency and test-retest reliability on all dimensions of abuse and neglect before and after approximately 3 months of treatment for the substance-dependence problem. As well, the CTQ demonstrated sensitivity and specificity when compared to assessments based on clinical interviews.
Paivio (2001) and Paivio and Patterson (1999) also examined the CTQ in a study of 33 adults undergoing outpatient psychotherapy for child abuse issues. These studies used the complete 70-item questionnaire (Bernstein & Fink, 1993) and the four factor scales that emerged in the original study of substance-dependent patients described above (Bernstein et al., 1994). Paivio (2001) found that, despite significant reductions in psychopathology following therapy, reports of abuse and neglect on the CTQ remained stable from pre- to posttherapy. As well, CTQ factor scales demonstrated good internal consistency and convergent validity with trauma-specific measures of distress. In another study, Paivio and Patterson (1999) found that CTQ scores predicted therapeutic alliance development in the same sample of outpatient survivors. Severity of childhood physical/emotional abuse, physical neglect, and emotional neglect were associated with early alliance difficulties but, as predicted, these associations disappeared over the course of therapy. These findings support the predictive validity of the CTQ factor scales and sensitivity of the measure to changes over time in an adult clinical sample.
Bernstein, Ahluvalia, Pogge, and Handelsman (1997) also evaluated the 70-item CTQ in a study of 298 adolescent in-patients. Results using this adolescent sample largely replicated those in the initial study of adult substance-dependent patients (Bernstein et al., 1994), except that emotional and physical abuse items loaded on separate factors. The latter permitted researchers to establish separate criteria for defining “cases” of physical, emotional, and sexual abuse. Lipschitz, Bernstein, Winegar, and Southwick (1999) also assessed the consistency of these hospitalized adolescents’ self-reports of sexual and physical abuse via the CTQ and the Traumatic Experiences Questionnaire for Adolescents (TEQ-A; Winegar & Lipschitz, 1997) which uses a multiple-choice format. They found higher rates of consistency on the CTQ compared to the TEQ-A. These researchers suggested that, because CTQ items assess a range of both severity and frequency of childhood experiences, the CTQ questionnaire may be more sensitive to less severe experiences of abuse and neglect. Again, this makes it particularly suitable for use with nonclinical groups.
Overall, the above findings support the utility of the 70-item CTQ (Bernstein et al., 1994; Bernstein & Fink, 1993) in assessing maltreatment histories, at least in clinical samples. However, understanding the factors associated with child abuse and neglect also requires large-scale correlational studies of nonreferred and community groups. Much of this research is conducted using college samples. The CTQ seems ideally suited for use in such nonreferred contexts because it is easily administered, yields continuous data, and likely is sensitive to less severe types of childhood maltreatment. However, a review of the literature yielded only one study on the psychometric properties of the complete 70-item CTQ in a nonreferred context. Rosen and Lee (1996) assessed the reliability and validity of the CTQ in a sample of 1365 male and female US Army soldiers. Their findings showed psychometric properties similar to other studies that support the reliability and validity of the measure. Principal components analyses resulted in four factor scales of emotional/physical abuse, sexual abuse, physical neglect, and emotional neglect. Thus, exploratory analyses replicated the factor structure reported in the original validation study of adult substance-dependent patients (Bernstein et al., 1994), rather than the five-factor solution found for adolescents (Bernstein et al., 1997). Different factor structures found for adult compared to adolescent samples suggest that experiences of emotional and physical abuse, in particular, may have somewhat different meanings for these different age groups.
Recently, a short version (28 items) of the CTQ has been developed based on data from seven samples, including 1187 nonreferred female HMO members and 92 male and female college undergraduates (Bernstein & Fink, 1998). This consists of five subscales of emotional, physical, and sexual abuse, and emotional, and physical neglect. Paivio (Paivio & McCullough 2004; Turner & Paivio, 2002) used the CTQ short-form to examine the links between childhood abuse and neglect and dimensions of psychopathology. CTQ scores predicted alexithymia, interpersonal functioning, and self-injurious behavior among university students, thus supporting the predictive validity of the CTQ short-form in nonclinical samples. Two recent studies of the CTQ short-form with nonclinical samples, however, found somewhat conflicting results concerning the stability of the five-factor model. Scher, Stein, Asmundson, McCreary, and Forde (2001) found that the five-factor model best described the data for a community sample, but a study of university students (Wright et al., 2001) found that the five-factor model was significantly less stable for female compared to male students. Thus, different factor structures appear to underlie the CTQ data in studies of different groups. Just as somewhat different structures were found in studies of adults (Bernstein et al., 1994; Rosen & Lee, 1996) compared to adolescents (Bernstein et al., 1997), it is possible that various experiences of maltreatment have different meanings for male and female students or for college students compared to patients. Different meanings attributed to maltreatment experiences has implications for use of the CTQ factor scales for assessing histories of different types of abuse and neglect in different contexts.
