The Pediatric Anxiety Rating Scale (PARS):

Development and Psychometric Properties

THE RESEARCH UNITS ON PEDIATRIC PSYCHOPHARMACOLOGY ANXIETY STUDY GROUP

ABSTRACT
Objective: To describe the development and psychometric properties of the Pediatric Anxiety Rating Scale (PARS), a clin-
ician-rated instrument for assessing the severity of anxiety symptoms associated with common DSM-IV anxiety disorders
(social phobia, separation anxiety disorder, and generalized anxiety disorder) in children. Method: As part of a multisite
study of the efficacy of fluvoxamine, 128 children (aged 6–17) and their parents were interviewed weekly with the PARS.
Data from multiple raters on a subsample of children (using live and videotaped interviews) were used to evaluate inter-
rater reliability. Internal consistency, test-retest reliability, and validity (convergent, divergent) also were evaluated. Results:
The PARS showed high interrater reliability, adequate test-retest reliability, and fair internal consistency. Convergent and
divergent validity were satisfactory. PARS scores were sensitive to treatment and paralleled change in other measures
of anxiety symptoms and global improvement. Conclusions: The PARS is a useful clinician-rated instrument for assess-
ing pediatric anxiety symptoms, severity, and impairment, particularly in treatment studies. Further study of the psycho-
metric properties is warranted. J. Am. Acad. Child Adolesc. Psychiatry, 2002, 41(9):1061–1069. Key Words: anxiety,
children, adolescents, rating scales, assessment.

Anxiety disorders represent the most prevalent childhood light of the high prevalence of these disorders and the
psychiatric disorders, with estimates of point prevalence associated short- and long-term impairment, there is a
generally falling in the 3% to 13% range, depending on critical need for developing reliable and valid methods
the degree of associated impairment (see Costello and for measuring and monitoring the severity of these dis-
Angold, 1995, for review). These disorders can severely orders and their symptoms. The primary objective in
impair the social, academic, familial, and personal func- developing the Pediatric Anxiety Rating Scale (PARS)
tioning of affected children and negatively impact their was to address this need, particularly for clinician-rated
families (e.g., Beidel et al., 1991; Francis et al., 1992; assessments in treatment studies.
Strauss et al., 1988). In addition, existing longitudinal Currently, the most common methods for identifying
data, coupled with adults’ retrospective reports, suggest and measuring anxiety in youths are structured and semi-
some continuity between child and adult disorders (Abe, structured diagnostic interviews, self-report rating scales,
1972; Costello and Angold, 1995; Pine et al., 1998). In and clinician-rated instruments. Each of these methods
is useful for specific purposes. For instance, structured
diagnostic interviews, such as the Diagnostic Interview
Accepted April 10, 2002.
The study sites and authors are listed at the end of the text. The PARS and Schedule for Children (Shaffer et al., 2000), were pri-
instructions for its administration are available via the Article Plus feature on marily developed for use in epidemiological research to
the Journal’s Web site: www.jaacap.com. make diagnoses and are most useful for reliably ascer-
Supported by contracts from the NIMH to Johns Hopkins University
(N01MH60016; PI: Dr. Riddle) and the Research Foundation for Mental Hygiene taining “externalizing” diagnoses, e.g., conduct disorder.
(N01MH60005; PI: Dr. Greenhill) and Solvay Pharmaceuticals; NIMH grants Semistructured interviews, such as the Anxiety Disorders
MH57503 and MH-01391 (Dr. Pine); NIMH RUPP contract N01MH70010 Interview Schedule for Children (ADIS-C) (Silverman
(Dr. McCracken); and National Center for Research Resources/NIH General Clinical
Research Center grant M01 RR00052 (Johns Hopkins University School of Medicine).
and Albano, 1996), the Schedule for Affective Disorders
Reprint requests to Dr. Riddle, Division of Child and Adolescent Psychiatry, and Schizophrenia for School-Age Children (K-SADS)
Johns Hopkins Children’s Center, Room 346, 600 North Wolfe Street, Baltimore, (Ambrosini, 2000), and the Diagnostic Interview for
MD 21287-3325; e-mail: [email protected].
Children and Adolescents (Welner et al., 1987), are used
0890-8567/02/4109–1061䉷2002 by the American Academy of Child and
Adolescent Psychiatry. predominantly in clinical research and are particularly
DOI: 10.1097/01.CHI.0000020259.43550.F4 useful for making diagnoses. These instruments also facil-

J . A M . A C A D . C H I L D A D O L E S C . P S YC H I AT RY, 4 1 : 9 , S E P T E M B E R 2 0 0 2 1061
RUPP ANXIETY STUDY GROUP

