Psychological Assessment Copyright 2000 by the American Psychological Association, Inc.

2000, Vol. 12, No. 3, 304-318 1040-3590/00/S5.00 DOI: 10.I037//1040-3590.12.3.304

Child-Adolescent Suicidal Potential Index (CASPI):

A Screen for Risk for Early Onset Suicidal Behavior

Cynthia R. Pfeffer, Hong Jiang, and Tatsuyuki Kakuma

Weill Medical College of Cornell University

This study's purpose was to develop a reliable and valid self-report questionnaire, the Child-Adolescent
Suicidal Potential Index (CASPI), to screen for risk for suicidal behavior in children and adolescents.
Four hundred twenty-five child and adolescent psychiatric patients and nonpatients completed the CASPI
and other research instruments to rate suicidal and assaultive behavior and symptoms of depression,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

anxiety, and hopelessness. The 30-item CASPI involves 3 factors (anxious-impulsive depression,
This document is copyrighted by the American Psychological Association or one of its allied publishers.

suicidal ideation or acts, family distress) that contributed to a unidimensional 2nd-order factor accounting
for 59% of the total variance. Internal consistency (alpha) for the total score was .90, and test-retest
reliability (ICC) for the total score was .76. Total score distinguished between children and adolescents
with different severity of psychopathology and different levels of suicidal and assaultive behavior. Each
of the 3 factors had different contributions to discriminating between levels of suicidal status. CASPI
total score of 11 distinguished suicidal ideation or acts from nonsuicidal behavior, with sensitivity 70%
and specificity 65%. CASPI total score positively correlated with symptom severity of depression,
anxiety, and hopelessness.

Rates of youth suicide are high despite dissemination of re-
search results on youth suicidal behavior and the premise that
identification of youth at risk will facilitate prevention of youth
suicidal acts (Alcohol, Drug Abuse, and Mental Health Adminis-
tration, 1989; Pfeffer, 1989). An important issue not sufficiently
addressed is the development and utilization of screening strate-
gies that reliably detect youth who are at risk for suicidal behavior.
Empirical research suggests that risk factors for youth suicidal
behavior can be classified as (a) psychiatric symptoms and psy-
chiatric disorders, especially impulsivity, mood, anxiety, disrup-
tive behavior, substance abuse, and personality disorders (Brent et
al., 1988, 1993; Harrington et al., 1994; Kovacs, Goldston, &
Gatsonis, 1993; Ohring et al., 1996; Pfeffer, 1986; Pfeffer,
Plutchik, Mizruchi, & Lipkins, 1986; Pfeffer et al., 1993; Rao,
Weissman, Martin, & Hammond, 1993; Shaffer, 1988; Shaffer et
al., 1996); (b) family discord and psychopathology, including
violence, depression, substance abuse, and personality disorders
(Myers, Burke, & McCauley, 1982; Pfeffer, Normandin, &
Kakuma, 1994); (c) stressful experiences, particularly abuse and
Cynthia R. Pfeffer, Hong Jiang, and Tatsuyuki Kakuma, Department of
Psychiatry, Weill Medical College of Cornell University.
This study was supported by United States Public Health Service Grant
MN 142120 from the National Institute of Mental Health, a fund estab-
lished in the New York Community Trust by DeWitt-Wallace, an Estab-
lished Investigator Award from the National Alliance for Schizophrenia
and Depression.
We thank Robert Plutchik who collaborated in the early phase of study
and Denise Fryberg-Bissacia who assisted with data collection.
Correspondence concerning this article should be addressed to Cynthia
R. Pfeffer, who is now at New York-Presbyterian Hospital-Westchester
Division, 21 Bloomingdale Road, White Plains, New York 10605. Elec-
tronic mail may be sent to pfeffer2@rsl.med.cornell.edu.
losses of emotionally important people (Cohen-Sandier, Berman,
& King, 1982; Deykin, Alpert, & McNamarra, 1985; Gould,
Fisher, Parides, Flory, & Shaffer, 1996; Pfeffer et al., 1993); and
(d) cognitive factors involving competence in academic and social
activities and perceptions of hopelessness (Asarnow, Carlson, &
Guthrie, 1987; Cotton & Range, 1993; Marciano & Kazdin, 1994;
Pfeffer, Hurt, Peskin, & Siefker, 1995; Rotheram-Borus & Traut-
man, 1990; Spirito, Overholser, & Hart, 1991).
Widely used clinician-rated instruments, such as the Beck Scale
for Suicide Ideation (Beck, Kovacs, & Weissman, 1979; Steer,
Kumar, & Beck, 1993) and the Suicide Intent Scale (Beck,
Schuyler, & Herman, 1974), and self-report instruments, such as
the Scale for Suicide Ideation (Beck, Steer, & Ranieri, 1988), the
Suicide Probability Scale (Cull & Gill, 1982), the Suicide Behav-
iors Questionnaire (Linehan, 1981), the Beck Hopelessness Scale
(Beck, Weissman, Lester, & Trexler, 1974), and the Reasons for
Living Inventory (Cole, 1989; Linehan, Goodstein, Nielsen, &
Chiles, 1983), were developed for adults to measure aspects of
suicidal risk. However, few reliable and valid clinician-rated or
self-report instruments have been developed for the measurement
of suicidal risk in children and adolescents (Eyman, Mikawa, &
Eyman, 1990; Garrison, Lewinsohn, Marsteller, Langhinrichsen,
& Lann, 1991; Range & Knott, 1997). Notably, instruments de-
veloped for adults may not have the necessary developmental
specificity to measure suicidal risk in children and adolescents.
Interview techniques, such as the Kiddie-SADS (Ambrosini &
Dixon, 1996) and the Diagnostic Interview Schedule for Children
(King et al., 1997), primarily rate severity of symptoms of psy-
chiatric disorders and include measures of suicidal ideation and
suicidal acts. However, such techniques do not measure multifac-
toral elements of suicidal risk.
The Child Suicide Potential Scales (CSPS), administered as a
semi-structured interview to children and adolescents, is among

