ArticlePDF Available

Confirmatory Factor Analysis of the Childhood Anxiety Sensitivity Index: A Gender Comparison

Authors:

Abstract and Figures

The Childhood Anxiety Sensitivity Index (CASI) is an 18-item self-report tool designed to measure the construct of anxiety sensitivity (i.e. the belief that anxiety may have harmful consequences such as sickness, embarrassment, or loss of control) in children and adolescents. Previous factor analytic examinations of the CASI have produced varied results. Gender may play a role in this observed variability. In an effort to confirm the factor structure of the measure across gender, CASI items for 671 children and adolescents were subjected to confirmatory factor analysis. Results indicated that for boys two-, three-, and four-factor structures provided a relatively good fit to the data, with the three-factor structure emerging as having the best fit overall. In contrast, for girls only the three-factor structure fitted the data well. Direct comparison of fit of the three-factor model across gender provided evidence to support the notion that childhood anxiety sensitivity is similar in structure across gender.
Content may be subject to copyright.
Confirmatory Factor Analysis of the Childhood Anxiety
Sensitivity Index: A Gender Comparison
Kristi D. Wright
1
, Gordon J. G. Asmundson
2
, Donald R. McCreary
3,4
,
Sherry H. Stewart
5,6
, Elizabeth McLaughlin
7
, M. Nancy Comeau
6
and
Trudi M. Walsh
8
1
Department of Psychology, University of Regina, Regina, Saskatchewan, Canada;
2
Faculty
of Kinesiology and Health Studies, University of Regina, Regina, Saskatchewan, Canada;
3
Department of Psychology, Brock University, St. Catharines, Ontario, Canada;
4
Department of Psychology, York University, Toronto, Ontario, Canada;
5
Department of
Psychiatry, Dalhousie University, Halifax, Nova Scotia, Canada;
6
Department of
Psychology, Dalhousie University, Halifax, Nova Scotia, Canada;
7
Pediatric Health
Psychology, IWK Health Centre, Halifax, Nova Scotia, Canada;
8
IWK Community Mental
Health Services Bedford/Sackville Branch, Halifax, Nova Scotia, Canada
Abstract. The Childhood Anxiety Sensitivity Index (CASI) is an 18-item self-report tool designed
to measure the construct of anxiety sensitivity (i.e. the belief that anxiety may have harmful
consequences such as sickness, embarrassment, or loss of control) in children and adolescents.
Previous factor analytic examinations of the CASI have produced varied results. Gender may play a
role in this observed variability. In an effort to confirm the factor structure of the measure across
gender, CASI items for 671 children and adolescents were subjected to confirmatory factor analysis.
Results indicated that for boys two-, three-, and four-factor structures provided a relatively good fit
to the data, with the three-factor structure emerging as having the best fit overall. In contrast, for girls
only the three-factor structure fitted the data well. Direct comparison of fit of the three-factor model
across gender provided evidence to support the notion that childhood anxiety sensitivity is similar in
structure across gender. Key words: child; anxiety sensitivity; gender; confirmatory factor analysis.
Received 16 October, 2009; Accepted 9 April, 2010
Correspondence address: Kristi D. Wright, PhD, Department of Psychology, Faculty of Arts, 3737
Wascana Parkway, Regina, SK S4S 0A2, Canada. Tel: 306-585-4180; Fax: 306-585-5429. E-mail:
kristi.wright@uregina.ca
Factor analytic investigations of the Child-
hood Anxiety Sensitivity Index (CASI) have
yielded mixed results, not unlike findings from
examinations the Anxiety Sensitivity Index
(ASI; for reviews, see Lilienfield, Turner, &
Jacob, 1993; Zinbarg, Mohlman, & Hong,
1999). Factor analytic studies of the CASI
have provided evidence to suggest that anxiety
sensitivity (AS) can be represented by factor
structures involving two (i.e. Autonomic
Concerns and Nonautonomic Concerns),
three (Physical Concerns, Psychological Con-
cerns, Social Concerns), and four (i.e. Disease
Concerns, Unsteady Concerns, Mental Incapa-
citation Concerns, Social Concerns) lower order
factors (Chorpita & Daleiden, 2000; Silverman,
Goedhart, Barrett, & Turner, 2003; Walsh,
Stewart, McLaughlin, & Comeau, 2004). Exist-
ing research has established that the factor
structure of the CASI is hierarchical: that is, the
lower order factors are shown to consistently
load on a higher order factor of global AS.
Several studies have yielded findings
suggesting that girls and boys may experience
AS differently (e.g. Muris, Schmidt, Merck-
elbach, & Schouten, 2001; Silverman, Gins-
q2010 Taylor & Francis ISSN 1650-6073
DOI: 10.1080/16506073.2010.486840
Cognitive Behaviour Therapy Vol. 39, No. 3, pp. 225–235, 2010
burg, & Goedhart, 1999; van Widenfelt,
Siebelink, Goedhart, & Treffers, 2002), just
as research findings in adult samples have
suggested that women and men may experi-
ence AS differently (see review by Stewart,
Taylor, & Baker, 1997). In line with these
findings, Walsh and colleagues (2004) exam-
ined the role of gender in AS dimension score
variability as measured by the CASI in a large
nonclinical sample of children and adolescents.
Principal-components analyses (PCA) on the 18
CASI items for the total sample, as well as
separately for boys and girls, revealed similar
three-factor lower order structures for all
groups (i.e. Physical, Psychological, and Social
Concerns factors). PCAs on the lower order
factor scores revealed similar unidimensional
higher order solutions for all three groups.
Other investigators have used confirmatory
factor analysis (CFA) to test competing models
of CASI factor structure (e.g. Adornetto et al.,
2008; Dehon, Weems, Stickle, Costa, & Ber-
man, 2005; Silverman et al., 2003). Differing
conclusions as to which model best fits the CASI
data have arisen. For example, Silverman and
colleagues (2003) identified the best-fitting
model for CASI data in a large nonclinical
sample of children and adolescents as being
composed of a single higher order AS factor and
four lower-order factors (Disease Concerns,
Unsteady Concerns, Mental Incapacitation
Concerns, Social Concerns). As with the
Walsh et al. (2004) findings, factor structure
did not differ across gender. Dehon and
colleagues (2005) tested two sets of nested
models in adolescent and adult samples (using
both the CASI and ASI): one based on the work
of Silverman and colleagues (2003), the other
based on the findings of Zinbarg, Barlow, and
Brown (1997). Results suggested that a hier-
archical model with three lower-order factors
(i.e. Zinbarg et al.’s model composed of
physical, psychological, and social factors)
provided the best fit to the data. Results
supported the hypothesis that factor structure
of the ASI is invariant across age and gender.
Adornetto and colleagues’ (2008) finding was
consistent with that of Silverman and colleagues
(2003) in that the four-factor hierarchical model
was found to best fit the data (using a 13-item
version of the CASI) in four nonclinical
adolescent samples; however, noninvariant
factor loadings between children and adoles-
cents and between boys and girls were observed.
Given the variability of the CASI structure
across studies, the goal of the present study was
to determine statistically which structure
provides the best fit to boys’ and girls’
responses to the CASI. To make this determi-
nation, we conducted a series of CFAs. Two-,
three-, and four-factor models were tested
(Chorpita & Daleiden, 2000; Walsh et al., 2004;
Silverman et al., 2003, respectively). Walsh and
colleagues’ (2004) three-factor model was
chosen, as opposed to a three-factor model
examined by Silverman et al. (2003), because it
is a close approximation to the three-factor
model established with adults (e.g. Stewart
et al., 1997; Taylor, Koch, Woody, & McLean,
1996; Zinbarg et al., 1997). All analyses were
conducted separately for boys and girls.
