Korean Anxiety Sensitivity Index—Revised: its factor structure, reliability, and validity in clinical and nonclinical samples

DEPRESSION AND ANXIETY 24:331–341 (2007)

Research Article

KOREAN ANXIETY SENSITIVITY INDEX—REVISED:ITS FACTOR STRUCTURE, RELIABILITY, AND VALIDITY

IN CLINICAL AND NONCLINICAL SAMPLES

Young-Jin Lim, M.A., Bum-Hee Yu, M.D., Ph.D., and Ji-Hae Kim, Ph.D.�

The factor structure and convergent and discriminant validity of the AnxietySensitivity Index—Revised (ASI-R) were examined in a community sample, astudent sample, and patients with panic disorder in Korea. Results from aconfirmatory factor analysis (CFA) comparing our data to factor solutionscommonly reported as representative of European American populationsindicated a poor fit. A subsequent exploratory factor analysis (EFA) indicatedthat a four-factor solution provided the best fit. Correlations between the ASI-Rand anxiety measures were moderately high, providing evidence of convergentvalidity. Implications for assessment with Koreans are discussed. Depression andAnxiety 24:331–341, 2007. & 2006 Wiley-Liss, Inc.

Key words: anxiety sensitivity; factor structure; Anxiety Sensitivity Index—Revised

INTRODUCTIONAnxiety sensitivity (AS) refers to the fear of anxiety-related sensations that arises from beliefs that thesesymptoms have harmful physical, psychological, orsocial consequences [Reiss and McNally, 1985]. AS is adispositional trait that amplifies fear and other anxietyreactions, and plays an important role in the etiologyand maintenance of anxiety disorders, particularlypanic disorder [Reiss, 1991].

An intense debate has raged in the AS literatureconcerning the factor structure of AS. However, aconsensus may be realized soon, because there isconverging evidence that the Anxiety Sensitivity Index[ASI; Reiss et al., 1986], the most commonly usedmeasure of AS, is hierarchically structured, consistingof multiple, lower order factors: (1) fear of somaticsensations, (2) fear of cognitive dyscontrol, and (3) fearof publicly observable anxiety symptoms, all of whichload on a single, higher order factor [see Zinbarg et al.,1999, for a review]. The factor-analytic structure of ASis a concept that allows researchers to study the basicmechanisms of AS [Taylor and Cox, 1998a], becausedistinct factors may correspond to distinct mechanisms[Cattell, 1978]. Furthermore, it is possible that theseindividual mechanisms have distinct causes that lead tospecific anxiety reactions.

Most factor-analytic studies are based on the ASI.However, because the ASI contains a relatively smallnumber of items, most of which measure fears ofsomatic sensations [e.g., Stewart et al., 1997], the scaleis too abbreviated to measure adequately the major ASfactors. For example, the ASI has too few itemsdesigned to ascertain whether the factor labeled ‘‘fearof somatic symptoms’’ consists of one or multiplefactors, such as fears of cardiac symptoms or fears ofgastrointestinal symptoms [Taylor and Cox, 1998a].Furthermore, some ASI items do not target any specificfactor. For example, it is unclear which factor isassessed by the item: ‘‘It is important for me to stayin control of my emotions.’’

Published online 13 October 2006 in Wiley InterScience (www.

interscience.wiley.com).

DOI 10.1002/da.20210

Received for publication 17 May 2005; Revised 20 December

2005; Accepted 7 February 2006

�Correspondence to: Ji-Hae Kim, Ph.D., Department of Psychia-

try, Samsung Medical Center, Seoul 135-710, Korea.

E-mail: [email protected]

Department of Psychiatry, Samsung Medical Center,

Sungkyunkwan University School of Medicine, Seoul, Korea

rr 2006 Wiley-Liss, Inc.

Thus, the Anxiety Sensitivity Index—Revised [ASI-R; Taylor and Cox, 1998b] was developed to measuremore comprehensively the lower order factors of AS.The 36-item ASI-R retains the same instructions andresponse format as the ASI, and contains 10 of the ASI’soriginal 16 items. Drawing on domains identified in theASI factor-analytic literature, the authors constructedthe ASI-R to measure fears of cardiovascular, respira-tory, gastrointestinal, publicly observable, dissociativeand neurological, and cognitive dyscontrol anxietysymptoms. In a study of 155 psychiatric outpatientsusing the ASI-R, Taylor and Cox [1998b] identifiedfour lower order AS factors (in addition to a higherorder, general AS factor): (1) fear of respiratorysymptoms, (2) fear of publicly observable anxietyreactions, (3) fear of cardiovascular symptoms, and (4)fear of cognitive dyscontrol.

Some researchers have examined the ASI factorstructure in non-English speakers and U.S. ethnicgroups, and have demonstrated differences in thenumber and item mix of the factors. A factor analysisin Native Americans [Zvolensky et al., 2001] revealed athree-factor structure similar to that found in previousinvestigations of Western populations [e.g., Zinbarget al., 1999]. A factor analysis in Native Americans[Norton et al., 2004] and a Spanish [Sandin et al., 1996]population revealed a one-factor solution to beoptimal, and in a study of African Americans, a four-factor solution [Carter et al., 1999]: mental incapacita-tion fear, unsteadiness fear, emotional control con-cerns, and cardiovascular concerns. A study of a PuertoRican population revealed a two-factor solution asoptimal, in which one factor—comprising all but twoof the ASI items—represented fear of physical andmental symptoms, and the other factor, comprising twoitems, represented fear of loss of emotional control[Cintron et al., 2005]. In a study in Cambodianrefugees that used an Augmented ASI (the ASIsupplemented with a nine-item addendum that assessesadditional Cambodian concerns about anxiety-relatedsensations), Hinton et al. [2005, 2006] found both a3-factor solution: weak heart concerns, control concerns,and social concerns, and a four-factor solution: weakheart concerns, control concerns, wind attack concerns,and social concerns. In an examination of the ASI-R insix Western countries (American–European samples),Zvolensky et al. [2003] found a two-factor solution:fear of somatic symptoms (cardiovascular, respiratory,and gastrointestinal) and of publicly observable arousalsensations and fear of cognitive dyscontrol. However,Zvolensky et al. [2003] hypothesized that if they hadexamined a non-Western group, for example, anAfrican or Asian sample, different factors might haveemerged.

