Psychiatry Research 245 (2016) 297302

Contents lists available at ScienceDirect

Psychiatry Research
journal homepage: www.elsevier.com/locate/psychres

Factorial validity of the Childhood Trauma Questionnaire in Italian

psychiatric patients
Marco Innamorati a, Denise Erbuto b, Paola Venturini b, Francesca Fagioli c, Federica Ricci b,
David Lester d, Mario Amore e, Paolo Girardi b, Maurizio Pompili b,n
a
Department of Human Sciences, European University of Rome, Rome, Italy
b
Department of Neurosciences, Mental Health and Sensory Organs, Suicide Prevention Center, SantAndrea Hospital, Sapienza University of Rome,
Rome, Italy
c
Mental Health Department, ASL RME, Rome, Italy
d
Stockton University, Galloway, NJ, United States
e
Section of Psychiatry, Department of Neuroscience, Ophthalmology, Genetics and Infant-Maternal Science, University of Genoa, Genova, Italy

art ic l e i nf o a b s t r a c t

Article history: Early adverse experiences are associated with neurobiological changes and these may underlie the in-
Received 23 May 2016 creased risk of psychopathology. The Childhood Trauma Questionnaire (CTQ-SF) is the most commonly
Received in revised form used instrument for assessing childhood maltreatment. Thus, the aim of our study was to investigate the
4 August 2016
factorial validity of an Italian version of the CTQ-SF in a sample of psychiatric inpatients by means of
Accepted 14 August 2016
conrmatory and exploratory factor analyses. The sample was composed of 471 psychiatric in-patients
Available online 16 August 2016
and out-patients (206 males and 265 females) aged 1680 years (mean age34.4 years [SD 16.3])
consecutively admitted to two psychiatric departments. All patients were administered the Italian ver-
sion of the CTQ-SF. We tested ve different factor models which lacked good t, while the exploratory
factor analysis supported the adequacy of a solution with three factors (Emotional Neglect/Abuse, Sexual
Abuse, Physical Neglect/Abuse). The three factors had satisfactory internal consistency (ordinal Cronbach
alphas 40.90). Our study supports results from previous research indicating the lack of structural in-
variance of the CTQ-SF in cross-cultural adaptations of the test, and the fact that, when measuring dif-
ferent types of childhood maltreatment, the difference between abuse and neglect may be not valid.
& 2016 Elsevier Ireland Ltd. All rights reserved.

1. Introduction abuse, was associated with diurnal cortisol regulation (Kuhlman

Research has suggested that early adverse experiences may be
associated with neurobiological changes in children and adults,
and these may underlie the increased risk of psychopathology
(Heim and Nemeroff, 2001; Baumeister et al., 2016). Furthermore,
several studies suggest that different types of childhood adverse
experience may be differentially associated with different patterns
of physiological reactivity to stress and psychopathological tra-
jectories (Heins et al., 2011; van Veen et al., 2013; Huh et al., 2014;
Kuhlman et al., 2015). For example, Kuhlman et al. (2015) tried to
characterize the unique associations between exposure to differ-
ent types of childhood maltreatment (physical and emotional
abuse, and non-intentional trauma) and HPA-axis functioning, in
121 121 youths aged 916 years old, the authors found that non-
intentional trauma exposure, but not physical and emotional
n
Corresponding author at: Department of Neurosciences, Mental Health and
Sensory Organs, Suicide Prevention Center, SantAndrea Hospital, Via di Grottarossa
1035/1039, 00189 Rome, Italy.
E-mail address: maurizio.pompili@uniroma1.it (M. Pompili).
et al., 2015). The situation was different when the children had to
face an acute stress (cold pressure task). In this situation, exposure
to physical abuse was positively associated with a steeper accel-
eration of cortisol toward a peak, while exposure to emotional
abuse was associated with a slower deceleration of cortisol during
the recovery phase, suggesting that youths exposed to emotional
abuse exhibited elevated cortisol levels longer than their peers.
Finally, the three types of trauma were not independently asso-
ciated with the cortisol awakening response. Heins et al. (2011), in
a sample of patients diagnosed with a psychotic disorder, reported
that childhood abuse, more than childhood neglect, was associated
with positive psychotic symptoms within the rst 10 years of their
illness, while childhood neglect, more than abuse, was associated
with general psychopathology. van Veen et al. (2013) investigated
whether different types of adversities (childhood life events,
childhood trauma, and recent life events) may have different ef-
fects on general distress, anhedonic depression, and anxiety
arousal. In 2615 individuals from the Netherlands Study of De-
pression and Anxiety (NESDA), the authors found that, in-
dependent of diagnostic groups, emotional neglect was

http://dx.doi.org/10.1016/j.psychres.2016.08.044
0165-1781/& 2016 Elsevier Ireland Ltd. All rights reserved.
298 M. Innamorati et al. / Psychiatry Research 245 (2016) 297302