Method
Participants
The study sample consisted of 671 children and
adolescents (333 boys and 338 girls). Both
gender subsamples ranged in age from 7 to 17
years (boys: M¼13.45, SD ¼2.81; girls:
M¼13.31, SD ¼2.90). Participants were
primarily Caucasian, lived on the East Coast
of Canada, were from a cross-section of
socioeconomic levels, and spoke English
fluently. The study sample was randomly
selected from a subset of a larger dataset,
which was composed of two smaller datasets:
one for children (McLaughlin, Stewart, &
Taylor, 2007) and one for adolescents
(Comeau, Stewart, & Loba, 2001). We chose
a smaller sample size (i.e. random subset of the
larger dataset) to avoid duplication effects and
the possibility of the null hypotheses being
rejected (i.e. null hypotheses may be rejected in
large samples with small effect sizes). Our
sample size is approximately twice the mini-
mum number of cases required for a model
with the number of estimated parameters in
the examined models (i.e. approximately nine
cases per estimated parameter; Kline, 1998).
The total participation rate was 81.7%.
Measure
The CASI (Silverman et al., 1991) is an
18-item self-report measure designed for use
with school-age children and adolescents.
Respondents rate their fear of anxiety-related
sensations on a 3-point scale (1 ¼none,
2¼some, 3¼a lot). The CASI has been
226 Wright et al. COGNITIVE BEHAVIOUR THERAPY
determined to have good internal consistency
(a¼.87; Silverman et al., 1991). Test retest
reliability (2 weeks) of the CASI has been
determined to be acceptable, ranging from
r¼.62 to .78 (Silverman et al., 1991). The
CASI has also been found to be useful in
distinguishing youth who have experienced
panic attacks from those who have not (Lau,
Calamari, & Waraczynski, 1996) and those
with panic disorder from those with other
anxiety disorders (Kearney, Albano, Eisen,
Allan, & Barlow, 1997). The CASI has shown
to be concurrently associated with panic
symptoms, fears, and negative cognitive errors
(Weems, Berman, Silverman, & Saavedra,
2001) and to be predictive of anxiety-related
responses to behavioral-stress challenge tasks
(Rabian, Embry, & MacIntyre, 1999). Orig-
inally, the CASI was tested on children from
Grades 7 to 9, but subsequent investigations
have used the measure with children from
the ages of 6 through 18 years (e.g. Chorpita &
Lilienfeld, 1999; see review in Silverman &
Ollendick, 2005).
Procedure
The study information and consent forms
were distributed to parents or guardians of
students in the participating schools before
data collection day. Consent forms indicated
that participation was voluntary and that
confidentiality would be maintained. Written
parental consent and voluntary child assent
were required for participation in the child
sample (McLaughlin et al., 2007) and passive
parental consent (i.e. parents were provided
consent forms and parental consent was
inferred unless researchers were informed
differently) and voluntary youth assent were
required in the adolescent sample (Comeau
et al., 2001). Data collection was conducted on
a class-by-class basis. The items were read
aloud to classes with younger students (ages
79 years) and the students followed along on
their written versions. Participants were
permitted to ask questions during adminis-
tration of questionnaires.
Data analysis
The structural modeling program EQS (Ben-
tler, 2003), along with the procedures outlined
by Byrne (2006) and Kline (1998), was used to
conduct the series of CFAs. For all analyses,
raw data and a maximum-likelihood solution
were used. Because the goal was to determine
which of the three proposed factor structures
provided the best fit to the observed data, no
correlated residuals or cross-loadings were
allowed. Each CFA model was tested separ-
ately for boys and girls. Finally, multiple
group comparisons were used to determine
whether the factor loadings (both lower and
higher order) differed significantly between
boys and girls.
As recommended by Hu and Bentler (1999)
and Browne and Cudeck (1993), multiple fit
indices were used to determine how well the
proposed factor structures fitted the observed
data. The following fit indices were evaluated:
(a) chi-square (values should not be signifi-
cant, but in larger samples this is often not
feasible); (b) chi-square/degree of freedom
(x
2
/df) ratio (values should be ,3.0; Kline,
1998); (c) comparative fit index (CFI-R; values
should be close to .95); (d) root-mean-square
error of approximation (RMSEA; values
should be close to .05); and (e) the standar-
dized root-mean-square residual (SRMR;
values should be close to .08). On the principle
of parsimony, a more complex model would
have to reveal a large improvement over a less
complex model in order to be accepted as the
best-fitting model to the data.
Results
Confirmatory factor analyses: boys
Initial analyses of the CASI data for boys
revealed a high degree of multivariate kurtosis
(Mardia’s normalized estimate ¼35.37). To
minimize the effects of the multivariate
kurtosis (i.e. typically a reduced likelihood of
finding a good-fitting model), the Satorra-
Bentler robust scaling procedure was used.
This procedure provides scaled chi-square
(S-B x
2
) and CFI (CFI-R) statistics. The other
fit indices are not corrected for this effect.
Factor loadings for all three CASI models are
shown in Table 1 and fit statistics for all three
CASI models are shown in Table 2.
Chorpita and Daleiden’s(2000) two-factor
model. For the model with two lower order
factors loading onto a single higher order
factor, all items loaded significantly onto their
proposed lower order factors, and all item
residuals and latent factor variances also were
statistically significant ( p,.01). Standardized
factor loadings for the first factor (Autonomic
VOL 39, NO 3, 2010 Confirmatory factor analysis of the CASI 227
Table 1. Standardized factor loadings for two-factor, three-factor, and four-factor models
Item Factor 2-factor model
a
Factor 3-factor model
a
Factor 4-factor model
a
1. I don’t want other people to know when I feel afraid. NA .33/.38 SC .35/.51 SO .31/.48
2. When I cannot keep my mind on my schoolwork, I worry
that I might be going crazy.
NA .62/.53 PY .68/.73 MI .62/.67
3. It scares me when I feel “shaky.” AU .54/.51 PH .57/.56 DI .59/.59
4. It scares me when I feel like I am going to faint. AU .43/.41 PH .43/.43 US .52/.54
5. It is important for me to stay in control of my feelings. NA .20/.25 SC .31/.36 SO .32/.36
6. It scares me when my heart beats fast. AU .65/.54 PH .66/.55 DI .67/.55
7. It embarrasses me when my stomach growls (makes noise). AU .51/.40 PH .50/.39 US .58/.41
8. It scares me when I feel like I am going to throw up. AU .51/.38 PH .51/.40 US .54/.44
9. When I notice that my heart is beating fast, I worry that
there might be something wrong with me.
AU .59/.58 PH .61/.58 DI .61/.58
10. It scares me when I have trouble getting my breath. AU .52/.46 PH .53/.49 US .53/.52
11. When my stomach hurts, I worry that I might be really sick. AU .62/.55 PH .62/.54 DI .62/.55
12. It scares me when I can’t keep my mind on my schoolwork. NA .57/.54 PY .59/.52 MI .57/.54
13. Other kids can tell when I feel shaky. NA .29/.29 PH .30/.30 DI .30/.29
14. Unusual feelings in my body scare me. AU .58/.65 PH .58/.66 DI .59/.67
15. When I am afraid, I worry that I might be crazy. AU .56/.46 PY .64/.66 MI .65/.63
16. It scares me when I feel nervous. AU .52/.42 PH .52/.42 DI .52/.43
17. I don’t like to let my feelings show. NA .37/.35 SC .57/.56 SO .54/.54
18. Funny feelings in my body scare me. AU .65/.59 PH .66/.60 DI .67/.60
Note. NA, Nonautonomic Concerns; AU, Autonomic Concerns; PH, Physical Concerns; SC, Social Concerns; PY, Psychological Concerns; DI, Disease Concerns;
US, Unsteady Concerns; MI, Mental Incapacitation Concerns; SO, Social Concerns.
a
Loadings are represented as boys’ values/girls’ values.