itate the gathering of information on a broad range of for review) highlight the limitation of this method for
symptoms, impairment associated with specific diagnoses, assessing the overall severity of anxiety symptoms. This
duration of illness, and lifetime occurrences of illness. issue is particularly relevant with respect to assessing global
Newer versions of these instruments, e.g., the ADIS-C, changes in anxiety severity in the context of treatment
assess severity of individual anxiety disorders and have outcome studies that include children with more than
been used as an outcome measure in treatment studies one anxiety disorder (Greenhill et al., 1998).
(Kendall et al., 1997; Silverman et. al., 1999). In summary, there are several methods for assessing
Self-report rating scales are most useful for screening and anxiety symptoms in youths, each useful for specific pur-
assessing the degree to which a child’s anxiety is normative poses. However, to date there are no clinician-rated instru-
in the context of developmental, gender, and cultural fac- ments for assessing both impairment and overall anxiety
tors. In addition, these rating scales are useful for moni- severity that cover the important domains of anxiety in
toring changes in anxiety symptoms over time or in relation youths. To fill this gap, the development of a new instru-
to specific treatments. Examples of anxiety scales that assess ment was undertaken. The PARS was designed to assess
a broad range of symptoms include the Revised Children’s the frequency, severity, and associated impairment of sep-
Manifest Anxiety Scale (Reynolds and Richmond, 1978), aration anxiety, social phobia, and generalized anxiety
the Multidimensional Anxiety Scale for Children (MASC) symptoms in youths aged 6–17 years. The purpose of the
(March et al., 1997), and the Screen for Child Anxiety present study is to describe this instrument and present
Related Emotional Disorders (SCARED) (Birmaher et al., initial data on its psychometric properties.
1997). Despite their advantages, self-report rating scales are
dependent on the child’s cognitive abilities (i.e., to read and METHOD
comprehend items), do not directly assess impairment, and
Construction and Description of the PARS
are vulnerable to response biases inherent in self-report mea-
The PARS and instructions for its administration are available via
sures (Anastasi and Urbina, 1997). the Article Plus feature on the Journal’s Web site (www.jaacap.com).
In contrast to parent reports or child-completed self- It was modeled after the CY-BOCS and the Yale Global Tic Severity
reports of anxiety, few clinician-rated instruments exist Scale (YGTSS) (Leckman et al., 1989). Item content was derived as
for assessing the severity of symptoms of pediatric anxi- follows. First, items were generated predominantly from the DSM-
IV criteria for anxiety disorders including social phobia (SoP), sepa-
ety disorders. Among the available instruments are the ration anxiety disorder (SAD), and generalized anxiety disorder (GAD).
Hamilton Anxiety Rating Scale (HAM-A) (Hamilton, Second, items were circulated among a group of expert child and ado-
1959) and the Children’s Yale-Brown Obsessive Compulsive lescent psychiatrists and psychologists at the participating sites of the
Research Units on Pediatric Psychopharmacology (RUPP) who were
Scale (CY-BOCS) (Scahill et al., 1997). The HAM-A asked to comment on (1) face validity; (2) comprehensiveness; (3)
assesses a broad range of anxiety symptoms; however, overlap with other constructs, for example, disruptive behavior or
they are predominantly physiological or somatic in nature. depression; and (4) readability and clarity. After several iterations of
this process, a 50-item “final version” was adopted for use in the RUPP
More important, this measure was originally developed fluvoxamine anxiety trial. This 50-item symptom checklist grouped
for use with adults; the psychometric properties have been items into the following categories: Social Interactions or Performance
examined in adolescents (Clark and Donovan, 1994), but Situations (9 items), Separation (10 items), Generalized (8 items),
it is unclear to what extent it accurately reflects current Specific Phobia (4 items), Physical Signs and Symptoms (13 items),
and Other (6 items). Each symptom is scored by the interviewing
concepts regarding anxiety in pediatric samples. clinician as present or absent (yes/no) during the previous week.
Although semistructured and structured diagnostic inter- Information about symptoms is gathered from interviews with both
views, such as the ADIS-C, are clinician-administered the child and parent(s). These interviews may be conducted with child
and parent together, as was done in the present study, or separately.
and may assess severity of individual anxiety symptoms Endorsed symptoms are then collectively (i.e., integrating both child
or disorders, these instruments typically do not provide and parent information) rated by the clinician on seven dimensions
an index of anxiety severity across disorders. Rather, these of severity, using a 6-point scale (0 for none, and 1–5 for minimal to
extreme) for each dimension. These dimensions are (1) number of
instruments link severity to one specific symptom or dis- symptoms (none to more than 10), (2) frequency (none to several
order (e.g., social phobia) based on the current (or a past) hours per day), (3) severity of distress associated with anxiety symp-
version of the DSM. Changes in DSM criteria for child- toms (none to extreme), (4) severity of physical symptoms (none to
hood anxiety disorders, the overlap of symptoms among extreme), (5) avoidance (none to extreme), (6) interference at home
(no interference to totally or almost totally unable to function at
anxiety disorders, and the high rates of comorbidity among home), and (7) interference out of home (no interference to totally
anxiety disorders in youths (see Curry and Murphy, 1995, or almost totally unable to function out of home such as with peers

1062 J . A M . A C A D . C H I L D A D O L E S C . P S YC H I AT RY, 4 1 : 9 , S E P T E M B E R 2 0 0 2
 PEDIATRIC ANXIETY RATING SCALE

