Assessment of Psychiatrically Hospitalized Suicidal

Adolescents: Self-Report Instruments as Predictors

of Suicidal Thoughts and Behavior
ALISSA C. HUTH-BOCKS, PH.D., DAVID C.R. KERR, PH.D., ASHA Z. IVEY, M.A.,
ANNE C. KRAMER, M.S.W., AND CHERYL A. KING, PH.D.

ABSTRACT
Objective: The validity and clinical utility of the Reynolds Adolescent Depression Scale, Beck Hopelessness Scale,
Suicidal Ideation Questionnaire-Junior, and Suicide Probability Scale (SPS) were examined longitudinally among suicidal
adolescents. Method: Between 1998 and 2000, 289 psychiatrically hospitalized, suicidal youth, ages 12 to 17 years,
participated in this study. Self-report measures were completed at baseline. Clinician-rated suicidality and suicide attempt
were collected at baseline and 6-month follow-up. Results: Baseline self-reports were internally consistent and strongly
intercorrelated within male, female, white, and black subsamples. All of the measures predicted follow-up suicidality and
suicide attempts. Using published cutoff scores, the Beck Hopelessness Scale and SPS were moderately to highly
sensitive predictors of subsequent suicide attempts, as was the Suicidal Ideation Questionnaire-Junior for predicting
suicide attempts and broad suicidality. Alternative cutoff scores that predicted outcomes with moderate and high sensitivity
also were examined, with attention to resultant sacrifices in specificity. Conclusions: Baseline self-report scores predicted
follow-up suicidality. SPS contributed uniquely to prediction of future suicidality and suicide attempt. SPS may supplement
other sources of information when assessing suicide risk with this population. J. Am. Acad. Child Adolesc. Psychiatry,
2007;46(3):387Y395. Key Words: suicide, assessment, gender, race, adolescent.

Suicidal thoughts and behaviors are relatively common
among adolescents and can have devastating personal
and family consequences. National Center for Health
Statistics (2005) data indicate that 8.5% of high school
students reported attempting suicide in the past year,
Accepted October 12, 2006.
Dr. Huth-Bocks is with the Department of Psychology, Eastern Michigan
University, Ypsilanti. Dr. Kerr is with Oregon Social Learning Center, Eugene.
Ms. Ivey (Department of Psychology), Ms. Kramer (Department of Psychiatry),
and Dr. King (Departments of Psychology and Psychiatry) are with the
University of Michigan, Ann Arbor.
This research was supported by a Ronald McDonald House Foundation grant
awarded to Dr. King. The authors are grateful for the contributions of
participants, staff interventionists, evaluators, research assistants, and the
advisory team. We acknowledge Mohammad Ghaziuddin for his support in
launching this study, Paul Quinlan and Sanjeev Venkataraman for their
ongoing facilitation, Lois Weiss and Lesli Preuss for their assistance with
evaluations, and Jean Pletcher for her administrative assistance.
Correspondence to Dr. Cheryl A. King, University of Michigan Medical
School, 1500 E Medical Center Dr, Ann Arbor, MI 48109-0295; e-mail:
kingca@umich.edu.
0890-8567/07/4603-0387Ó2007 by the American Academy of Child
and Adolescent Psychiatry.
DOI: 10.1097/chi.0b013e31802b9535
and 16.9% reported having seriously considered it.
Suicidal thoughts and behaviors occur across gender,
age, and racial groups; however, adolescent girls more
often contemplate and attempt suicide but less
frequently die by suicide than adolescent boys. In
addition, the suicide rate among black adolescent boys,
although lower than for white youth, nearly doubled in
the 1970s to 1990s; rates in both groups have since
dropped slightly and stabilized.
Definitional and Conceptual Issues Related to Suicidal
Thoughts and Behaviors
To facilitate comparisons across studies and commu-
nication among researchers and clinicians, a standard-
ized nomenclature was proposed by O_Carroll et al.
(1996) and advocated for use by others (e.g., Goldston,
2003). Accordingly, a suicide attempt (with or without
injuries) is a self-inflicted behavior with intent (however
ambivalent) to kill oneself; a suicide threat is a verbal or
nonverbal interpersonal action (but not a direct act to
oneself) that communicates that a suicide-related action