Differing definitions of abuse and neglect partly account for the variability in community prevalence estimates (Pilkington & Kremer, 1995). Sexual abuse, for example, can be restricted to contact experiences and adult perpetrators, or can include exposure to pornography or voyeurism and nonadult perpetrators. Emotional abuse and neglect can be particularly difficult to define resulting in few studies of these childhood experiences. However, even when identical definitions are employed, prevalence rates for different types of abuse and neglect continue to vary across countries, cultures, and regions. For example, there is evidence for higher prevalence of sexual abuse on the west coast of the United States compared to other regions (Dearwater et al., 1998). A recent study also found higher prevalence of physical abuse for women in rural Ontario compared to urban regions of the province (MacMillan et al., 1997). On the other hand, Yamamoto et al. (1999) tested assumptions about the infrequency of childhood physical and emotional abuse in Japan. They found much higher rates than previously estimated; rates that were comparable to those reported in the United States. Continued research on local and regional prevalence of different types of childhood trauma using standardized measures permits further study of the demographic factors that could account for regional differences or moderate the occurrence of childhood maltreatment. As well, baseline data using uniform standards can help to gage the effectiveness of local interventions and allow comparisons across communities (Wynkoop, Capps, & Priest, 1995). The present study, therefore, assessed the prevalence of different types of abuse and neglect among university students in a relatively underpopulated mid-western region of Canada.
Section snippets
Sample
The sample was comprised of 470 students in a single introductory Psychology class that was part of the research participant pool for the Psychology Department at the University of Saskatchewan. The University of Saskatchewan is located in a medium-sized (population: 200,000) city in the Canadian mid-west. It is one of two universities in a rural province of less than one million people and its mandate is to meet the post-secondary educational needs of students in the province. Demographic
Factor analyses
Confirmatory analysis. For the four-factor model, results showed that the four-factor oblique model did not adequately fit the observed data: χ2 (1312, N = 470) = 5455, p < .001, χ2/df = 4.16, NNFI = .765, CFI = .777, AGFI = .629, RMSEA = .082. Similarly, the five-factor model oblique model did not adequately fit the observed data: χ2 (1308, N = 470) = 5243, p < .001, χ2/df = 4.01, NNFI = .777, CFI = .788, AGFI = .650, RMSEA = .080. As such, there was no evidence to support either the
Discussion
When examined empirically through confirmatory factor analyses, the structure of the 70-item CTQ in the present sample of undergraduates was not comparable to that reported for clinical samples (Bernstein et al., 1994, Bernstein et al., 1997). Nonetheless, the five-factor solution extracted in the present study most closely replicated the five-factor solution reported for a sample of adolescent inpatients (Bernstein et al., 1997). Those factors are described as emotional, physical, and sexual
References (35)
- et al.
Validity of the Childhood Trauma Questionnaire in an adolescent psychiatric population
Journal of the American Academy of Child and Adolescent Psychiatry
(1997) - et al.
Differential adult symptomatology associated with three types of childhood abuse histories
Child Abuse & Neglect
(1990) - et al.
Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: The Adverse Childhood Experiences (ACE) Study
American Journal of Preventive Medicine
(1998) The international epidemiology of child sexual abuse
Child Abuse & Neglect.
(1994)Stability of self-reports of child abuse and neglect before and after therapy for child abuse issues
Child Abuse & Neglect
(2001)- et al.
Alexithymia as a mediator between childhood trauma and self-injurious behaviors
Child Abuse & Neglect
(2004) Comparative fit indexes in structural models
Psychological Bulletin
(1990)- Bentler, P. M. (1995). EQS: Structural equations program manual. Encino, CA: Multivariate Statistical...
- Bernstein, D., & Fink, L. (1993). Manual for the Childhood Trauma Questionnaire. Unpublished manuscript. Bronx, NY: VA...
- et al.
Manual for the Childhood Trauma Questionnaire
(1998)