Based on the existing research in Caucasians, anxietysensitivity may be a vulnerability factor for panic andother typical, as well as culturally specific, psycho-pathology in Koreans. Before researchers can begin toaddress these issues, it is first necessary to evaluate the

psychometric properties of the ASI-R in Korean peopleto inform researchers about the parameters of theanxiety sensitivity construct in this group [Zvolenskyet al., 2003]. Toward this end, our goal in this study wasto (1) determine the factor structure of the ASI-Ramong Koreans, and (2) investigate the psychometricproperties of the ASI-R in this group.

METHODS—STUDY 1Study 1 was conducted to investigate the factor

structure and psychometric properties of ASI-R in aKorean community sample.

PARTICIPANTS

The study population comprised a communitysample of 245 individuals from the Seoul metropolitancommunity. These participants were solicited viaadvertisements requesting volunteers for various psy-chological and medical assessments. No data areavailable on the clinical history of these participants.Participants ranged in age from 18 to 71 years, and51% were female (mean age 5 38.47 years,SD 5 10.70). All participants provided written in-formed consent.

MEASURES

Anxiety Sensitivity Index—Revised. The ASI-R[Taylor and Cox, 1998b] is a 36-item, expanded versionof the original ASI [Reiss et al., 1986]. Respondentsindicated their level of agreement with each item on ascale ranging from very little (coded as 0) to very much(coded as 4). Total scores range from 0 to 144. Theinternal consistency coefficient of the Korean versionof the ASI-R is .92 [Kim et al., 2004], and test–retestreliability (r 5 .82) is for a 3-week period [Kim et al.,2004].

Anxiety Sensitivity Index. The ASI [Reiss et al.,1986] is a 16-item questionnaire in which respondentsindicated their level of agreement with each item on ascale ranging from very little (coded as 0) to very much(coded as 4). The internal consistency coefficient of theKorean version of the ASI is .89 [Won et al., 1995], andtest–retest reliability (r 5 .90) is for a 2-week period[Won et al., 1995].

Beck Depression Inventory (BDI). The BDI [Beckand Steer, 1984] is a well-performing, 21-item, self-report questionnaire designed to assess and evaluate thefrequency of depressive symptoms over a 1-weekperiod. We administered a Korean version of the BDI[Lee and Song, 1991], which has demonstrated goodpsychometric properties. The internal consistencycoefficient of the Korean version of the BDI is .92[Lee and Song, 1991].

Beck Anxiety Inventory (BAI). The BAI [Becket al., 1988] is a well-performing, 21-item, self-reportquestionnaire designed to assess and evaluate thefrequency of anxiety symptoms over a 1-week period.

332 Lim et al.

Depression and Anxiety DOI 10.1002/da

We administered a Korean version of the BAI [Kwon,1992] that has demonstrated good psychometricproperties. The internal consistency coefficient of theKorean version of the BAI is .93, with test–retestreliability at r 5 .84 [Kwon, 1992].

PROCEDURE

Participants were requested to fill out each measurethemselves (paper and pencil). Some participantsvisited the clinic to fill out a battery of questionnaires;others were delivered the questionnaires at home andrevisited the clinic to submit them. Administration ofthe questionnaire took approximately 30 minutes tocomplete.

DATA ANALYSES

Prior to analysis, data were evaluated for possibledepartures from normality. We used the Kolomogor-ov–Smirnov test to examine skewness and kurtosisof each variable distribution. The results suggested thatseveral of the indicators departed from normaldistribution (e.g., for ASI-R item 5, skewness 5 4.20and kurtosis 5 21.96). Due to the non-normality ofsome indicators, we conducted latent variable analysesusing robust maximum likelihood (MLM) in Mplus2.02 [Muthen and Muthen, 2002].

ASSESSMENT OF MODEL FIT

We used four types of fit indices in this study.The goodness of fit of the CFA models was evaluatedusing the root-mean-square error of approximation[RMSEA; Steiger, 1990], the Tucker–Lewis index[TLI; Tucker and Lewis, 1973], and the ComparativeFit Index [CFI; Bentler, 1990]. We evaluated thegoodness of fit of the EFA models using the RMSEAand the root-mean-square residual (RMSR). Each typeof incremental fit index used in this study is based ona different rationale and describes somewhat differentaspects of fit [e.g., Maruyama, 1998]. An acceptablemodel fit was defined as follows: RMSEA (r.08), CFI(Z.90), TLI (Z.90), and RMSR (r.05). In addition, todetermine the internal consistency reliability of theASI-R total scale and subscales, we used Cronbach’s aand examined item–total correlations, with thresholdsof a at or above .70, and item–total correlationsexceeding the minimum acceptable value of .30[Nunnally and Bernstein, 1994]. Last, to explore therelationship between the ASI-R and the remainingmeasures, we used Spearman’s r correlations. Giventhe number of correlations, P values were set at .006 tocontrol for experiment-wise error (the Bonferroniadjustment was used, so an initial a of .05 was dividedby the number of measures, or .05/8).

RESULTS AND DISCUSSION—STUDY 1

CONFIRMATORY FACTOR ANALYSIS

The first set of analyses examined the one-, two-,three-, and four-factor models of the ASI-R, using thebaseline models examined by Taylor and Cox [1998b]and Zvolensky et al. [2003]. The two-factor modelcomprised fear of somatic sensations and fear of social-cognitive concerns [Zvolensky et al., 2003]. The three-factor model broke down fear of social-cognitiveconcerns into social concerns and cognitive concerns.Although Taylor and Cox [1998b] and Zvolensky et al.[2003] did not test a three-factor model, we tested athree-factor model in which all of Taylor and Cox’s[1998b] cardiovascular and respiratory symptom itemswere classified as somatic sensations. The four-factormodel comprised (1) fear of respiratory symptoms,(2) fear of publicly observable anxiety reactions, (3) fearof cardiovascular symptoms, and (4) fear of cognitivedyscontrol. These models proved to be a poor fit(Table 1).