304
 CHILD-ADOLESCENT SUICIDAL POTENTIAL INDEX (CASPI)
the few clinician-rated measures of multiple risk domains of sui-
cidal behavior (Ofek, Weizman, & Apter, 1998; Pfeffer, 1986;
Pfeffer, Conte, Plutchik, & Jerrett, 1979; Pfeffer, Conte, Plutchik,
& Jerrett, 1980; Pfeffer et al., 1986, 1993; Pfeffer, Zuckerman,
Plutchik, & Mizruchi, 1984). Internal consistency (alpha) for each
of the eight scales assessed in prepubertal psychiatric inpatients
and outpatients (Pfeffer et al., 1979, 1980) and Israeli adolescent
psychiatric inpatients (Ofek et al., 1998) ranged from .57 to .98.
Interrater reliability (ICC) for each of the scales in prepubertal
psychiatric inpatients and nonpatients ranged from .54 to 1.00
(Pfeffer et al., 1984) and in adolescent psychiatric inpatients from
.77 to .89 (Ofek et al., 1998). Discriminant validity was suggested
by significantly different scores (p = .01) for each of the CSPS
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Scales for prepubertal psychiatric patients compared to prepubertal
This document is copyrighted by the American Psychological Association or one of its allied publishers.
nonpatients (Pfeffer et al., 1984). Convergent validity was sug-
gested by significant positive correlations between scales of the
CSPS and other standard research instruments in a sample of
Israeli adolescent psychiatric inpatients (Ofek et al., 1998). For
example, significant positive correlations were identified between
the CSPS Assaultive Behavior Scale and the Overt Aggression
Scale (r = .52), the CSPS Suicidal Behavior Scale and the Overt
Aggression Scale (r = .37), the CSPS Suicidal Behavior Scale and
a Suicidal Risk Scale (r = .45), the CSPS Depression Scale and the
Beck Depression Inventory (r = .21), and the CSPS Anxiety Scale
and the Spielberger State-Trait Anxiety Scale (r - .30; Ofek et al.,
1998). Predictive validity of the CSPS Suicidal Behavior Scale
was suggested in a prospective study of prepubertal children
identified on the CSPS Suicidal Behavior Scale as being suicidal
or not having a history of suicidal behavior (Pfeffer et al., 1993).
At 6-year follow-up, those children who reported a prepubertal
suicide attempt had a sixfold relative risk of attempting suicide.
Other clinician-rated instruments primarily measure specific
features of suicidal tendencies. For example, internal consistency
(alpha) for the clinician-rated Scale for Suicide Ideation was .88 in
rating the severity of suicidal ideation in prepubertal psychiatric
inpatients (Allan, Kashani, Dahlmeier, Taghizadeh, & Reid, 1997).
Convergent validity was suggested by significant positive corre-
lations between the total score of the Scale for Suicide Ideation and
two self-harm items (item 12 and 13) of the Children's Depression
Rating Scale (CDRS; r = .63, r = .63, respectively; Allan et al.,
1997). The prepubertal psychiatric inpatients scored significantly
higher on the Scale for Suicide Ideation when they endorsed items
indicating general preoccupations with death (p = .003) and
thoughts of their own death (p = .009) on the Diagnostic Inter-
view for Children and Adolescents. Significant positive correla-
tions were identified between the total score on the Scale for
Suicide Ideation and scores on the Hopelessness Scale for Children
(r = .39) and the CDRS (r = .55; Allan et al., 1997). Notably, in
this study of prepubertal psychiatric inpatients, a strong factor
structure was not identified. Clinical utility of this instrument was
not sufficiently evaluated for children because of limited empirical
data on children who had varied levels of suicidal states or with
less severe psychopathology (Allan et al., 1997).
Among the few clinician-rated measures to assess suicidal in-
tent, the Beck Suicide Intent Scale had internal consistency (alpha)
of .85 among 190 medically and psychiatrically hospitalized ado-
lescents who attempted suicide (Spitito, Sterling, Donaldson, &
Arrigan, 1996). Common factor analysis yielded a three-factor
305
solution that accounted for 44% of the variance. Construct validity
was suggested by significant positive correlations between the
Beck Suicide Intent Scale and the Reynolds Adolescent Depres-
sion Scale (r = .18), the Reynolds Suicide Ideation Questionnaire
(r = .26), and the Hopelessness Scale for Children (r = .20).
Additional samples of adolescents with diverse levels of psycho-
pathology are required to characterize the psychometric character-
istics of this measure.
Few studies exist about self-report measures of suicidal risk for
prepubertal children and adolescents. The self-report version of the
Scale for Suicide Ideation had internal consistency (alpha) of .95
and significant positive correlations with the Beck Depression
Inventory (r — .52), the Beck Hopelessness Scale (r = .63), and
the Beck Anxiety Inventory (r = .44) when administered to
adolescent psychiatric inpatients (Steer, Kumar, & Beck, 1993).
Convergent validity was suggested by a significant positive cor-
relation with the suicide item of the Beck Depression Inventory
(r = .69; Steer et al., 1993). Additional research among adoles-
cents with more varied levels of suicidal tendencies and psycho-
pathologies is needed to evaluate the clinical utility of this
measure.
The 30-item self-report Suicidal Ideation Questionnaire, stan-
dardized initially in a school-based sample of 2,400 adolescents
(Reynolds, 1987), had internal consistency (alpha) of .97 in an
adolescent psychiatric inpatient sample (Pinto, Whisman, & Mc-
Coy, 1997). An exploratory principal-components analysis with
varimax rotation suggested one major factor that accounted for
58% of the variance (Pinto et al., 1997). The greatest clinical utility
of the Suicidal Ideation Questionnaire was obtained with a cutoff
score of .20, which distinguished either suicide attempters or
suicide ideators from nonsuicidal inpatients with sensitivity of
83% and 82%, respectively, and specificity of 58% for the suicide
attempters and suicide ideators (Pinto et al., 1997). The results of
this study suggested that the Suicidal Ideation Questionnaire may
be an adequate measure of suicidal cognition in adolescent psy-
chiatric inpatients, but additional studies are needed to evaluate the
clinical utility of this measure in adolescents with a range of
psychopathology.
The 15-item self-report Suicidal Ideation Questionnaire-Junior
Version, evaluated in an inner-city sample of young adolescents
(W. M. Reynolds & Mazza, 1999), had internal consistency (al-
pha) of .90 and test-retest reliability (ICC) within a 2- to 4-week
period of .89. Construct validity was suggested by significantly
higher scores on this measure for subjects who reported a history
of a suicide attempt compared to those without a history of a
suicide attempt (p = .001). More extensive research is needed
with samples of young adolescents to evaluate the reliability and
validity of this measure.
The Suicide Probability Scale, administered to high school
students, had internal consistency (alpha) for the subscales of this
instrument that ranged from .59 to .90 and a range of item-subscale
correlations (r) of .02 to .21 (Tatum, Greene, & Karr, 1993). This
study failed to replicate the original four-factor structure, a result
suggesting that the items of this instrument may not have adequate
developmental specificity for adolescents.
Convergent validity for the Reasons for Living Inventory, eval-
uated in high school and delinquent adolescents, was suggested by
significant negative correlations of its subscales with the Chil-
 306 PFEFFER, JIANG, AND KAKUMA
dren's Depression Inventory (r = —.13 to —.57) and the Chil-
dren's Hopelessness Scale (r = -.14 to -.57; Cole, 1989). Con-
struct validity was suggested by significant negative correlations
with the Suicide Behaviors Questionnaire (r = —.19 to —.62).
Studies are needed of the reliability and validity of this measure in
adolescents with a wide range of psychiatric symptoms and
disorders.
Few instruments have been evaluated for predictive validity.
Larzelere, Smith, Batenhorst, and Kelly (1996) reported that the
Suicide Probability Scale administered to children and adolescents
in residential treatment and who were followed up for 6 years had
a sensitivity of 48% and specificity of 80%. Keane, Dick, Bech-
told, and Manson (1996) reported that the Suicide Ideation
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Questionnaire-Junior High School Version administered to Amer-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
ican Indian adolescents who were followed up 2 months later
during an outbreak of suicide attempts had a sensitivity of 80% and
specificity of 86%.
The present article reports on the operationalization of multiple
risk domains for child and adolescent suicidal behavior that have
been suggested by empirical studies, including those utilizing the
CSPS. A 30-item self-report questionnaire, the Child-Adolescent
Suicidal Potential Index (CASPI), has been developed to be ad-
ministered to children and adolescents to screen for risk for sui-
cidal behavior. In this article, we describe the initial results of
CASPI's reliability and validity.
Study 1
Overview
This study's purpose was to generate items indicative of risk for
suicidal behavior in children and adolescents and to incorporate
them into a self-report questionnaire, the CASPI.
Method
Items were created to indicate the construct of risk for suicidal behavior
in children and adolescents. Items reflected factors that have been empir-
ically identified as associated with risk for child and adolescent suicidal
behavior, specifically psychiatric symptoms, suicidal ideation or suicidal
acts, family discord and psychopathology, and social stress. Many of these
items are included in the CSPS (Pfeffer, 1986; Pfeffer et al., 1979, 1980,
1986).
Pilot testing was conducted to evaluate the readability and comprehen-
sion of each item, to construct instructions for completing the question-
naire, and to include a response format for the items that is suitable for a
screening instrument that is to be administered in a brief period of time.
This process involved an initial discussion of the instructions, the items,
and the response format by three psychiatric professionals and subse-
quently trials of reading, comprehending, and answering the questions by 7
child and adolescent psychiatric patients and 7 children and adolescents
from the community. Subsequently, revisions were made in the instruc-
tions, the content and wording of items, and the response format. This
revised questionnaire was reviewed by 2 child and adolescent psychia-
trists, 2 child and adolescent psychologists, 2 psychiatric social workers,
and 10 child and adolescent psychiatric patients with suicidal ideation or
suicide attempts, 10 child and adolescent psychiatric patients without
suicidal ideation or suicide attempts, and 10 nonsuicidal children and
adolescents from the community.
Results
On the basis of this review and revision process, 98 items were
included in the questionnaire that reflected suicidal risk factors for
children and adolescents and included multiple items about each
risk domain involving psychiatric symptoms, suicidal ideation or
suicidal acts, family discord and psychopathology, and social
stress. Instructions were to "Please answer the questions about
yourself in the past 6 months." This is a time frame that is similar
to that on the CSPS and considered sufficient to highlight multiple
domains of suicidal risk. The response format is a yes (score = 1 )
or no (score = 0) format. This format was chosen to enable brevity
in completing this questionnaire.
All children and adolescents, age 8 through 17 years, read the 98
items. Their minimal reading level was estimated to be second
grade. Younger children generally required the questions to be
read to them as a means of assisting the children to maintain
attention, to enhance their motivation to complete the items, and to
clarify issues about the intent of an item. It is recommended that
younger children, those who are 6 through 7 years old, have the
questions read to them and older children and adolescents read and
respond to the questions on their own. The children and adoles-
cents understood and could respond to the questions with a yes
(score = 1) or no (score = 0) format. Higher scores on this
questionnaire were intended to indicate higher levels of risk for
suicidal behavior.
Study 2
Overview
In this study, we had multiple aims. First, we refined the items
and determined the factor structure of the instrument with a sample
of children and adolescents with varied risk for suicidal behavior.
Once we determined the factor structure of the instrument, we
examined features of reliability, including the internal consistency
and test-retest reliability of the instrument. We also evaluated the
discriminant validity of the instrument by determining whether its
scores differed with regard to levels of suicidal behavior, levels of
assaultive behavior, and patient status. In addition, we evaluated
the convergent validity of the instrument by comparing the instru-
ment's scores with those of standard reliable and valid self-report
questionnaires that measure factors associated with suicidal risk,
such as depression, hopelessness, and anxiety.
Method
Sample
Most of the 425 children and adolescents were consecutively admitted to
psychiatric services or were randomly selected as community controls from
school rosters. All parents, children, and adolescents were told that they
would participate in studies to identify risk for suicidal behavior (Pfeffer et
al., 1986). Ninety percent of those invited agreed to participate. Parents,
children, and adolescents provided written informed consent and assent,
respectively. The mean age of the total sample was 11.88, SD = 3.54 years
(range = 6 through 18 years). The 274 (64%) children, who were less
than 13 years old, had a mean age of 9.73, SD = 1.90 years, and the 151
(36%) adolescents, who were at least 13 years old, had a mean age
of 16.13, SD = 2.19 years. There were approximately 278 (65%) boys. The
 CHILD-ADOLESCENT SUICIDAL POTENTIAL INDEX (CASPI)
racial or ethnic distribution was 175 (41%) White, 83 (19%) Black, 60
(14%) Hispanic, 7 (2%) other racial or ethnic background, and 100 (24%)
whose racial or ethnic background was not identified. Social status distri-
bution (Hollingshead & Redlich, 1958) was I = 19 (5%), II = 39 (9%),
ID = 61 (15%), IV = 56 (13%), V = 23 (5%), and 227 (53%) whose social
status was not identified. The children and adolescents who were psychi-
atric patients were either prepubertal and adolescent psychiatric inpatients
(n = 196; 46%), prepubertal psychiatric day hospital patients (n = 21;
5%), prepubertal psychiatric outpatients (n = 35, 8%), or adolescents (n =
47; 11%) who were followed up after they were prepubertal psychiatric
inpatients. The community nonpatient controls included 126 (30%) prepu-
bertal children and adolescents.
The 129 nonpatients (M = 13, SO = 3.43 years) were significantly older
than the 299 patients (M = 11.35, SD = 3.45 years), f(418) = 4.94,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
p £ 0.0001. There were significantly more males among the patients (68%)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
than the nonpatients (57%), /(I, N = 428) = 5.28, p < .02. The
significantly higher rates of White, Black, and Hispanic patients (White =
45%, Black = 26%, Hispanic = 19%, other = 2%, unknown = 8%) may
be related to the significantly higher rates of unknown race or ethnicity
among the nonpatients (White = 32%, Black = 4%, Hispanic = 2%,
other = 2%, unknown = 60%), x*(4, N = 428) = 141.55, p < .0001.
There was a significantly higher rate of unknown social status backgrounds
among nonpatients (I = 2%, II = 4%, in = 13%, IV = 14%, V = 1%,
unknown = 66%) than patients (I = 6%, II = 11%, in = 15%, IV = 13%,
V = 7%, unknown = 48%), x*(5, N = 428) = 19.82, p s .001.
Subsamples of children and adolescents had data collected to determine
the test-retest reliability of the instrument, the relations of demographic
variables to the scores of the instrument, and the convergent validity of the
instrument. Specifically, for the determination of test-retest reliability, 74
children and adolescents participated. They were predominately males
(71%) and 99% were less than 13 years old. Fifty-four percent were White,
37% were Black, 6% were Hispanic, and 3% were from other racial or
ethnic backgrounds. Social status distribution included 4% I, 23% II, 27%
III, 35% IV, and 11% V. Approximately 62% were prepubertal and
adolescent psychiatric inpatients, 14% were prepubertal day hospital pa-
tients, 10% were prepubertal psychiatric outpatients, and 14% were pre-
pubertal and adolescent normal controls.
One hundred eighty-seven children and adolescents had available data
on all demographic variables involving age, gender, race or ethnicity, and
social status. These data were utilized to evaluate the relations between
demographic variables and the scores for the instrument under develop-
ment. Among these 187 children and adolescents, 45% were younger
than 13 years, 65% were boys, 67.5% were White, and 50% were from
middle social status backgrounds.
Assaultive behavior was rated for a subsample of 312 children and
adolescents. Among these, 65% were at least 13 years old and 67% were
males. Approximately 62% were White, 25% were Black, 11% were
Hispanic, and 2% were from other racial or ethnic backgrounds. The social
status distribution included approximately 10% 1,20% II, 31% III, 28% IV,
and 11% V. Patient status included 50% prepubertal and adolescent psy-
chiatric inpatients, 7% prepubertal psychiatric day hospital patients, 7%
prepubertal psychiatric outpatients, 15% adolescents with ,a history of
prepubertal psychiatric hospitalization, and 21% prepubertal and adoles-
cent normal controls.
Relationships of the instrument's scores with scores on self-report ques-
tionnaires measuring symptoms of depression, anxiety, and hopelessness
were evaluated in subsamples of children and adolescents to whom the
questionnaires were administered. The depression questionnaire was ad-
ministered to 97 children and adolescents among whom 59% were less
than 13 years old, 73% were male, 58% were White, 33% were Black, 8%
were Hispanic, and 1% were from other racial or ethnic backgrounds. .
Thirty-five percent were from middle social status background. The chil-
dren and adolescents were psychiatric inpatients or day hospital patients
307
(45%), adolescents with a history of prepubertal psychiatric hospitalization
(18%), and normal controls (37%).
The anxiety questionnaire was administered to 138 children and adoles-
cents among whom 46% were less than 13 years old, 70% were male, 62%
were White, 28% were Black, 8% were Hispanic, and 2% were from other
racial or ethnic backgrounds. Thirty-two percent were from a middle social
status background. These children and adolescents included psychiatric
inpatients or day hospital patients (36%), adolescents with a history of
prepubertal psychiatric hospitalization (28%), and normal controls (36%).
The hopelessness questionnaire was administered to 62 children and
adolescents among whom 40% were less than 13 years old, 73% were
male, 68% were White, 22% were Black, 8% were Hispanic, and 2% were
from other racial or ethnic backgrounds. Thirty-four percent were from a
middle social status background. Children and adolescents were psychiatric
inpatients or day hospital patients (36%), adolescents with a history of
prepubertal psychiatric hospitalization (26%), and normal controls (38%).
Procedures: Assessment Instruments
Six- and 7-year old children answered questions on the 98-item instru-
ment that was read to them by trained masters level research assistants.
Children, age 8 years or older, read questions on this instrument and
responded to them. All assessment instruments and the 98-item instrument
were administered at the medical center in one session of approxi-
mately 1.5 hr.
Ratings of suicidal ideation and suicide attempts within 6 months of
entering this study were obtained using the Spectrum of Suicidal Behavior
Scale (Pfeffer, 1986; Pfeffer et al., 1979,1980; Pfeffer, Solomon, Plutchik,
Mizruchi, & Weiner, 1982; Pfeffer et al., 1986) in separate semistructured
interviews of the children, adolescents, and their parents. Interviews were
conducted by masters level research assistants who were trained to have
high interrater reliability (kappa a 0.85; Cohen, 1960) and who were blind
to the children's and adolescents' responses on the instrument under
development. The Spectrum of Suicidal Behavior Scale rated suicidal
ideation or suicide attempts for prepubertal children and adolescents along
a 3-point spectrum indicating no suicidal ideation or suicide attempt
(rating = 1), suicidal ideation or threat (rating = 2), and suicide attempt
(rating = 3). The rating ascribed to the child or adolescent was the highest
rating reported by the child, adolescent, or parent. In the study, agreement
of parent-child ratings (kappa) was .40 (p £ .0001) with children and
adolescents reporting more severe suicide attempts or suicidal ideation.
In addition, interviewer interrater reliability (ICC) on the Spectrum of
Suicidal Behavior Scale was 1. Discriminant validity for the Spectrum
of Suicidal Behavior Scale was suggested by results that a higher
percentage of prepubertal children, \2(2, N = 133) = 28.4, p < .0001,
and adolescents, x*(2, N = 133) = 20.9, p < .0001, who were rated on
the Spectrum of Suicidal Behavior Scale as having a history of suicidal
ideation or suicide attempt, had at least one psychiatric disorder com-
pared to the percentage of children and adolescents who were rated as
not reporting suicidal ideation or a suicide attempt. It is hypothesized in
the present study that scores on the instrument under development
would be positively associated with scores on the Spectrum of Suicidal
Behavior Scale.
Ratings of assaultive ideation and assaultive acts within 6 months of
entering this study were obtained using the Spectrum of Assaultive Be-
havior Scale (Pfeffer, 1986; Pfeffer et al., 1987) in separate semistructured
interviews of the children and adolescents and their parents. The Spectrum
of Assaultive Behavior Scale rated assaultive tendencies along a 3-point
spectrum similar to the Spectrum of Suicidal Behavior Scale. The rating
assigned to a child and adolescent was the highest rating reported by the
child and adolescent or parent. In the study, agreement of parent-child
ratings (kappa) was .46 (p s .0001) with children reporting more severe
assaultive acts. Interviewer interrater reliability (ICC) on the Spectrum of
 308 PFEFFER, JIANG, AND KAKUMA
Assaultive Behavior Scale was .97. Discriminant validity for the Spectrum
of Assaultive Behavior Scale was suggested by results that a higher
percentage of children and adolescents who were rated on the Spectrum of
Assaultive Behavior Scale as having a history of assaultive behavior had a
psychiatric diagnosis of conduct disorder or oppositional defiant disorder
than the percentage of prepubertal children who were rated as not having
a history of assaultive behavior, x*(2, N = 133) = 26.5, p < .0001.
Because suicidal behavior is often positively associated with assaultive
behavior (Pfeffer et al., 1984), it was hypothesized that scores of the
instrument under development would be positively associated with more
severe assaultive behavior.
The Child Depression Inventory (GDI; Kovacs, 1980, 1992), a 27-
item questionnaire, rated severity of depressive symptoms involving
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
negative mood, interpersonal problems, ineffectiveness, anhedonia, and
This document is copyrighted by the American Psychological Association or one of its allied publishers.
negative self-esteem. It had internal consistency (alpha) of .86 and
test-retest reliabilities (ICC) for periods of 2-3 week intervals of .74 to
.83 for prepubertal children and adolescents with varied levels of
depression and psychopathology (Kovacs, 1980, 1992). It distinguished
between children and adolescents with mood disorders and those with-
out mood disorders (Kovacs, 1980, 1992). Among children and adoles-
cents in this study, internal consistency (alpha) for the GDI was .84.
Convergent validity of the GDI in this sample was suggested by
significantly higher GDI scores for children and adolescents who were
rated on the Spectrum of Suicidal Behavior Scale as having suicidal
ideation or a history of a suicide attempt than for children and adoles-
cents without a history of suicidal ideation or a suicide attempt, F(l,
98) = 20.38, p s .0001,/= .49.
The Revised Children's Manifest Anxiety Scale (R-CMAS; Reynolds &
Richmond, 1985) rated severity of anxiety symptoms. It had internal
consistency (alpha) of .83 and test-retest reliability (ICC) for a period of 2
weeks of .68 for prepubertal children and adolescents with varied levels of
anxiety and psychopathology (C. R. Reynolds & Richmond, 1985). It
distinguished between children and adolescents with anxiety disorders
from those without anxiety disorders (C. R. Reynolds & Richmond, 1985).
In this study, internal consistency (alpha) for the R-CMAS was .80.
Convergent validity of the R-CMAS in this sample was suggested by
significantly higher R-CMAS scores for children and adolescents who
were rated with the Spectrum of Suicidal Behavior Scale as reporting a
history of suicidal ideation or a suicide attempt than for children and
adolescents who did not report suicidal ideation or a suicide attempt, F(l,
139) = 4.72, p < .03, / = .38.
The Hopelessness Scale for Children and Adolescents (HS; Kazdin,
French, Unis, Esveldt-Dawson, & Sherick, 1983; Kazdin, Rodgers, &
Colbus, 1986; Marciano & Kazdin, 1994) rated severity of hopelessness. It
had internal consistency (alpha) of .75 to .97 and test-retest reliability
(ICC) over a 6-week interval of .52 for prepubertal children with varied
levels of psychopathology (Kazdin et al., 1983, 1986). It had internal
consistency (alpha) of .69 and .84, respectively for normal adolescents and
adolescent suicide attempters and test—retest reliabilities (ICC) of .49 in
normal adolescents over 10-weeks (Spirito, Williams, Stark, & Hart, 1988).
It distinguished between children and adolescents with suicidal ideation or
suicide attempt, and those without suicidal tendencies and a high score on
the HS were significantly associated with more severe symptoms of de-
pression and poorer self-esteem (Kazdin et al., 1983, 1986; Spirito et al.,
1988). Internal consistency (alpha) for the HS was .76 for the children and
adolescents in this study. Convergent validity of the HS in this sample was
suggested by significantly higher HS scores for children and adolescents
rated on the Spectrum of Suicidal Behavior Scale as reporting a history of
suicidal ideation or a suicide attempt than for children and adolescents who
did not report suicidal ideation or a suicide attempt, F(l, 60) = 4.42, p £
.04,/= .27.
Statistical Analyses
Determination of factor structure. A series of common factor analyses
with maximum likelihood extraction method and oblique (promax) rotation
was used to identify clinically meaningful items and the factor structure of
the instrument under development (Gorsuch, 1983; Hatcher, 1994; SAS
Institute, Inc., 1989). These analyses utilized the item scores (i.e., 0 or 1).
Factors were retained if they satisfied a combination of criteria (Hatcher,
1994) including: If the factors with relatively large eigenvalues were
demarcated in a scree test plot from those with relatively small eigenvalues,
if a factor accounted for more than 10% of the common variance, if there
were at least 3 items on each factor, if items that loaded on a specific factor
shared a conceptual meaning, if items that loaded on different factors
measured different constructs, if most items had high loading on only one
factor and low loadings on other factors, and if factors had relatively high
loadings for some items and low loadings for the other items. Second-order
common factor analysis with maximum likelihood extraction method and
oblique (promax) rotation was conducted on the retained factors of the
initial factor analysis to determine if retained factors contributed to a
unidimensional result.
Internal consistency. Once the factor structure of the instrument was
identified, internal consistency of the instrument was calculated with
coefficient alpha (Spector, 1992).
Test-retest reliability. Once the factor structure of the instrument was
identified, the test-retest reliability of the instrument was calculated using
intraclass correlation coefficients (ICC; Shrout & Fleiss, 1979) to compare
the instrument's scores for a subsample of participants who had completed
the items initially and 2 weeks later.
Relationships between demographic variables and scores on the instru-
ment. Once the factor structure of the instrument under development was
identified, the relationship between each demographic variable (i.e., age,
gender, race or ethnicity, and social status) and the total score of the
instrument was evaluated with analysis of covariance (ANCOVA) to
account for effects of the other demographic variables. In addition, multi-
variate analysis of covariance (MANCOVA) using Wilks's Lambda was
utilized to evaluate the instrument's factors, when they were included in the
analysis simultaneously, with regard to specific demographic variables
when the effects of the other demographic variables were controlled. If
significant effects were identified, ANCOVA was utilized to identify if
there were significant effects for each of the instrument's factors with
regard to specific demographic variables when the effects of the other
demographic variables were controlled.
Discriminant validity. Differences in the total scores of the instrument
were evaluated with regard to ratings on the Spectrum of Suicidal and
Assaultive Behavior Scales and patient status utilizing ANCOVA to ac-
count for effects of demographic covariates, such as age and gender.
MANCOVA using Wilks's Lambda was utilized to evaluate the effects of
the instrument's factors, included simultaneously, with regard to suicidal
behavior, assaultive behavior, and patient status while controlling for
effects of demographic variables that were found to have significant
relations with the instrument's scores. If significant effects were identified,
ANCOVA was utilized to identify if there were significant effects for the
scale scores for each of the instrument's factors with regard to suicidal
behavior, assaultive behavior, and patient status when the effects of de-
mographic variables that were found to have significant relations with the
instrument's scores were controlled. In these analyses, Bonferroni correc-
tions were applied to ensure a comparison-wise error rate of .05 for
multiple comparisons. Effect sizes (/) were reported (Cohen, 1998). Non-
parametric statistics (Kruskal-Wallis and Wilcoxon test) were used for
ordinal and categorical variables, respectively, and for scores that were not
normally distributed. If parametric and nonparametric results were similar,
parametric results were reported.
Discriminant analysis with cross-validation (Lachenbruch & Mickey,
1968) was used to determine the degree that the instrument distinguished
 CHILD-ADOLESCENT SUICIDAL POTENTIAL INDEX (CASPI) 309