or at school). A score of 3 on each of these 5-point scales indicates a TABLE 1

clinically significant level of severity, avoidance, or interference. Sociodemographic and Clinical Characteristics
A PARS Total Score was calculated by summing five of the seven of Entire Sample (N = 128)
items. The “Number of Symptoms” item and the “Physical Symptoms”
Mean age (years) 10.8
item were excluded. The “Number of Symptoms” item was excluded
Gender (female) 63 (49.2)
because of concern that it might not be related to severity of anxiety
Ethnicity
and might be highly skewed in this sample. The “Physical Symptoms”
White 81 (63.3)
item was excluded because of concern that outcome data included rat- Hispanic 24 (18.8)
ings on subjects who were being treated with a selective serotonin reup- African American (non-Hispanic) 9 (7.0)
take inhibitor (SSRI). Common adverse events of SSRIs include Other 14 (10.9)
symptoms that are identical with some of the symptoms in the PARS Total family income
Symptom Checklist: “sweating”; “feels sick to stomach, nausea or <$25,000 19 (14.8)
abdominal distress”; “restlessness or feeling keyed-up or on edge”; “sleep $25,000–39,999 20 (15.6)
disturbance, especially difficulty falling asleep.” Thus, throughout this $40,000–59,000 18 (14.1)
article, PARS Total Score refers to a sum of the remaining five items. >$60,000 55 (43.0)
Refused/unknown 16 (12.5)
Participants Diagnosis (based on K-SADS)
The current study was conducted in two stages. The first stage SoP only 26 (20.3)
examined the interrater reliability of the PARS, while the second stage SAD only 18 (14.1)
examined test-retest reliability and validity. To examine interrater reli- GAD only 5 (3.9)
ability, 16 children aged 5–15 years (63% males; mean age 10.5 years) SoP and SAD only 11 (8.6)
were interviewed with the PARS. Of these children, 11 were white, SoP and GAD only 21 (16.4)
1 was African American, 2 were Hispanic American, and 2 were Asian SAD and GAD only 21 (16.4)
American. Seven of these children also participated in the RUPP anx- SoP, SAD, and GAD 26 (20.3)
iety treatment study (described below). Note: With the exception of age, values represent n (%). K-SADS =
The second stage of this study was conducted within the context of Schedule for Affective Disorders and Schizophrenia for School-Age
a double-blind, placebo-controlled clinical trial of fluvoxamine for Children; SoP = social phobia; SAD = separation anxiety disorder;
youths with SoP, SAD, and GAD (RUPP, 2001). Five RUPP sites were GAD = generalized anxiety disorder.
involved in that study: Duke University, Johns Hopkins University,
New York State Psychiatric Institute/Columbia University, New York
University, and University of California, Los Angeles. Participants scale ranges from 1 (lowest) to 100 (highest). Ample evidence exists sup-
included 128 children, aged 6–17 years (mean = 10.8 years; 49% female; porting the psychometric soundness of the scale (Green et al., 1994).
63% white), with a current DSM-IV diagnosis of SoP, SAD, or GAD, Clinician’s Global Impression of Severity Scale. The Clinician’s Global
based on child and parent K-SADS interview with a trained clinician. Impression of Severity Scale (CGI-S) (Guy, 1976) was used to assess
Sites recruited participants from a variety of sources, including men- global severity of anxiety using a 7-point scale, ranging from normal
tal health settings and newspaper advertisements. Other inclusion cri- to most extreme. This measure was completed at major assessment
teria were as follows: clinically significant impairment as assessed with points by a clinician to assess changes in anxiety severity.
a Children’s Global Assessment Scale (CGAS) score less than 60; and Clinician’s Global Impression of Improvement Scale. The Clinician’s
availability of both a parent and the child for weekly visits. Exclusion Global Impression of Improvement Scale (CGI-I) was used to assess
criteria included current use of any illicit or prescribed psychoactive overall global change at each visit using data from all other measures
substance; current diagnoses of either major depressive disorder, Tourette’s in the study. An 8-point scale (Abikoff and Gittelman, 1985; Klein
disorder, obsessive-compulsive disorder, posttraumatic stress disorder, et al., 1992) that is similar to the traditional 7-item scale (Guy, 1976),
conduct disorder, or panic disorder; any past or current history of but allows for greater discrimination among levels of response in treat-
mania, psychosis, or pervasive developmental disorder; current sui- ment responders, was used (1 = completely well, recovered; 2 = much
cidal ideation; mental retardation as assessed with the Kaufman Brief improved; 3 = improved; 4 = slightly improved; 5 = unchanged; 6 =
Intelligence Test (IQ < 70); previous use of an SSRI in appropriate slightly worse; 7 = worse; 8 = much worse).
doses; and a diagnosis of attention-deficit/hyperactivity disorder that Multidimensional Anxiety Scale for Children. The MASC (March
required pharmacological treatment. Table 1 summarizes the charac- et al., 1997) is a 39-item, 4-point, Likert, self-report rating scale that
teristics of participants in the entire sample. has shown robust psychometric properties in studies of clinical and
nonclinical samples. The MASC includes four factors derived to match
Measures the population factor structure of anxiety: physical symptoms (tense/
restless and somatic/autonomic subfactors), social anxiety (humiliation/
Schedule for Affective Disorders and Schizophrenia for School-Age rejection and public performance subfactors), harm avoidance (anxious
Children. Inclusion/exclusion criteria for psychiatric diagnoses were coping and perfectionism subfactors), and separation/panic anxiety.
assessed with the K-SADS (Ambrosini, 2000; Kaufman et al., 1997), Three-week test-retest reliability for the MASC is good as shown by
administered by experienced clinicians. The K-SADS possesses good intraclass correlation coefficient (ICC) = 0.79 in clinical (March et al.,
reliability and validity (Ambrosini, 2000). The entire K-SADS inter- 1997) and 0.88 in school-based samples (March and Sullivan, 1999).
view was administered during intake into the treatment study. Screen for Child Anxiety Related Emotional Disorders. The SCARED
Children’s Global Assessment Scale. The CGAS (Bird et al., 1987; Shaffer (Birmaher et al., 1997, 1999) is a 41-item, child- and parent-report
et al., 1983) is a modification of the adult GAS and provides a measure instrument that assesses DSM-IV symptoms of panic, SAD, SoP, and
of global impairment and functioning over the previous month. The GAD and symptoms of school refusal. The SCARED has shown good

J . A M . A C A D . C H I L D A D O L E S C . P S YC H I AT RY, 4 1 : 9 , S E P T E M B E R 2 0 0 2 1063
RUPP ANXIETY STUDY GROUP