EXPLORATORY FACTOR ANALYSIS

Given the lack of good fit to the previous models, wetested the structure of our data using exploratory factoranalysis (EFA). Because ASI-R subscales are generallymoderately correlated, we used an oblique (promax)rotation. We evaluated the number of factors to retainusing (1) Kaiser’s [1961] eigenvalue 4 1 factor extractionrule, (2) the scree test [Cattell, 1966], (3) the inter-pretability of the factor structures [Gorsuch, 1983], and(4) model fit indices [Muthen and Muthen, 2002]. Wealso used Thurstone’s [1947] criteria, which include(1) a minimum number of items with salient loadings(Z0.30) on more than one factor, (2) a minimumnumber of items that do not have salient loadings onany factor, and (3) each factor is well-defined (i.e., hasthree or more salient loadings per factor).

Six factors possessed eigenvalues greater than one(13.53, 3.38, 2.48, 2.14, 1.14, and 1.06). From the screetest, we estimated that one to four factors werenecessary to explain the data, but the one-, two-, andthree-factor models were not sufficient to explainthe data (Table 2). An acceptable model fit was foundfor a four-factor solution (w2 (492) 5 906.047, RMSEA 5.059, RMR 5 .042). In addition, Thurstone’s criteriaand Gorsuch’s criterion showed that the four-factorsolution had the best simple structure.

Table 3 shows the rotated factor loadings for thisfour-factor solution. The four-factor solutionaccounted for 59.8% of the ASI-R item variance. Thefour-factor solution had (1) a relatively small number ofcomplex items (only two items with salient loadings onmore than one factor); (2) a small number of hyper-plane items (zero items with no salient loading on anyfactor); and (3) a relatively high number of salientloadings per factor (i.e., Factor I had 9, Factor II had

333Research Article: Anxiety Sensitivity Index—Revised


13, Factor III had 10, and Factor IV had 6). Takingsalient loadings as thoseZ.30, Factor I pertains to fearof respiratory symptoms; Factor II pertains to fear

of physical diseases; Factor III pertains to fear ofpublicly observable anxiety reactions; and Factor IVpertains to fear of cognitive dyscontrol. These findingsgenerally replicated those reported by Taylor and Cox[1998b].

RELIABILITY AND ITEM-LEVEL ANALYSES

The mean ASI-R total score was 23.82 (SD 5 18.50).ASI-R total scores for women (M 5 23.12, SD 5 17.91)were similar to those for men (M 5 24.66, SD 5 19.16).Internal consistency tests gave a Cronbach’s a of .95 forthe entire scale, with an a of .90 (9 items) for the fear ofrespiratory symptoms subscale, .93 (11 items) for thefear of physical diseases subscale, .88 (9 items) for thefear of publicly observable anxiety reactions subscale,and .87 (5 items) for the fear of cognitive dyscontrolsubscale. Based on the criterion of .30 as an acceptablecorrected item–total correlation [Nunnally andBernstein, 1994], all 36 items performed adequately(range 5 .45–.69).

CONVERGENT AND DISCRIMINANTVALIDITY

To examine the relationship between the ASI-R andthe convergent and discriminant validity measures, we

TABLE 1. Goodness-of-fit indices for ASI-R models: confirmatory factor analysis

Model w2 df TLI CFI RMSEA

Community sample (N 5 245)One factor 2,347.183 594 .552 .578 .110Zvolensky et al. [2003]’s two factor 2,124.599 593 .608 .631 .103Taylor & Cox [1998b]’s correlated three factor 1,835.406 591 .680 .700 .093Taylor & Cox [1998b]’s hierarchical three factor 1,835.406 591 .680 .700 .093Taylor & Cox [1998b]’s correlated four factor 1,352.578 588 .803 .816 .073Taylor & Cox [1998b]’s hierarchical four factor 1,352.835 590 .804 .816 .073

Student sample (N 5 503)One factor 3,096.599 594 .523 .550 .092Zvolensky et al. [2003]’s two factor 2,565.557 593 .624 .646 .081Taylor & Cox [1998b]’s correlated three factor 1,960.257 591 .738 .754 .068Taylor & Cox [1998b]’s hierarchical three factor 1,960.257 591 .738 .754 .068Taylor & Cox [1998b]’s correlated four factor 1,666.835 588 .792 .806 .060Taylor & Cox [1998b]’s hierarchical four factor 1,692.273 590 .789 .802 .061Correlated four factor of Study 1 1,737.256 586 .778 .793 .062Hierarchical four factor of Study 1 1,768.837 588 .773 .788 .063

Patient sample (N 5 255)One factor 3,090.907 594 .562 .587 .128Zvolensky et al. [2003]’s two factor 2,561.651 593 .654 .675 .114Taylor & Cox [1998b]’s correlated three factor 2,072.025 591 .739 .755 .099Taylor & Cox [1998b]’s hierarchical three factor 2,072.025 591 .739 .755 .099Taylor & Cox [1998b]’s correlated four factor 1,765.594 588 .791 .805 .089Taylor & Cox [1998b]’s hierarchical four factor 1,788.153 590 .788 .802 .089Correlated four factor of Study 1 1,771.882 586 .789 .804 .089Hierarchical four factor of Study 1 1,802.468 588 .785 .799 .090Correlated four factor of Study 2 1,521.124 521 .808 .822 .087Hierarchical four factor of Study 2 1,543.390 523 .805 .818 .087

TLI, Tucker-Lews Index; CFI, Comparative Fit Index; RMSEA, root mean square error of approximation.

TABLE 2. Goodness-of-fit indices for ASI-R models:exploratory factor analysis

Model w2 df RMSEA RMSR

Community sample (N 5 245)One factor 1,800.842 594 .091 .115Two factor 1,507.622 559 .083 .084Three factor 1,166.152 525 .071 .063Four factor 906.047 492 .059 .042

Student sample (N 5 503)One factor 2,825.691 594 .086 .101Two factor 2,079.076 559 .074 .075Three factor 1,620.979 525 .064 .062Four factor 1,271.224 492 .056 .048

Patient sample (N 5 255)One factor 2,725.711 594 .119 .110Two factor 2,106.571 559 .104 .081Three factor 1,672.436 525 .093 .061Four factor 1,264.557 492 .078 .047

RMSEA, root mean square error of approximation; RMSR, rootmean square residual.

