The Liebowitz Social Anxiety Scale for Children and

Adolescents: An Initial Psychometric Investigation

CARRIE MASIA-WARNER, PH.D., ERIC A. STORCH, M.S., DONNA B. PINCUS, PH.D.,
RACHEL G. KLEIN, PH.D., RICHARD G. HEIMBERG, PH.D., AND MICHAEL R. LIEBOWITZ, M.D.

ABSTRACT
Objective: To examine the psychometric properties of a newly developed clinician rating scale, the Liebowitz Social
Anxiety Scale for Children and Adolescents (LSAS-CA). Method: A total of 154 children and adolescents participated in
an assessment consisting of a diagnostic interview, the LSAS-CA, and other measures of psychopathology and impair-
ment. Sixty-one of these children also participated ina second LSAS-CA administration, by a different rater blind to diag-
nosis, within 7 days of the initial assessment. Results: High internal consistency (a = .90-.97 for full sample and .83-.95
for social phobia group) and test-retest reliability (intraclass correlation coefficient = 0.89-0.94) were obtained for LSAS-
CA total and subscale scores. LSAS-CA scores had stronger associations with measures of social anxiety and general
impairment than with a measure of depression. Subjects with social anxiety disorder had significantly higher LSAS-CA
scores than subjects with other anxiety disorders and healthy controls. A LSAS-CA cutoff score of 22.5 represented the
best balance of sensitivity and specificity when distinguishing between individuals with social phobia and normal controls,
whereas a cutoff of 29.5 was optimal for distinguishing social phobia from other anxiety disorders. Conclusion: Initial find-
ings suggest that the LSAS-CA is a reliable and valid Instrument for the assessment of social anxiety disorder. J. Am. Acad.
ChildAdolesc. Psychiatry, 2003, 42(9):1076-1084. Key Words: clinician-rated, reliability, social phobia.

Fear of social or performance situations in which the per- intimate relationships, and alcohol use (Wilttchen et al.,
son is exposed to unfamiliar people or to possible scrutiny 1999). Studies of social phobia suggest that there is sig-
is the hallmark of social phobia (American Psychiatric nificant stability into adulthood and that this disorder
Association, 1994). Situations commonly feared by chil- may contribute to increased risk for depression, suicide
dren and adolescents include interactions with peers, attempts, alcohol abuse, incomplete educational attain-
answering questions in class, public speaking, initiating ment, and severe social restrictions (Liebowitz et al., 1985;
conversations, attending parties or school, speaking to Pine et al., 1998; Schneier et al., 1992; Wittchen et al.,
authority figures, and performing in front of others (Albano, 1999). Such information is of concern in view of data sug-
1995; Beidel, 1991; Hofmann et al., 1999). Social anxi- gesting a rise in the prevalence of generalized social pho-
ety disorder in youth isassociated with relatively few friends, bia in younger cohorts (Heimberg et al., 2000).
disturbances in academic functioning, difficulties with Recognition of the importance of social phobia has
led to a surge of research on its etiology ahd treatment
AcceptedApril 15, 2003. in youth (Beidel et al., 1995, 2000a,b; Hayward et al.,
From New York University School of Medicine. Child Study Center (Drs. 2000; Masia et al., 2001; Spence et al., 2000). These
Masia-W1ZarnerandVein),Columbia University (Mr.Storch), CenterforAnxiety
andRelated DisordersatBoston Univensity (Dr. Pincus), Temple Universiy (Dr. efforts have created the need for assessment instruments.
Heimberg), and Columbia University, New York StatePsychiatricInstitzute (Dr. Several psychometrically sound self-rating instruments
Liebowitz). for children and adolescents have been developed
This rsearch uasfundedbyagrantfromSmithKline Beecham to Drs. Liebouwitz
andMasia-Warner. The authorsthankjoseAlvirandMark Daviesforstatisti-
(Beidel et al., 1995; Birmaher et al., 1999; La Greca and
cal consultationand Sharon Davieaforher contribution. Lopez, 1998; March et al., 1999; Muris and Steerneman,
Peprintrequests to Dr.Masia-IWarner,NYU ChildStudy Center, 215 Lexi ngton 2001), but no clinician-rating scales exist to assess the
Avenue, 13th Floor, New York, NY10016; e-mail: carrie.masia@med.nyu.edu. severity of social anxiety in children and adolescents.
0890-8567/0314209-1076©D2003 by the American Academy of Child and
Adolescent Psychiatry. Although self-rating scales are an economical strategy for
DOI: 10.1097/0I.CHI.0000070249.24125.89 assessing large groups of children for prevention programs