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:3, MARCH 20 07 387

Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
HUTH-BOCKS ET AL.
may occur in the near future; and suicidal ideation refers
to self-reported thoughts of suicide-related behavior.
Research within adolescent samples has found these
constructs to be related but conceptually distinct.
Specifically, in community samples, suicidal ideation
correlates moderately to strongly with suicide-related
behaviors, such as suicide planning, giving away
possessions, and suicide attempts (Keane et al., 1996;
Reynolds and Mazza, 1999). Similarly, in psychiatri-
cally hospitalized adolescents, self-reported ideation
relates to self-report of suicidal behaviors and clinician
ratings of suicidal ideation (King et al., 1997b;
Prinstein et al., 2001), as well as future suicide attempts
(King et al., 1995, 1997a).
Suicidal Thoughts and Behaviors and the Related
Constructs of Depression and Hopelessness
Although a range of psychiatric disorders have been
linked to suicidality, defined here as suicidal ideation,
threats, and/or attempts, one of the clearest relation-
ships has been between suicidality and mood dis-
turbances. Within adolescent community samples,
moderate to strong correlations exist between depressive
symptom severity and suicidal ideation (Hovey and
King, 1996; Mazza and Reynolds, 1998; Reynolds,
1987). Depressive symptoms also correlate highly with
suicidal behaviors and intent and suicide risk estimates
(e.g., Cole, 1989). Several studies with adolescent
clinical samples have reported comparably strong links
between depressive symptomatology and self-reported
suicidal ideation (Kumar and Steer, 1995; Steer et al.,
1993), prior suicide attempts (Cole, 1989; Reinecke
et al., 2001), and future ideation and attempts (Goldston
et al., 1999; King et al., 1997a; Spirito et al., 2003).
Hopelessness (negative views or expectancies about
the future) also is associated with suicidality. Specifi-
cally, hopelessness has been related to suicidal ideation,
verbally expressed intent, and suicidal behaviors in
adolescent community (Cole, 1989; Mazza and
Reynolds, 1998; Osman et al., 1998) and psychiatric
inpatient samples (Kumar and Steer, 1995; Reinecke
et al., 2001) and is associated with adolescents_ past and
future suicide attempts (Goldston et al., 2001; Reinecke
et al., 2001). Importantly, although depression and
hopelessness are related, often co-occurring constructs,
each may contribute uniquely to suicidality. However,
whereas most studies with adults show that hopelessness
more strongly predicts suicidality, findings with
388
Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
adolescents have not converged. For example, several
studies in which both constructs have been considered
have found that adolescents_ depression, rather than
hopelessness, more strongly predicts suicide ideation
and attempts (Mazza and Reynolds, 1998; Spirito et al.,
2003), yet others have found hopelessness to be a
stronger predictor of suicidal ideation and to predict
suicidality beyond the effects of depression (Cole, 1989;
Kumar and Steer, 1995; Steer et al., 1993). Given these
mixed findings, both constructs require further con-
sideration as predictors of adolescent suicidality.
Measurement of Suicidality and Related Constructs
Although suicidality and related constructs can be
assessed through self-report ratings, structured inter-
views, and clinical judgment, agreement among
methods is only low to moderate (King et al., 1997b;
Prinstein et al., 2001). Several studies suggest that self-
report measures identify adolescents at risk for depres-
sion and suicidality as well as or better than other
methodologies (Prinstein et al., 2001; Shain et al.,
1990), perhaps because adolescents may more often
self-disclose on measures they perceive as Bprivate.[
Self-report measures also are of value for screening and
monitoring purposes because they can be rapidly
completed, scored, and interpreted. Self-report scales
differ in part by whether they are directed at tapping a
specific construct, such as suicidal ideation (e.g.,
Suicidal Ideation Questionnaire, Junior [SIQ-JR],
Reynolds, 1988), or whether they draw on multiple
domains of risk or adaptation (e.g., Suicide Probability
Scale [SPS], Cull and Gill, 1988, and Reasons for
Living Inventory for Adolescents, Osman et al., 1998).
Unfortunately, little is known about the ability of
specific self-report instruments to predict future
suicidality in clinical populations.
Present Study
Stemming from the need to accurately assess suicidal
youth, the convergent and predictive validity of several
well-known measures of depression, hopelessness, and
suicidalityVthe Reynolds Adolescent Depression Sur-
vey (RADS; Reynolds, 1987), Beck Hopelessness Scale
(BHS; Beck and Steer, 1988), SIQ-JR (Reynolds,
1988), and SPS (Cull and Gill, 1988)Vwas examined
in a sample of psychiatrically hospitalized, suicidal
adolescents. The incremental value of measures asses-
sing depression and hopelessness (RADS and BHS) in
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:3, MARCH 2007