signicantly associated with increased general distress and anhe- trauma) and on one of the two factors dened by the items po-
donic depression, while sexual abuse was signicantly associated larity (positive and negative polarity). If conrmatory factor ana-
with increased general distress and anxious arousal. lysis indicated the t of one of the competing models, we reported
To provide a reliable measure of traumatic experiences in the statistics for this model. Otherwise we planned to investigate the
childhood, Bernstein et al. (1994) developed the Childhood Trau- factorial structure of the CTQ-SF by means of exploratory factor
ma Questionnaire (CTQ), a 70-item self-report questionnaire analysis.
which assesses, retrospectively, experiences of abuse and neglect
in the childhood. The CTQ is the most commonly used ques-
tionnaire for assessing different types of childhood maltreatment 2. Methods
(DiLillo et al., 2006). The rst factor analytic studies on the CTQ
suggested a four-factor structure (Physical and Emotional Abuse, 2.1. Participants and procedure
Emotional Neglect, Sexual Abuse, and Physical Neglect) (Bernstein
et al., 1994) or a ve-factor structure (with the split of Physical and The nal sample was composed of 471 inpatients (206 males
Emotional Abuse into two separate factors) (Bernstein et al., 1997). and 265 females) consecutively admitted between 2013 and 2015
Subsequently, Bernstein et al. (2003) developed a 28-item version either to the Department of Psychiatry of the SantAndrea Uni-
of the CTQ (CTQ-SF) in order to provide more rapid screening of versity Hospital in Rome, Italy, or to the Via Plinio Adolescent
abuse and neglect histories. The CTQ-SF has 25 items measuring Psychiatric Outpatient Clinic, part of the Rome E Territorial Psy-
ve dimensions of abuse (Physical, Emotional, and Sexual Abuse) chiatric Services (see Table 1). Inclusion criteria were age of 16 and
and neglect (Emotional Neglect, and Physical Neglect), and 3 items higher and admission in the time period indicated. Exclusion cri-
included in a Minimization/Denial validity scale. In their study, the teria were the presence of major disorders of the central nervous
authors used exploratory and conrmatory factor analyses (with system (e.g., epilepsy, dementia, or Parkinson disease), and any
maximum likelihood estimators) to derive ve independent di- condition affecting the ability to take the assessment, including
mensions of abuse and neglect, but they were forced to add some delirium or denial of informed consent. Twenty-nine patients
non-hypothesized covariances among the error residuals of some (5.5%) denied their informed consent and were excluded from the
items included in the same dimension, reecting associations study. Furthermore, thirty patients (6.0%) provided missing in-
between items that are not accounted for by the latent factors formation for one to three items of the CTQ-SF and were excluded
(Bernstein et al., 2003; Thombs et al., 2007; Thombs et al., 2009). from the nal sample. Patients excluded from the study did not
To date, the ve-factor structure of the CTQ-SF has been only differ in sex and age from those included in the nal sample. The
partially conrmed in cross-cultural studies (Thombs et al., 2009; mean age of the participants in the study was 34.4 years
Hernandez et al., 2013; Karos et al., 2014). Some studies have re- (SD 16.3; Range: 1680 years). Most of the patients had been
ported that not all the items loaded signicantly on the hy- diagnosed with a mood disorder (48.5%; 19.5% bipolar disorder-I,
pothesized dimension (Thombs et al., 2009; Karos et al., 2014), 10.2% bipolar disorder-II, and 18.8% major depressive disorder) or
while others reported different factor solution for cross-cultural psychosis (20.6%). Other diagnoses were anxiety disorders (8.6%),
versions of CTQ-SF (Gerdner and Allgulander, 2009; Grassi-Oli- personality disorders (13.2%), substance abuse (1.2%), and other
veira et al., 2014; Dudeck et al., 2015). Recently, Grassi-Oliveira DSM-IV-TR specied disorders (including eating disorders, and
et al. (2014) have suggested that the lack of structural invariance of dissociative disorders: 5.6%). Only, 2.3% of the sample had received
the CTQ-SF in cross-cultural adaptations of the questionnaire no diagnosis.
could be related to problems in the original CTQ construction and Subjectes were approached by senior fully qualied psychia-
not a weakness of the different language versions. trists and psychologists in training who informed them about the
Considering the results presented above, and the fact that Ita- aim of the study and explained how to ll-in the questionnaire.