1076 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:9, SEPTEMBER 2003

 LIEBOWITZ SOCIAL ANXIETY SCALE

and epidemiological studies, clinician ratings are more 1 = occasionally, 2 = often, 3 = usually) are given for each
valuable for documenting clinical severity and treatment of the 24 items for a total of 48 ratings. Thus, the LSAS-
response (Klein and Pine, 2002). CA provides seven scores: (1)anxiety related to social inter-
The Liebowitz Social Anxiety Scale (LSAS) (Liebowitz, action, (2)performance anxiety, (3)total anxiety, (4)avoidance
1987), a clinician-rating scale for the assessment of social of social interaction, (5)avoidance of performance situa-
phobia in adults, has been widely used in studies of phar- tions, (6) total avoidance, and (7)a total LSAS-CA score.
macological and cognitive-behavioral treatment of social The purpose of the present study isto examine the psy-
phobia. The 24 LSAS items are divided into two types: chometric properties of the LSAS-CA. This measure is
11 questions assessing anxiety in social interactions (e.g., currently used in many clinical settings and in national
giving a party) and 13 questions reflecting anxiety in per- pharmacological trials of social phobia in children and
formance situations (e.g., speaking up at a meeting). The adolescents, and Lthus an examination of its reliability and
clinician asks for current severity of anxiety and avoid- validity is required. We addressed the following questions:
ance on 0 to 3 Likert scales. The clinician adjusts the rat- (1)What is the internal consistency of the LSAS-CA items?
ings based on clinical judgment and direct behavioral (2) What is the relationship between the subscale and total
observations. Initial evidence suggests that the LSAS isa scores? (3) Are LSAS-CA scores consistent when admin-
reliable, valid, and treatment-sensitive measure of adult istered at two separate times by different clinicians? (4)
social phobia (Heimberg et al., 1999; Safren et al., 1999). Does the LSAS-CA correlate with other social anxiety and
More specifically, the LSAS subscales have excellent inter- impairment measures? (5) Is the LSAS-CA more strongly
nal consistency and are significantly correlated to other associated with social anxiety than depression measures?
measures of social anxiety and avoidance. An exploratory (6) Is the LSAS-CA able to provide meaningful differen-
factor analysis yielded four factors: (1)social interaction, tiation between subjects with and without a diagnosis of
(2) public speaking, (3) observation by others, and (4) social phobia? (7) What threshold scores may indicate
eating and drinking in public (Safren et al., 1999). Finally, clinically significant levels of social anxiety and avoidance?
the LSAS has been shown to be sensitive to effects of
pharmacological treatments for social phobia compared METHOD
with pill placebo (Heimberg et al., 1999).
Based on the strength of the LSAS and the need for a Participants
clinician-rating scale for youth, we developed a similar Participants were obtained from three sources: (1)consecutive refer-
instrument for children and adolescents, the Liebowitz rals to an anxiety clinic in Boston, (2)volunteers from a high school
in suburban New York, and (3) children and adolescents enrolled in
Social Anxiety Scale for Children and Adolescents (LSAS- a multisite psychopharmacological clinical trial of social phobia.
CA) (Masia et al.,1999). The LSAS-CA items were gen- Written consent and assent to participate were provided by parents
erated from two sources. First, items were formed from and children. Consent forms, however, were specific to the research
study and institution.
reports of the 10 most feared situations of a group of 33 Anxiety clinic subjects (n = 30) had the following primary diag-
adolescents with social phobia (Hofmann et al., 1999). noses: social phobia (n = 5), panic disorder with agoraphobia (n =10),
Second, items on the adult version that were not gener- separation anxiety disorder (n = 8), agoraphobia without panic (n =
ated by the first method were included with slight word- 2), obsessive-compulsive disorder (n = 1), specific phobia (n = 1), gen-
eralized anxiety disorder (n = 1), oppositional defiant disorder (n =
ing changes if they were considered developmentally 1), and depersonalization disorder (n =1). There were no comorbid
appropriate. For example, the LSAS question, "Participating diagnoses of social phobia. All participants recruited from the mul-
in small groups," was modified to "Participating in work tisite trial had a primary diagnosis of social phobia (n = 92). To obtain
a normal comparison group, high school students enrolled in psy-
groups in the classroom." The majority of the LSAS-CA chology and sociology classes (n = 122) were offered participation.
items (15 items) were consistent across both sources. Thirty-four students (28%) returned consent forms. Of these, 32 had
The resulting measure consists of 24 items: 12 social no diagnosis, as assessed by a diagnostic interview. Two students were
excluded from participation due to diagnoses of generalized anxiety
interaction situations (e.g., "looking at people you don't disorder and posttraumatic stress disorder.
know well in the eye") and 12 performance situations (e.g., The final sample consisted of 154 children and adolescents (93
"asking questions in class"). The administration procedure, female and 61 male). Overall, the mean age of the sample was 13.4
rating scales, and scoring structure from the adult LSAS years (SD 3.2 years, range 7-18 years). The ethnic composition was
as follows: 85.7% white, 5.2% African American, 4.6% biracial, 3.9%
were retained. Clinician ratings of anxiety (0= none, 1 = Latin American, and 0.6% Asian American. The sample was classi-
mild, 2 = moderate, 3 = severe) and avoidance (0 =never, fied into three diagnostic groups: group I, 97 subjects with a primary

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:9. SEPTEMBER 2003 1077

MASIA-WARNER ET AL.