228 Wright et al. COGNITIVE BEHAVIOUR THERAPY
Table 2. Fit statistics for two-factor, three-factor, and four-factor models
Hierarchical model S-B x
2
df S-B x
2
/df CFI-R RMSEA RMSEA CI SRMR
Boys’ model (N¼333)
Chorpita & Daleiden (2000) two-factor 225.77 134 1.69 .89 .05 .035– .055 .06
Walsh et al. (2004) three-factor
a
199.43 134 1.49 .92 .04 .027– .049 .05
Silverman et al. (2003) four factor 204.85 134 1.53 .91 .04 .028– .050 .06
Girls’ model (N¼338)
Chorpita & Daleiden (2000) two-factor 319.15 134 2.38 .79 .06 .055–.073 .06
Walsh et al. (2004) three-factor
a
259.88 134 1.94 .86 .05 .043– .062 .06
Silverman et al. (2003) four factor 271.48 134 2.03 .85 .06 .046– .064 .06
Note. S-B, Satorra-Bentler; CFI-R, robust comparative fit index; RMSEA, root-mean-square error of approximation; CI, confidence interval; SRMR, squared-root-
mean residual.
a
Best-fitting model.
VOL 39, NO 3, 2010 Confirmatory factor analysis of the CASI 229
Concerns; 12 items) ranged from .43 to .65.
Examination of the R
2
value associated with
each of the standardized residuals for this
factor showed that the items contributed
between 18% and 43% of their variance to
the autonomic concerns factor. Standardized
factor loadings for the second factor (Non-
autonomic Concerns; six items) ranged from
.20 to .62. Examination of the R
2
value
associated with each of the standardized
residuals showed that the items contributed
between 8% and 39% of their variance to the
Nonautonomic Concerns factor. The corre-
lation between the latent Autonomic and
Nonautonomic Concerns factors was .81.
When the higher order structure was imposed
on the data, the path coefficients from both of
the two lower order factors to the higher order
factor were significant: CASI to Autonomic
Concerns, .93 (R
2
¼.87); CASI to Nonauto-
nomic Concerns, .87 (R
2
¼.75). Three of the
four main fit statistics suggested a good fit,
with only the CFI-R being slightly below its
recommended value.
Walsh et al.’s (2004) three-factor model. For
this model with three lower order factors
loading onto a single higher order factor, all
items loaded significantly onto their proposed
lower order factors, and all item residuals and
latent factor variances also were statistically
significant ( p,.01). Standardized factor
loadings for the first factor (Physical Concerns;
12 items) ranged from .43 to .66. Examination
of the R
2
value associated with each of the
standardized residuals for this factor showed
that the items contributed between 9% and
43% of their variance to the Physical Concerns
factor. Standardized factor loadings for the
second factor (Social Concerns; three items)
ranged from .31 to .57. Examination of the R
2
value associated with each of the standardized
residuals showed that the items contributed
between 10% and 32% of their variance to the
Social Concerns factor. The standardized
factor loadings for the third factor (Psycho-
logical Concerns; three items) ranged from
.59 to .68. Examination of the R
2
value
associated with each of the standardized
residuals showed that the items contributed
between 35% and 46% of their variance to the
Psychological Concerns factor. The corre-
lations among the three lower order factors
were moderately high: Physical versus Social
Concerns, .49; Physical versus Psychological
Concerns, .78; and Social versus Psychological
Concerns, .40. When the higher order structure
was imposed on the data, the path coefficients
from each of the three lower order factors to
the higher order factor were significant: CASI
to Physical Concerns, .83 (R
2
¼.70); CASI to
Social Concerns, .62 (R
2
¼.38); CASI to
Psychological Concerns, .87 (R
2
¼.76). As
before, only the CFI-R fit statistic was below
the recommended value.
Silverman et al.’s (2003) four-factor model.
For the model with four lower order factors
loading onto a single higher order factor, all
items loaded significantly onto their proposed
lower order factors, and all item residuals
and latent factor variances also were statisti-
cally significant ( p,.01). Standardized factor
loadings for the first factor (Disease Concerns;
eight items) ranged from .30 to .67. Examin-
ation of the R
2
value associated with each of
the standardized residuals for this factor
showed that the items contributed between
9% and 45% of their variance to the Disease
Concerns factor. Standardized factor loadings
for the second factor (Unsteady Concerns; four
items) were between .52 and .58. Examination
of the R
2
value associated with these two
standardized residuals showed that they con-
tributed between 27% and 33% of their
respective variance to the Unsteady Concerns
factor. The standardized factor loadings for
the third factor (Mental Incapacitation Con-
cerns; three items) ranged from .57 to .65.
Examination of the R
2
value associated with
each of the standardized residuals showed that
the items contributed between 32% and 42%
of their variance to the Mental Incapacitation
Concerns factor. The standardized factor
loadings for the fourth factor (Social Con-
cerns; three items) ranged from .31 to .54.
Examination of the R
2
value associated with
each of the standardized residuals showed that
the items contributed between 10% and 29%
of their variance to the Social Concerns factor.
The correlations among the four lower order
factors were high: Disease versus Unsteady
Concerns, .91; Disease versus Mental Incapa-
citation Concerns, .78; Disease versus Social
Concerns, .47; Unsteady versus Mental Inca-
pacitation Concerns, .71; Unsteady versus
Social Concerns, .52; and Mental Incapacita-
tion versus Social Concerns, .40. The high
correlation between Disease and Unsteady
Concerns factors suggests redundancy between
230 Wright et al. COGNITIVE BEHAVIOUR THERAPY
these two factors. When the higher order
structure was imposed on the data, the path
coefficients from each of the four lower order
factors to the higher order factor were
significant: CASI to Disease Concerns, .89
(R
2
¼.76); CASI to Unsteady Concerns, .90
(R
2
¼.80), CASI to Mental Incapacitation
Concerns, .88 (R
2
¼.77);CASI to Social
Concerns, .62 (R
2
¼.38). As before, only the
CFI-R fit statistic was of concern.
Confirmatory factor analyses: girls
Initial analyses of the CASI data for girls
revealed a high degree of multivariate kurtosis
(Mardia’s normalized estimate ¼23.19). To
minimize the effects of the multivariate
kurtosis, the Satorra-Bentler robust scaling
procedure was used. Factor loadings for all
three CASI models are shown in Table 1 and
fit statistics for all three CASI models are
shown in Table 2.
Chorpita and Daleiden’s (2000) two-factor
model. For the model with two lower order
factors loading onto a single higher order
factor, all items loaded significantly onto their
proposed lower order factors, and all item
residuals and latent factor variances also were
statistically significant ( p,.01). Standardized
factor loadings for the first factor (Autonomic
Concerns; 12 items) ranged from .38 to .65.
Examination of the R
2
value associated with
each of the standardized residuals for this
factor showed that the items contributed
between 16% and 42% of their variance to
the Autonomic Concerns factor. Standardized
factor loadings for the second factor (Non-
autonomic Concerns; six items) ranged from
.25 to .54. Examination of the R
2
value
associated with each of the standardized
residuals showed that the items contributed
between 6% and 29% of their variance to the
Nonautonomic Concerns factor. The corre-
lation between the latent Autonomic and
Nonautonomic Concerns factors was .94.
When the higher order structure was imposed
on the data, the path coefficients from both of
the two lower order factors to the higher order
factor were significant: CASI to Autonomic
Concerns, .98 (R
2
¼.96); CASI to Nonauto-
nomic Concerns, .96 (R
2
¼.92). Only two of
the four main fit statistics suggested an
adequate fit, with the x
2
/df and CFI-R both
out of their range for acceptable fit.
Walsh et al.’s (2004) three-factor model. For
this model with three lower order factors
loading onto a single higher order factor, all
items loaded significantly onto their proposed
lower order factors, and all item residuals
and latent factor variances also were statisti-
cally signifi cant ( p,.01). Standardized factor
loadings for the first factor (Physical Concerns;
12 items) ranged from .30 to .66. Examination
of the R
2
value associated with each of the
standardized residuals for this factor showed
that the items contributed between 9% and
43% of their variance to the Physical Concerns
factor. Standardized factor loadings for the
second factor (Social Concerns; three items)
ranged from .36 to .56. Examination of the R
2
value associated with each of the standardized
residuals showed that the items contributed
between 13% and 31% of their variance to the
Social Concerns factor. The standardized
factor loadings for the third factor (Psycho-
logical Concerns; three items) ranged from .52
to .73. Examination of the R
2
value associated
with each of the standardized residuals showed
that the items contributed between 27% and
54% of their variance to the Psychological
Concerns factor. The correlations among the
three lower order factors were moderate:
Physical versus Social Concerns, .57; Physical
versus Psychological Concerns, .65; and Social
versus Psychological Concerns, .34. When the
higher order structure was imposed on the
data, the path coefficients from each of
the three lower order factors to the higher
order factor were significant: CASI to Physical
Concerns, .96 (R
2
¼.92); CASI to Social
Concerns, .88 (R
2
¼.77); CASI to Psycho-
logical Concerns, .79 (R
2
¼.63). With regard
to model fit, only CFI-R was below the
recommended value.