between levels of suicidal status rated on the Spectrum of Suicidal Behav- Table 1
ior Scale. This analysis included the instrument's scale scores for factors Factor Loadings for 30-Item CASPI
that were included as simultaneous classifier variables and suicidal status
as the categorical criterion. To make results of the discriminant analysis Factor 1 Factor 2
clinically meaningful, the instrument's total score was evaluated also with Anxious- Suicidal Factor 3
discriminant analysis with cross-validation. Sensitivities and specificities Impulsive Ideation/ Family
obtained from these analyses were reported. In addition, results of canon- Abbreviated item (item number) Depression Acts Distress
ical discriminant analysis were reported to indicate the relative contribution
People talk about you (21) .65 .03 -.14
of each factor to the discriminatory task. The results included the canonical .04
People didn't like you (13) .62 .01
correlation, total canonical structure, and the total-sample standardized Concentration (1) .60 -.05 .00
canonical coefficients for the scores of each factor. Sadness (8) .56 .02 .04
Receiver operating characteristics (ROC) analyses (Metz, 1978; Metz, Not worth much (12) .55 .22 -.04
Goodenough, & Rossmann, 1973; Somoza, Steer, Beck, & Clark, 1994) Nervousness (3) .52 .00 .08
were used to identify multiple pairs of test sensitivities (true positives) and Impatience (2) .50 .01 .03
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