diagnosis of social phobia (Social Phobia Group), group 11, 23 par- TABLE 1
ticipants with anxiety disorders other than social phobia (Mixed Demographics of Total Sample and Each Diagnostic Group
Anxiety Group), and group III,32 subjects with no psychiatric diag-
nosis (Nonpsychiatric Control Group). Two participants recruited Total Social Mixed Nonpsychiatric
from the Boston anxiety clinic, one with a primary diagnosis of oppo- Demographics Sample Phobia Anxiety Comparison
sitional defiant disorder and the other with a diagnosis of deperson- No. 154 97 23 32
alization disorder, did not fit into any of the diagnostic groups. See % Female 60.4 59.8 60.9 62.5
Table I for demographics of the toral sample and each diagnostic % White 85.7 86.6 100 71.9
group. No significant differences were found in gender or ethnicity Mean age (SD) 13.4 (3.2) 12.4 (2.9) 13.0 (3.0) 16.8 (1.0)
across groups. The social phobia and mixed anxiety groups were sig-
nificandy younger than the control group, F2,149 = 34 .5,p < .001. Note: Social phobia and mixed anxiety groups are significantly
younger than the control group, F2,149 = 34 .5,p < .001.
Measures
DiagnosticAssessment. The Anxiety Disorders Interview Schedule
for DSM-IV: Child and Parent Versions (ADIS-CP) (Silverman and Procedures
Albano, 1996) assess anxiety and mood disorders and screen for the Training in the administration of the LSAS-CA consisted of an
presence of disruptive behavior, psychotic, and eating disorders. instructional meeting with the first author. The meeting consisted of
Interviews are conducted with parents and children separately by one
a review of the scale's content and structure, as well as viewing a video-
clinician, and diagnoses are assigned based on both informants. The rape that included administration directions and two practice inter-
interview has demonstrated good to excellent test-retest reliabiliry for views, one with a child and the other with an adolescent, conducted
diagnoses (Silverman et al., 2001). The validity of the ADIS-CP has by Drs. Masia-Warner and Liebowitz. Clinicians completed the LSAS-
also been supported (Wood et al., 2002). CA for each videotaped interview, and disagreements in scoring were
The LSAS-CA (Masia et al., 1999) assesses anxiey and avoidance discussed and clarified. All clinicians had prior training and experi-
in 24 situations. The independent evaluator asks the child to provide ence with the administration of the other measures.
separate ratings for anxiety and avoidance but is given latitude to An inidal evaluation was scheduled with all participants that induded
adjust ratings based on clinical judgment and direct behavioral obser-
administration of the ADIS-Child and Parent Versions, LSAS-CA,
vations. The LSAS-CA can be obtained from the first author. SPAI or SPAI-C (depending on the child's age), CDRS-R, CGI, and
Participants age 14 years and older completed the Social Phobia GAF. Assessments began with the diagnostic interview followed by
andAnxiety Inventory (SPAI) (Turner et al., 1989).The SPAI is a 45- the completion of the other study measures. The anxiety clinic and
item self-report instrument that assesses behavioral, physiological, healthy volunteer participants were scheduled for a second adminis-
and cognitive symptoms associated with social phobia. Items are scored tration of the LSAS-CA by a different rater who was blind to the first
on a 7-point Likert scale from 1 (never) to 7 (always). It consists of
assessments, 3 to 7 days following the first evaluation. All assessments
social phobia and agoraphobia subscales that are used to obtain a dif- were conducted via in-person interviews. One child from the anxiety
ference (total) score; the authors recommend the use of the difference
clinic did not participate in the secotid administration, leaving a total
score (Turner et al., 1989). The SPAI has demonstrated strong psy- of 61 participants for the readministration of the LSAS-CA.
chometric characteristics (for review see Clarke et al., 1997).
The Social Phobia and Anxiety Inventory for Children (SPAI-C)
(Beidel et al., 1995), a self-report measure, was completed by partic- Data Analysis
ipants younger than 14 years. The 26 items assess somatic symptoms, Psychometric analyses were conducted using the initial completion
cognitions, and behavior across fear-producing situations. Questions of the LSAS-CA. Calculations were conducted both for the full sam-
are answvered on a 3-point Likert-type scale ranging from 0 (never) to ple (n = 154) and social phobia group (n = 97). Internal consistency
2 (most of the time or always). The SPAI-C is internally consistent of the LSAS-CA was evaluated with the Cronbach a coefficient.
(Cronbach a = .95) and has adequate test-retest reliability (Beidel Reliability across the first and second ISAS-CA assessments was exam-
et al., 1995). It discriminates social phobia from other anxiety disor- ined by calculating one-way random effects intraclass correlation coef-
ders, disruptive behavior disorders, and no psychiatric disorders ficients (Shrout and Fleiss, 1979) between the LSAS-CA subscale and
(Beidel et al., 1996, 2000a). total scores across the two administrations (n = 61). Descriptive sta-
The Childrens Depression Rating Scale-Revised (CDRS-R) (Poznanski tistics for time I and time 2 were conducted, and paired sample t tests
and Mokros, 1995) is a clinician-rated inventory of depressive symp- were used to compare mean values between the two administrations.
toms. Ratings for each symptom range from I to 7. Pearson product-moment correlations examined the relation between
The Clinical Global Impression-Severity of Illness (CGI) (National LSAS-CA subscale and total scores, as well as between LSAS-CA scores
Institute of Menral Health, 1985) is a clinician rating of the severity and other measures of social anxiety, impairment, and depression.
of the participant's mental illness. The categories include the follow- Differences in correlations between the LSAS-CA total and other mea-
ing: not assessed, normal, borderline ill, mild, moderate, marked, sures for the full sample and social phobia group were examined using
severe, and among the most extremely ill patients. This instrument t tests for dependent correlations. One-way analysis of variance was
has been extensively used in treatment studies of children and ado- used to investigate differences in LSAS-CA scores across diagnostic
lescents (Cook et al., 2001; Pliszka et al., 2000) groups. All statistically significant analyses of variance were further
The Global Assessment of Functoning (GAF) (American Psychiatric examined usingTukey honestly significant difference follow-up tests.
Association, 1994) scale is a clinician rating of a patient's function- Receiver operating characteristic (ROC) analysis (Kraemer, 1992;
ing considering psychological, social, and occupational functioning. Swets et al., 2000) was performed on LSAS-CA total scores to identify
Scores range from 0 to 100, with higher ratings indicating better func- the most useful cutoffs for discriminating individuals with social pho-
tioning. Good psychometric data exist for the GAF (Hilsenroth et al., bia from those with other anxiety disorders, and from those with no
2000; Jones et al., 1995). diagnosis. Parametric ROC analysis is based on logistic regression with