psychometric properties in two different large clinical samples and 2 clinical child psychologists (this group of raters participated in
(Birmaher et al., 1997, 1999) and in a community sample (Muris a training session on the use of the PARS and completed all assess-
et al., 1998, 1999). For instance, data obtained on an outpatient clinic ments described above); however, 5 of the 11 raters were unable to
sample of 341 youths (9–18 years, mean = 14.5) indicated test-retest rate all subjects. Of these five, two raters were available for seven video-
reliability coefficients ranging from 0.70 to 0.90 (over a 4-day to 15- taped interviews only; one rater was available for the seven videotaped
week period) and internal consistency coefficients of the subscales interviews and two live interviews; and one rater was available for
ranged from 0.74 to 0.89 (Birmaher et al., 1997). The SCARED seven videotaped interviews and six live interviews.
demonstrated good sensitivity and specificity in a study of children During the treatment phase, the treating child and adolescent psy-
with anxiety, depressive, and disruptive disorders, and it was sensitive chiatrist, who was blind to treatment condition, saw each child and
to treatment effects (Birmaher et al., 1999). a parent on a weekly basis for the first 6 weeks of the trial and again
Children’s Depression Rating Scale. The Children’s Depression Rating at week 8. At each visit, psychiatrists completed three tasks: (1) they
Scale (CDRS) (Poznanski et al., 1984) is a clinician-administered administered the supportive psychoeducational therapy to children
screening tool that assesses 17 symptom areas of depression (e.g., sleep, and families, (2) they conducted a clinical assessment of anxiety symp-
social withdrawal, school). The CDRS is widely used and has estab- toms using the PARS and other instruments, and (3) they assessed
lished psychometric properties. For instance, test-retest reliability over adverse events. The rating scales described above were also completed
a 2-week period was 0.86, interrater reliability was 0.92 (for the sum- at week 4 and week 8 treatment visits.
mary score), and the Cronbach α was .85. Moreover, the CDRS dis-
criminates between depressed and nondepressed psychiatric patients Primary Statistical Analyses
and is correlated with other measures of depression and suicidal behav-
ior (Poznanski et al., 1984). Reliability. Internal consistency was calculated using the Cronbach
Hamilton Anxiety Rating Scale. The HAM-A (Hamilton, 1959) is α for the five items (see below) that comprise the PARS Total Score.
a 14-item, clinician-administered measure of anxiety symptoms with ICCs (Shrout and Fleiss, 1979) were calculated for the Total Score
a focus on somatic symptoms. The measure has demonstrated relia- and individual severity items to assess interrater and test-retest relia-
bility and construct validity with adults and adolescents. Specifically, bility, which was examined over the 1- to 4-week period between
in a sample of adolescents, interrater reliability was 0.92 and the screening and baseline.
Cronbach α was .86 (Clark and Donovan, 1994). The HAM-A was Validity. Convergent validity was evaluated by calculating Pearson
also correlated with self-report measures of anxiety and fear and dis- correlation coefficients for the PARS with other anxiety measures includ-
criminated between nonclinical, clinical, and clinically anxious youths. ing the MASC, SCARED, HAM-A, and CBCL (i.e., Anxious/Depressed
Child Behavior Checklist. The Child Behavior Checklist (CBCL) and total Internalizing scores). To examine divergent validity, we exam-
(Achenbach, 1991), a widely used parent-completed measure, assesses ined the correlation between the PARS Total Score and scores on the
a broad range of child behavior problems. The measure yields a Total CDRS and the CBCL Externalizing and Inattention subscale scores.
Problem Behavior score, two broad-band scores (Internalizing and Statistically significant and relatively large correlations were expected
Externalizing), and several narrow-band scale scores (e.g., Anxious/ for convergent, but not divergent, validity. Differences between depen-
Depressed, Inattention, etc.). dent correlations were tested using the method described by Cohen and
Cohen (1975). To assess the PARS sensitivity to change, we examined
Procedure the correspondence of changes in PARS scores to changes in the CGI-
I, CGI-S, CGAS, and other measures of anxiety before and after the 8-
Prior to participation in this study, all parents signed informed con- week, double-blind, placebo-controlled study of fluvoxamine.
sents; children over age 6 signed assents. A comprehensive pretreat- All children were included in each analysis as long as the data were
ment evaluation was conducted which included the diagnostic interview not missing. Systematic missing data were not a problem, although
and additional measures to assess inclusion/exclusion criteria and out- missing data were evident for most measures. Statistical significance
come measures. The pretreatment evaluation was completed over three is reported at the 5% level.
visits, which spanned a 3- to 4-week period of time. Generally, at the
first visit the K-SADS, PARS, CGI, CGAS, and HAM-A were admin-
istered. On the second visit, parents completed the CBCL and med- RESULTS
ical assessments were conducted (e.g., medical history, physical
examination, blood draw). On the third visit, demographics, family Internal Validity
history, and intellectual assessments were conducted. At baseline, the
CDRS, K-SADS, PARS, HAM-A, CGI, CGAS, SCARED, MASC, Descriptive Statistics. The means and standard devia-
other self-report and medical procedures were completed. Thus, prior tions of the PARS scores for the seven individual severity
to treatment, the PARS was administered on two separate occasions
by the same pediatric psychiatrist or psychologist: at the screening items and the five-item Total Score for boys, girls, and the
evaluation (the beginning of the 3-week period) and at the baseline total sample are presented in Table 2. In general, the means
(randomization) evaluation (the end of the 3- to 4-week period). At (and modes) for the PARS individual items were above
each visit, the PARS was conducted prior to all other rating scales.
For the interrater reliability part of the study, live interviews were
the mid-point (2.5) and slightly positively skewed on the
conducted and videotaped. Clinicians who were present during the 6-point (0–5) scale. The PARS Total Score was unimodal
interview provided ratings. All other ratings were from clinicians who and normally distributed. A comparison of PARS Total
watched the videotaped session. During the interviews, children were Scores between younger and older children using the total
administered the PARS first, then parents. Parents and children were
brought together to sort out significant discrepancies. There were a sample showed no significant differences. The t test com-
total of 11 raters from the 5 sites, 9 child and adolescent psychiatrists parisons between boys and girls on the PARS Total revealed

1064 J . A M . A C A D . C H I L D A D O L E S C . P S YC H I AT RY, 4 1 : 9 , S E P T E M B E R 2 0 0 2
 PEDIATRIC ANXIETY RATING SCALE

TABLE 2 from 0.08 (Interference-Home and Interference-Other)

Mean Scores (and Standard Deviations) on the PARS at Baseline and 0.50 (Avoidance and Interference-Other).
Boys Girls Total
PARS Item (n = 64) (n = 63) (n = 127) Reliability

Number of Symptoms 4.8 (0.6) 4.8 (0.5) 4.8 (0.6) Interrater Reliability. The ICC for the PARS Total Score
Symptom Frequency 3.9 (1.0) 3.9 (1.0) 3.9 (1.0) across raters was 0.97. Table 4 displays the ICCs for each
Severity-Distress 4.0 (0.8) 4.0 (0.7) 4.0 (0.7)
Severity-Somatic 2.8 (1.2) 2.6 (1.5) 2.7 (1.4)
of the seven severity items.
Avoidance 3.8 (1.0) 4.0 (0.8) 3.9 (0.9) Test-Retest Reliability. The test-retest reliability coeffi-
Interference-Home 3.4 (1.2) 3.4 (1.0) 3.4 (1.1) cient between screen and baseline (mean = 24.7 ± 14.7
Interference-Other 3.4 (1.2) 3.1 (1.1) 3.5 (1.2) days) for the PARS Total Score was 0.55 for the total sam-
Total (5-item)a 18.5 (3.5) 18.8 (3.0) 18.7 (3.21)
ple. ICCs for each item ranged from 0.35 (Avoidance)
Note: Items range from 0 (none) to 5 (extreme); Total Score range to 0.59 (Interference-Home). These coefficients are pre-
is 0 to 25. PARS = Pediatric Anxiety Rating Scale. sented in Table 5.
a
Two items, Number of Symptoms and Physical Symptoms, were
excluded from five-item Total. Validity: Convergent and Divergent