334 Lim et al.


calculated Spearman’s correlations (Table 4). Becausenone of the measures was normally distributed(Kolomogorov–Smirnov test), we conducted the cor-relation analyses using Spearman’s correlations. TheASI-R significantly correlated with all measures ofanxiety and depression. As expected, significant corre-lations between the ASI-R and the ASI and BAIprovided strong evidence for convergent validity. It washypothesized that the correlation existing between theASI-R and the BDI would be weaker than thecorrelation existing between the ASI-R and theBAI. We used the method suggested by Meng et al.[1992] to test the difference in magnitude betweencorrelation coefficients in this study. This predictionwas not supported by testing in accordance with themethod suggested by Meng et al. (Z 5�1.59, notsignificant).

METHODS—STUDY 2Our aims in Study 2 were to (1) determine whether

the ASI-R’s factor structure, as demonstrated in Study1, would generalize to a different sample of Koreancollege students and (2) further explore its concurrentand discriminant validity. The measures were identicalto those used in Study 1.

PARTICIPANTS

Participants were 503 college students recruited fromintroductory psychology courses at a University inSeoul, between 18 to 42 years of age; 74% of themwere female (mean age 5 22.44 years, SD 5 2.69). Nodata were available on the clinical history of thesestudents.

TABLE 3. Promax rotated loadings (four factor model: community sample)a

Item Factor I Factor II Factor III Factor IV

1. When I feel like I’m not getting enough air I get scared that I might suffocate. .673 �.023 .064 �.0342. Smothering sensations scare me. .419 .164 .122 �.0323. It scares me when I become short of breath. .778 .004 .053 �.0604. When my chest feels tight, I get scared that I won’t be able to breathe properly. .907 �.022 �.004 �.0605. It scares me when I feel faint. .778 �.058 �.182 .2116. When my throat feels tight, I worry that I could choke to death. .843 .006 �.077 .0387. It scares me when my heart beats rapidly. .611 .046 .146 �.0018. When my breathing becomes irregular, I fear that something bad will happen. .550 .141 .175 .0489. It scares me when I feel ‘‘shaky’’ (trembling). .272 .013 .468 .152

10. When I have trouble swallowing, I worry that I could choke. .333 .204 �.028 .16711. It frightens me when my surroundings seem strange or unreal. .116 .097 .347 .20712. It scares me when my body feels strange or different in some way. .121 .335 .385 .00413. It is important for me not to appear nervous. �.029 �.034 .699 �.01414. I believe it would be awful to vomit in public. �.087 .168 .580 �.17615. I think it would be horrible for me to faint in public. �.053 .134 .606 �.13616. I worry that other people will notice my anxiety. �.008 �.044 .811 .05717. When I tremble in the presence of others, I fear what people might think of me. �.016 �.153 .937 �.01618. When I begin to sweat in a social situation, I fear people will think negatively of me. .044 �.011 .714 .10619. It scares me when I blush in front of people. �.022 �.068 .699 .16520. When I feel a strong pain in my stomach, I worry it could be cancer. .061 .580 .209 �.06921. When my head is pounding I worry I could have a stroke. �.093 .846 �.072 �.06322. When my heart is beating rapidly, I worry that I might have a heart attack. .200 .787 �.111 �.07623. When my face feels numb, I worry that I might be having a stroke. .009 .764 �.048 .02924. When I feel pain in my chest, I worry that I’m going to have a heart attack. .172 .871 �.147 �.08325. When I feel dizzy, I worry there is something wrong with my brain. �.016 .769 �.026 .00726. When my stomach is upset, I worry that I might be seriously ill. .017 .712 .065 �.00327. When I notice my heart skipping a beat, I worry seriously wrong with me. .037 .787 .064 �.07328. When I get diarrhea, I worry that I might have something wrong with me. �.144 .694 .050 .07329. It scares me when I am nauseous. �.139 .621 .134 .09730. It scares me when I feel tingling or prickling sensations in my hands. .042 .543 .064 .17531. When I feel ‘‘spacey’’ or spaced out, I worry that I may be mentally ill. �.176 .594 �.102 .40932. When my thoughts seem to speed up, I worry that I might be going crazy. �.060 .091 .016 .70833. When I have trouble thinking clearly, I worry there is something wrong with me. .009 .168 .015 .62834. When I cannot keep my mind on a task, I worry that I might be going crazy. .078 �.036 �.012 .85135. It scares me when I am unable to keep my mind on a task. .062 �.159 .165 .70436. When my mind goes blank, I worry there is something terribly wrong with me. .042 �.027 .001 .745

aBold numbers denote salient loadings (Z0.30).



PROCEDURE

We first obtained informed consent, then studentscompleted the battery in a classroom setting, duringclass time. Researchers were available during ques-tionnaire completion to answer any questions.



The alternative conceptual models of the ASI-R usedin Study 1 and the models derived from the EFA resultin Study 1 are included. Each model produced a poorfit according to all criteria except the RMSEA,suggesting that the models did not adequately accountfor the data (Table 1).


Because CFA results suggested that previous factorsolution models did not provide a good fit for thepresent data, we conducted an EFA with obliquerotation. An examination of the scree plot andeigenvalues greater than one (in descending order:9.81, 3.09, 2.23, 1.91, 1.51, 1.28, 1.13, 1.05) initiallysuggested the presence of one to five factors. Table 2shows that the one-, two-, and three-factor models didnot sufficiently explain the data, because these modelsyielded poor fit indices. According to Thurstone’scriteria, the four-factor model had fewer items withsalient loadings (Z.30) on more than one factor, andfewer items that did not have salient loadings on anyfactor other than the five-factor model. The adequacyof the four-factor solution was evaluated throughconsideration of simple structure [Gorsuch, 1983].