lian versions of the CTQ-SF have been used in previous clinical
studies (Pompili et al., 2014; Imperatori et al., 2015; Imperatori Table 1
et al., 2016) without a proper investigation of their psychometric Characteristics of the sample (n 471).
properties, the aim of our study was to investigate the factorial
validity of an Italian version of the CTQ-SF in a sample of psy- Adult in- Adolescent Test Signicance
patients outpatients
chiatric inpatients as a rst step for a cross-cultural validation of (n342) (n 129)
the questionnaire. Local information on reliability and validity of a
cross-cultural adapted test is necessary to avoid the limitations Sex 0.01a
associated with a use of the instrument in an applied context Males 47.4% 34.1%
Females 52.6% 65.9%
without the availability of local reliability and validity estimates
Age M(SD) 40.99 (14.62) 17.38 (1.31) t347.72 29.25 0.001
(Gudmundsson, 2009). The lack of estimates for local reliability Diagnosis 28 159.35 0.001
and validity of a measure reduces the possibility of reliably inter- None 0.0% 7.8%
preting the ndings from studies which use this instrument and BD-I 23.5% 10.1%
the impossibility of comparing ndings from international studies. BD-II 14.2% 0.8%
MDD 17.5% 21.7%
In line with two previous studies (Grassi-Oliveira et al., 2014; Psychosis 27.8% 3.9%
Dudeck et al., 2015), we compared different models able to explain Anxiety 1.7% 24.8%
the structure of the CTQ-SF by means of conrmatory factor ana- disorders
lysis: (1) the original ve-factor model reported originally by Substance 1.3% 0.8%
abuse
Bernstein et al. (2003); (2) a ve-factor model proposed by
Other other- 6.6% 3.1%
Gerdner and Allgulander (2009), in which the items CTQ2 and wise speci-
CTQ26 of the Physical neglect subscale were allocated to the ed
Emotional neglect subscale; (3) a four-factor solution combining diagnoses
emotional and physical abuse (Lundgren et al., 2002); (4) a uni- Personality 7.3% 27.1%
disorders
dimensional model with all the items loading on a single factor
(childhood trauma); and (5) a one constructtwo method bi-factor a
One-way Fisher-exact test. BD-I = Bipolar Disorder type I; BD-II = Bipolar
model, in which each item loads on the same construct (childhood Disorder type II; MDD = Major Depressive Disorder.
 M. Innamorati et al. / Psychiatry Research 245 (2016) 297302
They participated in the study voluntarily and received no pay-
ment. Each patient provided written, informed consent prior to
data collection. For patients under the age of 18 (Italian legal age),
consent was provided by both parents. The study protocol received
ethics approval from the local research ethics review board.
2.2. Measures
At entry into the study, all patients were administered the
Italian version of the CTQ-SF. The original version of the CTQ-SF is
a 28-item self-report measure used to assess the severity of ve
different types of childhood maltreatment (Emotional Abuse EA,
Physical Abuse-PA, Sexual Abuse SA, Emotional Neglect EN,
Physical Neglect PN). Each factor is composed of 5 items rated on
a 5-point Likert type scale (from 1 never true to 5 very often
true) (Bernstein et al., 2003), plus 3 items included as a validity
scale. In the present study, we used an Italian adaptation of this
scale. One author of the present study (DE) translated the CTQ-SF
from the original English version and a second researcher (MI)
back-translated to English. Finally, an English-speaking person,
uent in the Italian language, supervised the procedure and
amended any minor discrepancies in the nal version.
2.3. Statistical analysis
All the analyses were performed with the Statistical Package for
the Social Sciences (SPSS) 19.0 for Windows, Mplus 6.0 (Muthn
and Muthn, 19982010), and the statistical software Factor v. 10.3
(Lorezo-Seva and Ferrando, 2013). In order to carry out con-
rmatory and exploratory factor analyses independently, the
sample was randomly divided in two subsamples. The con-
rmatory factor analysis was performed on a subsample of 271
patients (109 males and 162 females; mean age 34.06 years
[SD 15.88]), and the exploratory factor analysis on a subsample of
200 patients (97 males and 103 females; mean age 34.91 years
[SD 16.95]). The number of patients to include in each one of the
subsample was decided in order to guarantee the maximum sta-
tistical power possible for the analyses, approaching the standards
suggested in the literature (Floyd and Widaman, 1995; Fabrigar
et al., 1999). For example, the number of two hundred participants