To assess convergent validity, we calculated Pearson

no significant differences—both for the total sample and correlations between the PARS Total Score at baseline
within younger (ages 6–12) or older (ages 13–17) age and other measures of anxiety, including (1) CGI-S, (2)
groups. No differences were found on any PARS items HAM-A, (3) MASC, (4) SCARED-child and -parent
between whites and Hispanics (the largest minority group versions, and (5) CBCL Anxious/Depressed subscale and
in the sample). The correlation between PARS Total Scores Internalizing subscales. These results are presented in
and income was nonsignificant (r = –0.15). Table 6. The PARS Total score was moderately correlated
Internal Consistency. The Cronbach α coefficient for with the other clinician- and parent-rated measures of
the five items comprising the PARS Total Score at base- anxiety, such as the CGI-S (clinician rating of anxiety
line was .64, indicating that the PARS items are associ- severity), HAM-A, and SCARED-P. Correlations were
ated but not redundant. Correlations among the PARS generally weakest with the two child self-report measures
seven items and the PARS five-item Total Score severity of anxiety (MASC and SCARED-C). In additional analy-
items at baseline are presented in Table 3. Correlation ses, each PARS severity item was correlated with the above
coefficients among the seven severity items ranged from measures (correlations available from the authors). Using
0.08 (between total number of symptoms and interfer- a minimum criterion of r = 0.40, results indicated that
ence out of home) and 0.45 (between interference at the CGI-S was correlated with PARS Distress (0.41) and
home and distress). Most of the seven individual sever- Interference-Out of Home (0.55), and the HAM-A was
ity items were moderately correlated with the PARS Total correlated with Interference-Home (0.55), Physical
Score (range, 0.24–0.68). Correlations among the indi- Symptoms (0.46), and Distress (0.43). No other corre-
vidual PARS severity items were quite varied, ranging lations surpassed this criterion.

TABLE 3
Correlations Among PARS Seven Items and PARS Five-Item Total Score at Baseline
1 2 3 4 5 6 7 8

1. Number of Symptoms 1 0.34*** 0.41*** 0.18* 0.20* 0.36*** 0.08 0.41***

2. Frequency 1 0.37*** 0.13 0.13 0.39*** 0.15 0.63***
3. Distress 1 0.32*** 0.18* 0.45*** 0.34*** 0.68***
4. Physical Symptoms 1 0.06 0.14 0.19* 0.24*
5. Avoidance 1 0.19* 0.50*** 0.62***
6. Interference-Home 1 0.08 0.65***
7. Interference-Other 1 0.66***
8. Total Score 1

Note: PARS = Pediatric Anxiety Rating Scale.

* p < .05; *** p < .001.

J . A M . A C A D . C H I L D A D O L E S C . P S YC H I AT RY, 4 1 : 9 , S E P T E M B E R 2 0 0 2 1065
RUPP ANXIETY STUDY GROUP

TABLE 4 TABLE 5
ICC Interrater Reliability Coefficients on the PARS ICC Test-Retest Coefficients and Means at Screen
and Baseline on the PARS
PARS ICC a
PARS ICC a Screen (SD) Baseline (SD)
Number of Symptoms 0.78
Frequency 0.92 Number of Symptoms 0.48 4.6 (0.8) 4.8 (0.6)
Distress 0.96 Frequency 0.42 3.7 (1.1) 3.9 (1.0)
Physical Symptoms 0.89 Distress 0.37 3.8 (0.6) 4.0 (0.7)
Avoidance 0.92 Physical Symptoms 0.45 2.8 (1.3) 2.7 (1.4)
Interference-Home 0.89 Avoidance 0.35 3.7 (0.9) 3.9 (0.9)
Interference-Other 0.92 Interference-Home 0.59 3.4 (1.0) 3.4 (1.1)
Total (5-item) 0.97 Interference-Other 0.54 3.3 (1.2) 3.5 (1.2)
Total Score (5-item) 0.55 18.0 (3.1) 18.7 (3.21)
Note: PARS = Pediatric Anxiety Rating Scale; ICC = intraclass cor-
relation coefficient. Note: PARS = Pediatric Anxiety Rating Scale; ICC = intraclass cor-
a
All ICC values significant at p < .001. relation coefficient.
a
All ICC values significant at p < .001.

Divergent validity was assessed in a preliminary man- clinician-rated anxiety scale (HAM-A), and (4) self-reports
ner by correlating the PARS Total Score at baseline with and parent reports of anxiety (i.e., MASC, SCARED-C,
the baseline CDRS and the CBCL Inattention and and SCARED-P). Change in the PARS Total Score was
Externalizing subscales. These correlations, which appear significantly correlated with pre-post treatment changes
in Table 6, were relatively small (0.18–0.30). A test of in the CGAS (0.78), CGI-S (0.53), and HAM-A (0.41),
the difference in correlations between the PARS Total but uncorrelated with changes in the MASC (0.03),
Score and the CBCL Internalizing (0.40) and CBCL SCARED-C (0.02), or SCARED-P (0.03).
Externalizing (0.22) scales revealed that the correlation
between the PARS and the CBCL Internalizing scale was DISCUSSION
significantly larger (t119 = 2.23, p = .03). The purpose of this study was to describe the devel-
Sensitivity to Treatment. We also examined correlations opment and initial psychometric properties of the PARS,
between change in the PARS Total Score and change after a clinician-rated instrument for assessing the severity of
treatment in (1) global assessment of functioning (CGAS), symptoms common to a broad range of pediatric anxi-
(2) clinician-rated global anxiety severity (CGI-S), (3) ety disorders. The PARS was designed to fill a gap among

TABLE 6
Correlations Between PARS Total Score and Measures of Internalizing and Externalizing Behaviors at Baseline
Scale Construct Source Mean (SD) R
Convergent validity
CGI-S Anxiety Clinician 5.06 (0.81) 0.61***
HAM-A Anxiety Clinician 17.25 (7.67) 0.49***
SCARED-P Anxiety Parent 36.60 (13.38) 0.46***
CBCL-I Internalizing Parent 20.36 (9.77) 0.41***
CBCL-A Anxiety Parent 11.30 (5.48) 0.36***
SCARED-C Anxiety Child 31.52 (13.79) 0.32***
MASC Anxiety Child 54.27 (18.58) 0.22*
Divergent validity
CDRS-P Depression Clinician 30.44 (10.52) 0.33***
CBCL-In Inattention Parent 6.93 (4.60) 0.27**
CBCL-E Externalizing Parent 13.16 (9.68) 0.22*
CDRS-C Depression Clinician 28.49 (9.16) 0.18*