1078 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:9, SEPTEMBER 2003

 LIEBOWITZ SOCIAL ANXIETY SCALE

TABLE 2 phobia group. All correlations were statistically signifi-

Cronbach a Coefficients for the LSAS-CA cant,p < .001. The total score was highly correlated with
Total and Subscale Scores
all other subscales (r= 0.93-0.99 for full sample and r =
Total Sample Social Phobia 0.88-0.98 for social phobia group). Anxiety and avoid-
LSAS-CA Scores (n = 154) (n = 97)
ance ratings were highly correlated at the total level (r =
Total score; .97 .95 0.96 for full sample and r= 0.92 for social phobia group),
Total anxiety .95 .91
Interaction anxiery .92 .86
aswell as within performance (r= 0.93 for full sample and
Performance anxiety .91 .85 r = 0.89 for social phobia group) and within social inter-
Total avoidance .95 .90 action situations (r= 0.97 for full sample and r = 0.94).
Avoidance of interaction .91 .83 Correlations between performance and social interaction
Avoidance of performance .90 .85
subscales were significant but somewhat lower (r =0.78-0.84
Note: LSAS-CA = Liebowitz Social Anxiety Scale for Children and for full sample and r= 0.63-0.71 for social phobia group).
Adolescents.
a Total Score = Total anxiety t Total avoidance. Reliability Across Assessments
One-way random effects intraclass correlation coeffi-
a continuous predictor variable (LSAS-CA total score) and with a
dichotomous criterion variable (positive or negative for social phobia). cient values were uniformly high for LSAS-CA scores,
Probabilities of each value of the predictor are plotted on a graph with ranging from 0.89 to 0.94. Table 4 shows intraclass cor-
their associated sensitivity on the y-axis and I - specificity (the false- relation 'coefficientvalues, 95% confidence intervals, and
positive rate) on the x-axis. This curve is referred to as the test ROC.
The greater the distance between the ROC curve of the measure and descriptive data for LSAS-CA scores at each administra-
the random ROC diagonal line (0.50 or chance probability), the more tion. No significant mean differences were found between
accurate the test. The area between the random ROC and the test ROC, the two administrations in the LSAS-CA total score, total
called the area under the curve, provides a summary index of a test's
ability to classify individuals correcdy. The score that maximizes bodh
anxiety, performance anxiety, interaction anxiety, and inter-
sensitivity and specificity isconsidered an optimal cutoff value. action avoidance. Significant differences between time 1
and time 2 were found for performance avoidance (t6o =
RESULTS -3.2, p = .002) and total avoidance (t6o = -2.1,p = .04),
with higher scores reported at the second administration.
Internal Consistency
Convergent and Divergent Validity
As shown in Table 2, oa coefficients for LSAS-CA scores
were high (.83-.97). Internal consistency for performance Table 5 presents correlations between'the LSAS-CA
and social interaction was comparable for the overall sam- scores and measures of social anxiety, depression, general
ple and the social phobia group. impairment, and functioning. Correlations were highest
between the LSAS-CA scores and social phobia measures
Correlations Among LSAS-CA Total and Subscale Scores (r= 0.69-0.82 for full sample and r= 0.23-0.70 for social
Table 3 presents the correlations between the LSAS- phobia group). The LSAS-CA was only moderately cor-
CA total score and subscales for the full sample and social related with measures of general impairment and func-

TABLE 3
Correlations of LSAS-CA Scores for all Participants (and Social Phobia Group)
Total Total Interaction Performance Total Avoidance of Avoidance of
LSAS-CA Scores Score Anxiety Anxiety Anxiety Avoidance Interaction Performance
Totalscore 1.0 .99 (.98) .95 (.91) .95 (.90) .99 (.98) .96 (.92) .93 (.88)
Total anxiety 1.0 .97 (.93) .96 (.92) .96 (.92) .94 (.87) .89 (.81)
Interaction anxiety 1.0 .84 (.71) .92 (.85) .97 (.94) .78 (.63)
Performance anxiety 1.0 .92 (.85) .83 (.68) .93 (.89)
Total avoidance 1.0 .96 (.92) .95 (.92)
Avoidance of interaction 1.0 .82 (.68)
Avoidance of performance 1.0
Note: All correlations are statistically significant at the .001 level. LSAS-CA = Liebowitz Social Anxiety Scale for Children
and Adolescents.

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:9, SEPTEMBER 2003 1079

MASIA-WARNER ET AL.

TABLE 4
Test-Retest Reliability of LSAS-CA Scores (n = 61)"
Time I Time 2
LSAS-CA Scores ICC 95% CI Mean (SD) Mean (SD)
Total score 0.94 0.91-0.97 18.6 (18.4) 19.5 (19.7)
Total anxiety 0.93 .88-0.96 10.5 (9.5) 10.4 (10.3)
Interaction anxiety 0.89 0.82-0.93 5.8 (5.7) 5.3 (5.6)
Performance anxiety 0.90 0.84-0.94 4.7 (4.4) 5.1 (5.4)
Total avoidance 0.92 0.88-0.95 8.1 (9.2) 9.1 (9.7)*
Avoidance of interaction 0.91 0.85-0.94 4.5 (5.2) 4.6 (5.0)
Avoidance of performance 0.89 0.83-0.93 3.6 (4.5) 4.5 (5.4)*
Note: LSAS-CA = Liebowitz Social Anxiety Scale for Children and Adolescents; ICC = intraclass correlation coefficient; C=
confidence interval.
* Statistically significant difference between means,p < .05.
a Anxiety clinic n = 29; healthy subjects n = 32.

tioning (r= 0.61-0.68 for full sample and r= 0.22-0.27 higher scores for the social phobia group than for the mixed
for social phobia group), and more weakly correlated with anxiety and nonpsychiatric comparison groups. Scores in
the measure of depression (r = 0.34-0.39 for full sample the mixed anxiety group were significantly more elevated
and r= 0.02-0.14 for social phobia group). compared with the nonpsychiatric comparison group.
For both the full sample and social phobia group, cor- Using ROC analysis, we examined cutoff values that
relations between the LSAS-CA total and SPAI-C were distinguished individuals with social phobia from those
significantly greater than those between the LSAS-CA with no psychiatric diagnosis and those with other anxi-
total and CDRS-R (t56 =7.5,p < .001; t50 =6. 4 ,p < .001) ety disorders. First, an ROC analysis was performed on the
and more general measures of impairment (CGI: ts6 = social phobia group (n = 97) and the nonpsychiatric con-
3.7,p < .001; t5o =3.5,p < .001 and GAF: t56 =3.5,p < trol group (n = 32). As shown in Figure 1, the area under
.001; t5 0 = 3.2, p < .05). The correlation of LSAS-CA the curve was 0.99 and was significant versus the chance
total with the SPAI was only statistically higher than the or random ROC line (p < .0001). An LSAS-CA total score
LSAS-CA total with the CDRS-R (t80 =3.1,p < .05) for of 22.5 provided optimal sensitivity and specificity. That
the full sample. is, 95.9% of individuals with social phobia were correctly
classified and none of the nonpsychiatric comparisons was
Construct Validity misclassified. As shown in Table 7, decreasing the LSAS-
As shown in Table 6,LSAS-CA mean values differed sig- CA total score sacrifices specificity with minimal gain in
nificantly across groups. Tukey tests indicated significantly sensitivity, whereas increasing the LSAS-CA total score