predicting future suicidality, beyond measures of
suicidal ideation (SIQ-JR) or what is conceptualized
as general suicide risk (SPS), also was examined.
Comparisons among the measures of clinical utility
for predicting suicidality were based on published and
alternative cutoff scores. Finally, given the lack of data
on the assessment of suicidal racial minority youth,
measurement properties of these study measures were
examined within matched subsamples of white and
black adolescents.
METHOD
Participants
Participants were 289 psychiatrically hospitalized adolescents
(12Y17 years of age; mean 15.3, SD = 1.5 years) recruited from
either a private or university psychiatric hospital in the U.S.
Midwest to participate in a randomized clinical trial (King et al.,
2006) between 1998 and 2000. Inclusion criteria were a suicide
attempt (54%) or expression of suicidal intent or severe suicidal
ideation (46%) during the past month and a score (Q20) indicating
moderate to serious impairment on the Child and Adolescent
Functional Assessment Scale Self-Harm Subscale (Hodges, 1996).
Youth who were unable to understand study information or to
complete self-report measures because of severe mental retardation
or incapacitating psychosis were excluded.
Of the eligible adolescents approached for the study, 35% agreed
to participate. Most participants were girls (68%) and white (82%);
10% were black; and 8% were identified as another ethnic minority
group. Median annual family income was $50,000 to $59,999.
Parental education ranged from less than a high school education to
a graduate degree; 64% and 57% of mothers and fathers,
respectively, had obtained education beyond high school. Partici-
pant lifetime history of suicidality was as follows: suicide ideation
only (34%), single suicide attempt (47%), or multiple attempts
(19%).
A case-control design was used to examine race differences. Each
black participant (n = 29) was matched with up to two white
participants (n = 50) to the extent that all of the match criteria
(gender, lifetime attempt history, maternal education, and family
income) were met. Groups did not differ significantly on age.
Procedures
After Institutional Review Board approval was obtained,
adolescent assent and parent/guardian written informed consent
were obtained after hospital admission. Baseline assessments
required $1 hour for adolescents and 45 minutes for parents/
guardians. Participants were randomized to treatment as usual (n =
138) or treatment as usual plus supplemental intervention (n =
151). This supplemental intervention provided psychoeducation to
youth-nominated support people, who then supported adolescents_
care after hospital discharge. Baseline data on study measures were
available for 286 participants. Subject numbers varied somewhat
because measures were added when a larger battery became feasible.
Follow-up assessments occurred 5.0 to 8.7 months (mean 6.5,
SD = 0.7) after hospital discharge. Follow-up data from both
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:3, MARCH 20 07
Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
ASSESSMENT OF SUICIDALITY
treatment groups were considered because groups did not differ
on primary outcome measures (King et al., 2006) and the
supplemental intervention was not associated with the predictive
power of baseline measures in the present study. Of the 286
participants with baseline data, 36 were completely missing at
follow-up (28 actively withdrew, 4 were not located, 2 moved out of
state, 2 repeatedly failed to attend appointments), and 18 were
assessed outside the follow-up window, leaving 232 participants
available for analyses. There were no demographic differences
between study completers and participants with missing data.
Adolescents and their parents/guardians each were compensated for
completing the follow-up assessment ($25).
Measures
The RADS. The RADS (Reynolds, 1987) is a 30-item self-report
questionnaire that assesses frequency and severity of depressive
symptoms on a 4-point Likert scale (Balmost never[ to Bmost of the