Note: CGI-S = Clinician’s Global Impression of Severity Scale; HAM-A = Hamilton Anxiety Rating Scale; SCARED-P = Screen
for Child Anxiety Rated Emotional Disorders-Parent; CBCL-I = Child Behavior Checklist-Internalizing; CBCL-A = Child Behavior
Checklist-Anxiety; SCARED-C = Screen for Anxiety Related Emotional Disorders-Child; MASC = Multidimensional Anxiety
Scale for Children; CDRS-P = Children’s Depression Rating Scale-Parent; CBCL-In = Child Behavior Checklist-Inattention;
CBCL-E = Child Behavior Checklist-Externalizing; CDRS-C = Children’s Depression Rating Scale-Child.
* p < .05; ** p < .01; *** p < .001.

1066 J . A M . A C A D . C H I L D A D O L E S C . P S YC H I AT RY, 4 1 : 9 , S E P T E M B E R 2 0 0 2
 PEDIATRIC ANXIETY RATING SCALE

existing child anxiety assessment tools, and our findings culating the PARS Total Score. Alternatively, because the
suggest it holds considerable promise as both a research PARS Total Score contains the other five items, it is pos-
and clinical instrument. sible that the other five correlations are artificially elevated.
As noted earlier, correlations among the individual PARS
Descriptive Statistics and Selection of Severity Items
severity items were quite varied, ranging from 0.08
Modeled after the CY-BOCS and YGTSS, the current (Interference-Home and Interference-Other) and 0.50
version of the PARS has 50 symptom items and 7 severity/ (Avoidance and Interference-Other). The pattern of cor-
impairment items. A total score was obtained by sum- relations suggests that children who report a high number
ming five of the severity items. For reasons presented in of symptoms are more likely to experience them more often
the “Method” section, two items, “Number of Symptoms” and report more distress and interference at home. In addi-
and “Severity of Physical Symptoms,” were not included tion, avoidance due to anxiety was most closely linked to
in the calculation of the total score. Although the distri- higher levels of distress and interference outside of the
bution of the PARS Total Score was “normal,” most of home (e.g., with friends, at school). Finally, the absence of
the individual severity items were positively skewed. For a relation between the two interference items suggests that
instance, the modal score for the “Number of Symptoms” children’s impairment may be domain-specific.
item was 5.0, indicating that most youths with SoP, SAD,
and/or GAD in the present sample had greater than 10 Reliability
symptoms. Consequently, this item should be dropped In general, the PARS demonstrated acceptable psy-
from the PARS or new anchors need to be established chometric properties with respect to interrater reliability
and this item then needs to be reevaluated. The “Severity (high, ICC = 0.97), internal consistency (adequate, α =
of Physical Symptoms” item was dropped from the total .64), and test-retest reliability (fair, ICC = 0.55). Regarding
score because many of the physical symptoms of anxiety test-retest reliability, to some extent these lower coeffi-
are similar to adverse events of psychotropic medications, cients reflect expected instability of anxiety symptoms
particularly the SSRIs (e.g., “trembling and shaking,” and associated impairment in children. However, during
“sweating,” “feels sick to stomach, nausea or abdominal the testing interval psychoeducational instruction in cop-
distress”). Obviously, for assessment of anxiety in chil- ing was administered, which may have contributed to the
dren not receiving medication, this would not be an issue. instability in these scores. Overall, the PARS reliability fares
Thus further work is needed to clarify the value of includ- well in comparison with other self-, parent-, and clinician-
ing physical symptoms in the PARS Total Score. rated measures of anxiety and suggests that the instru-
Comparisons between boys and girls and younger and ment will produce consistent scores when used by different
older children on the PARS Total Score revealed no sig- (trained) examiners and will yield a relatively stable assess-
nificant differences. The absence of gender and age dif- ment of anxiety over time.
ferences may reflect the nature of the present sample. Both
Validity
boys and girls in this study met strict inclusion/exclusion
criteria and thus were selected to surpass a prespecified Convergent/Divergent. Correlations between the PARS
level of severity of anxiety. Total Score and indices of anxiety, depression, and exter-
Correlations among the PARS severity items and PARS nalizing behaviors provide preliminary support for the
Total Score varied, ranging from 0.08 (between total num- instrument’s convergent and divergent validity. The PARS
ber of symptoms and interference out of home) to 0.45 Total Score showed the strongest correlation with the
(between interference at home and anxiety severity). The CGI-S—a clinician-rated measure of global anxiety sever-
five individual severity items that comprise the Total Score ity and the measure closest to the underlying construct
were uniformly correlated with the PARS Total Score assessed by the PARS (i.e., global severity of anxiety). As
(range, 0.62–0.68). Of note, the two omitted items, expected, the PARS was also correlated with other clini-
“Number of Symptoms” and “Severity of Physical cian- and parent-rated measures of anxiety such as the
Symptoms,” showed the lowest correlation with the PARS HAM-A, SCARED-P, and CBCL Internalizing score. In
Total Score, suggesting that they may be a poor proxy of contrast, correlations between the PARS and measures
overall severity or impairment. The low correlations for of externalizing behavior, inattention, and depression
these two items support the decision to omit them in cal- (CDRS-C) were low, supporting divergent validity.