TABLE 5
Correlations Between LSAS-CA and Other Measures of Psychopathology for Total Sample (and Social Phobia Group)
LSAS-CA scores SPAI SPAI-C CGI GAF CDRS-R
Nof sample 83 (40) 59 (53) 152 (97) 153 (97) 153 (97)
Total score .80* (.48)** .75** (.70)** .68** (.27)** -.68** (-.26)** .38** (.10)
Total anxiety .79** (.45)** .75** (.70)** .67** (.26)* -. 66** (-.25)* .39** (.07)
Interaction anxiety .82** (.57)** .69** (.63)** .66** (.24)* -. 65** (-.22)* .37** (-.10)
Performance anxiety .69** (.23) .73** (.68)** .63** (.23)* -. 62** (-.24)* .38** (.02)
Total avoidance .80** (.49)** .73** (.68)** .68** (.27)** -.67** (-.27)** .37** (.14)
Avoidance of interaction .80** (.53)** .70** (.64)** .68** (.26)* -.67** (-.22)' .36** (.12)
Avoidance of performance .71** (.31) .71** (.65)** .61** (.24)* -.61** (-.27)** .34** (.09)
Note. LSAS-CA = Liebowitz Social Anxiety Scale for Children and Adolescents; SPAI = Social Phobia and Anxiety Inventory;
SPAI-C = Social Phobia and Anxiety Inventory for Children; CGI = Clinical Global Impression-Severity of Illness; GAF = Global
Assessment of Functioning; CDRS-R = Children's Depression Rating Scale-Revised.
* Correlation is significant at the .05 level (two-tailed); **correlation is significant at the .01 level (two-tailed).

1080 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:9, SEPTEMBER 2003

 LIEBOWITZ SOCIAL ANXIETY SCALE

TABLE 6
Means (SD) of LSAS-CA Scores by Diagnostic Group
Social Mixed Nonpsychiatric
Phobia Anxiety Comparison
LSAS-CA Scores (n = 97) (n = 23) (n = 32) Fvalue* Tukey**
Totalscore 68.6 (28.9) 27.4 (15.5) 7.8 (5.8) 88.1 a>b>c
Total anxiety 35.3 (14-7) 14.7 (8.1) 5.2 (3.8) 83.1 a>b>c
Interaction anxiety 19.3 (8.2) 8.0 (4.8) 2.9 (2.9) 77.8 a> b> c
Performance anxiety 16.1 (7.7) 6.7 (4.2) 2.3 (1.7) 63.1 a>b>c
Total avoidance 33.3 (14.8) 12.7 (7.8) 2.6 (2.3) 85.7 a> b> c
Avoidance of interaction 18.9 (8.1) 7.0 (4.3) 1.5 (1.9) 90.8 a >b >c
Avoidance of performance 14.4 (8.0) 5.6 (4.5) 1.1 (1.4) 54.1 a > b >c
Note: LSAS-CA = Liebowitz Social Anxiety Scale for Children and Adolescents.
4 Social phobia group.
b Mixed anxiety group.
vNonpsychiatric comparison group.
*All Fvalues are significant at thep <.001 level; **all Tukey follow-up tests are significant at thep <.01 level.

lowers sensitivity with no gain in specificity. Another ROC classified (false positives). As seen in Table 7, decreasing
analysis was conducted on the social phobia group (n = the cutoff score increases sensitivity only slightly while sac-
97) and the mixed anxiety disorders group (n = 23). As rificing specificity, whereas increasing the cutoff score low-
illustrated in Figure 2, the area under the curve was 0.89 ers sensitivity with no gain in specificity.
and was significantly different from the random ROC line
(p < .0001). Sensitivity and specificity were maximized at
DISCUSSION
a cutoff value of 29.5. The majority of the social phobia
participants (91.8%) were correctly classified (true posi- This article describes the psychometric characteristics
tives), and 34.8% of the mixed anxiety group was mis- of a new clinician-rated measure to assess social anxiety
in children and adolescents. Findings suggest that the
ROC Curve LSAS-CA is a reliable and valid instrument for assessing
social phobia in youth. The LSAS-CA scores had excel-
1.00
lent internal consistency for the total score as well as the
specific subscale scores. In addition, the LSAS-CA demon-
strated excellent test-retest reliability within a 7-day retest
.75 - interval with the stringent test of using different clini-
cians at each administration.
The overall pattern of higher correlations of LSAS-CA
scores with social phobia measures than with measures
.50-
of depression and general impairment provides support
for the convergent and divergent validity of the LSAS-
CA. Strong support for the construct validity of the ISAS-
~.25 CA and its utility for assessing the severity of social phobia
symptoms is evident by its ability to discriminate indi-
viduals with social phobia from healthy individuals, as
i 0.00 well as to differentiate between social phobia and other
0.00 .25 .50 .75 1.00 anxiety disorders. ROC findings indicate that an LSAS-
CA total score of 22.5 discriminates well between par-
I - Specificity ticipants with social phobia and healthy nonpsychiatric
volunteers, whereas a cutoff of 29.5 distinguishes par-
Fig. 1 Receiver operating characteristic (ROC) curve for social phobia (group
I) and nonpsychiatric control (group III) participants. Diagonal segments are ticipants with social anxiety disorder from those with
produced by ties. other anxiety diagnoses. Of course, the appropriate cut-

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:9, SEPTEMBER 2003 1081

 MASIA-WARNER ET AL.