Loadings of the variances on each factor are shown inTable 5. The four-factor solution accounted for 47.3%of the ASI-R item variance. As can be seen in Table 5,the pattern of items showing salient loadings on each ofthese four factors was very similar to that reported byTaylor and Cox [1998b] and Study 1 (see Table 3). Thefour-factor solution had (1) a relatively small number of

complex items (zero items with salient loadings onmore than one factor); (2) a small number of hyper-plane items (two items with no salient loading on anyfactor); and (3) a relatively high number of salientloadings per factor. Factor I had 11 items with salient(Z.30) loadings and assessed fear of respiratorysymptoms. Factor II had seven items with salientloadings and was labeled ‘‘fear of publicly observableanxiety reactions.’’ Factor III contained six items withsalient loadings and was labeled ‘‘fear of cognitivedyscontrol.’’ Factor IV comprised 10 items with salientloadings and was labeled ‘‘fear of physical diseases.’’


The mean ASI-R total score was 27.64 (SD 5 16.16).ASI-R total scores for women (M 5 29.05, SD 5 16.16)were higher than those for men (M 5 23.63,SD 5 15.55; Mann–Whitney test, Po.001). MeanASI-R scores were largely similar to those reportedfor European American college students (M 5 25.4,SD 5 18.7) [Deacon et al., 2003]. The entire scaledemonstrated excellent internal consistency (a5 .92).Cronbach’s a was .81 (11 items) for the fear ofrespiratory symptoms subscale, .85 (7 items) for thefear of publicly observable anxiety reactions subscale,.86 (6 items) for the fear of cognitive dyscontrolsubscale, and .86 (10 items) for the fear of physicaldiseases subscale. Each item had an adequate correcteditem–total correlation (range 5 .32–.59).


Table 6 shows that the ASI-R had large correlationswith the ASI and BAI, and moderately large correla-tions with the BDI. The correlation between the ASI-Rand the BAI was significantly higher than that betweenthe ASI-R and the BDI, providing some evidence ofdiscriminant validity (Z 5 4.314, Po0.01).

TABLE 4. Zero-correlations between the factors of ASI-R, ASI, BAI, and BDI (N 5 245)a

ASI-R Total score ASI-R Factor I ASI-R Factor II ASI-R Factor III ASI-R Factor IV ASI BAI

ASI-R Factor I .75ASI-R Factor II .88 .58ASI-R Factor III .86 .58 .60ASI-R Factor IV .69 .46 .62 .48ASI .92 .69 .79 .79 .71BAI .45 .38 .43 .34 .36 .51BDI .53 .35 .49 .40 .51 .46 .57

aAll correlations are significant at the .001 level (two-tailed).Note. ASI-R Factor I, fear of respiratory symptoms; ASI-R Factor II, fear of physical diseases; ASI-R Factor III, fear of publicly observablesymptoms; and ASI-R Factor IV, fear of cognitive dyscontrol.ASI, Anxiety Sensitivity Index; BAI, Beck Anxiety Inventory; BDI, Beck Depression Inventory.

336 Lim et al.



ASI-RTotal score

ASI-RFactor I

ASI-RFactor II

ASI-RFactor III

ASI-RFactor IV ASI BAI


aAll correlations are significant at the .001 level (two-tailed).Note. ASI-R Factor I, fear of respiratory symptoms; ASI-R Factor II, fear of publicly observable symptoms; ASI-R Factor III, fear of cognitivedyscontrol; and ASI-R Factor IV, fear of physical diseases.ASI, Anxiety Sensitivity Index; BAI, Beck Anxiety Inventory; BDI, Beck Depression Inventory.

TABLE 5. Promax rotated loadings (four factor model)a

Student sample Patients with panic disorder

Item Factor I Factor II Factor III Factor IV Factor I Factor II Factor III Factor IV

1 .632 �.172 �.059 .013 .789 �.039 �.083 .0352 .426 .032 .103 �.069 .295 .284 .091 .0743 .623 .067 �.009 �.129 .911 �.095 .009 �.0814 .735 .014 �.018 �.071 .944 �.053 �.111 �.0555 .549 �.087 .019 �.077 .508 .140 �.004 .1966 .613 �.118 �.090 .087 .771 �.050 �.028 .0217 .542 .190 �.114 �.034 .590 .276 .098 �.1408 .523 .116 �.028 .093 .628 .255 .076 �.0409 .313 .261 .024 .007 .325 .053 .377 .057

10 .356 �.026 �.017 .227 .572 .020 �.053 .13511 .339 .038 .236 .101 .228 .079 .111 .44912 .251 .078 .188 .248 .263 .378 .017 .19113 .002 .705 .000 .004 �.006 .018 .758 .05014 .200 .331 .077 �.052 .105 �.027 .539 .01215 .155 .405 .045 �.067 �.030 .076 .533 .15016 .029 .791 .045 �.075 �.050 .000 .861 .08117 �.038 .825 �.031 .075 �.085 �.027 .992 �.06118 �.097 .720 .005 .078 .023 .011 .776 �.06519 .025 .625 .061 .117 �.015 �.039 .705 .06120 �.135 .138 �.097 .727 �.075 .636 .045 �.00621 .010 �.033 �.062 .638 �.132 .835 �.074 .04722 .245 �.036 �.058 .498 .156 .889 �.055 �.19123 .202 �.038 .000 .476 �.144 .814 .000 .04224 .241 �.007 �.068 .522 .155 .838 .002 �.16125 .042 �.068 .100 .624 �.073 .757 .004 .12526 �.177 .052 �.044 .835 �.012 .639 �.039 .08827 .251 .000 .011 .534 .219 .631 .097 �.02028 �.109 .132 .026 .515 �.023 .414 .077 .08829 .072 �.004 .173 .461 .228 .183 .005 .26630 .237 �.066 .214 .250 .029 .382 .005 .29131 �.028 �.110 .592 .155 .023 .156 �.057 .74532 �.018 �.022 .665 .017 .026 �.037 .019 .84833 .014 .027 .780 �.001 �.033 .068 .091 .75934 �.088 .025 .799 �.004 �.064 �.016 .042 .90435 �.011 .174 .679 �.095 �.036 �.164 .241 .78136 .003 .040 .788 �.051 �.007 .073 .006 .825

aBold numbers denote salient loadings (Z0.30).