J . A M . A C A D . C H I L D A D O L E S C . P S YC H I AT RY, 4 1 : 9 , S E P T E M B E R 2 0 0 2 1067
RUPP ANXIETY STUDY GROUP

Correlations between each PARS individual severity symptoms of SoP, he or she is likely also to have symptoms
item and measures assessing convergent validity revealed associated with SAD and GAD. Thus the PARS will enable
that the PARS Total Score had the strongest correlation researchers to assess the impact of treatment on symptoms
for four of the seven measures (CGI-S, SCARED-P, CBCL of several anxiety disorders. Finally, because the PARS is
Internalizing, CBCL Anxious/Depressed). However, the easy to administer and time-efficient, it will be useful for
HAM-A was more strongly correlated with Interference- clinicians in a variety of settings (e.g., psychiatry, psychol-
Home (r = 0.55) than the PARS Total Score. Finally, the ogy, primary care, schools, etc.). Administration of the PARS
SCARED-C and the MASC, the two child-report mea- generally takes about 30 minutes.
sures of anxiety, had a stronger association (r = 0.38 and
Limitations and Directions of Future Research
r = 0.27, respectively) with physical symptoms compared
with the PARS Total Score. This finding has several inter- The promising findings of this study must be inter-
pretations. For instance, it may be that children are more preted in the context of several methodological limita-
able to endorse somatic complaints, which are more con- tions. First, the study was done with a clinically anxious
crete, than report “cognitive” items related to severity or sample that was selected on the basis of stringent inclu-
impairment. Alternatively, this finding may reflect that sion/exclusion criteria. Consequently, they may have been
these self-report measures contain more somatic symp- more severely affected than other clinically anxious sam-
tom items relative to interference or severity. Related, the ples given their interest in participating in a medication
lack of correspondence between the PARS Total Score treatment study, with fewer comorbid nonanxiety disor-
and the MASC may reflect that the clinician-raters relied ders and higher socioeconomic status than families seen
more on parent than child report in determining anxiety in community mental health centers. Therefore, whether
severity, particularly for younger children. Additional the results are generalizable to community or other clin-
research is needed to test these hypotheses. ical samples is unknown. Second, there is a lack of data
Sensitivity to Treatment. Finally, the PARS also showed on discriminant validity. The next step in the develop-
sensitivity to treatment effects. Specifically, the PARS ment of the PARS is to establish appropriate norms for
showed a reduction in severity from pre- to posttreatment community and nonanxious clinical samples, such as
(RUPP, 2001). Moreover, the change in PARS scores par- youths suffering from depression or disruptive behavior
alleled changes in other measures of global severity (CGI- disorders. Such data are necessary to determine scores that
S) and specific anxiety symptoms (HAM-A), although fall in the clinical range and to determine whether the
this may be due, in part, to shared rater effects. The absence PARS can successfully discriminate between youths from
of a significant correlation between changes in PARS Total nonclinical, clinical nonanxious, and anxious groups.
Scores and changes in the MASC and SCARED from pre- With respect to future directions, additional research
to posttreatment was also unexpected. This finding may and revisions will be needed on the instrument’s individ-
reflect the omission of the physical symptoms item from ual items. These revisions will likely address the follow-
the PARS Total Score, given that these measures were ing issues: (1) examining the utility of the individual
found to be more highly associated with this item than severity items and reexamining the anchors associated with
the PARS Total Score. each item in light of possible ceiling effects; (2) examin-
ing the utility of the list of physical symptoms in order to
Clinical Applications
determine how they should be grouped and to differen-
The PARS can be used for monitoring the course of anx- tiate medication side effects from physical symptoms asso-
iety symptoms and assessing the impact of treatments on ciated with an anxiety disorder; and, perhaps, (3) expanding
the severity and interference of anxiety symptoms in youths. the symptom list to include symptoms of obsessive-
One advantage of the PARS relative to other clinician-rated compulsive disorder, posttraumatic stress disorder, and
pediatric anxiety instruments is that it allows the clinician/ acute stress disorder, to include all anxiety disorders in one
researcher to assess the severity and impairment of anxiety instrument. Nevertheless, the PARS represents a useful
symptoms across a broad spectrum of anxiety disorders. measure for the assessment of childhood anxiety.
This is significant given the high degree of comorbidity and
symptom overlap among anxiety disorders in youths. For The authors of this report, listed by study site: Mark A. Riddle, M.D., Golda
example, although a child may have most difficulty with S. Ginsburg, Ph.D., John T. Walkup, M.D., Michael J. Labelarte, M.D.,

1068 J . A M . A C A D . C H I L D A D O L E S C . P S YC H I AT RY, 4 1 : 9 , S E P T E M B E R 2 0 0 2
 PEDIATRIC ANXIETY RATING SCALE