Silverman et al.’s (2003) four-factor model.
For the model with four lower order factors
loading onto a single higher order factor, all
items loaded significantly onto their proposed
lower order factors, and all item residuals and
latent factor variances also were statistically
significant ( p,.01). Standardized factor load-
ings for the first factor (Disease Concerns; eight
items) ranged from .29 to .67. Examination of
the R
2
value associated with each of the
standardized residuals for this factor showed
that the items contributed between 9% and
44% of their variance to the Disease Concerns
factor. Standardized factor loadings for the
VOL 39, NO 3, 2010 Confirmatory factor analysis of the CASI 231
second factor (Unsteady Concerns; four items)
were .41 and .54. Examination of the R
2
value
associated with these two standardized
residuals showed that they contributed
between 17% and 29% of their respective
variance to the Unsteady Concerns factor. The
standardized factor loadings for the third
factor (Mental Incapacitation Concerns; three
items) ranged from .54 to .67. Examination of
the R
2
value associated with each of the
standardized residuals showed that the items
contributed between 30% and 44% of their
variance to the Mental Incapacitation Con-
cerns factor. The standardized factor loadings
for the fourth factor (Social Concerns; three
items) ranged from .36 to .54. Examination of
the R
2
value associated with each of the
standardized residuals showed that the items
contributed between 13% and 30% of their
variance to the Social Concerns factor. The
correlations among the four lower order
factors ranged from moderate to high: Disease
versus Unsteady Concerns, .89; Disease versus
Mental Incapacitation Concerns, .70; Disease
versus Social Concerns, .55; Unsteady versus
Mental Incapacitation Concerns, .42,
Unsteady versus Social Concerns, .60, and
Mental Incapacitation versus Social Concerns,
.35. The high correlation between Disease and
Unsteady Concerns factors suggests redun-
dancy between these two factors. When the
higher order structure was imposed on the data,
the path coefficients from each of the four lower
order factors to the higher order factor were
significant: CASI to Physical Concerns, .86
(R
2
¼.74); CASI to Unsteady Concerns, .83
(R
2
¼.69); CASI to Mental Incapacitation
Concerns, .79 (R
2
¼.63); CASI to Social
Concerns, .69 (R
2
¼.48). The fit statistics
were mixed. The x
2
/df ratio was slightly out of
range while the CFI-R was well below its
recommended value. The SRMR and the
RMSEA, however, suggested good fit.
Measurement equivalence across
gender
To determine whether the lower order and
higher order factor loadings differed signifi-
cantly as a function of gender, we conducted a
series of multiple-group comparisons that
constrained these loadings to be equal
(Byrne, 2006). We did this for each of the
two-, three-, and four-factor models. In EQS, a
chi-square difference test is used to determine
whether there is a significant difference
between the loading that is allowed to vary
versus the loading that is constrained. In each
model, only one constraint reached statistical
significance: Item 14 (“Unusual feelings in my
body scare me”). The factor loading for girls
was significantly stronger in the two-factor,
x
2
(1, N¼671) ¼4.61, p,.05, three-factor,
x
2
(1, N¼671) ¼5.26, p,.05, and four-
factor, x
2
(1, N¼671) ¼4.40, p,.05, models.
None of the other factor loadings differed
significantly as a function of gender.
Discussion
Disagreement exists regarding the best repre-
senting factor structure of AS in children and
adolescents. Evidence has been put forth for a
variety of models (i.e. two-, three-, four-factor
hierarchical models). Both Walsh et al. (2004)
and Silverman et al. (2003) have demonstrated
that gender plays a role in the levels of concern
for the constructs that represent the lower
order CASI factors. The purpose of the present
study was to statistically determine which
factor structure provided the best fit to the
CASI data in boys and girls. In order to do so,
we used CFA. We initially compared CASI
data from boys and girls separately against
three models of the CASI factor structure that
have been discussed in the literature (i.e. two-,
three-, and four-factor structures). We sub-
sequently conducted a direct comparison of the
CASI factor loadings across gender in each of
the three models.
For boys, the fit statistics indicated that all
models fittted the CASI data well. On closer
examination of model fit statistics, Walsh and
colleagues’ (2004) three-factor model (com-
posed of Physical Concerns, Psychological
Concerns, and Social Concerns) emerged as
the best-fitting model to the boys’ CASI data
in comparison to the other factor structures
examined. Conversely, for the girls, the fit
statistics indicated that only one model fitted
the CASI data particularly well (i.e. Walsh
and colleagues’ three-factor model). In
addition, when we examined the correlations
between lower order and higher order factors
across models, Walsh and colleagues’ three-
factor model remained superior. Specifically,
the three-factor model produced fewer redun-
dant factors.
232 Wright et al. COGNITIVE BEHAVIOUR THERAPY
Our subsequent analyses included a series of
multiple-group comparisons in order to
determine whether the lower order and higher
order factor loadings differed significantly as a
function of gender. Our findings suggested
that the factor structure is relatively invariant
across gender. There was only one constraint
that was significantly different across gender:
Item 14 (“Unusual feelings in my body scare
me”). The factor loading of this item was
significantly stronger for girls across models,
while the remaining factor loadings did not
differ significantly as a function of gender. The
latter may suggest that being scared by
unusual bodily sensations is a more of a part
of the physical concerns construct for girls
than it is for boys. This appears consistent
with the notion that girls have more physical
concerns than boys (Walsh et al., 2004), yet
takes it further and suggests that girls and
boys experience physical concerns somewhat
differently at least in terms of the contribution
of fear of unusual bodily sensations.
The results provide important information
regarding gender differences in the factorial
validity of the CASI. Our results suggest that
for boys all factor structures were statistically
valid, but based on observed fit indices and the
principle of parsimony, the three-factor
structure was determined to provide the best
fit over all. For girls, on the other hand, the
three-factor structure provided the “best fit.”
When we directly compared the factor
structures across boys and girls, we deter-
mined that the CASI factor structure was
statistically the same for boys and girls. Our
results are consistent with EFA findings of
Walsh and colleagues (2004) in that their
three-factor model provided the best fit to the
data, irrespective of gender. Given our
findings, it appears that the utilization of
three AS dimensions would prove useful in
assessment and treatment planning for both
boys and girls.
Our findings are inconsistent with those of
Silverman et al. (2003) and Adornetto et al.
(2008). Both these groups of researchers
identified a four-factor structure as providing
the best fit to the CASI data (although
Adornetto et al. only found good fit for the
four-factor model for a 13-item version of the
CASI, while Silverman et al. found good fit for
this more complex model for both the 13- and
18-item CASI versions). Silverman et al., but
not Adornetto et al., found the factor
structure to be invariant across gender. In
contrast, we found the best-fitting structure to
be the three-factor structure identified by
Walsh et al. (2004), although we did find the
factor structure to be relatively invariant
across gender. The two physical concerns
(i.e. disease vs. unsteady) factors identified in
Silverman et al.’s four-factor model were
highly correlated in both girls and boys in the
present study, suggesting redundancy between
these two factors. The Social Concerns factor
has been noted to have poor reliability (e.g.