specificities (true negatives) at multiple cutoff scores for the total score of Like to be alone (5) .48 .00 .07
This document is copyrighted by the American Psychological Association or one of its allied publishers.

the instrument. Area under the ROC curve with standard deviation pro- Things not get better (9) .46 .11 .10
vided an index of accuracy of discrimination provided by this instrument's Decisions (14) .45 -.01 .04
total score (Somoza & Mossman, 1991) and represented the probability
Daydream (10) .44 -.10 .10
Feel angry (6) .43 .09 .11
that a subject who was identified by the instrument as having suicidal Self-blame (7) .43 .08 .02
tendencies, rated on the Spectrum of Suicidal Behavior Scale, ranked Cause family problems (16) .43 .16 -.04
higher on this instrument than another subject, who was randomly identi- Angry when did not get way? (30) .42 -.05 .22
fied and was not suicidal (Hanley & McNeil, 1982). Sensitivity and Sad when did not get way (22) .42 .02 .20
specificity values on the ROC curves were calculated from corresponding Thoughts to kill self (25) -.02 .82 .05
cutoff scores thought to be clinically useful in discriminating between two Try to kill self (26) -.16 .76 .12
levels of suicidal status. Try to hurt self (24) -.03 .75 .06
Convergent validity. Once the factor structure of the instrument was Want to hurt self (23) .11 .73 .04
Want to die (11) .18 .69 -.09
identified, analyses of convergent validity were undertaken by correlating Do dangerous things (4) .41 .22
.23
the total scores on the instrument with total scores on the CDI, the Parents hit you (29) -.03 .02 .65
R-CMAS, and the HS. Since depression, anxiety, and hopelessness are risk Father hit mother (18) -.16 .02 .63
factors for suicidal behavior, it was hypothesized that total scores on these Parents sad (19) .19 -.08 .53
questionnaires would positively correlate with the total score of the Parents yell at you (28) .09 .04 .52
instrument. Parents not talking (27) .02 -.02 .50
Punished (15) .15 -.02 .50
Arguments between parents (17) .08 .09 .49
Results Parents drink alcohol (20) -.04 .02 .46

Determination of the Instrument's Factor Structure
A three-factor solution with 30 items was selected, it satisfied all
criteria for retention of factors, and it accounted for 37% of the
total variance. Table 1 shows the CASPI factor loading values for
the three-factor solution for the 30 items. Factor 1 (16 items, total
variance explained = 13%) rated psychiatric symptoms indicative
of anxious-impulsive depression; Factor 2 (6 items, total variance
explained = 16%) rated predominantly symptoms of suicidal
ideation or acts; and Factor 3 (8 items, total variance explained =
8%) rated family distress arising from family discord and family
psychopathology. Significant positive correlations were identified
between Factors 1 and 2 (r = .54), Factors 1 and 3 (r = .40), and
Factors 2 and 3 (r = .33). Second-order common factor analysis
utilizing the three primary factors indicated that Factor 1 (load-
ing = .86), Factor 2 (loading = .71), and Factor 3 (loading = .56)
each were retained on one second-order factor, which explained
59% of the total variance.
Factor analyses conducted separately on data for the 274 chil-
dren younger than 13 years and for the 151 adolescents at least 13
years old yielded three-factor solutions that accounted for 37% and
41% of the total variance, respectively. Factor 1 (younger children:
total variance explained = 11%; adolescents: total variance ex-
plained = 16%), Factor 2 (younger children: total variance ex-
plained = 18%; adolescents: total variance explained = 10%), and
Factor 3 (younger children: total variance explained = 8%; ado-
Note. Bold type indicates that the item loads highest on that specific
factor. CASPI = Child-Adolescent Suicidal Potential Index. Total vari-
ance for Factor 1 = 13%, Factor 2 = 16%, Factor 3 = 8%, and CASPI
total = 37%.
lescents: total variance = 16%) contained items that loaded in the
same way as in the total sample. Significant positive correlations
were identified between Factors 1 and 2 (younger children: r =
.46; adolescents: r = .41), Factors 1 and 3 (younger children: r =
.36; adolescents: r = .57), and Factors 2 and 3 (younger children:
r = .29; adolescents: r = .35). Second-order factor analyses
indicated that Factor 1 (younger children: loading = .83; adoles-
cents: loading = .87), Factor 2 (younger children: loading = .64;
adolescents: loading = .54), and Factor 3 (younger children:
loading = .53; adolescents: loading = .74) loaded on one second-
order factor, which explained 52% and 61% of the total variance
for the children and adolescents, respectively.
Internal Consistency
Internal consistency (alpha) for the total 30-item CASPI was .90
(N = 425). For Factor 1, Factor 2, and Factor 3, internal consis-
tency (alpha) was .86, .85, and .77, respectively. Similarly, when
the 30-item CASPI was administered to 279 children less than 13
years old, internal consistency (alpha) for the total 30-item CASPI
was .88, and for Factor 1, Factor 2, and Factor 3, it was .83, .85,
 310 PFEFFER, JIANG, AND KAKUMA
and .74, respectively. When the 30-item CASPI was administered
to 146 adolescents older than 13 years, internal consistency (alpha)
for the total 30-item CASPI was .92 and for Factor 1, Factor 2, and
Factor 3, it was .89, .86, and .83, respectively.
Test-Retest Reliabilities
Test-retest reliabilities (ICC) for the subsample of predomi-
nantly prepubertal children for the total score of the 30-item
CASPI was .76 (p < .0001) and for the scale scores for Factor 1,
Factor 2, and Factor 3, it was .76 (p < .0001), .59 (p =£ .0001),
and .71 (p < .0001), respectively.
Relations Between Demographic Variables and
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
CASPI Scores
Results of parametric analyses were reported because they were
similar to those of nonparametric results. ANCOVA controlling
for the effects of age, gender, race or ethnicity, and social status
suggested that there were significant differences in CASPI total
scores for age, F(l, 186) = 4.88,/>< .02,/= .17, and gender, F(l,
186) = 4.77, p < .03, / = .18. CASPI total scores for the 88
children younger than 13 years (M = 12.53, SD = 6.76) were
significantly greater than for the 109 adolescents at least 13 years
old (M = 9.62, SD = 7.25). CASPI total scores for the 69 females
(M = 12.68, SD = 7.40) were significantly greater than for the 128
males (M = 9.98, SD = 6.88). Several MANCOVAs, controlling
for the effects of age, gender, race or ethnicity, and social status,
were utilized to identify if there were significant differences in the
scale scores for the three CASPI factors simultaneously. These
analyses suggested that scale scores for CASPI factors were sig-
nificantly different with respect to age, F(3, 184) = 2.94, p < .03,
and gender, F(3, 184) = 2.58, p < .05. Several ANCOVAs
controlling the effects of age, gender, race or ethnicity, and social
status were utilized to evaluate differences in scale scores for each
factor regarding demographic variables. There were significant
differences in scale scores for CASPI Factor 1 for age, F(l,
186) = 4.13, p < .04, / = .16, with the 88 children younger
than 13 years (M = 7.76, SD = 4.20) having significantly higher
scale scores for CASPI Factor 1 than the 109 adolescents at
least 13 years old (M = 6.18, SD = 4.67) and gender, F(l,
186) = 6.68, p < .01,/= .21, with the 69 females (M = 8.13,
SD = 4.85) having significantly higher scale score for CASPI
Factor 1 than the 128 males (M = 6.22, SD = 4.23).
Discriminant Validity
Since results of parametric and nonparametric analyses were
similar, results of parametric analyses were reported. Because age
and gender were significantly different with regard to some CASPI
scores, analyses utilized covariates of age and gender to evaluate
relations between CASPI scores and suicidal behavior, assaultive
behavior, and patient status. Table 2 shows results of ANCOVAs
controlling for the effects of age and gender for the CASPI total
score and scale scores for each CASPI factor regarding psychiatric
symptoms and patient status.
Suicidal Behavior
Ratings on the Spectrum of Suicidal Behavior Scale indicated
247 (58%) children and adolescents had no history of suicidal
ideation or suicide attempt, 108 (25%) children and adolescents
reported suicidal ideation only, and 70 (17%) children and ado-
lescents reported at least one suicide attempt.
MANCOVA, controlling for the effects of age and gender and
including the three CASPI factors simultaneously, suggested sig-
nificant differences between children and adolescents with either
suicidal ideation or suicidal attempts and those without suicidal
ideation or suicidal attempts, F(3, 409) = 40.09, p < .0001.
CASPI total scores and scale scores for Factor 1, Factor 2, and
Factor 3 were significantly higher for the 175 children and ado-
lescents who reported either suicidal ideation or suicide attempts
compared with the 242 children and adolescents who reported no
suicidal ideation or attempts (see Table 2). There was a significant
suicidal status by age interaction for scale scores for CASPI
Factor 2.
With regard to each of the three child and adolescent groups
(i.e., 70 children and adolescents reported a suicide attempt, 108
children and adolescents reported suicidal ideation, and 247 chil-
dren and adolescents reported no history of suicidal ideation or
suicide attempts) MANCOVA controlling for the effects of age
and gender and including the three CASPI factors simultaneously
indicated significant differences between the three groups of chil-
dren and adolescents, F(6, 816) = 24.55, p < .0001. ANCOVA
indicated that CASPI total scores and scale scores for each of the
CASPI factors were significantly different (see Table 2). CASPI
total and scale scores for Factor 2 were significantly different for
each of the three suicidal status groups. Scores for children and
adolescents with a history of a suicide attempt had the highest
CASPI scores. Scale scores for CASPI Factor 1 were significantly
higher for children and adolescents who attempted suicide and
those who reported suicidal ideation than those without suicidal
behavior. Scale scores for Factor 3 were significantly higher for
children and adolescents with suicide attempts compared to those
who had no history of suicidal behavior. There was a suicidal
status by age interaction with respect to scale scores for CASPI
Factor 1.
Assaultive Behavior
MANCOVA controlling for the effects of age and gender and
including the three CASPI factors simultaneously suggested sig-
nificant differences between children and adolescents with assaul-
tive ideation, those with assaultive acts, and those without assaul-
tive behavior, F(6, 608) = 7.95, p s .0001. The 148 children and
adolescents with assaultive acts (AA) and the 51 children and
adolescents with assaultive ideation (AI) had significantly higher
CASPI total scores and scale scores for Factor 1 and Factor 3 than
the 113 children and adolescents without assaultive acts or assaul-
tive ideation (NAIA; see Table 2).
Patient Status
MANCOVA controlling for the effects of age and gender and
including the three CASPI factors simultaneously suggested sig-
nificant differences for children and adolescents with different
patient status, F(3, 1074.5) = 4.93, p < .0001. Among the patient
status groups for the 425 children and adolescents, there were
 CHILD-ADOLESCENT SUICIDAL POTENTIAL INDEX (CASPI) 311