TABLE 7
LSAS-CA Cutoff Scores and Their
Sensitivity and Specificity Values
ROC Curve
1.00
LSAS-CA
Total Score Sensitivity Specificity
SocialPhobia vs. Nonpsychiatric Comparisons
18 0.959 0.937
20 .75
0.959 0.969
22.5 0.959 1.0
25 0.938 1.0
SocialPhobia vs. OtherAnxiety Disorders
.50,
27 0.928 0.522
28.5 0.928 0.609
29.5 0.918 0.652
30.5 0.907 0.652
.25
Note: LSAS-CA = Liebowitz Social Anxiety Scale for Children and .5
Adolescents.
a)
off for the LSAS-CA depends on its purpose. For exam- U) 0.00 1
ple, if the LSAS-CA were used in medical or school set- 0. 00 .25 .50 .75 1.00
tings to assist with the early identification of social anxiety
disorder, a lower cutoff might be indicated, which would I - Specificity
maximize identifying more affected cases but at the cost Fig. 2 Receiver operating characteristic (ROC) curve for social phobia (group
of a higher rate of false positives. I) and mixed anxiety disorders (group 11) participants. Diagonal segments
are produced by ties.
An uncertain issue is what scoring structure for the
LSAS-CA will be most appropriate and economical. That Participants were recruited from different sources and
is, do separate ratings of anxiety and avoidance provide were largely white. In addition, the nonpsychiatric group
important information or are they redundant? Are some was composed of volunteers with a low consent rate and
performance anxiety items better categorized as social was significantly older than the two diagnostic groups.
interaction anxiety and vice versa? Is the distinction Because of our concern that the older age of the normal
between performance and social interaction situations, comparison group may have influenced our results, we
as well as between anxiety and avoidance, valid for con- examined the possibility of age differences in reporting
structing subscales? Similar to the findings of Heimberg within each diagnostic group using the lower age range
et al. (1999) with the LSAS, the LSAS-CA anxiety and of the nonpsychiatric comparison group (age 14 years)
avoidance subscales were highly associated, and correla- as the cutoff. There were no significant differences in
tions between performance and social interaction sub- LSAS-CA values of participants younger than 14 years
scales were noticeably lower than associations between and those 14 years of age and older in either diagnostic
subscales in a given domain (e.g., performance anxiety group, indicating that the observed differences between
and performance avoidance). We also found that the diagnostic groups were not a function of age.
LSAS-CA social interaction scores, as compared to its Moreover, the DSM-IVsocial phobia criteria allow for
performance scores, had a stronger association with self- a "generalized" classifier if the fears include most situa-
ratings on the SPAI. Correlations with other measures tions (American Psychiatric Association, 1994). Altematively,
were similar across LSAS-CA subscales. Further investi- a "specific" descriptor has been used if the fears are in one
gations incorporating factor analysis will inform on the or a limited number of discrete social situations (e.g.,
best format and scoring structure for the LSAS-CA. public speaking, only performance situations) (Boone
Limitations
et al., 1999; Heimberg et al., 1993; Kessler et al., 1998).
The pharmaceutical clinical trial, from which the major-
Although our findings support the reliability and valid- ity of social phobia participants were obtained, did not
ity of the LSAS-CA, some limitations should be noted. include these descriptors. Therefore, it was not possible
First, the generalizability of the findings is not certain. to examine the association of social phobia subtypes with