was suggested as the minimum for exploratory factor analysis in
those situation when latent factors are not overdetermined and
the communalities are moderately high (Fabrigar et al., 1999), and
two hundred seventy-one participants for the conrmatory factor
analysis guaranteed the suggested standard of 10 participants per
variable measured (Floyd and Widaman, 1995). The two sub-
samples did not differ in sex (two-sided Fisher's exact test
p 0.08), mean age (t469 0.55; p 0.58), diagnosis (28 6.00;
p 0.65), and the proportions of patients enrolled from each one of
the two psychiatric centers (two-sided Fisher's exact test p 0.53).
A conrmatory-exploratory approach is not the only that may
be used to rene factor models. Sometimes, researchers prefer to
use modication indices in a conrmatory factor analysis frame-
work for identifying localized areas of strain and poor t. However,
this process is data driven and error prone, and the nal model
Table 2
Fit indices for the competing factor models.
Model
Original ve-factor model (Bernstein et al., 2003)
Five-factor model proposed by Gerdner and Allgulander (2009)
Four-factor model (Lundgren et al., 2002)
One constructtwo method bi-factor model
One-factor model
*
p o 0.001.
299
may be not generalizable to other samples or to the population
(MacCallum et al., 1992). Nevertheless, even an exploratory factor
procedure is data driven and error prone, despite the fact that
optimal results may result from the correct choice of the extrac-
tion and rotation methods to use and the use of tests for selecting
the number of factors to rotate (Costello and Osborne, 2005). We
also decided not to use a multi-group conrmatory factor analysis
despite the fact that it could have permitted us to investigate
structural invariance between adolescents and adult patients. In
fact, the subjects-to-variables ratio was not satisfactory for this
kind of analysis, especially when considering the small number of
adolescents included in our sample.
The conrmatory factor analysis was performed using Mplus
6.0 (Muthn and Muthn, 19982010). We used a Mean- and
Variance-adjusted Weighted Least Square (WLSMV) estimator
with a polychoric correlation matrix, after calculating Mardia's
multivariate asymmetry skewness and kurtosis statistics (Mardia,
1970). Model t was assessed using the following indices: (1) the
Root Mean Square Error of Approximation (RMSEA), with values
between 0.05 and 0.08 indicative of good adequacy of the model,
and values below 0.05 indicating evidence of absolute t (Browne
and Cudek, 1993; Hu and Bentler, 1999); (2) the Comparative Fit
Index (CFI), with values greater than 0.95/0.96 indicating good t
of the model; (3) the Weighted Root Mean Square Residual
(WRMR), with values of less than 1.0 indicating good t (Yu,
2002); and (4) the chi-square (2) test and the normed 2 (2/
degrees of freedom). P-values for the 2 test greater than 0.05 and
a normed 2 less than 2 (Tabachnick and Fidell, 2013) indicate that
the model is an adequate t to the data, although the 2 test over-
rejects true models for large samples.
The exploratory factor analysis was performed using Factor v.
10.3 (Lorezo-Seva and Ferrando, 2013). Polychoric correlations
matrix and unweighted least squares (ULS) estimator were used.
The number of factors to retain was determined by results of the
Hull method based on the RMSEA (Lorenzo-Seva, 2013). The
Goodness of Fit Index (GFI) and root mean square of residuals
(RMSR) are reported. An acceptable model should have mean va-
lue of the RMSR of 0.071 or lower (Kelley, 1935; Bollen and Long,
1993), and a GFI greater than 0.95/0.96 (Hu and Bentler, 1999). As
measures of reliability, we reported ordinal Cronbach's alpha ()
(Zumbo et al., 2007). The association with sociodemographic
variables was evaluated by means of a series of t-tests and Pear-
son's r correlation coefcients.
3. Results
3.1. Conrmatory factor analysis
Multivariate Kurtosis was signicant in both subsamples (sub-
sample 1 [n 271]: coefcient 1127.03; statistic 101.26;
po 0.001; subsample 2 [n 200]: coefcient 1009.45;
statistic 64.36; p o0.001), while skewness was not (subsample 1:
coefcient 315.34; statistic 14,242.92; df 2925; p 1.00; sub-
sample 2: coefcient 312.21; statistic 10,406.84; df 2925;
2 (df) Normed 2 RMSEA (90%CI) CFI WRMR
*
656.64 (265) 2.48 0.074 (0.067/0.081) 0.92 1.40
633.27* (265) 2.39 0.072 (0.064/0.079) 0.93 1.35
771.50* (269) 2.87 0.083 (0.076/0.090) 0.90 1.58
542.95* (250) 2.17 0.066 (0.058/0.073) 0.94 1.11
1343.62* (275) 4.89 0.120 (0.113/0.126) 0.78 2.31
300 M. Innamorati et al. / Psychiatry Research 245 (2016) 297302