METHODS—STUDY 3In Study 3, we sought to (1) test whether the ASI-R’s

factor structure as demonstrated in Study 1 and Study 2would fit a different sample of Korean patients withpanic disorder with or without agoraphobia, and (2)report its reliability and validity. The self-reportquestionnaires were exactly like those used in Study 1.

PARTICIPANTS

Clinical sample subjects (59% males; meanage 5 39.61 years, SD 5 9.90, range 5 15–66) com-prised 255 patients with panic disorder, with or withoutagoraphobia. Patients were recruited from psychiatricoutpatients attending the Samsung Medical Center inSeoul. Diagnosis was made by a psychiatrist with anunstructured interview and independently confirmedby a Master’s-level psychologist with the AnxietyDisorders Interview Schedule for DSM-IV [ADIS-IV;Brown et al., 1994]. A doctoral-level psychologistsupervised all interviews. Patients with no majormedical or psychiatric illnesses besides panic disorderwere enrolled in this study. All participants signedwritten informed consent.

PROCEDURE

Diagnostic interviews were completed before thecommencement of treatment. Information from thequestionnaires was not used to make diagnoses. Theself-report measures were completed before treatmentin most cases.



We conducted a third set of CFAs to assess the fit ofour data to the alternative models of the ASI-R used inStudy 2. The models derived from EFA results in Study2 indicated a poor fit with the current data (Table 1).


Because the alternative models produced no fitindices meeting criteria standards, we used EFA toexamine ASI-R factor structure. Although seven factorshad eigenvalues greater than 1.0 (15.10, 3.25, 2.39,1.74, 1.34, 1.10, and 1.06), the scree test indicated one-to five-factor solutions. However, the one-, two-, andthree-factor models were not sufficient to explain thedata (Table 2). According to Thurstone’s criteria, thefive-factor model had more items with salient loadings(Z.30) on more than one factor and more itemswithout salient loadings on any factor other than thefour-factor model. Using Gorsuch’s criterion, we againobtained the four-factor solution as the best choice.

Table 5 provides the factor loadings for the lowerorder, rotated, four-factor solution for the ASI-R. The

four-factor solution accounted for 62.4% of the ASI-Ritem variance. Taking salient loadings as those Z.30,the four-factor lower order solution showed excellentsimple structure [Thurstone, 1947] with a smallnumber of hyperplane items (only one item with nosalient loading on any factor), a small number ofcomplex items (only one item with salient loadings onmore than one factor), and a relatively high number ofsalient loadings per factor (i.e., Factor I had 9, Factor IIhad 11, Factor III had 8, and Factor IV had 7). Thepattern of loadings in Table 5 suggests the followingfactor labels: ‘‘fear of respiratory symptoms’’ (Factor I),‘‘fear of publicly observable anxiety reactions’’ (FactorII), ‘‘fear of physical diseases’’ (Factor III), and ‘‘fear ofcognitive dyscontrol’’ (Factor IV). As can be seen, theseresults are highly comparable to those from Taylor andCox [1998b], Study 1, and Study 2.


Mean ASI-R total scores were 54.08 (SD 5 28.6), andmarginally higher for women (M 5 57.97, SD 5 27.88)than for men (M 5 51.32, SD 5 28.91; Mann–Whitneytest, P 5.070). The internal consistency estimates ofreliability were good for the ASI-R total scale (.96), thefear of respiratory symptoms subscale (.91), the fear ofphysical diseases subscale (.92), the fear of publiclyobservable anxiety reactions subscale (.91), and the fearof cognitive dyscontrol subscale (.94). All individualscale items were positively correlated with the totalscale score (range 5 .48–.74).


Table 7 shows that ASI-R correlated with each of thesymptom measures. As would be expected, the correla-tion between the ASI-R and the BAI was greater thanthe correlation between the ASI-R and the BDI(Z 5 6.194, Po.01).

GENERAL DISCUSSIONResults from our CFAs revealed that neither solution

provided a suitable fit for the data set. Although ourEFAs extracted a four-factor solution that demon-strated good fit, the EFA does not imply that the modelis confirmed. It is possible that poor fit of our CFAs isattributable to cultural differences. However, consis-tencies in the factor structure of the EFA are evident inthe three studies and between these studies and theoriginal study by Taylor and Cox [1998b]. Thus, poorfit of our CFAs does not appear to be attributable toour use of a Korean version of the ASI-R or to culturaldifferences. Instead, it is possible that this problem ofthe ASI-R derives from the fact that the four-factorstructure of the ASI-R does not correspond to the sixfactors the scale was designed to measure. Thus, itappears that further refinements could be made byremoving some items from each scale that fail to show

338 Lim et al.


‘‘correct’’ loadings. However, it may be premature todo this, because another factor analysis (with larger andmore diverse samples) might spotlight different itemsas weak or inappropriate.

As expected, significant correlations between theASI-R and the ASI and BAI provided strong evidencefor convergent validity. The significant correlationbetween the ASI-R and the BDI suggests poordiscriminant validity. Contrary to the expectation, thecorrelation between the ASI-R and the BAI was notstronger than the correlation between the ASI-R andthe BDI in the community sample. The previousstudies consistently demonstrated that AS has specifi-city regarding its relationship with anxiety as opposedto depression [Schmidt et al., 1998; Taylor et al., 1996;Zinbarg et al., 2001]. According to the tripartite model[Clark and Watson, 1991], depression is specificallycharacterized by anhedonia, anxiety is specificallycharacterized by physiological hyperarousal, and gen-eral negative affect is a nonspecific factor that relates toboth depression and anxiety. Thus, AS may be morerelated to general negative affect or anhedonia, but lessrelated to physiological hyperarousal among Koreancommunity samples. More research is needed to betterspecify the nature of this relationship.