1082 J. AAM.ACAD. CHILD ADOLESC. PSYCHIATRY, 42:9, SEPTEMBER 2003

 LIEBOWITZ SOCIAL ANXIETY SCALE

performance and social interaction subscales or potential Hayward C, Vardy S,Albano AM, Thienemann M, Henderson L,Schatzberg
AF (2000), Cognitive-behavioral group therapy for social phobia in female
subtype differences in LSAS-CA mean values and cutoff adolescents: results of a pilot study.JAmAcad ChildAdolescPsychiatry
scores, as has been studied for the adult LSAS (Mennin 39:721-726
HeimbergRG, Holt CS, Schneier FR, Spitzer RC, Liebowim MR (1993),The
et al., 2002). issue of subtypes in the diagnosis of social phobia JAnxiay Disoni7:249-269
Finally, we do not have information on the interrater Heimberg RG, Horner KJ, Juster HR et al. (1999), Psychometric properties
reliability of the study interviewers. One indication that of the Liebowitz Social Anxiety Scale. PsycholMed 29:199-212
Heimberg RG, Stein MB, Hiripi E, Kessler RC (2000), Trends in the preva-
rater agreement was high isthe strong test-retest reliability lence of social phobia in the United States: a synthetic cohort analysis of
values even with the use of different raters. In other words, changes over four decades. EirPsychiatry 15:29-37
Hilsenroth MJ, Ackerman SJ, Blagys MD et al. (2000), Reliability and valid-
strong consistency isshown between two LSAS-CA admin- ity of DSM-IVAxis V.AmJPsychiatry157:1858-1863
istrations separated by time and completed by different Hofmann SG, Albano AM, Heimberg RG, Tracey S, Chorpita BF, Barlow
DH (1999), Subtypes of social phobia in adolescents. DepressAnxiety
interviewers, which isa stringent test of reliability. 9:15-18
Jones SH, Thornicroft G, Coffey M, Dunn G (1995), A brief mental health
Clinical Implications outcome scale: reliabiity and validity ofthe Global Assessment ofFunctioning
(GAF). BrJPsychiatry166:654-659
The LSAS-CA was found to be a reliable and valid Kessler RC, Stein MB, Berglund P (1998), Social phobia subtypes in the
clinician-rating scale for assessing social anxiety in chil- national comorbidity survey. AmJPsychiatry 155:613-619
Klein RG, Pine DS (2002). Anxiety disorders. In: ChildandAdolscentPsrhiatry:
dren and adolescents. This instrument is a quantitative Modern Approaches, Rutter M, Taylor E, eds. London: Blackwell Science,
measure of the severity and pervasiveness of anxiety and pp 48 6 -509
Kraemer HC (1992), Eval/natingMedicalTests. Newbury Park, CA: Sage
avoidance, which is a reflection of impairment or dis- La Greca AM, Lopez N (1998), Social anxiety among adolescents: linkages
tress. The specificity of LSAS-CA items may be helpful with peer relations and friendships.JAbnorm Psychol26:83-94
Liebowitz MR (1987), Social phobia. ModProblPychopharmacol22:141-173
in treatment planning and monitoring progress. The Liebowitz MR, GormanJM, FyerAJ, Klein DF (1985), Social phobia: review
LSAS-CA is not meant to be used to establish a diagno- of a neglected disorder. Arrh Gen Psychiatry42:729-736
MarchJS, Conners C, Arnold E et al. (1999), The Multidimensional Anxiety
sis. Further work on the LSAS-CA needs to address its Scale for Children (MASC): confirmatory factor analysis in a pediatric
treatment sensitivity, factor structure, and, possibly, the ADHD sample. JAttention Disord3:85-89
development of a briefer version. Masia CL, Klein RG, LiebowitzMR (1999), TheLiebowitzSocialAnxiesyScale
fir ChildrenandAdolescents (LSAS-CA) (available from Carrie Masia-
Warner, NYU Child Study Center, 215 Lexington Avenue, 13th floor,
NewYork, NY 10016)
REFERENCES Masia CL, Klein RG, Storch EA, Corda B (2001), School-based behavioral
treatment for social anxiety disorder in adolescents: results of a pilot study.
Albano AM (1995), Treatment of social anxiety in adolescents. Cogn Behav JAm Acad ChildAdolesc Psychiatry40:780-786
Pract2:271-298 Mennin DS, Fresco DM, Heimberg RG, Schneier FR, Davies SO, Liebowitz
American Psychiatric Association (1994), DiagnosticandStatisticalManual of MR (2002), Screening for social anxiety disorder in the clinical setting:
MentalDisorders, 4th edition (DSM-JV). Washington, DC: American using the Liebowitz Social Anxiety Scale.JAnxiety Disord16:661-673
Psychiatric Association Muris P,Steerneman P (2001), The revised version of the Screen for Child
Beidel DC (1991), Social phobia and overanxious disorder inschool age chil- Anxiety Related Emotional Disorders (SCARED-R): first evidence for.its
dren.JAmAcad ChildAdolescPsychiary 30:545-552 reliability and validity in a clinical sample. BrJClin Pychol40:35-44
Beidel DC,Turner SM, Fink CM (1996),Assessment of childhood social pho- National Institute of Mental Health (1985), Special feature: ratingscales and
bia: construa, convergent, and discriminative validity of the Social Phobia assessment instruments for use in pediatric psychopharmacology research.
and Anxiety Inventory for Children (SPAI-C). PsycholAssess 8:235-240 PsychopharmacolBull21:839-843
Beidel DC, Turner SM, Morris TL (1995), A new inventory to assess child- Pine DS, Cohen P,Gurley D, BrookJ, MaY (1998), The risk for early-adult
hood social anxiety and phobia: the Social Phobia and Anxiety Inventory anxiety and depressive disorders in adolescents with anxiety and depres-
for Children. PsycholAssess 7:73-79 sive disorders. Arch Gen Psychiatry55:56-64
Beidel DC, Turner SM, Hamlin K, Morris TL (2000a), The Social Phobia Pliszka SR, Browne RG, Olvera RL, Wynne SK (2000), A double-blind,
and Anxiety Inventory for Children (SPAI-C): external and discrimina- placebo-controlled study of Adderall and methylphenidate in the treat-
tive validity. Behav Ther31:75-87 ment of attention-deficit/hyperactivity disorder.JAmAcadChildAdolesc
Beidel DC, Turner SM, Morris TL (2000b), Behavioral treatment of child- Psychiatry39:619-626
hood social phobia. J Consul Clin Psychol68:1072-1080 Poznanski EO, Mokros HB (1995), ChildrenDepressionRating Scale-Revised.
Birmaher B,Brent DA, Chiappetta L,BridgeJ, Monga S, Baugher M (1999), Los Angeles: Western Psychological Services
Psychometric properties of the Screen for Child Anxiety Related Emotional Safren SA, Heimberg RG, Horner KJ,Juster HR, Schneier FR, Liebowitz MR
Disorders (SCARED): a replication study.jAmAcad CildAdolacPsychiary (1999), Factor structure of social fears: the Liebowitz Social Anxiety Scale.
38:1230-1236 JAnxiety Disord13:253-270
Boone ML, McNeil DW, Masia CL et al. (1999), Multimodal comparisons Schneier F,Johnson J, Hornig C, Liebowitz M, Weissman M (1992), Social
of social phobia subtypes and avoidant personality disorder. JAnxiety phobia: comorbidity and morbidity in an epidemiologic sample. Arrh Gen
Disord13:271-292 Psychiatry49:282-288
Clark DB, Feske U, Masia CL et al. (1997), Systematic assessment of social Shrout PE, FleissJL (1979), Intraclass correlations: uses in assessing rater reli-
phobia in clinical practice. DepressAnxiety 6:47-61 ability. PsycholButl 86:420-428
Cook EH, Wagner KD, March JS et al. (2001), Long-term sertraline treat- Silverman WK, Albano AM (1996), Anxiety DisordersIntervietwSchedutlefor
ment of children and adolescents with obsessive-compulsive disorder. J DSM-IV-ChildandParentVesions. San Antonio,1T: Gray wind (Psychological
Am Acad ChildAdolesc Psychiatry 40:1175-1189 Corporation)

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:9, SEPTEMBER 2003 1083

 MASIA-WARNER ET AL.