p 1.00). The second factor may be considered a measure of Sexual Abuse (5

In the rst subsample (n 271), no models tted the data well
(see Table 2), and, for the ve-factor models, MPLUS also warned
that the residual covariance matrix (theta) was not positive de-
nite. Among all the models, the one constructtwo method bi-
factor model had the best tting indices. In this model, all the
items negatively worded (CTQ2, CTQ5, CTQ7, CTQ13, CTQ19,
CTQ26, and CTQ28) loaded signicantly on the corresponding
group factor (standardized loadings ranged from 0.36 for CTQ26 to
0.69 for CTQ19), while six items positively worded (CTQ1, CTQ3,
CTQ4, CTQ8, CTQ14, and CTQ18) did not load signicantly on their
group factor (loadings ranging from  0.02 for CTQ8 to 0.20 for
CTQ1), suggesting the absence of a method effect for these items.
All the items except for two (CTQ5, CTQ21) also loaded sig-
nicantly on the general factor (loadings ranging from 0.22 for
CTQ26 to 0.85 for CTQ11). Finally, three thresholds (items CTQ2,
CTQ19, and CTQ14) were nonsignicant.
Modication indices indicated several areas of local strain ei-
ther in the BY statements (i.e., statements dening which items
are supposed to measure latent factors; n 7), suggesting the need
for a renement of the model to permit several items to load on a
different group factor, and in the WITH statements (i.e., statements
dening correlational relationships among latent factors; n 2),
suggesting the necessity for permitting correlations between the
two group factors, as well as between the negative and the general
factors. All these results suggest that some of the characteristics of
the bi-factor model (independence between the latent factors, all
items loading on a group factor and on the general factor, pre-
ferably with higher loadings on the general factor) are not met in
the data and the necessity for dening the structure of the CTQ-SF
in an exploratory way.
3.2. Exploratory factor analysis
In the verication sample (n 200), the Kaiser-Meyer-Olkin
(KMO) measure of sampling adequacy ( 0.88) indicated the
adequacy of the correlation matrix for factor analysis. Exploratory
factor analysis resulted in four factors with eigenvalues 4 1 (ei-
genvalues between 11.04 and 1.27). However, the Hull method
suggested retaining only the rst three factors (44.1%, 13.8%, and
5.8% of real-data variance, respectively for F1, F2, and F3). The GFI
(0.99) and the RMSR (0.054) suggested a good t for this model.
Factor loadings are reported in Table 3. Considering the pattern of
loadings on each factor, the rst factor may be considered a
measure of Emotional Neglect/Abuse (4 items from the original EA
dimension, 5 items from the original EN dimension, and 1 item
from the original PN dimension loaded on this factor) (see Table 3).
items from the original SA dimension, and 1 item from the original
PA dimension loaded on this factor), and the third factor a measure
of Physical Neglect/Abuse (3 items from the original PN dimension,
and 4 items from the original PA dimension loaded on this factor).
The three factors correlated signicantly with one another (Pear-
son's r: F1F2 0.22; F1F3 0.46; F2F3 0.35; p o0.001). For the
whole sample (n 471), ordinal Cronbach's alpha coefcients for
the three dimensions of the CTQ were satisfactory (0.95, 0.99, and
0.93, respectively, for Emotional Neglect/Abuse, Sexual Abuse, and
Physical Neglect/Abuse). Males and females did not differ sig-
nicantly on Emotional Neglect/Abuse (t469  1.96, p 0.051;
mean score on the whole sample 21.22 [SD 8.61]) and Physical
Neglect/Abuse (t469  0.21, p 0.84; mean score on the whole
sample 9.33 ([SD 3.90]). Sexual Abuse mean scores
(t464.15 2.43, p=0.015; mean scores of 7.51 [SD 3.38] and 8.44
[SD 4.84], respectively, for males and females) did differ sig-
nicant by sex. Scores were not signicantly correlated with age
(r's ranging between  0.003 and 0.06; p 40.05).
4. Discussion
In our sample of psychiatric patients, the ve-factor model
originally proposed for the CTQ-SF by Bernstein et al. (2003) did
not t the data. This lack of t was also true for all the competing
models we tested (a ve-factor model in which two items of the
physical neglect subscale were allocated to load on the Emotional
neglect subscale, a four factor model combining emotional and
physical abuse, one constructtwo method bi-factor model, and a
one-factor model).
The factor structure of the questionnaire for the present sample
was derived by means of an exploratory factor analysis which
indicated a three-factor solution. The ve dimensions proposed by
the authors of the CTQ were collapsed into fewer dimensions.
Almost all the items originally measuring Emotional Neglect and
Abuse were collapsed into one common dimension, with the ex-
ception of one item (Emotionally abused) originally measuring
emotional abuse which loaded signicantly on two factors (Emo-
tional Neglect/Abuse and Sexual Abuse) and was excluded from all
the dimensions. Seven items of the original dimensions measuring
physical abuse and neglect loaded signicantly on the same factor,
while two items loaded on a different factor and the last item
(Take me to the doctor) had low loadings on all the three factors.
Last, only Sexual Abuse remained a separate dimension of child-
hood trauma; all the items originally loading on the corresponding
factor loaded on this factor plus one more item (Physically

Table 3
Factor loadings after normalized Promax rotation for the CTQ items (n 200).

Items Emotional Neglect/Abuse Sexual Abuse Physical Neglect/Abuse Items Emotional Neglect/Abuse Sexual Abuse Physical Neglect/Abuse

CTQ1PN  0.16 0.28 0.42 CTQ15PA 0.03 0.71 0.10

CTQ2*PN 0.67 0.09 0.03 CTQ17PA 0.17 0.21 0.60
CTQ3EA 0.45 0.02 0.26 CTQ18EA 0.60 0.15 0.07
CTQ4PN 0.18 0.09 0.73 CTQ19*EN 0.78  0.07  0.29
CTQ5*EN 0.56  0.08 0.11 CTQ20SA 0.01 0.97  0.12
CTQ6PN  0.18 0.26 0.62 CTQ21SA  0.07 0.78 0.23
CTQ7*EN 0.74  0.16 0.20 CTQ23SA  0.15 0.73 0.24
CTQ8EA 0.49 0.21 0.10 CTQ24SA 0.02 0.92  0.05
CTQ9PA 0.16  0.04 0.78 CTQ25EA 0.42 0.49  0.07
CTQ11PA 0.28 0.01 0.67 CTQ26*PN 0.39 0.24 0.02
CTQ12PA 0.12  0.15 0.83 CTQ27SA 0.002 1.00  0.11
CTQ13*EN 0.81  0.18  0.02 CTQ28*EN 0.90 0.04  0.20
CTQ14EA 0.61 0.06 0.18