The results of this study have several importantimplications. The most interesting validity findingspertain to the ‘‘fear of physical diseases’’ dimension.The results of Study 1 and Study 3 were similar toTaylor and Cox’s [1998b] results. Four of the fivehighest loadings in Study 1 and the five highestloadings in Study 3 were for items assessing fear ofcardiovascular symptoms. In Study 2, only two of thefive highest loadings were for items assessing fear ofcardiovascular symptoms, but two of the highestloadings were for items assessing fear of gastrointest-inal symptoms. Therefore, we chose to label this factoras ‘‘fear of physical diseases’’ rather than fear ofcardiovascular symptoms. In addition, Cox et al.[2001] found supportive evidence that AS might notbe restricted to a fear of anxiety, but might extend to abroader catastrophic style concerning bodily symptoms

and health that go beyond anxiety symptoms per se.The fear of physical diseases, a more specific dimen-sion, might represent this catastrophic thinking style[Silverman et al., 2003].

One possible explanation for higher loading of fearof gastrointestinal symptoms in students is that thosesymptoms have a greater relative impact in that groupcompared with cardiovascular concerns in the othergroups. In fact, according to the Korean NationalStatistics Office [2000], the leading cause of death inKorea is circulatory diseases (cerebrovascular accidentand coronary heart disease), especially for Koreansover 40 years old. For Koreans under age 40 years,however, stomach cancer and circulatory disease havea similar impact.

This study also indicates that highly replicable factorsolutions were obtained in diverse samples of under-graduates, community adults, and patients with panicdisorder. Previous studies involving the original ASIsupport the conclusion that the factor structures of theASI-R do not vary appreciably across clinical andnonclinical populations [e.g., Zinbarg et al., 1997]. Wedemonstrated the same characteristics for the ASI-R.

This study, however, has several important limita-tions. First, Study 3 included only patients with panicdisorder. Therefore, we should be cautious aboutgeneralizing these findings to other clinical popula-tions, and they should be replicated with other anxietyand depressive patients. Second, only self-reportingdata were included; thus, relationships between studyvariables may have been inflated by questionnaire-specific method variance. Third, the reliability (stabi-lity) of our four-factor solutions remains to beestablished given the inconsistent factor solutionsobtained thus far in factor-analytic studies of the ASI-R. According to Guadagnoli and Velicer [1988],stability problems occur when sample sizes are small(N 5 50 to 300). Although our sample size wassufficiently large in the case of the student sample(N 5 503), the size of our community sample (N 5 245)and patient sample (N 5 255) may not have beensufficiently large to obtain a stable lower order factor


ASI-RTotal score

ASI-RFactor I

ASI-RFactor II

ASI-RFactor III

ASI-RFactor IV ASI BAI


aAll correlations are significant at the .001 level (two-tailed).Note. ASI-R Factor I, fear of respiratory symptoms; ASI-R Factor II, fear of physical diseases; ASI-R Factor III, fear of publicly observablesymptoms; and ASI-R Factor IV, fear of cognitive dyscontrol.ASI, Anxiety Sensitivity Index; BAI, Beck Anxiety Inventory; BDI, Beck Depression Inventory.



solution. Last, the sex ratio of Study 3 is unusual inKorea, where panic disorder is much more common inwomen, at least in the community [Lee et al., 1987].However, the female:male ratio in patients with panicdisorder at our clinic was almost equal. Thesedifferences may be due to cultural factors, becausewomen suffering panic attacks are considered chehada,or ‘‘feeling heavy in the stomach,’’ and are treated bytraditional medicine at home, whereas men with thesame symptoms are treated medically [Udomratn,2001]. Thus, these findings should be replicated inthe community with patients with panic disorder.

Although the ASI-R would benefit from furtherrefinement, it appears to be a promising measure thathas a strong theoretical application as a measure of AS.The ASI-R was highly internally consistent andcomposed of psychometrically acceptable items. Itappears to measure reliably four lower order factorsassessing fears of somatic, social, and cognitive anxietysymptoms. However, we relabeled the fear of cardio-vascular symptoms reported by Taylor and Cox[1998b]. Korean participants distinguished the twosomatic factors not only according to the type ofsensations they assessed (e.g. respiratory) but alsoaccording to whether experiencing somatic sensationsmay refer to being diseased. Our results suggest thatthe ASI-R is superior to the original ASI with respectto its content validity and breadth. Given that each ofthe four factors was internally consistent and had asubstantial number of items with salient factor load-ings, the ASI-R appears to assess these domains moreadequately than the ASI. Future studies of this scalemight provide a more detailed picture of the role of ASin the development and maintenance of anxietydisorders in Korea.

REFERENCESBentler PM. 1990. Comparative fit indexes in structural models.

Psychol Bull 107:238–246.Beck AT, Epstein N, Brown G, Steer RA. 1988. An inventory

for measuring clinical anxiety: Psychometric properties. J ConsultClin Psychol 56:893–897.

Beck AT, Steer RA. 1984. Internal consistencies of the original andrevised Beck Depression Inventory. J Clin Psychol 40:1365–1367.

Brown TA, DiNardo PA, Barlow DA. 1994. Anxiety disordersinterview schedule for DSM-IV. Boulder, CO: Graywind.

Carter MC, Miller O, Sbrocco T, Suchday S, Lewis EL. 1999. Factorstructure of the Anxiety Sensitivity Index among African Americancollege students. Psychol Assess 11:525–533.

Cattell RB. 1966. The scree test for the number of factors.Multivariate Behav Res 1:245–276.

Cattell RB. 1978. The scientific use of factor analysis in thebehavioral and life sciences. New York: Plenum.

Cintron JA, Carter MC, Suchday S, Sbrocco T, Gray J. 2005. Factorstructure and construct validity of the Anxiety Sensitivity Indexamong island Puerto Ricans. J Anxiety Disord 19:51–68.

Clark LA, Watson D. 1991. Tripartite model of anxiety anddepression: Psychometric evidence and taxonomic implications.J Abnorm Psychol 100:316–336.