Chorpita & Lilienfeld, 1999; Silverman et al.,
2003; Walsh et al., 2004). The poor reliability
of this factor may lead to inconsistency in the
measurement of childhood AS and, in turn,
could play a role in the disagreement observed
across studies. It has been suggested that the
modification and/or removal of existing CASI
items as well as the addition of new items may
aid in improving the reliability of the Social
Concerns factor. Changes to the items that
comprise the CASI (specifically to those items
that comprise the Social Concerns factor) may
eliminate the inconsistencies that have been
observed across studies. We are not suggesting
that Social Concerns items be entirely
removed, because these items have shown to
be predictive of anxiety symptoms (e.g. Social
Concerns factor scores predictive of social
phobia symptoms; McLaughlin et al., 2007).
Disparate findings may also be the result of
the variability in the items that comprise each
individual CASI factor across models and
studies (e.g. use of 13- and 18-item versions of
the CASI; see Adornetto et al., 2008; Silver-
man et al., 2003). The present findings,
coupled with those of Silverman et al. and
Adornetto et al., highlight the importance of
modifying the CASI in an effort to improve its
clinical utility (i.e. help understand the
mechanisms that serve to initiate develop-
ment, exacerbate, and maintain anxiety
disorders in children and adolescents; Muris,
2002) as well as to clarify and confirm its
underlying factor structure. Further, in an
effort to understand the inconsistent findings
in the literature in terms of CASI factor
structure, researchers may also consider
using factor mixture modeling, a statistical
approach that integrates mixture modeling
and factor analysis in order to model the fit of
various competing, latent structural models
VOL 39, NO 3, 2010 Confirmatory factor analysis of the CASI 233
(Lubke & Neale, 2006, as cited in Bernstein
et al., in press).
Acknowledgements
This project was completed by Kristi
D. Wright under the supervision of Gordon
J. G. Asmundson, Donald R. McCreary, and
Sherry H. Stewart, in partial completion of her
comprehensive requirement for the PhD
degree at Dalhousie University. This research
was supported in part by the Nova Scotia
Health Research Foundation and the Cana-
dian Institutes of Health Research. The
second author is supported through a CIHR
Investigator Award and the fourth author
through a Killam Research Professorship from
the Dalhousie University Faculty of Science.
The authors wish to thank Susan Button,
Brent Conrad, Pamela Loba, Grant
MacLeod, Paula MacPherson, Meredith
McLaughlin, Heather Lee Loughlin, Ellen
Rhyno, Kathy Silver, and Jennifer Komar for
their assistance with data collection and data
entry. This study would not have been possible
without the cooperation of the principals,
teachers, parents, and students of the partici-
pating schools.
References
Adornetto, C., Hensdiek, M., Meyer, A., In-Albon,
T., Federer, M., & Schneider, S. (2008). The
factor structure of the Childhood Anxiety
Sensitivity Index in German children. Journal
of Behavior Therapy and Experimental Psychia-
try,39, 404–416.
Bentler, P. M. (2003). EQS: Structural equation
program manual. Los Angeles: Multivariate
Software, Inc.
Bernstein, A., Stickle, T. R., Zvolensky, M. J.,
Taylor, S., Abramowitz, J., & Stewart, S. H.
(in press). Dimensional, categorical, or dimen-
sional-categories: Testing the latent structure of
anxiety sensitivity among adults using factor-
mixture modeling. Behavior Therapy.
Browne, M. W., & Cudeck, R. (1993). Alternate
ways of assessing model fit. In K. A. Bollen &
J. S. Long (Eds.), Testing equation models
(pp. 136–162). Newbury Park, CA: Sage.
Byrne, B. M. (2006). Structural equation modeling
with EQS: Basic concepts, applications, and
programming (2nd ed.). Mahwah, NJ: Erlbaum.
Chorpita, B. F., & Daleiden, E. L. (2000). Proper-
ties of the Childhood Anxiety Sensitivity Index
in children with anxiety disorders: Autonomic
and nonautonomic factors. Behavior Therapy,
31, 327–349.
Chorpita, B. F., & Lilienfeld, S. O. (1999). Clinical
assessment of anxiety sensitivity in children and
adolescents: Where do we go from here?
Psychological Assessment,11(2), 212–224.
Comeau, N., Stewart, S. H., & Loba, P. (2001). The
relations of trait anxiety, anxiety sensitivity, and
sensation seeking to adolescents’ motivations
for alcohol, cigarette, and marijuana use.
Addictive Behaviors,26, 803–825.
Dehon, C., Weems, C. F., Stickle, T. R., Costa,
N. M., & Berman, S. L. (2005). A cross sectional
evaluation of factorial invariance of anxiety
sensitivity in adolescents and young adults.
Behavior Research and Therapy,43, 799–810.
Hu, L., & Bentler, P. M. (1999). Fit indices in
covariance structure modeling: Sensitivity to
underparameterized model mis-specification.
Psychological Methods,3, 424–453.
Kearney, C. A., Albano, A. M., Eisen, A. R., Allan,
W. D., & Barlow, D. H. (1997). The phenom-
enology of panic disorder in youngsters: An
empirical study of a clinical sample. Journal of
Anxiety Disorders,11, 49–62.
Kline, R. B. (1998). Principles and practice of
structural equation modeling. New York: Guil-
ford Press.
Lau, J. J., Calamari, J. E., & Waraczynski, M.
(1996). Panic attack symptomatology and
anxiety sensitivity in adolescents. Journal of
Anxiety Disorders,10, 355–364.
Lilienfield, S. O., Turner, S. M., & Jacob, R.G.
(1993). Anxiety sensitivity: An examination of
theoretical and methodological issues. Advances
in Behaviour Research and Therapy,15,
147–183.
Lubke, G. H., & Neale, M. C. (2006). Distinguish-
ing between latent classes and continuous
factors: Resolution by maximum likelihood?
Multivariate Behavioral Research,41, 499–532.
McLaughlin, E., Stewart, S. H., & Taylor, S. (2007).
Childhood Anxiety Sensitivity Index factors
predict unique variance in DSM-IV anxiety
disorder symptoms. Cognitive Behaviour
Therapy,36, 210–219.
Muris, P. (2002). An expanded Childhood Anxiety
Sensitivity Index: Its factor structure, reliability,
and validity in a non-clinical adolescent sample.
Behavior Research and Therapy,40, 299–311.
Muris, P., Schmidt, H., Merckelbach, H., &
Schouten, E. (2001). Anxiety sensitivity in
adolescents: Factor structure and relationships
to trait anxiety and symptoms of anxiety
disorders and depression. Behaviour Research
and Therapy,39, 89–100.
Peterson, R. A., & Reiss, S. (1992). Anxiety
Sensitivity Index manual (2nd ed.). Worthing-
ton, OH: International Diagnostic Systems.
Rabian, B., Embry, L., & MacIntyre, D. (1999).
Behavioral validation of the Childhood Anxiety
Sensitivity Index in children. Journal of Clinical
Child Psychology,28, 105–112.
Reiss, S., & McNally, R. J. (1985). The expectancy
model of fear. In S. Reiss & R. R. Bootzin
(Eds.), Theoretical issues in behavior therapy
(pp. 107–121). New York: Academic Press.
234 Wright et al. COGNITIVE BEHAVIOUR THERAPY
Silverman, W. K., Fleisig, W., Rabian, B., &
Peterson, R. (1991). Childhood Anxiety Sensi-
tivity Index. Journal of Clinical Child Psychol-
ogy,20, 162–168.
Silverman, W. K., Ginsburg, G. S., & Goedhart,
A. W. (1999). Factor structure of the Childhood
Anxiety Sensitivity Index. Behaviour Research
and Therapy,37, 903–917.
Silverman, W. K., Goedhart, A. W., Barrett, P., &
Turner, C. (2003). The facets of anxiety
sensitivity in the Childhood Anxiety Sensitivity
Index: Confirmatory analyses of factor models
from past studies. Journal of Abnormal Psychol-
ogy,112, 364–374.
Silverman, W. K., & Ollendick, T. H. (2005).
Evidence-based assessment of anxiety and its
disorders in children and adolescents. Journal of
Clinical Child and Adolescent Psychology,34,
380–411.