*Items reversed; In bold items included in each factors. Original dimensions: EA Emotional Abuse; PA Physical Abuse; SASexual Abuse; ENEmotional Neglect;
PN Physical Neglect.
 M. Innamorati et al. / Psychiatry Research 245 (2016) 297302
abused) originally loading on the Physical Abuse dimension.
The lack of t of the original ve-factor model is not new in the
research. For example, Villano et al. (2004) examined the factor
structure of the CTQ-SF in a sample of 171 female street-based sex
workers and reported a poor-t for the original ve-factor model,
whereas a four-factor solution best represented the data. The au-
thors also reported several cross-loadings of items on more than
one factor and hypothesized that abuse and neglect were not two
conceptually distinct constructs in their sample. This observation
is also true for our sample, where two out of the three dimensions
in the factor analysis had loadings from items assessing either
abuse or neglect. However, the results from Villano et al.'s (2004)
study may also be related to the fact that, in their study, the CTQ-
SF was administered face-to-face.
The lack of t of the original ve-factor model has been also
reported by several other studies (Thombs et al., 2009; Hernandez
et al., 2013; Karos et al., 2014). Grassi-Oliveira et al. (2014) sug-
gested that the lack of structural invariance of the CTQ-SF in cross-
cultural adaptations of the questionnaire could be related to pro-
blems in the original CTQ construction.
Supporting the validity of the three-factor solution in the pre-
sent study, all three dimensions had satisfactory internal con-
sistency (ordinal Cronbach alpha 4 0.90), in line with what some
authors consider desirable (Nunnally and Bernstein, 1994).
Nevertheless, this is the rst study on the CTQ-SF which used a
Cronbach's alpha coefcient corrected for ordinal measures, while
previous studies using the standard Cronbach's alpha coefcient
indicated problems with the homogeneity in content of the items
included in the physical neglect dimension (Karos et al., 2014).
In our sample, males and females did not differ in their scores
on Emotional Neglect/Abuse and Physical Neglect/Abuse, but they
did differ in their Sexual Abuse mean scores. Differences on the
CTQ-SF between males and females have been reported in the
past. For example, differences between females and males are
reported in the manual of the CTQ-SF for all ve dimensions of
maltreatment in a sample of adolescent psychiatric inpatients
(Bernstein and Fink, 1998). Pompili et al. (2014), in a sample of 163
Italian psychiatric inpatients, reported gender differences, but only
for emotional abuse scores. Thombs et al. (2007), assessing mea-
surement equivalence of the CTQ-SF across sex and race in a
sample of drug-abusing adults, reported that the intercept
equivalence for three items (emotionally abused, wear dirty
clothes, and physically abused) was not supported, indicating
that men and women had different scores on these items. Speci-
cally, women reported higher scores for the two items measuring
abuse, one of which (physically abused) loaded on the factor
Sexual Abuse in our sample. Furthermore, the structural invariance
across males and females has been questioned in the past. For
example, Wright et al. (2001) investigated the factor structure of
the CTQ-SF in 916 university students, and reported that their
results support a different factor structure for males (a ve-factor
model with the deletion of two items and with some cross-load-
ings) and for females (a four-factor model obtained by dropping
items assessing physical abuse). Nevertheless, Dudeck et al. (2015)
found that measurement invariance of the CTQ-SF across males
and females may be invalid only in nonclinical samples, but valid
in clinical samples.
Our results must be considered in light of some issues referred
to the design of the study. First, we administered only the CTQ-SF,
and some aspects of validity were not investigated (e.g., con-
vergent validity with independent measures of childhood mal-
treatment, and test-retest reliability). Second, our results may not
be generalizable for other groups of patients, for example, patients
with different medical conditions, or for nonclinical samples
(Dudeck et al., 2015). Third, our sample is composed of a hetero-
geneous group of psychiatric patients, and the structural
301
invariance of the CTQ-SF should be assessed in future studies
across diagnoses, although Spinhoven et al. (2014) found structure
invariance of the CTQSF across the presence of anxiety and/or
depressive disorder, while Bernstein et al. (2003) reported mea-
surement invariance among substance-abusing patients, adoles-
cent psychiatric inpatients, and community controls. In conclusion,
our study may be considered only a rst necessary step in the
process of the cross-cultural validation of the CTQ-SF.
5. Conclusions
Our results indicate that, when measuring different type of
childhood maltreatment, the difference between abuse and ne-