Table 2
Comparisons of Child-Adolescent Suicidal Potential Index (CASPI) Scores for Suicidal Behavior,
Assaultive Behavior, and Patient Status

Group ANOVA Pair-wise

Groups compared CASPI score M SD M SD M SD df F / Difference (p £)

N vs. SIA Total SIA N

14.67 6.6 9.19 6.4 1,409 58.4**** .43
N vs. SIA Factor 1 SIA N
9.23 4.1 6.21 4.24 1,409 47.03**** .38
N vs. SIA Factor 2 SIA N
2.74 2.18 0.87 1.42 1,409 79.29**** .5
(N vs. SIA) X Age Factor 2 SIA N 1,409 4.77* .52 C-SIA > C-N (.0001)°
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Age < 13 2.8 2.16 0.74 1.26 C-SIA > A-N (.0001)'
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Age > 13 2.6 2.23 1.06 1.63 A-SIA >C-N (.0001)"

A-SIA > A-N (.0001)"
A-N > C-N (.04)
N vs. SIA Factor 3 SIA N
2.71 2.26 2.12 2.14 1,409 5.78* .13
N vs. SI vs. SA Total SA SI N 2,405 33.78**** .67 SA > SI (.009)a
16.6 6.35 13.39 6.44 9.19 6.44 SA > N (.0001)a
SI > N (.0001)"
N vs. SI vs. SA Factor 1 SA SI N 2,405 26.02**** .56 SA > N (.0001)"
9.8 3.98 8.87 4.11 6.21 4.24 SI > N (.0001)a
(N vs. SI vs. SA) X Age Factor 1 SA SI N 2,405 3.15* .49 A-SA > A-SI (.03)
Age < 13 9.31 4.2 9.07 3.96 6.72 4.02 A-SA > A-N (.0001)"
Age > 13 2.7 8.46 4.43 5.39 4.47 11.3 A-SA > C-N (.0001)a
C-SA > A-N (.0001)"
C-SA > C-N (.001)a
A-SI > A-N (.0006)"
A-SI > C-N (.03)
P-SI > A-N (.0001)"
C-SI > C-N (.0001 )a
C-N > A-N (.04)
(N vs. SI vs. SA) X Sex Factor 1 SA SI N 2,405 3.51* .55 M-SA > M-SI (.01)
Sex
F 9.89 3.89 10.9 3.57 6.97
6.97 4.5 M-SA > M-N
M-N (.0001)'
(.0001)'
M 9.74 4.08 7.67 3.95 6.97
6.97 4.5 M-SA > F-N (.001)"
F-SA > M-SI (.005)
F-SA > M-N (.0001)"
F-SA > F-N (.0004)'
M-SI > M-N (.009)
F-SI > M-SI (.0001)"
F-SI > M-N (.0001)a
F-SI > F-N (.0001)"
F-N > M-N (.05)
N vs. SI vs. SA Factor 2 SA SI N 2,405 48.63**** .84 SA > SI (.0001)"
3.58 2.21 2.19 1.98 0.87 1.42 SA > N (.0001)"
SI > N (.0001)"
N vs. SI vs. SA Factor 3 SA SI N 2,405 4.15* .26 SA > N (.005)"
3.26 2.24 2.35 2.21 2.12 2.14
N vs. AI vs. AA Total AA AI N 2,300 17.32**** .43 AA > N (.0001)"
12.8 7 11.96 7 8.33 6.42 AI > N (.0003)"
N vs. AI vs. AA Factor 1 AA AI N 2,300 11.08**** .35 AA > N (.0001)"
8.09 4.26 7.61 4.8 5.81 4.19 AI > N (.004)a
N vs. AI vs. AA Factor 2 AA AI N 2,300 6.07** .25 AA > N (.0006)'
1.98 2.04 1.57 1.85 1.22 1.95
N vs. AI vs. AA Factor 3 AA AI N 2,300 18.57**** .46 AA > N (.0001)a
2.75 2.28 2.78 2.24 1.3 1.53 AI > N (.0001)'
Patient status Total 4,410 10.85**** .35 IP > FIP (.0001)'
IP 13.48 6.56 IP > NP (.0001)"
DHP 12.43 8.01 DHP > FIP (.02)
OP 11.76 5.48 DHP > NP (.004)"
FTP 9.26 7.41 OP > NP (.008)
NP 8.93 6.86
(table continues)
 312 PFEFFER, JIANG, AND KAKUMA

Table 2 (continued)

Group ANOVA Pair-wise

Groups compared CASPI score M SD M SD M SD df F f Difference (p <)

Patient status Factor 1 4,410 9.33**** .36 IP > FIP (.0001)'
IP 8.62 4.19 IP > NP (.0001)a
DHP 8.81 4.47 DHP > FIP (.004)a
OP 7.56 3.54 DHP > NP (.0009)"
FIP 5.89 4.64 OP > NP (.03)
NP 6.02 4.36
Patient status Factor 2 4,410 9.26**** .29 IP > FIP (.0005)"
IP 2.19 2.19 IP > NP (.0001)"
DHP 1.52 2.02
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

OP 1.12 1.43
This document is copyrighted by the American Psychological Association or one of its allied publishers.

FIP 1.3 1.8

NP 1.11 1.66
Patient status Factor 3 4,410 3.81** .23 IP > NP (.0005)a
IP 2.69 2.12 OP > NP (.003)"
DHP 2.1 2.28
OP 3.09 2.59
FIP 2.06 2.28
NP 1.82 2.07

Note. ANOVA = analysis of variance; N = nonsuicidal/nonassaultive participants; SIA = participants with suicidal ideation or suicidal attempt; SI =
participants with suicidal ideation; SA = participants with suicidal attempt; C = children < 13 years old; A = adolescents a 13 years old; AI = participants
with assaultive ideation; AA = participants with assaultive attempt; IP = inpatients; DHP = day hospital patients; OP = outpatients; FIP = former
inpatients; NP = nonpatients; M = male; F = female; / = effect size.
a
Significant after Bonferroni correction was applied.
*p£.05. **/?£.01. * * * / ) < . 001. ****p £ .0001.

significant differences in the CASPI total scores and scale scores
for CASPI Factor 1, Factor 2, and Factor 3 (see Table 3).
Multivariate discriminant analyses with cross-validation sug-
gested that the CASPI total score for children and adolescents in
the total sample classified different levels of suicidal status with
sensitivities ranging from 58% to 76% and specificities ranging
from 58% to 72%. When the CASPI factors were included as
simultaneous classifier variables in the multivariate discriminant
analyses for the total sample, the scale scores for the CASPI
factors classified different levels of suicidal status with sensitivi-
ties ranging from 50% to 69% and specificities ranging from 62%
to 85%. For children less than 13 years old, the total CASPI score
and the scale scores for the three factors as simultaneous classifier
variables had sensitivities ranging from 60% to 74% and 50% to
78%, respectively, and specificities of 58% to 73% and 68% to
88%, respectively, in discriminating different levels of suicidal
status. For adolescents at least 13 years old, the CASPI total score
and the scale scores for the three factors as simultaneous classifier
variables had sensitivities ranging from 25% to 54% and 25% to
56%, respectively, and specificities of 78% to 96% and 74% to
96%, respectively, in discriminating different levels of suicidal
status.
For example, in discriminating children and adolescents in the
total sample who attempted suicide form those without suicidal
ideation or a history of a suicidal attempt, the CASPI total score
had a sensitivity of 76% and a specificity of 72%. The sensitivity
was 74% and specificity was 73% for the CASPI total score in
discriminating children less than 13 years old who attempted
suicide from those without a history of suicidal ideation or suicidal
attempt. For adolescents at least 13 years old, the CASPI total
score had a sensitivity of 33% and a specificity of 94% in dis-
criminating adolescents who attempted suicide from those without
a history of suicidal ideation or a suicidal attempt.
Canonical discriminant analysis indicated the relative contribu-
tion of the scale scores for each factor in discriminating groups of
suicidal and nonsuicidal children and adolescents. In general, scale
scores for Factor 2 followed by scale scores for Factor 1 for the
total sample of children and adolescents, for children less than 13
years, and for adolescents at least 13 years old consistently con-
tributed most in discriminating nonsuicidal children and adoles-
cents from those with suicidal ideation and/or a history of a
suicidal attempt. For example, scale scores for Factor 2 (canonical
structure = .97, standardized canonical coefficient = .34) and
Factor 1 (canonical structure = .7, standardized canonical coeffi-
cient = .34) contributed more than scale scores for Factor 3
(canonical structure = .27, standardized canonical coefficient =
-.15) in discriminating children and adolescents in the total sam-
ple with a history of suicidal ideation or a suicidal attempt from
those without a history of suicidal ideation or suicidal attempt
(canonical correlation = .48), F(3, 419) = 42.6, p < .001. In
contrast, scale scores for Factor 2 (canonical structure = .92,
standardized canonical coefficient = .93) followed by scale scores
for Factor 3 (canonical structure = .6, standardized canonical
coefficient = .44) compared with scale scores for Factor 1 (ca-
nonical structure = .31, standardized canonical coefficient = —.2)
contributed most in discriminating children and adolescents in
total sample with suicidal ideation from those with a history of a
suicide attempt (canonical correlation = .32), F(3, 173) = 7.36,
p < .001.
 CHILD-ADOLESCENT SUICIDAL POTENTIAL INDEX (CASPI) 313