Silverman, W. K., & Weems, C. F. (1999). Anxiety
sensitivity in children. In S. Taylor (Ed.),
Anxiety sensitivity: Theory, research, and treat-
ment of the fear of anxiety (pp. 239–268).
Mahwah, NJ: Erlbaum.
Stewart, S. H., Taylor, S., & Baker, J. M. (1997).
Gender differences in dimensions of anxiety
sensitivity. Journal of Anxiety Disorders,11,
179–200.
Taylor, S., Koch, W. J., Woody, S. R., & McLean,
P. (1996). Anxiety sensitivity and depression:
How are they related? Journal of Abnormal
Psychology,105, 474–479.
van Widenfelt, B. M., Siebelink, B. M., Goedhart,
A. W., & Treffers, P. D. A. (2002). The Dutch
Childhood Anxiety Sensitivity Index: Psycho-
metric properties and factor structure. Journal
of Clinical Child and Adolescent Psychology,31,
90–100.
Walsh, T. M., Stewart, S. H., McLaughlin, E., &
Comeau, N. (2004). Gender differences in
Childhood Anxiety Sensitivity Index (CASI)
dimensions. Journal of Anxiety Disorders,18,
695–706.
Weems, C. F., Berman, S. L., Silverman, W. K., &
Saavedra, L. S. (2001). Cognitive errors in
youth with anxiety disorders: The linkages
between negative cognitive errors and anxious
symptoms. Cognitive Therapy and Research,25,
559–575.
Zinbarg, R. E., Barlow, D. H., & Brown, T. A.
(1997). Hierarchical structure and general
factor saturation of the Anxiety Sensitivity
Index: Evidence and implications. Psychological
Assessment,9, 277–284.
Zinbarg, R. E., Mohlman, J. F., & Hong, N. N.
(1999). Dimensions of anxiety sensitivity. In
S. Taylor (Ed.), Anxiety sensitivity: Theory,
research and treatment of the fear of anxiety
(pp. 83–144). Mahwah, NJ: Erlbaum.
VOL 39, NO 3, 2010 Confirmatory factor analysis of the CASI 235
Copyright of Cognitive Behaviour Therapy is the property of Routledge and its content may not be copied or
emailed to multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.
... Using CFAs, Wright et al. (2010) compared three factor models previously identified in the literature: (1) a two-factor model proposed by , (2) a three-factor model reported by Walsh et al. (2004), and (3) a four-factor model suggested by Silverman et al. (2003) while also assessing factor invariance across sex. Wright et al. (2010) observed acceptable model fit for all three of the tested structures for boys, although the three-factor structure provided the best fit for the data. ...
... Using CFAs, Wright et al. (2010) compared three factor models previously identified in the literature: (1) a two-factor model proposed by , (2) a three-factor model reported by Walsh et al. (2004), and (3) a four-factor model suggested by Silverman et al. (2003) while also assessing factor invariance across sex. Wright et al. (2010) observed acceptable model fit for all three of the tested structures for boys, although the three-factor structure provided the best fit for the data. Conversely, in girls, only the three-factor solution provided a good fit for the data. ...
... Citing the previous lack of agreement over the AS factor structure in youth and suggesting that genetic methods might help to identify the optimal number of AS factors, Brown et al. (2012) used data from a large-scale longitudinal study of adolescent twin and sibling pairs to conduct a confirmatory analysis of the CASI. Specifically, previously identified two- , three- (Walsh et al. 2004;Wright et al. 2010), and four-factor (Silverman et al. 1999(Silverman et al. , 2003 models were used to compare their data, with subsequent analyses employed to examine genetic and environmental contributions to AS. Although three-and four-factor solutions provided comparable fits to the data, a three-factor solution was favored due to considerations of "greater interpretability and parsimony" (Brown et al. 2012). ...
Article
Full-text available
Background Child anxiety sensitivity (AS) is measured almost exclusively using the Childhood Anxiety Sensitivity Index (CASI). Yet, in the context of significant discrepancies regarding the CASI factors and how they are scored and reported, it remains unclear whether the CASI reliably and validly assesses the purported multifactorial AS construct. Objective This paper will: (1) provide a comprehensive summary of previous CASI factor analyses by which these factor structures were identified, (2) evaluate evidence regarding the multifactorial nature of AS in youth, and (3) discuss potential directions for continued research in this area. Method In a PsycInfo search, peer-reviewed studies published between 1991 and 2018 were identified for inclusion if they examined the factor structure of the CASI or reported data on the CASI subscales as administered to child participants. Results Findings from the 50 studies reviewed suggested that (1) the 18-item CASI does not consistently yield internally reliable assessments of specific AS facets, (2) significant discrepancies exist regarding the CASI subscales identified, the items comprising these scales, and their predictive validity in terms of anxiety, and (3) alternatives to assess the multifactorial construct of AS in youth do exist, but they have not been systematically examined in the literature. Conclusions Directions for future study include further examining expanded scales for AS in youth, continued study of shorter scales assessing more consistently reliable AS content, and evaluating the utility of an expanded response scale for the CASI.
... Anxiety sensitivity was first noted by Reiss and McLen (1). Anxiety sensitivity is a stable tendency variable that indicates a desire to interpret the physical, psychological, and social consequences of anxiety as distressing and dangerous issues (16). ...
... It assesses the insomnia during the last month and the dimensions evaluated are: sleep maintenance, severity of sleep, onset and early morning awakening problems, interference of sleep, sleep dissatisfaction, noticeability of sleep problems by others, difficulties with daytime functioning, and distress caused by the sleep difficulties. A total score ranges from 0 to 28 and it is interpreted as follows: absence of insomnia (0-7), sub-threshold insomnia (8)(9)(10)(11)(12)(13)(14), moderate insomnia (15)(16)(17)(18)(19)(20)(21), and severe insomnia (22)(23)(24)(25)(26)(27)(28) (20). ...
Article
Full-text available
Background and Objective: Insomnia is a universal phenomenon that many people experience and is characterized by difficulty in sleep initiation, maintaining sleep, waking up early in the morning, and inability to return to sleep. Due to its high prevalence and the effect of insomnia on the mental and physical performance of individuals, especially stu-dents, the present study aimed to investigate the relationship between transdiagnostic structures and insomnia disorder in students. Materials and Methods: This was a cross-sectional study. Population of this study was 400 medical students of Zanjan University of Medical Sciences and Zanjan University, Zanjan, Iran, who were selected by non-random sampling method. Research tools included demographic characteristics questionnaire, Insomnia Severity Index (ISI), 12-question Intolerance of Uncertainty Scale (IUS), Difficulties in Emotion Regulation Scale (DERS), Anxiety Sensitivity Index-Third Edition (ASI-3), and Repetitive Thinking Questionnaire (RTQ). Research data were analyzed using LISREL software. Results: The average age of students was 21.54 years. Transdiagnostic constructs explained a total of 9% of insomnia disorders in students. Insomnia disorders in women were significantly higher than in men; moreover, in the single group was higher than in the married group. Conclusion: Transdiagnostic constructs can predict insomnia in students.
... Ranney et al. (2021) found maternal and paternal anxiety to be associated with adolescent anxiety more strongly for female offspring. There are also gender differences related to AS, such that adolescent girls experience more physical and social concerns than adolescent boys (Walsh et al., 2004;Wright et al., 2010). Additionally, youth gender may moderate the association between parent and child AS, as Graham and Weems (2015) found a positive association between female offspring and parent AS and a negative association between male offspring and parent AS. ...