glect may be not clear. Future studies are needed to assess the
measurement invariance of the Italian version of the CTQ-SF across
sex and type of sample, and to investigate further the validity of
this measure. Nevertheless, the CTQ-SF may be used by re-
searchers investigating the association between different types of
childhood maltreatment and psychopathology or reactivity to
stress. The CTQ-SF may be also useful for clinicians. It could be
used for assessing the severity of a history of childhood mal-
treatment in order to identify patients with potentially worse
psychopathology and outcomes. However, considering the differ-
ent factor structure reported for the Italian CTQ-SF and the ab-
sence of research supporting the use of a specic cutoff for clas-
sifying individuals who have experienced maltreatment versus
those who have experienced none or little maltreatment, the CTQ-
SF should not be used at the present time for diagnostic or
screening purposes.
Disclosure of potential conicts of interest
All the authors declare no conict of interest.
Ethical approval
All procedures performed in studies involving human partici-
pants were in accordance with the ethical standards of the in-
stitutional and/or national research committee and with the 1964
Helsinki declaration and its later amendments or comparable
ethical standards.
Informed consent
Informed consent was obtained from all individual participants
included in the study.
References
Baumeister, D., Akhtar, R., Ciufolini, S., Pariante, C.M., Mondelli, V., 2016. Childhood
trauma and adulthood inammation: a meta-analysis of peripheral C-reactive
protein, interleukin-6 and tumour necrosis factor-alpha. Mol. Psychiatry 21,
642649.
Bernstein, D.P., Fink, L., Handelsman, L., Foote, J., Lovejoy, M., Wenzel, K., Sapareto,
E., Ruggiero, J., 1994. Initial reliability and validity of a new retrospective
measure of child abuse and neglect. Am. J. Psychiatry 151, 11321136.
Bernstein, D.P., Ahluvalia, T., Pogge, D., Handelsman, L., 1997. Validity of the
Childhood Trauma Questionnaire in an adolescent psychiatric population. J.
Am. Acad. Child Adolesc. Psychiatry 36, 340348.
Bernstein, D.P., Fink, L., 1998. Childhood Trauma Questionnaire: A Retrospective
Self-Report. Harcourt Brace & Co., San Antonio, TX.
Bernstein, D.P., Stein, J.A., Newcomb, M.D., Walker, E., Pogge, D., Ahluvalia, T.,
Stokes, J., Handelsman, L., Medrano, M., Desmond, D., Zule, W., 2003. Devel-
opment and validation of a brief screening version of the Childhood Trauma
302 M. Innamorati et al. / Psychiatry Research 245 (2016) 297302
Questionnaire. Child Abus. Negl. 27, 169190.
Bollen, K.A., Long, J.S., 1993. Testing Structural Equation Models. Sage Publications,
Newbury Park.
Browne, M.W., Cudek, R., 1993. Alternative ways of assessing model t. In: Long, J.S.
(Ed.), Testing Structural Equation Models. Sage, Newbury Park, CA, pp. 136162.
Costello, A.B., Osborne, J.W., 2005. Best practices in exploratory factor analysis: Four
recommendations for getting the most from your analysis. Pract. Assess. Res.
Eval. 10, 19.
DiLillo, D., Fortier, M.A., Hayes, S.A., Trask, E., Perry, A.R., Messman-Moore, T.,
Fauchier, A., Nash, C., 2006. Retrospective assessment of childhood sexual and
physical abuse: a comparison of scaled and behaviorally specic approaches.
Assessment 13, 297312.
Dudeck, M., Vasic, N., Otte, S., Streb, J., Wingenfeld, K., Grabe, H.J., Freyberger, H.J.,
Schroder, T., Von Schonfeld, C.E., Driessen, M., Barnow, S., Spitzer, C., 2015.
Factorial validity of the short form of the Childhood Trauma Questionnaire
(Ctq-Sf) in German psychiatric patients, inmates, and university students.
Psychol. Rep. 116, 685703.
Fabrigar, L.R., Wegener, D.T., MacCallum, R.C., Strahan, E.J., 1999. Evaluating the use
of exploratory factor analysis in psychological research. Psychol. Methods 4,
272299.
Floyd, F.J., Widaman, K.F., 1995. Factor analysis in the development and renement
of clinical assessment instruments. Psychol. Assess. 7, 286299.
Gerdner, A., Allgulander, C., 2009. Psychometric properties of the Swedish version
of the Childhood Trauma Questionnaire-Short Form (CTQ-SF). Nord. J. Psy-
chiatry 63, 160170.
Grassi-Oliveira, R., Cogo-Moreira, H., Salum, G.A., Brietzke, E., Viola, T.W., Manfro, G.
G., Kristensen, C.H., Arteche, A.X., 2014. Childhood Trauma Questionnaire (CTQ)
in Brazilian samples of different age groups: ndings from conrmatory factor
analysis. PloS One 9, e87118.
Gudmundsson, E., 2009. Guidelines for translating and adapting psychological in-
struments. Nord. Psychol. 61, 2945.
Heim, C., Nemeroff, C.B., 2001. The role of childhood trauma in the neurobiology of
mood and anxiety disorders: preclinical and clinical studies. Biol. Psychiatry 49,
10231039.
Heins, M., Simons, C., Lataster, T., Pfeifer, S., Versmissen, D., Lardinois, M., Marcelis,
M., Delespaul, P., Krabbendam, L., van Os, J., Myin-Germeys, I., 2011. Childhood
trauma and psychosis: a case-control and case-sibling comparison across dif-