Table 3
Cutoff Scores for CASPI Total Scores and Sensitivity and Specificity With Regard to Suicide Status

N vs. SIA N + SI vs. SA N vs. SI N vs. SA SI vs. SA

Cutoff score Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity

Total sample
9 78 53 89 45 71 53 89 45 89 29
10 72 61 83 53 65 61 83 61 83 35
11 70 65 80 56 64 65 80 65 80 36
12 66 69 77 61 58 69 77 69 77 42
13 63 72 76 64 56 72 76 72 76 44
14 58 77 71 69 49 77 71 77 71 51
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Children < 13 years

9 79 48 85 40 75 48 85 48 85 25
10 71 58 77 50 67 58 77 58 77 33
11 70 63 75 54 67 63 75 63 75 33
12 67 69 75 59 60 69 75 69 75 40
13 63 73 74 63 56 72 74 73 74 44
14 58 76 70 68 49 76 70 76 70 51
Adolescents a 13 years
9 75 61 1 54 63 61 1 61 1 37
10 73 66 1 59 6 66 1 66 1 4
11 69 68 94 61 57 68 94 68 94 43
12 63 70 82 64 54 70 82 70 82 46
13 63 71 82 64 54 71 82 71 82 46
14 58 78 76 71 49 78 76 78 76 51
Note. CASPI = Child-Adolescent Suicidal Potential Index; N = nonpatients; SIA = participants with suicidal ideation or suicidal attempt; SI
participants with suicidal ideation; SA = participants with suicidal attempt. Sensitivity and Specificity are given as percentages.

Figure 1 illustrates ROC curves for the total sample that plot
sensitivity versus 1-specificity and area under the curve data for
comparisons of the 30-item CASPI total scores throughout its
entire range of possible cutoff scores between children and ado-
lescents with varied levels of suicidal tendencies.
Table 3 shows cutoff scores for the CASPI total score and
sensitivities and specificities with regard to suicidal status. Opti-
mum cutoff scores for the CASPI total score were suggested by
these data. Specifically, a CASPI total score of 11 identified 70%
of children and adolescents in the total sample and 70% of younger
children who truly had a history of suicidal ideation or a suicide
attempt (SIA; sensitivity). This score distinguished the 65% of
children and adolescents in the total sample and 63% of younger
children who truly do not have a history of suicidal ideation or
suicide attempt (N; specificity). A CASPI total score of 11 had a
sensitivity of 80% and a specificity of 65% in distinguishing
children and adolescents in the total sample and a sensitivity of
75% and a specificity of 63% in distinguishing younger children
with a history of a suicide attempt (SA) from children and ado-
lescents without a history of either suicidal ideation or suicide
attempt (N). Children and adolescents in the total sample and those
in the subsample of children younger than 13 years without a
history of suicidal ideation or a suicide attempt (N) were distin-
guished on the CASPI total score from children and adolescents
with suicidal ideation (SI) at a score of 11 with a sensitivity of 64%
and 67%, respectively, and a specificity of 65% and 63%, respec-
tively. Higher CASPI total cutoff scores for the total sample and
the younger subsample were necessary to distinguish nonsuicidal
children and adolescents plus children and adolescents with sui-
cidal ideation (N + SI) from children and adolescents with a
history of a suicide attempt (SA; i.e., total sample CASPI total
score = 14, sensitivity = 71%, specificity = 69%; younger
children total CASPI score = 14, sensitivity = 70%, specificity =
68%) and children and adolescents with suicidal ideation (SI) from
children and adolescents with a suicide attempt (SA; total sample
CASPI total score = 14, sensitivity = 71%, specificity = 51%,
younger children total CASPI score = 14, sensitivity = 70%,
specificity = 51%).
With regard to adolescents who were at least 13 years old, a
CASPI total score of 11 distinguished adolescents with suicidal
ideation or a suicidal attempt (SIA) from nonsuicidal adolescents
(N) with sensitivity of 69% and specificity of 68%. In general, a
CASPI total score of 11 was an optimal cutoff score.
Convergent Validity of the CASPI
As hypothesized, the CASPI total score (r = .63, p < .0001) and
scale scores for Factor 1 (r = .60, p < .0001), Factor 2 (r = .57,
p =£ .0001), and Factor 3 (r = .35, p < .0005) were significantly
positively correlated with scores on the GDI. The CASPI total
score and the scale scores for CASPI Factors 1, 2, and 3 for
children younger than 13 years were significantly positively cor-
related with scores on the GDI (CASPI total score: r = .57, p ^
.0001; Factor 1: r = .54, p < .0001; Factor 2: r = .50, p s .0001;
 314 PFEFFER, JIANG, AND KAKUMA

1.0

0.9

f 0.8
2 7
8 °-
|o,
£ 0.5
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

s
This document is copyrighted by the American Psychological Association or one of its allied publishers.

8 0.4

N vs. SIA (ADC = .72, SE = .02, P <= .0001)

SI vs. SA (AUC = .64, SE = .04, £ = .0003)
Nvs. SA (ADC = .78, SE = .03, P <= .0001)
0.2 N + SI vs. SA (AUC = .74, ££ = .03, P <= .0001)
Nvs. SI (AUC = .69, SE = .03, £ <= .0001)
0.1

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
1-False Positive Rate (Specificity)
Figure 1. ROC curves for CASPI scores for suicidal status. CASPI = Child-Adolescent Suicidal Potential
Index; ROC = receiver operator characteristics; AUC = area under the curve; N vs. SIA = nonsuicidal
participants versus participants with suicidal ideation or attempts; N vs. SA = nonsuicidal participants versus
participants with suicidal attempts; N vs. SI = nonsuicidal participants versus participants with suicidal ideation;
SI vs. SA = participants with suicidal ideation versus participants with suicide attempts; N + SI vs. SA =
nonsuicidal participants plus participants with suicidal ideation versus participants with suicide attempts.