Article
Full-text available
Anxiety disorders often onset in adolescence and continue into adulthood. Multiple factors contribute to anxiety disorder development, such as parent emotional availability (EA) and adolescent anxiety sensitivity (AS). Previous research determined attachment is a similar construct to EA, and attachment is related to anxiety disorders. Moreover, EA is a more specific construct than attachment that can be targeted in therapy. It has been determined that adolescent attachment to parental figures contributes to anxiety, with AS mediating this relationship. The present research sought to determine if AS is one mechanism underlying the relationship between parent EA and adolescent anxiety, while exploring the moderating roles of parent and adolescent gender. In a Midwestern sample (13–19 years old), this study examined adolescent AS as a mediator of the relationship between perceived parent EA and adolescent anxiety, and gender as a potential moderator. Self-report data were collected from adolescents via Qualtrics. The findings supported the mechanism of adolescent AS as a mediator between perceived parent EA and adolescent anxiety. As AS emerges during adolescence, it is a pertinent treatment target for youth anxiety. Furthermore, this study underscores the significance of parent and adolescent gender and parental EA as treatment targets for adolescent anxiety. While targeting adolescent AS and enhancing parental EA can be beneficial for both male and female adolescents, focusing on parental EA may offer greater benefits for male adolescents. In sum, these findings can increase the efficacy of current parenting programs and further increase treatment outcomes for youth and families.
... The ASI-3, which I codeveloped and validated, is a psychometrically sound and now commonly used measure of AS which reliably taps each of the AS lower order dimensions with a set of three sixitem subscales . My former graduate student Kristi Wright has shown that these same three dimensions are evident in children/adolescents (Wright et al., 2010). ...
Article
Full-text available
Le présent article résume notre programme de recherche sur la sensibilité à l’anxiété (SA) – un facteur dispositionnel cognitif et affectif impliquant des craintes de sensations liées à l’anxiété en raison de croyances selon lesquelles ces sensations entraînent des conséquences catastrophiques. La SA et ses dimensions d’ordre inférieur sont considérées comme des facteurs transdiagnostiques de risque ou de maintien des troubles émotionnels et des troubles addictifs. La compréhension des mécanismes par lesquels la SA exerce ses effets peut révéler des cibles d’intervention clés pour les programmes de prévention et de traitement axés sur la SA. Dans le présent article, je passe en revue les recherches fondamentales que nous avons menées pour comprendre les mécanismes qui relient la SA à ces troubles et à leurs symptômes. Je décris également les interventions transdiagnostiques ciblées sur la SA et j’illustre la manière dont la recherche fondamentale a permis d’orienter le contenu de ces interventions. Enfin, je passe en revue les projets en cours dans mon laboratoire et je souligne les orientations futures importantes dans ce domaine. Bien que des progrès considérables aient été réalisés au cours des trois dernières décennies et que la recherche ait considérablement fait avancer notre compréhension de la SA en tant que facteur transdiagnostique, de nombreuses questions restent en suspens. Les réponses devraient nous aider à affiner les interventions afin d’en faire bénéficier au maximum les personnes qui ont une grande peur d’avoir peur.
... The Childhood Anxiety Sensitivity Index (CASI) is an 18-item measure that was used to assess adolescent anxiety sensitivity. The CASI has a hierarchical structure, with one global factor and three subfactors (i.e., physical concerns, social concerns, and cognitive concerns; Silverman et al., 1991;Wright et al., 2010). The current study employed the global factor, which represents a generalized fear of the consequences of anxiety related sensations, all of which may pose problems for adolescents trying to fall asleep. ...
Article
Introduction Sleep disturbances are common among adolescents and are associated with elevated anxiety, and difficulties managing affect. Familial conflict is associated with both anxiety sensitivity and adolescent sleep disturbances. No work to date has examined how adolescent sleep disturbances may interact with anxiety sensitivity in relation to adolescent affective responding to parent-adolescent conflict. The current study was designed to address this gap in the literature by examining how adolescent sleep disturbances, anxiety sensitivity, conflict elicited anger, and conflict avoidance are associated. Method Seventy-two American adolescents (n = 39 males) between the ages of 12 and 16 years (Mage = 13.84, SD = 1.38) completed a baseline assessment as well as a well-validated mother-adolescent laboratory-based conflict task. Results For youth low in anxiety sensitivity, greater sleep disturbance related positively to conflict-elicited anger, which in turn predicted higher conflict avoidance. In contrast, this indirect effect was not significant for adolescents relatively higher in anxiety sensitivity. Instead, for these adolescents, increased sleep disturbances were associated with lower levels of conflict elicited anger. Conclusions Results suggest that the effects of sleep disturbances on conflict elicited anger may vary as a function of adolescent anxiety vulnerability. These findings highlight the importance of considering the unique effects of sleep disturbances on adolescent affect as a function of adolescent anxiety vulnerability.
... The CASI differs from the original Anxiety Sensitivity Index (ASI) 8 for adults in a few ways: It consists of two more items altogether (18 CASI items vs. 16 ASI items), 43 and it has a somewhat different factor composition in terms of the proportion of items per factor, with the CASI being more strongly slanted toward physical concerns relative to the ASI (CASI: 12 physical concerns, 3 psychological concerns, and 3 social concerns items; ASI: 8 physical concerns, 4 psychological concerns, and 4 social concerns items). 44 A sum score for the CASI items was computed, with higher scores indicating higher levels of anxiety sensitivity. Cronbach's alpha for the CASI was .88 in the current sample. ...
Article
Full-text available
Theoretically, anxiety sensitivity—fear of anxiety symptoms—enhances perception of and emotional reactivity to autonomic arousal and mental distress, thereby increasing negative affect and motivation to use substances for negative reinforcement. Because no prior study of adolescents has tested if anxiety sensitivity is indirectly associated with substance use problems through symptoms of emotional disorders (i.e., disorders involving high levels of negative affect), the current cross-sectional study examined this theoretical pathway. Participants included 9th-grade students from 10 different high schools in the Los Angeles metropolitan area (N = 3005; 54.3% female). Self-report measures of anxiety sensitivity, emotional disorder symptoms, tobacco dependence, and alcohol and other drug problems were administered. Controlling for sex, race/ethnicity, parental education, school, and impulsiveness, we tested the associations of anxiety sensitivity with tobacco, alcohol, and other drug use problems as well as the indirect effects of anxiety sensitivity on each domain of substance use problems through emotional disorder symptoms. Anxiety sensitivity was associated with more severe tobacco dependence and greater alcohol problems and other drug problems, and anxiety sensitivity further was indirectly associated with all three domains of substance use problems through emotional disorder symptoms. Current findings suggest that adolescents high in anxiety sensitivity tend to experience emotional disorder symptoms, which may increase risk for substance use problems. Interventions that target anxiety sensitivity and enhance negative-affect coping skills may assist in preventing and reducing adolescent substance use problems.
Article
High and rising rates of smoking and drinking among Chinese adolescents are contributing to increasingly serious physical and mental health issues. While impulsivity has been demonstrated to be significantly related to adolescent cigarette and alcohol use, little is known about the mechanisms behind this association. The current study focused on resilience and depressive symptoms as potential mediators of this link, exploring the indirect pathways connecting impulsivity to teenage tobacco and alcohol use. Possible gender differences in this indirect pathway were also explored. Participants were secondary school students from southern China ( N = 3466; 49.2% were female; M age = 14.18; SD age = 1.57). Results revealed that adolescents who were more impulsive used cigarettes and alcohol more frequently, and that this effect was partially mediated by lower resilience and more depressive symptoms. It is noteworthy that there were gender differences in this mediating effect, with the effect of impulsivity on cigarette use for girls being mediated by resilience and depressive symptoms, whereas this statistically significant association was not identified for boys. These findings show how adolescents’ impulsivity drives them to smoke and drink, and also emphasize gender as a crucial consideration for intervening with adolescents’ drinking and smoking behaviors.