ferent levels of genetic liability, psychopathology, and type of trauma. Am. J.
Psychiatry 168, 12861294.
Hernandez, A., Gallardo-Pujol, D., Pereda, N., Arntz, A., Bernstein, D.P., Gaviria, A.M.,
Labad, A., Valero, J., Gutierrez-Zotes, J.A., 2013. Initial validation of the Spanish
childhood trauma questionnaire-short form: factor structure, reliability and
association with parenting. J. Interpers. Violence 28, 14981518.
Hu, L., Bentler, P.M., 1999. Cutoff criteria for t indexes in covariance structure
analysis: conventional criteria versus new alternatives. Struct. Equ. Model.:
Multidiscip. J. 6, 155.
Huh, H.J., Kim, S.Y., Yu, J.J., Chae, J.H., 2014. Childhood trauma and adult inter-
personal relationship problems in patients with depression and anxiety dis-
orders. Ann. Gen. Psychiatry 13, 26.
Imperatori, C., Innamorati, M., Bersani, F.S., Imbimbo, F., Pompili, M., Contardi, A.,
Farina, B., 2015. The association among childhood trauma, pathological dis-
sociation and gambling severity in casino gamblers. Clin. Psychol. Psychother.
Imperatori, C., Innamorati, M., Lamis, D.A., Farina, B., Pompili, M., Contardi, A.,
Fabbricatore, M., 2016. Childhood trauma in obese and overweight women with
food addiction and clinical-level of binge eating. Child Abus. Negl. 58, 180190.
Karos, K., Niederstrasser, N., Abidi, L., Bernstein, D.P., Bader, K., 2014. Factor
structure, reliability, and known groups validity of the German version of the
Childhood Trauma Questionnaire (Short-form) in Swiss patients and non-
patients. J. Child Sex. Abus. 23, 418430.
Kelley, T.L., 1935. Essential Traits of Mental Life. Harvard University Press,
Cambridge.
Kuhlman, K.R., Geiss, E.G., Vargas, I., Lopez-Duran, N.L., 2015. Differential associa-
tions between childhood trauma subtypes and adolescent HPA-axis function-
ing. Psychoneuroendocrinology 54, 103114.
Lorenzo-Seva, U., 2013. How to Determine the Number of Common Factors Using
the Hull Method. Universitat Rovira i Virgili, Tarragona.
Lorezo-Seva, U., Ferrando, P.J., 2013. Manual of the Program Factor v.9.20. Depart-
ment de Psicologia, Universitata Rovira i Virgili, Tarragona.
Lundgren, K., Gerdner, A., Lundqvist, L.O., 2002. Childhood abuse and neglect in
severely dependent female addicts: homogeneity and reliability of a Swedish
version of the Childhood Trauma Questionnaire. Int. J. Soc. Welf. 11, 219227.
MacCallum, R.C., Roznowski, M., Necowitz, L.B., 1992. Model modications in cov-
ariance structure analysis: the problem of capitalization on chance. Psychol.
Bull. 111, 490504.
Mardia, K.V., 1970. Measures of multivariate skewnees and kurtosis with applica-
tions. Biometrika 57, 519530.
Muthn, L.K., Muthn, B.O., 1998. Mplus User's Guide, 6th ed. Muthn & Muthn,
Los Angeles, CA.
Nunnally, J.C., Bernstein, I.H., 1994. Psychometric Theory, 3rd ed. McGraw-Hill, New
York.
Pompili, M., Innamorati, M., Lamis, D.A., Erbuto, D., Venturini, P., Ricci, F., Serani,
G., Amore, M., Girardi, P., 2014. The associations among childhood maltreat-
ment, "male depression" and suicide risk in psychiatric patients. Psychiatry Res.
220, 571578.
Spinhoven, P., Penninx, B.W., Hickendorff, M., van Hemert, A.M., Bernstein, D.P.,
Elzinga, B.M., 2014. Childhood Trauma Questionnaire: factor structure, mea-
surement invariance, and validity across emotional disorders. Psychol. Assess.
26, 717729.
Tabachnick, B.G., Fidell, L.S., 2013. Using Multivariate Statistics, 6th ed. Pearson
Education, Boston.
Thombs, B.D., Lewis, C., Bernstein, D.P., Medrano, M.A., Hatch, J.P., 2007. An eva-
luation of the measurement equivalence of the Childhood Trauma Ques-
tionnaireShort Form across gender and race in a sample of drug-abusing
adults. J. Psychosom. Res. 63, 391398.
Thombs, B.D., Bernstein, D.P., Lobbestael, J., Arntz, A., 2009. A validation study of the
Dutch Childhood Trauma Questionnaire-Short Form: factor structure, relia-
bility, and known-groups validity. Child Abus. Negl. 33, 518523.
van Veen, T., Wardenaar, K.J., Carlier, I.V., Spinhoven, P., Penninx, B.W., Zitman, F.G.,
2013. Are childhood and adult life adversities differentially associated with
specic symptom dimensions of depression and anxiety? Testing the tripartite
model. J. Affect. Disord. 146, 238245.
Villano, C.L., Cleland, C., Rosenblum, A., Fong, C., Nuttbrock, L., Marthol, M., Wallace,
J., 2004. Psychometric utility of the childhood trauma questionnaire with fe-
male street-based sex workers. J. Trauma Dissociation 5, 3341.
Wright, K.D., Asmundson, G.J., McCreary, D.R., Scher, C., Hami, S., Stein, M.B., 2001.
Factorial validity of the Childhood Trauma Questionnaire in men and women.
Depress Anxiety 13, 179183.
Yu, C.Y., 2002. Evaluating Cutoff Criteria of Model Fit Indices for Latent variable
Models with Binary and Continuous Outcomes. University of California, Los
Angeles.
Zumbo, B.D., Gadermann, A.M., Zeisser, C., 2007. Ordinal versions of coefcients
alpha and theta for likert rating scales. J. Mod. Appl. Stat. Methods 6, 2129.