Cox BJ, Fuentes K, Borger SC, Taylor S. 2001. Psychopathological

correlates of anxiety sensitivity: evidence from clinical interviews

and self-report measures. J Anxiety Disord 15:317–332.Deacon BJ, Abramowitz JS, Woods CM, Tolin DF. 2003. The

Anxiety Sensitivity Index — Revised: psychometric properties and

factor structure in two nonclinical samples. Behav Res Ther 41:

1427–1449.Gorsuch R. 1983. Factor analysis. Hillsdale. NJ: Erlbaum.Guadagnoli E, Velicer WF. 1988. Relation of sample size to the

stability of component patterns. Psychol Bull 103:265–275.Hinton DE, Pich V, Safren SA, Pollack MH, McNally RJ. 2005.

Anxiety sensitivity in traumatized Cambodian refugees: A discri-

minant function and factor analytic investigation. Behav Res Ther 43:

1631–1643.Hinton DE, Pich V, Safren SA, Pollack MH, McNally RJ. 2006.

Anxiety sensitivity among Cambodian refugees with panic dis-

order: A factor analytic investigation. J Anxiety Disord 20:281–295.Kaiser H. 1961. A note on Guttman’s lower bound for the number

of common factors. Multivariate Behav Res 1:249–276.Kim JH, Yu BH, Oh KS, Yang JC, Kim Y, Lee SY, Lim YJ. 2004.

A validation study of Korean Anxiety Sensitivity Index—Revised

(ASI-R). J Korean Neuropsychiatr Assoc 43:54–61.Korean National Statistics Office. 2000. Deaths and death rates by

cause. Available online http://www.nso.go.kr/Kwon SM. 1992. Differential roles of dysfunctional attitudes and

automatic thoughts in depression: An integrated model of

depression. Unpublished doctoral dissertation, University of

Queensland, Australia.Lee CK, Kwak YS, Rhee H, Kim YS, Choi JO, Lee YH. 1987. The

nationwide epidemiological study of mental disorders in Korea.

J Korean Med Sci 2:19–34.Lee YS, Song JY. 1991. Study of reliability and validity of the BDI,

the SDS, and the MMPI depression scale. Korean J Clin Psychol

10:98–113.Maruyama GM. 1998. Basics of structural equation modeling.

Thousand Oaks, CA: Sage.Meng XL, Rosenthal R, Rubin DB. 1992. Comparing correlated

coefficients. Psychol Bull 111:172–175.Muthen LK, Muthen BO. 2002. Mplus 2.02 [Computer software].

Los Angeles: Author.Norton PJ, De Coteau TJ, Hope DA, Anderson J. 2004. The factor

structure of the Anxiety Sensitivity Index among Northern Plains

Native Americans. Behav Res Ther 42:241–247.Nunnally J, Bernstein I. 1994. Psychometric theory. New York:

McGraw-Hill.Reiss S. 1991. Expectancy model of fear, anxiety, and panic. Clin

Psychol Rev 11:141–153.Reiss S, McNally RJ. 1985. Expectancy model of fear. In: Reiss S,

Bootzin RR, editors. Theoretical issues in behavior therapy. San

Diego: Academic Press. p 107–121.Reiss S, Peterson RA, Gursky DM, McNally RJ. 1986. Anxiety

sensitivity, anxiety frequency, and the prediction of fearfulness.

Behav Res Ther 24:1–8.Sandin B, Chorot P, McNally RJ. 1996. Validation of the Spanish

version of the Anxiety Sensitivity Index in a clinical sample. Behav

Res Ther 34:283–290.Schmidt NB, Lerew DR, Joiner TE. 1998. Anxiety sensitivity and the

pathogenesis of anxiety and depression: Evidence for symptom

specificity. Behav Res Ther 36:165–177.Silverman WK, Goedhart AW, Barrett P, Turner C. 2003. The facets

of anxiety sensitivity represented in the Childhood Anxiety

Sensitivity Index: Confirmatory analyses of factor models from

past studies. J Abnorm Psychol 112:364–374.

340 Lim et al.


Steiger JH. 1990. Structural model evaluation and modification: Aninterval estimation approach. Multivariate Behav Res 25:173–180.

Stewart SH, Taylor S, Baker JM. 1997. Gender differences indimensions of anxiety sensitivity. J Anxiety Disord 11:179–200.

Taylor S, Cox BJ. 1998a. Anxiety sensitivity: Multiple dimensions andhierarchic structure. Behav Res Ther 36:37–51.

Taylor S, Cox BJ. 1998b. An expanded Anxiety Sensitivity Index:Evidence for a hierarchic structure in a clinical sample. J AnxietyDisord 12:463–483.

Taylor S, Koch WJ, Woody S, McLean P. 1996. Anxiety sensitivityand depression: How are they related? J Abnorm Psychol 105:474–479.

Thurstone LL. 1947. Multiple factor analysis. Chicago: University ofChicago Press.

Tucker LR, Lewis C. 1973. A reliability coefficient for maximumlikelihood factor analysis. Psychometrika 38:1–10.

Udomratn P. 2001. Epidemiology of panic disorder. J Psychiatr AssocThailand 46:247–261.

Won HT, Park HS, Kwon SM. 1995. A study on the development of

the Korean versions of panic scales. Kor J Clin Psychol 14:95–110.Zinbarg RE, Brown TA, Barlow DH, Rapee RM. 2001. Anxiety

sensitivity, panic, and depressed mood: A reanalysis teasing

apart the contributions of the two levels in the hierarchical

structure of the Anxiety Sensitivity Index. J Abnorm Psychol 110:

372–377.Zinbarg RE, Mohlman J, Hong NN. 1999. Dimensions of anxiety

sensitivity. In: Taylor S, editor. Anxiety sensitivity: Theory,

research, and treatment of the fear of anxiety. Mahwah, NJ:

Erlbaum. p 83–114.Zvolensky MJ, Arindell WA, Taylor S, Bouvard M, Cox BJ, Stewart

SH. 2003. Anxiety sensitivity in six countries. Behav Res Ther

41:841–859.Zvolensky MJ, McNeil DW, Porter CA, Stewart SH. 2001.

Assessment of anxiety sensitivity in young American Indians and

Alaska natives. Behav Res Ther 39:477–493.



Documents

Korean Anxiety Sensitivity Index—Revised: its factor structure, reliability, and validity in clinical and nonclinical samples