Article
Full-text available
Though anxiety sensitivity (AS)—fear of anxiety-related experiences—is primarily tied to anxiety vulnerability, AS has also been prospectively associated with general negative affect and depression. Furthermore, depression has been longitudinally associated with different forms of substance use, and some AS subfactors (e.g., cognitive concerns) have been associated more consistently with depression and substance use than others. However, no previous study has investigated if longitudinal associations of AS with substance use may be mediated by depression or whether aspects of AS may be prospectively associated with substance use among adolescents. Hence, the present study tested depressive affect (the negative affective aspect of depression) as a prospective mediator of AS associations with substance use and examined longitudinal AS subfactor associations with substance use and problems. High school 9th graders (N = 2,877; Mage = 14.1 years; 55.3% female) completed self-report measures at baseline and at 6 months and 1 year later. Depressive affect mediated AS associations with subsequent alcohol, cigarette, electronic cigarette, cannabis, benzodiazepine, and opioid use. Also, AS cognitive and social concerns (vs. physical concerns) were more consistently associated with later depressive affect and substance use and problems. Current findings suggest that adolescents high in anxiety sensitivity tend to prospectively experience greater depressive affect, which in turn is related to a higher likelihood of engaging in several different forms of substance use. Thus, it is possible that interventions which target AS (particularly AS cognitive concerns) may help to treat or prevent depression and substance use among adolescents.
Article
Full-text available
Anxiety sensitivity is a transdiagnostic construct that has been associated with the etiology of emotional disorders, especially panic and other anxiety disorders. Although it is well known that, for the adult population, the factor structure of anxiety sensitivity includes three separate facets (physical, cognitive, and social concerns), the facets of anxiety sensitivity for adolescents have not yet been established. The main goal of the present study was to examine the factor structure of the Spanish version of the Childhood Anxiety Sensitivity Index (CASI). A large sample of nonclinical adolescents completed the Spanish version of the CASI in school settings (N = 1655; aged 11–17 years; 800 boys and 855 girls). Exploratory and confirmatory factor analyses of the full scale (CASI-18) indicate that a three first-order factor solution represents appropriately the three anxiety sensitivity facets previously defined for the adult population. The 3-factor structure had a better fit and was more parsimonious than a 4-factor solution. Results also indicate that the 3-factor structure remains invariant across genders. Girls scored significantly higher than boys on the total scale and on all three dimensions of anxiety sensitivity. In addition, the present study provides information concerning normative data for the scale. The CASI holds promise as a useful tool for assessing general and specific facets of anxiety sensitivity. It could be helpful for the assessment of this construct in clinical and preventive settings. The limitations of the study and suggestions for further research are outlined.
Article
The conditions and populations for which anxiety sensitivity (AS; i.e., the tendency to interpret unpleasant physiological sensations as dangerous) relates to adolescent alcohol use is unclear. This study tested latent-variable cross-lagged panel modeling of AS-alcohol relations in a racially/ethnically heterogenous longitudinal youth cohort (N=3,396; 53.4% female, 45.8% Latinx) assessed annually across high school. Anxiety and race/ethnicity were tested as mediators and moderators, respectively, of AS-alcohol associations. AS prospectively predicted alcohol problems (β’s=0.05-0.07) but not alcohol consumption (β’s=0.02-0.04) across high school. Alcohol problems predicted AS at the end (β=0.09) but not beginning (β’s=0.01-0.03) of high school and alcohol consumption predicted lower AS at the beginning (β=-0.06) but not end (β’s=-0.02-0.01) of high school. Anxiety mediated AS’s predictive effects on alcohol problems (βindirect’s=0.01, 95% CI [0.003, 0.03]) across high school. Race/ethnicity did not moderate interrelations of AS, anxiety, and alcohol outcomes. These findings lend support to the risk factor model of AS on alcohol problems (but not consumption) through anxiety and the scar/complication model whereby alcohol problems may exacerbate AS later in adolescence. Addressing reciprocal risk processes between AS, anxiety, and alcohol problems warrant consideration in adolescent behavioral health promotion for various racial/ethnic populations.
Article
Full-text available
This study examined the linkages between negative cognitive errors and anxiety in a sample of children and adolescents referred for anxiety problems (N = 251). The Children's Negative Cognitive Error Questionnaire was used to assess cognitive errors and several self-report measures of anxiety were used to examine the specific linkages between cognitive errors and different aspects of children's anxiety phenomenology. Results indicated that each of the measures of anxiety (i.e., trait anxiety, manifest anxiety, and anxiety sensitivity) were significantly related to each of the cognitive errors examined (i.e., catastrophizing, overgeneralization, personalizing, and selective abstraction). The cognitive errors of catastrophizing, overgeneralization, and personalizing were still correlated with trait anxiety, anxiety sensitivity, and manifest anxiety while controlling for children's level of depression. In addition, the correlation between selective abstraction and anxiety sensitivity was still evident while controlling for children's level of depression. Stepwise regression analyses indicated that overgeneralization was the strongest predictor of trait anxiety, catastrophizing and personalizing were the strongest predictors of anxiety sensitivity and manifest anxiety, and overgeneralization and selective abstraction were the strongest predictors of depression. Results also indicated that age moderated the relation between some types of cognitive errors and anxiety. The results are discussed with respect to the development of cognitive models of anxiety in youth, future research directions, and potential treatment implications.
Article
Full-text available
Latent variable models exist with continuous, categorical, or both types of latent variables. The role of latent variables is to account for systematic patterns in the observed responses. This article has two goals: (a) to establish whether, based on observed responses, it can be decided that an underlying latent variable is continu- ous or categorical, and (b) to quantify the effect of sample size and class propor- tions on making this distinction. Latent variable models with categorical, continu- ous, or both types of latent variables are fitted to simulated data generated under different types of latent variable models. If an analysis is restricted to fitting con- tinuous latent variable models assuming a homogeneous population and data stem from a heterogeneous population, overextraction of factors may occur. Similarly, if an analysis is restricted to fitting latent class models, overextraction of classes may occur if covariation between observed variables is due to continuous factors. For the data-generating models used in this study, comparing the fit of different explor- atory factor mixture models usually allows one to distinguish correctly between categorical and/or continuous latent variables. Correct model choice depends on class separation and within-class sample size. Starting with the introduction of factor analysis by Spearman (1904), different types of latent variable models have been developed in various areas of the social sciences. Apart from proposed estimation methods, the most obvious differences between these early latent variable models concern the assumed distribution of the
Article
Full-text available
The literature on the assessment of anxiety sensitivity (AS) in children and adolescents is reviewed. Following a discussion of theoretical and developmental issues relevant to AS and anxiety disorders, the construct validity of measures of childhood AS is reviewed using J. Loevinger's (1957) threefold framework. The collective evidence suggests mixed support at present for the validity of the clinical assessment of AS in youths, owing to (a) a limited number of adequate measures, (b) insufficient data on construct validity, and (c) a deficiency of knowledge regarding the potential influence of developmental factors on AS and anxiety disorders. Limitations of the literature are summarized, and suggestions for future research are provided. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
Full-text available
The Anxiety Sensitivity Index (ASI) is one of the most widely used measures of the construct of anxiety sensitivity. Until the recent introduction of a hierarchical model of the ASI by S. O. Lilienfeld, S. M. Turner, and R. G. Jacob (1993), the factor structure of the ASI was the subject of debate, with some researchers advocating a unidimensional structure and others proposing multidimensional structures. In the present study, involving 432 outpatients seeking treatment at an anxiety disorders clinic and 32 participants with no mental disorder, the authors tested a hierarchical factor model. The results supported a hierarchical factor structure consisting of 3 lower order factors and 1 higher order factor. It is estimated that the higher order, general factor accounts for 60% of the variance in ASI total scores. The implications of these findings for the conceptualization and assessment of anxiety sensitivity are discussed. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
The Anxiety Sensitivity Index (ASI) is one of the most widely used measures of the construct of anxiety sensitivity. Until the recent introduction of a hierarchical model of the ASI by S. O. Lilienfeld, S. M. Turner, and R. G. Jacob (1993), the factor structure of the ASI was the subject of debate, with some researchers advocating a unidimensional structure and others proposing multidimensional structures. In the present study, involving 432 outpatients seeking treatment at an anxiety disorders clinic and 32 participants with no mental disorder, the authors tested a hierarchical factor model. The results supported a hierarchical factor structure consisting of 3 lower order factors and 1 higher order factor. It is estimated that the higher order, general factor accounts for 60% of the variance in ASI total scores. The implications of these findings for the conceptualization and assessment of anxiety sensitivity are discussed. (PsycINFO Database Record (c) 2012 APA, all rights reserved)