Factor 3: r = .44, p £ .0006). For adolescents at least 13 years old,
GDI scores were significantly positively correlated with CASPI
total scores (r = .73, p £ .0001) and scale scores for CASPI
Factor 1 (r = .67, p ^ .0001) and CASPI Factor 2 (r = .68, p ^
.0001).
The CASPI total score (r = .47, p < .0001) and scale scores for
Factor 1 (r = .53, p < .0001), Factor 2 (r = .29, p < .0005), and
Factor 3 (r = .19, p s .02) were significantly positively correlated
with scores on the R-CMAS. The CASPI total score and the scale
scores for CASPI Factors 1, 2, and 3 for children younger than 13
years were significantly positively correlated with scores on the
R-CMAS (CASPI total score: r = .70, p < .0001; Factor 1: r =
.75, ^s.0001; Factor 2: r= .57, p < .0001; Factor 3: r = .38, p <
.003). For adolescents at least 13 years old, R-CMAS scores were
significantly positively correlated with CASPI total score (r = .52,
p < .0001) and scale scores for CASPI Factor 1 (r = .63, p <
.0001).
The CASPI total scores (r = .60, p s .0001) and scale scores for
Factor 1 (r = .63, p ^ .0001), Factor 2 (r = .33, p < .008), and
Factor 3 (r — .31, p £ .02) were significantly positively correlated
with scores on the HS. Regarding children less than 13 years old,
there were significant positive correlations for the CASPI total
score and scale scores for Factor 1 with scores on the HS (CASPI
total score: r = .61, p ^ .001; Factor 1: r = .68, p := .0002). For
adolescents at least 13 years old, HS scores were significantly
positively correlated with CASPI total score (r = .50, p ^ .002)
and scale score CASPI Factor 1 (r = .54, p < .0006).
General Discussion
This article described the development of a 30-item self-report
questionnaire for children and adolescents, the CASPI, to rate risk
for suicidal ideation or a suicidal act. Psychometric properties of
the CASPI were evaluated in prepubertal children and adolescents
with varied levels of suicidal behavior and who were either psy-
chiatric patients or community controls and who were from diverse
racial or ethnic and social status backgrounds. Responses to
CASPI questions reflected the construct of risk for suicidal behav-
ior in the past 6 months. CASPI was scored in a yes-no format and
contained one second-order factor that accounted for 59% of the
total variance. There were three primary factors indicative of
anxious-impulsive depression (16 items), suicidal ideation or acts
(6 items), and family distress (8 items) involving family discord
and psychopathology of relatives. These factors reflected multiple
correlates of suicidal behavior that had been identified in empirical
studies of children and adolescents.
This study suggested that the internal consistency (alpha) for the
30-item CASPI was .90, and for the factors it ranged from .77 to
 CHILD-ADOLESCENT SUICIDAL POTENTIAL INDEX (CASPI)
.86. The test-retest reliability (ICC) for the CASPI total score was
.76, and for the three factors it ranged from .59 to .76.
The CASPI total score and scale scores for Factors 1, 2, and 3
were valid in discriminating severity of suicidal and assaultive
behaviors and types of patient status. In general, the canonical
discriminant analyses suggested that scale scores for Factor 2,
followed by scale scores for Factor 1, contributed most to discrim-
inating between levels of suicidal tendencies. However, scale
scores for Factor 3 were important in discriminating subjects with
suicidal ideation from those with a history of a suicide attempt.
These results pointed out that each of the three factors had impor-
tant and different degrees of contributions in distinguishing types
of suicidal tendencies. Sample questions are included in the
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Appendix.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
Criterion validity was evaluated with discriminant analyses and
ROC analyses that suggested a range of sensitivities and specific-
ities for distinguishing levels of suicidal tendencies at their respec-
tive cutoff scores. For example, a CASPI total score of 11 identi-
fied 80% of children or adolescents who made a suicide attempt
within the past 6 months (sensitivity) with a rate of 65% for
children or adolescents who truly had no history of suicidal ide-
ation or suicide attempts (specificity). A CASPI total score of 11
identified 70% of children or adolescents who were truly positive
for a history of suicidal ideation or suicide attempt within the
past 6 months (sensitivity) with a rate of 65% for children or
adolescents who were truly negative for suicidal ideation or sui-
cide attempts (specificity). Reports of criterion validity are rare for
other self-report measures of suicidal risk.
Convergent validity of the CASPI was suggested by significant
correlations of CASPI total scores and scale scores for the three
factors with scores of reliable and valid self-report questionnaires
that rated depression, anxiety, and hopelessness.
Strengths and Limitations
These results about the psychometric characteristics of the
CASPI suggested that the CASPI is a reliable and valid self-report
questionnaire that can be administered to children and adolescents
to identify multiple domains of risk for suicidal behavior. It was
sensitive and specific in distinguishing groups of children and
adolescents with varied levels of suicidal behavior, suggesting that
the CASPI total score can identify a sizable proportion of children
and adolescents who are truly positive and truly negative for
suicidal risk. In general, it is important that a screening measure
for suicidal risk be as sensitive as possible to identify children and
adolescents who are truly at risk for suicidal behavior and with as
high a specificity as possible to avoid rating children and adoles-
cents as falsely negative, thereby missing a child or adolescent
who may be at risk for suicidal behavior.
In seeking to screen children and adolescents for a condition,
such as risk for suicidal behavior, clinical inference is required in
interpreting the score results of the screening measure. Specifi-
cally, since a main purpose of a screening instrument, such as the
CASPI, is to identify children and adolescents who are likely to
manifest risk for suicidal behavior, a relatively high false positive
rate can be generally acceptable. However, an important issue with
respect to the predictive ability of a screening instrument is that the
clinical utility of an instrument is affected by the true prevalence
315
rate of a condition, such as suicidal behavior, in a representative
population (Mossman & Somoza, 1991). In general, when the
prevalence of a condition is low, the degree of false positive results
increases and the results of a screening measure should be inter-
preted suitably (Mossman & Somoza, 1991). In the case of the
CASPI, it may be expected that since the true prevalence of
suicidal behavior in the general population is relatively low, the
number of false positives will be relatively high when the CASPI
is utilized in a community sample of children and adolescents. In
this regard, it is recommended that if the CASPI is to be utilized
to screen children and adolescents for risk of suicidal behavior in
the general population, it be utilized in conjunction with additional
methods that are applied after the score of the CASPI is reviewed
to further evaluate the level of risk for suicidal behavior, such as
by means of a subsequent in-person interview. In this way, false
positive children and adolescents may be identified and not be
impacted with undue alarm about their psychological state or be
referred for unnecessary interventions.
Research Implications
Additional research to evaluate the validity of the CASPI should
include larger samples of children and adolescents with diverse
severity of psychiatric symptoms and behaviors. Such studies
should be designed to evaluate the discriminant validity of the
CASPI for specific psychiatric diagnoses as well as other variables
that may not usually be associated with suicidal risk. Larger
samples will enable further study of the gender and age distinc-
tions that were noted in this study. Future studies that utilize other
standard instruments that rate suicidal risk can further test the
validity of the CASPI. Additional research is needed to develop
normative scores based on results utilizing large samples of clin-
ical and nonclinical children and adolescents. Furthermore, the
predictive validity of the CASPI to identify children and adoles-
cents who will commit suicide was not assessed in this study.
Prospective studies are needed to elucidate the CASPI's predictive
validity for events, such as suicide, with a low base rate and for
which predictive errors can have serious consequences. Additional
research is required to test the predictive validity of the CASPI for
suicidal ideation or suicide attempts. It is recommended that the
use of the CASPI as a predictor of suicidal behavior should await
the results of such research.
References
Alcohol, Drug Abuse, & Mental Health Administration. (1989). Report of
the secretary's task force on youth suicide Volume I: An overview and
recommendations. (DHHS Publication No. ADM 89-1621). Washing-
ton, DC: U.S. Government Printing Office.
Allan, W. D., Kashani, J. H., Dahlmeier, J., Taghizadeh, P., & Reid, J. C.
(1997). Psychometric properties and clinical utility of the Scale for
Suicide Ideation with inpatient children. Journal Abnormal Child Psy-
chology, 25, 465-473.
Ambrosini, P. J., & Dixon, J. F. (1996). Schedule for Affective Disorders
and Schizophrenia for School-Age Children (Kiddie-SADS IVR)-Present
state and epidemiological version. Philadelphia: Medical College of
Pennsylvania And Hahneman University.
 316 PFEFFER, JIANG, AND KAKUMA
Asarnow, J. R., Carlson, G. A., & Guthrie, D. (1987). Coping strategies,
self-perceptions, hopelessness, and perceived family environments in
depressed and suicidal children. Journal of Consulting and Clinical
Psychology, 55, 361-366.
Beck, A. T., Kovacs, M., & Weissman, A. (1979). Assessment of suicidal
ideation: The Scale for Suicide Ideation. Journal of Consulting and
Clinical Psychology, 47, 343-352.
Beck, A. T., Schuyler, D., & Herman, I. (1974). Development of suicidal
intent scales. In A. T. Beck, H. L. P. Resnik, & D. J. Lettieri (Eds.), The
prediction of suicide (pp. 45-56). Bowie, MD: Charles Press.
Beck, A. T., Steer, R., & Ranieri, W. (1988). Scale for Suicide Ideation:
Psychometric properties of a self-report version. Journal of Clinical
Psychology, 44, 499-505.
Beck, A. T., Weissman, A., Lester, D., & Trexler, L. (1974). The mea-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
surement of pessimism: The Hopelessness Scale. Journal of Consulting
This document is copyrighted by the American Psychological Association or one of its allied publishers.
and Clinical Psychology, 42, 861-865.
Brent, D. A., Johnson, B., Bartle, C., Bridge, J., Rather, C., Matta, J.,
Connolly, J., & Constantino, D. (1993). Personality disorder, tendency to
impulsive violence, and suicidal behavior in adolescents. Journal of the
American Academy of Child and Adolescent Psychiatry, 32, 69—75.
Brent, D. A., Perper, J. A., Goldstein, C. E., Kolko, D. J., Allan, M. J.,
Altaian, C. J., & Zeienak, J. P. (1988). Risk factors for adolescent
suicide: A comparison of adolescent suicide victims with suicidal inpa-
tients. Archives of General Psychiatry, 45, 581-588.
Cohen, J. (1960). A coefficient of agreement for nominal scales. Educa-
tional Psychological Measures, 20, 37—46.
Cohen, J. (1988). Statistical power analysis for behavioral sciences (2nd
ed.). Hillsdale, NJ: Erlbaum.
Cohen-Sandier, R., Berman, A. L., & King, R. A. (1982). Life stress and
symptomatology: Determinants of suicidal behavior in children. Journal
of the American Academy of Child and Adolescent Psychiatry, 21,
178-186.
Cole, D. A. (1989). Validation of the Reasons for Living Inventory in
general and delinquent adolescent samples. Journal of Abnormal Child
Psychology, 17, 13-27.
Cotton, C. R., & Range, L. M. (1993). Suicidality, hopelessness, and
attitudes toward life and death in children. Death Studies, 17, 185-191.
Cull, J. G., & Gill, W. S. (1982). Suicide Probability Scale manual. Los
Angeles, CA: Western Psychological Services.
Deykin, E. Y., Alpert, J. J., & McNamarra, J. J. (1985). A pilot study of the
effect of exposure to child abuse or neglect on adolescent suicidal
behavior. American Journal of Psychiatry, 142, 1299-1303.
Eyman, J. R., Mikawa, J. K., & Eyman, S. K. (1990). The problem of
adolescent suicide: Issues and assessment. In P. McReynolds, J. C.
Rosen, & G. J. Chelune (Eds.), Advances In Psychological Assessment
(Vol. 7, pp. 165-201). New York: Plenum.
Garrison, C. Z., Lewinsohn, P. M., Marsteller, F., Langhinrichsen, J., &
Lann, I. (1991). The assessment of suicidal behavior in adolescents.
Suicide And Life-Threatening Behavior, 21, 217-230.
Gorsuch, R. L. (1983). Factor analysis. Hillsdale, NJ: Erlbaum.
Gould, M. S., Fisher, P., Parides, M., Flory, M., & Shaffer, D. (1996).
Psychosocial risk factors for child and adolescent completed suicide.
Archives of General Psychiatry, 53, 1155-1162.
Hanley, J. A., & McNeil, B. J. (1982). The meaning and use of the area
under a receiver operating characteristic (ROC) curve. Radiology, 143,
29-36.
Harrington, R., Bredenkamp, D., Groothues, C., Rutter, M., Fudge, H., &
Pickles, A. (1994). Adult outcomes of childhood and adolescent depres-
sion: III. Links with suicidal behaviors. Journal of Child Psychology and
Psychiatry, 35, 1309-1319.
Hatcher, L. (1994). A step-by-step approach to using the SAS system for
factor analysis and structural equation modeling. Gary, NC: SAS
Institute.
Hollingshead, A. B., & Redlich, P. (1958). Social class and mental illness.
New York: Wiley.
Kazdin, A. E., French, N. H., Unis, A. S., Esveldt-Dawson, K., & Sherick,
R. B. (1983). Hopelessness, depression, and suicidal intent among psy-
chiatrically disturbed inpatient children. Journal of Consulting and Clin-
ical Psychology, 51, 504-510.
Kazdin, A. E., Rodgers, A., & Colbus, D. (1986). The Hopelessness Scale
for Children: Psychometric characteristics and concurrent validity. Jour-
nal of Consulting and Clinical Psychology, 54, 241-245.
Keane, E. M., Dick, R. W., Bechtold, D. W., & Manson, S. M. (1996).
Predictive and concurrent validity of the Suicidal Ideation Questionnaire
among American Indian adolescents. Journal of Abnormal Child Psy-
chology, 24, 735-747.
King, C. A., Katz, S. H., Ghaziuddin, N., Brand, E., Hill, E., & McGovern,
L. (1997). Diagnosis and assessment of depression and suicidality using
the NIMH Diagnostic Interview Schedule for Children (DISC-2.3).
Journal of Abnormal Child Psychology, 25, 173-181.
Kovacs, M. (1980). Rating scales to assess depression in school-aged
children. Acta Paedopsychiatty, 46, 305-315.
Kovacs, M. (1992). Children's depression inventory manual. North
Tonawanda, NY: Multi-Health System.
Kovacs, M., Goldston, D., & Gatsonis, C. (1993). Suicidal behaviors and
childhood onset depressive disorders: A longitudinal investigation. Jour-
nal of the American Academy of Child and Adolescent Psychiatry, 32,
8-20.
Lachenbruch, P. A., & Mickey, M. A. (1968). Estimation of error rates in
discriminant analysis. Technometrics, 10, 1-10.
Larzelere, R. E., Smith, G. L., Batenhorst, L. M., & Kelly, D. B. (1996).
Predictive validity of the Suicide Probability Scale among adolescents in
group home treatment. Journal of the American Academy of Child and
Adolescent Psychiatry, 35, 166-172.
Linehan, M. M. (1981). Suicidal Behaviors Questionnaire. Seattle: Uni-
versity of Washington.
Linehan, M. M., Goodstein, J. L., Nielsen, S. L., & Chiles, J. A. (1983).
Reasons for staying alive when you are thinking of killing yourself: The
Reasons for Living Inventory. Journal of Consulting and Clinical Psy-
chology, 51, 276-286.
Marciano, P. L., & Kazdin, A. E. (1994). Self-esteem, depression, hope-
lessness, and suicidal intent among psychiatrically disturbed inpatient
children. Journal of Clinical Child Psychology, 23, 151-160.
Metz, C. E. (1978). Basic principles of ROC analysis. Seminars In Nuclear
Medicine, 8, 283-298.
Metz, C. E., Goodenough, D. J., & Rossmann, K. (1973). Evaluation of
receiver-operating characteristic curve data in terms of information
theory with applications in radiology. Radiology, 109, 297-303.
Mossman, D., & Somoza, E. (1991). Neuropsychiatric decision making:
The role of disorder prevalence in diagnostic testing. Journal ofNeuro-
psychiatry and Clinical Neurosciences, 3, 84-88.
Myers, K. M., Burke, P., & McCauley, E. (1982). Family history of
suicidal behavior among suicidal attempters. Journal of the American
Academy of Child and Adolescent Psychiatry, 24, 474-480.
Ofek, H., Weizman, T., & Apter, A. (1998). The Child Suicide Potential
Scale: Inter-rater reliability and validity in Israeli in-patient adolescents.
Israeli Journal of Psychiatry Related Science, 35, 253-261.
Ohring, R., Apter, A., Ratzoni, G., Weizman, R., Tyano, S., & Plutchik, R.
(1996). State and trait anxiety in adolescent suicide attempters. Journal
of the American Academy of Child and Adolescent Psychiatry, 35,
154-157.
Pfeffer, C. R. (1986). The suicidal child. New York: Guilford Press.
Pfeffer, C. R. (1989). Suicide among youth: perspectives on risk and
prevention. Washington, DC: American Psychiatric Press.
Pfeffer, C. R., Conte, H. R., Plutchik, R., & Jerrett, I. (1979). Suicidal
 CHILD-ADOLESCENT SUICIDAL POTENTIAL INDEX (CASPI)
behavior in latency-age children: An empirical study. Journal of the
American Academy of Child Psychiatry, 18, 679-692.
Pfeffer, C. R., Conte, H. R., Plutchik, R., & Jerrett, I. (1980). Suicidal
behavior in latency-age children: An empirical study: An outpatient
population. Journal of the American Academy of Child Psychiatry, 19,
703-710.
Pfeffer, C. R., Hurt, S. W., Peskin, J. R., & Siefker, C. A. (1995). Suicidal
children grow up: Ego functions associated with suicide attempts. Jour-
nal of the American Academy of Child and Adolescent Psychiatry, 34,
1318-1325.
Pfeffer, C. R., Klerman, G. L., Hurt, S. W., Kakuma, T., Peskin, J. R., &
Siefker, C. A. (1993). Suicidal children grow up: Rates and psychosocial
risk factors for suicide attempts during follow-up. Journal of the Amer-
ican Academy of Child and Adolescent Psychiatry, 32, 106-113.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Pfeffer, C. R., Normandin, L., & Kakuma, T. (1994). Suicidal children
This document is copyrighted by the American Psychological Association or one of its allied publishers.
grow up: Suicidal behavior and psychiatric disorders among relatives.
Journal of the American Academy of Child and Adolescent Psychia-
try, 33, 1087-1097.
Pfeffer, C. R., Plutchik, R., Mizruchi, M. S., & Lipkins, R. (1986). Suicidal
behavior in child psychiatric inpatients and outpatients and in nonpa-
tients. American Journal of Psychiatry, 143, 733—738.
Pfeffer, C. R., Plutchik, R., Mizruchi, M. S., & Lipkins, R. (1987).
Assaultive behavior in child psychiatric inpatients, outpatients, and
nonpatients. Journal of the American Academy of Child and Adolescent
Psychiatry, 26, 256-261.
Pfeffer, C. R., Solomon, G., Plutchik, R., Mizruchi, M. S., & Weiner, A.
(1982). Suicidal behavior in latency-age psychiatric inpatients: A repli-
cation and cross validation. Journal of the American Academy of Child
and Adolescent Psychiatry, 21, 564-569.
Pfeffer, C. R., Zuckerman, S., Plutchik, R., & Mizruchi, M. S. (1984).
Suicidal behavior in normal school children: A comparison with child
psychiatric inpatients. Journal of the American Academy of Child and
Adolescent Psychiatry, 23, 416-423.
Pinto, A., Whisman, M. A., & McCoy, K. J. M. (1997). Suicidal ideation
in adolescents: Psychometric properties of the Suicide Ideation Ques-
tionnaire in a clinical sample. Psychological Assessment, 9, 63-66.
Range, L. M., & Knott, E. C. (1997). Twenty suicide assessment instru-
ments: Evaluation and recommendations. Death Studies, 21, 25-58.
Rao, U., Weissman, M. M., Martin, J. A., & Hammond, R. W. (1993).
Childhood depression and risk for suicide: A preliminary report of a
longitudinal study. Journal of the American Academy of Child Adoles-
cent Psychiatry, 32, 21-27.
Reynolds, W. M. (1987). Suicidal Ideation Questionnaire. Odessa, FL:
Psychological Assessment Resources.
(Appendix follows)
317
Reynolds, C. R., & Richmond, B. O. (1985). Revised Children's Manifest
Anxiety Scale (RCMAS) manual. Los Angeles: Western Psychological
Services.
Reynolds, W. M., & Mazza, J. J. (1999). Assessment of suicidal ideation
in inner-city children and young adolescents: Reliability and validity of
the Suicidal Ideation Questionnaire—JR. School Psychology Review, 28,
17-30.
Rotheram-Borus, M. J., & Trautman, P. D. (1990). Cognitive style and
pleasant activities among female adolescent suicide attempters. Journal
of Consulting and Clinical Psychology, 58, 554-561.
SAS Institute, Inc. (1989). SAS/STAT users guide (Version 6, 4th ed.).
Cary, NC: SAS Institute.
Shaffer, D. (1988). The epidemiology of teen suicide: An examination of
risk factors. Journal of Clinical Psychiatry, 49, 36-41.
Shaffer, D., Gould, M. S., Fisher, P., Trautman, P., Moreau, D., Kleinman,
M., & Flory, M. (1996). Psychiatric disorder in child and adolescent
suicide. Archives of General Psychiatry, 53, 339-348.
Shrout, P., & Fleiss, I. (1979). Intraclass correlations: Uses in assessing
rater reliability. Psychological Bulletin, 86, 420-428.
Somoza, E., & Mossman, D. (1991). Neuropsychiatric decision making:
Designing non-binary diagnostic tests. Journal of Neuropsychiatry and
Clinical Neurosciences, 3, 197-200.
Somoza, E., Steer, R. A., Beck, A. T., & Clark, D. A. (1994). Differenti-
ating major depression and panic disorders by self-report and clinical
rating scales: ROC analysis and information theory. Behavior Research
Therapy, 32, 771-782.
Spector, P. E. (1992). Summated rating scale construction: An introduc-
tion. Newbury Park, CA: Sage.
Spirito, A., Overholser, J., & Hart, K. (1991). Cognitive characteristics of
adolescent suicide attempters. Journal of the American Academy of
Child and Adolescent Psychiatry, 30, 604-608.
Spirito, A., Sterling, C. M., Donaldson, D. L., & Arrigan, M. E. (1996).
Factor analysis of the suicide intent scale with adolescent suicide at-
tempters. Journal of Personality Assessment, 67, 90-101.
Spirito, A., Williams, C., Stark, L., & Hart, K. (1988). Hopelessness Scale
for Children: Psychometric properties with normal and emotionally
disturbed adolescents. Journal of Abnormal Child Psychology, 16, 445-
458.
Steer, R. A., Kumar, G., & Beck, A. T. (1993). Self-reported suicidal
ideation in adolescent psychiatric inpatients. Journal of Consulting and
Clinical Psychology, 61, 1096-1099.
Tatman, S. M., Greene, A. L., & Karr, L. C. (1993). Use of the Suicide
Probability Scale (SPS) with adolescents. Suicide and Life-Threatening
Behavior, 23, 188-203.
 318 PFEFFER, JIANG, AND KAKUMA