Silverman WK, Saavedra LM, Pina AA (2001), Test-retest reliability of anxi- Turner SM, Beidel DC, Dancu CV, Stanley MA (1989), An empirically derived
ety symptoms and diagnoses with the Anxiety Disorders Interview Schedule inventory to measure social fears and anxiety' the Social Phobia andAnxiety
for DSM-IV: child and parent versions.JAmAcad ChildAdolescPsychiatry Inventory. PsycholAssess 1:35-40
40:937-944 Wittchen H-U, Stein MB, Kessler RC (1999), Social fears and social phobia
Spence SH, Donovan C, Brechman-Toussaint M (2000), The treatment of in a communiry sample of adolescents and young adults: prevalence, risk
childhood social phobia: the effectiveness of asocial skills training-based, factors and comorbidity. PsycholMed 29:309-323
cognitive-behavioral intervention, with and without parental involvement. Wood J3,Piacentini JC, Bergman LR, McCracken J, Barrios V (2002),
J ChildPycholPsychiatry41:713-726 Concurrent validity of the anxiety disorders section of theAnxiety Disorders
SwetsJA, Dawes RM, Monahan J (2000), Psychological science can improve Interview Schedule for DSM-IV: child and parent versions. J Clin Child
diagnostic decisions. Psycho! SciPublicInterest1:1-26 AdolescPsychol31:335-342

Community Norms on Toy Guns. Tina L. Cheng, MD, MPH, Ruth A. Brenner, MD, MPH, Joseph L. Wright, MD, MPH,
Hari Cheryl Sachs, MD, Patricia Moyer, BS, Malla Rao, MEngg, MPH
Objective:Toy gun play has been associated with aggressive behavior, and it has been suggested that child health professionals coun-
sel families on limiting exposure. Effective violence prevention counseling requires an understanding of norms regarding parental
attitudes, practices, and influencing factors. Both theories of reasoned action and planned behavior emphasize that subjective norms
and attitudes affect people's perceptions and intended behavior. Few normative data exist on this issue from a cross-section of fam-
ilies. By establishing behavioral norms and understanding the spectrum of parental attitudes, community-sensitive and commu-
nity-specific interventions for violence prevention can be developed. The objective of this study was to assess community norms
on the topic of toy gun play from the perspective of parents. Methods: An anonymous self-report assisted survey was administered
to a convenience sample of parents/guardians who visited child health providers at 3 sites: an urban children's hospital clinic, an
urban managed care clinic, and a suburban private practice. The parent questionnaire included questions on child rearing attitudes,
practice, and sociodemographic information. Results: A total of 1004 eligible participants were recruited for the study; 922 surveys
were completed (participation rate 92%). The 830 (90%) respondents who were parents and had complete child data were the
focus of additional analysis. Regarding toy guns, 67% of parents believed that it was never 'OK for a child to play with toy guns,"
and 66% stated that they never let their children play with toy guns. Parents who thought that it was okay for children to play with
toy guns and allowed them to play with toy guns were more likely to be male parents, have male children, and be white. Conclusions:
There is variability in norms regarding toy gun play among parents, with most discouraging toy gun play. Norms varied based on
gender of the child, gender of the parent, and race. Understanding norms is a first step in designing effective community-sensitive
interventions. Pediatrics 2003;111:75-79.

"They're Too Smart for That": Predicting What Children Would Do in the Presence of Guns. Susan M. Connor, PhD, Kathryn
L. Wesolowski, BS
Objectives: We examined parents' beliefs about how children would react to finding guns, with particular emphasis on how parents
reasoned about the children's actions. Methods: Based on a randomized telephone survey of Northeast Ohio residents, we focused
on the 317 urban and 311 rural respondents who had children 5 to 15 years old in their homes. Respondents were asked about gun
ownership and their expectations of how children would react to finding guns. Analysis examined responses in relation to various
demographic and socioeconomic variables. Restltsr:All respondents-regardless of gun ownership, geography, race, gender, educa-
tion level, income, or child age-were equally likely (-87%) to believe that their children would not touch guns they found. Fifty-
two percent of those reasoned that children were "too smart" or 'knew better." Only 40% based their predictions on specific
instructions they had given their children. Only 12% (15/122) of owners stored guns locked and unloaded. Only 3 of 13 variables
tested were positively associated with safe storage: having a child 5 to 9 years old, having at least a 4-year college education, and
having an income 2 $65 000 per year. Conclusion: Results indicate that parental beliefs may effectively relieve adults of responsi-
biliry and place the burden on children to protect themselves. The implication for injury prevention is that caregivers' unrealistic
expectations of children's developmental levels and impulse control may influence storage decisions or the inclination to address
gun safety issues with children or other adults with whom children spend time (relatives, playmates' parents). Pediatrics
2003;I11:el09-el14.

1084 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:9, SEPTEMBER 2003

COPYRIGHT INFORMATION

TITLE: The Liebowitz Social Anxiety Scale for Children and

Adolescents: An Initial Psychometric Investigation
SOURCE: J Am Acad Child Adolesc Psychiatry 42 no9 S 2003
WN: 0324402114010

The magazine publisher is the copyright holder of this article and it

is reproduced with permission. Further reproduction of this article in
violation of the copyright is prohibited.

Copyright 1982-2003 The H.W. Wilson Company. All rights reserved.