Johns Hopkins University; Daniel S. Pine, M.D., Mark Davies, M.P.H., Hamilton M (1959), The assessment of anxiety states by rating. Br J Med
Laurence Greenhill, M.D., Michael Sweeney, Ph.D., Rachel Klein, Ph.D., Psychol 32:50–55
Columbia University and New York State Psychiatric Institute; Howard Kaufman J, Birmaher B, Brent D et al. (1997), The Schedule for Affective
Abikoff, Ph.D., Sabine Hack, M.D., Brian Klee, M.D., New York University; Disorders and Schizophrenia for School-Age Children-Present and Lifetime
James McCracken, M.D., Lindsey Bergman, Ph.D., John Piacentini, Ph.D., Version (K-SADS-PL): initial reliability and validity data. J Am Acad Child
Adolesc Psychiatry 36:980–988
University of California at Los Angeles; John March, M.D., M.P.H., Scott
Kendall PC, Flannery-Schroeder E, Panicelli-Mindel SM, Southam-Gerow
Compton, Ph.D., Duke University; James Robinson, M.E.D., Thomas MA, Henin A, Warman M (1997), Therapy for youths with anxiety dis-
O’Hara, M.B.A., Sherryl Baker, Ph.D., Nathan Kline Institute; Benedetto orders: a second randomized clinical trial. J Consult Clin Psychol 65:366–380
Vitiello, M.D., Louise Ritz, M.B.A., Margaret Roper, M.P.H., National Klein R, Koplewicz H, Kanner A (1992), Imipramine treatment of children
Institute of Mental Health. Dr. Riddle accepts responsibility for the overall with separation anxiety disorder. J Am Acad Child Adolesc Psychiatry 31:21–28
content and integrity of the manuscript. Leckman JF, Riddle MA, Hardin MT et al. (1989), The Yale Global Tic Severity
Scale (YGTSS): initial testing of clinician-rated scale of tic severity. J Am
Acad Child Adolesc Psychiatry 28:566–573
REFERENCES March J, Parker J, Sullivan K, Stallings P, Conners CKK (1997), The Multi-
Abe K (1972), Phobias and nervous symptoms in childhood and maturity: dimensional Anxiety Scale for Children (MASC): factor structure, relia-
persistence and associations. Br J Psychiatry 120:275–283 bility, and validity. J Am Acad Child Adolesc Psychiatry 36:554–565
Abikoff H, Gittleman R (1985), Hyperactive children treated with stimulants: March JS, Sullivan K (1999), Test-retest reliability of the Multidimensional
is cognitive training a useful adjunct? Arch Gen Psychiatry 42:953–961 Anxiety Scale for Children. J Anxiety Disord 13:349–358
Achenbach TM (1991), Manual for the 1991 CBCL/4–18 YSR and TRF Profiles. Muris P, Meesters C, Merckelbach H, Sermon A, Zwakhalen S (1998), Worry
Burlington: University of Vermont Department of Psychiatry in normal children. J Am Acad Child Adolesc Psychiatry 37:703–710
Ambrosini PJ (2000), Historical development and present status of the Schedule Muris P, Merckelbach H, Schmidt H, Mayer B (1999), The revised version of
for Affective Disorders and Schizophrenia for School-Age Children (K- the Screen for Child Anxiety Related Emotional Disorders (SCARED-R):
SADS). J Am Acad Child Adolesc Psychiatry 39:49–58 factor structure in normal children. Pers Individ Diff 26:99–112
Anastasi A, Urbina U (1997), Psychological Testing. Upper Saddle River, NJ: Pine DS, Cohen P, Gurley D, Brook J, Ma Y (1998), The risk for early adult-
Prentice Hall hood anxiety and depressive disorders in adolescents with anxiety and
Beidel DC, Christ MG, Long PJ (1991), Somatic complaints in anxious chil- depressive disorders. Arch Gen Psychiatry 55:56–64
dren. J Abnorm Child Psychol 19:659–670 Poznanski EO, Grossman JA, Buchsbaum Y, Banegas M, Freeman J, Gibbons
Bird HR, Canino G, Rubio-Stipec M, Ribera JC (1987), Further measures of R (1984), Preliminary studies of the reliability and validity of the Children’s
the psychometric properties of the Children’s Global Assessment Scale. Depression Rating Scale. J Am Acad Child Psychiatry 23:191–197
Arch Gen Psychiatry 44:821–824 Research Units on Pediatric Psychopharmacology Anxiety Study Group (2001),
Birmaher B, Brent DA, Chiappetta L, Bridge J, Monga S, Baugher M (1999), Fluvoxamine for the treatment of anxiety disorders in children and ado-
Psychometric properties of the Screen for Child Anxiety Related Emotional lescents. N Engl J Med 344:1279–1285
Disorders (SCARED): a replication study. J Am Acad Child Adolesc Psychiatry Reynolds CR, Richmond B (1978), What I Think and Feel: a revised measure
38:1230–1236 of children’s manifest anxiety. J Abnorm Child Psychol 6:271–280
Birmaher B, Khetarpal S, Brent D et al. (1997), The Screen for Child Anxiety Scahill L, Riddle MA, McSwiggin-Hardin M et al. (1997), Children’s Yale-
Related Emotional Disorders (SCARED): scale construction and psy- Brown Obsessive Compulsive Scale (CY-BOCS): reliability and validity.
chometric characteristics. J Am Acad Child Adolesc Psychiatry 36:545–553 J Am Acad Child Adolesc Psychiatry 36:844–852
Clark D, Donovan J (1994), Reliability and validity of the Hamilton Anxiety Shaffer D, Fisher P, Lucas CP, Dulcan MK, Schwab-Stone ME (2000), NIMH
Rating Scale in an adolescent sample. J Am Acad Child Adolesc Psychiatry Diagnostic Interview Schedule for Children Version IV (NIMH DISC-
33:354–360 IV): description, differences from previous versions, and reliability of some
Cohen J, Cohen P (1975), Applied Multiple Regression/Correlation Analysis for
common diagnoses. J Am Acad Child Adolesc Psychiatry 39:28–38
the Behavioral Sciences. Hillsdale, NJ: Erlbaum
Shaffer D, Gould MS, Brasic J et al. (1983), A children’s global assessment
Costello E, Angold A (1995), Epidemiology. In: Anxiety Disorders in Children
scale (CGAS). Arch Gen Psychiatry 40:1228–1231
and Adolescents, March J, ed. New York: Guilford, pp 109–124
Shrout PE, Fleiss JL (1979), Intraclass correlations: uses in assessing rater reli-
Curry JF, Murphy LB (1995), Comorbidity of anxiety disorders. In: Anxiety
Disorders in Children and Adolescents, March J, ed. New York: Guilford, ability. Psychol Bull 86:420–428
pp 301–320 Silverman WK, Albano AM (1996), The Anxiety Disorders Interview Schedule
Francis G, Last CG, Strauss CC (1992), Avoidant personality disorder and for Children-DSM-IV. San Antonio, TX: Psychological Corporation
social phobia in children and adolescents. J Am Acad Child Adolesc Psychiatry Silverman WK, Kurtines WM, Ginsburg GS, Weems CF, Rabian B, Serafini
31:1086–1089 LT (1999), Contingency management, self-control, and education sup-
Green B, Shirk S, Hanze D, Wanstrath J (1994), The Children’s Global port in the treatment of childhood phobic disorders: a randomized clini-
Assessment Scale in clinical practice: an empirical evaluation. J Am Acad cal trial. J Consult Clin Psychol 67:675–687
Child Adolesc Psychiatry 33:1158–1164 Strauss CC, Lahey BB, Frick P, Frame CL, Hynd GW (1988), Peer social sta-
Greenhill L, Pine D, March J, Birmaher B, Riddle MA (1998), Assessment tus of children with anxiety disorders. J Consult Clin Psychol 56:137–141
measures in anxiety disorders research. Psychopharmacol Bull 34:155–164 Welner Z, Reich W, Herjanic B, Jung KG, Amado H (1987), Reliability, valid-
Guy W (1976), ECDEU Assessment Manual for Psychopharmacology Revised ity, and parent–child agreement studies of the Diagnostic Interview for
(Publication ADM 76-338). Washington, DC: US Department of Health, Children and Adolescents (DICA). J Am Acad Child Adolesc Psychiatry
Education, and Welfare 26:649–653

J . A M . A C A D . C H I L D A D O L E S C . P S YC H I AT RY, 4 1 : 9 , S E P T E M B E R 2 0 0 2 1069