Appendix

Six Sample Items From the Child-Adolescent Suicidal Potential Index (CASPI)

Instructions: Please answer the questions about yourself in the past 6 months

Yes No
4. Did you feel like doing dangerous things?
6. Did you often feel angry?
8. Did you often feel sad? _ _
17. Did you see frequent arguments between your parents?
19. Was your father or mother sad a lot?
26. Did you ever try to kill yourself?
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Copyright 2000 by Cynthia R. Pfeffer.

Received October 18, 1999

Revision received May 9, 2000
Accepted May 12, 2000

AMERICAN PSYCHOLOGICAL ASSOCIATION

SUBSCRIPTION CLAIMS INFORMATION Today's Date:.
We provide this form to assist members, institutions, and nonmember individuals with any subscription problems. With the
appropriate information we can begin a resolution. If you use the services of an agent, please do NOT duplicate claims through
them and directly to us. PLEASE PRINT CLEARLY AND IN INK IF POSSIBLE.

PRINT FULL NAME OR KEY NAME OF INSTITUTION MEMBER OR CUSTOMERNUMBER(MAYBEFOUNDON ANYPASTISSUELABEL)

ADDRESS DATE YOUR ORDER WAS MAILED (OR PHONED)

PREPAID CHECK CHARGE

CHECK/CARD CLEARED DATE:_
CITY STATE/COUNTRY ZIP
(If possible, send a copy, front and back, of your cancelled check to help us in our research
__ of your claim.)
YOUR NAME AND PHONE NUMBER ISSUES: MISSING DAMAGED

TITLE VOLUME OR YEAR NUMBER OR MONTH

Thank you. Once a claim is received and resolved, delivery of replacement issues routinely takes 4-6 weeks.
—^————— (TO BE FILLED OUT BY APA STAFF) ———-^———
PATE RECEIVED: DATE OF ACTION: _
ACTION TAKEN: INV. NO. & DATE:
STAFF NAME: LABEL NO. & DATE:.

Send this form to APA Subscription Claims, 750 First Street, NE, Washington, DC 20002-4242

PLEASE DO NOT REMOVE. A PHOTOCOPY MAY BE USED.