time[). Scores range from 30 to 120. The published clinical cutoff
score is 77. Studies have reported high internal consistency (0.89 to
0.96), high test-retest reliability (0.79 to 0.87) coefficients in
community and clinical samples, and strong convergent and
discriminant validity (King et al., 1997b; Reynolds, 1987; Reynolds
and Mazza, 1998; Shain et al., 1990). Alphas in this study ranged
from .92 to .94 in the total sample and within gender and race
groups.
The BHS. The BHS (Beck and Steer, 1988) is a 20-item, true/
false questionnaire that measures negative expectations or attitudes
about the future. Total scores range from 0 to 20, with a published
clinical cutoff score of 9. The BHS, widely used with adolescents
(e.g., Goldston et al., 2001; Kumar and Steer, 1995; Mazza and
Reynolds, 1998; Osman et al., 1998; Reinecke et al., 2001), has
strong internal consistency and test-retest reliability (Goldston,
2000, as cited in Goldston, 2003) and convergent and discriminant
validity (Goldston et al., 2001; Kumar and Steer, 1995; Osman
et al., 1998; Reinecke et al., 2001; Steer et al., 1993). Internal
consistency ranged from 0.91 to 0.93 in the present total sample
and within gender and race groups.
The SIQ-JR. The SIQ-JR (Reynolds, 1988) is a 15-item self-
report questionnaire that assesses suicidal thoughts on a 7-point
scale (BI never had this thought[ to Balmost every day[). Scores
range from 0 to 90, with a published clinical cutoff score of 31.
Studies support its strong internal consistency and test-retest
reliability (0.89Y0.96) in community and clinical samples and in
different gender and race groups (Keane et al., 1996; Reynolds,
1988; Reynolds and Mazza, 1999) and its convergent and
discriminant validity in community and clinical samples (Keane
et al., 1996; King et al., 1997a, b; Prinstein et al., 2001; Reynolds
and Mazza, 1999). Alphas ranged from .93 to .95 in the present
total sample and within gender and race groups.
The SPS. The SPS (Cull and Gill, 1988) is a 36-item
questionnaire designed to assess four areas of general suicide risk
(hopelessness, suicidal ideation, negative self-evaluation, hostility)
using a 4-point Likert scale (Bnone or a little of the time[ to Bmost
or all of the time[). Total weighted scores range from 30 to 147; the
published clinical cutoff score is 78. Studies have indicated strong
internal consistency (0.90 and 0.93) and test-retest reliability
(0.92Y0.94; Cull and Gill, 1988; Tatman et al., 1993). Convergent
and discriminant validity has been supported in adolescent
community and clinical samples (Larzelere et al., 1996; Osman
et al., 1998). Alphas ranged from .90 to .93 for the present total
sample and within gender and race groups.
389
HUTH-BOCKS ET AL.
Spectrum of Suicide Behavior Scale. The Spectrum of Suicide
Behavior Scale (SSB; Pfeffer, 1986) is a 5-category clinician-rated
coding system used to assess history of suicidality. The instrument
yields five categories that are consistent with the nomenclature of
O_Carroll et al. (1996): nonsuicidal, suicidal ideation, suicidal
threat, mild suicide attempt (self-inflicted behavior with suicidal
intent that would not have necessitated intensive medical attention
or led to death), and serious suicide attempt (self-inflicted behavior
with suicidal intent that could have led to death or necessitated
intensive medical care). Prior studies support the predictive validity
of the instrument with respect to suicidal ideation, behavior, and
attempts (King et al., 1995, 1997a).
Trained research staff coded the SSB for suicidality during the
past month at baseline and follow-up. Staff used all available
information (parent and youth report, hospital records) to
determine the most serious applicable behavior. In addition, two
research staff independently coded recorded information concern-
ing possible suicide attempts during the entire 6-month follow-up
period and assigned participants to either the attempt (SSB criteria
for mild or serious suicide attempt) or nonattempt (including self-
harmful behavior with no intent) category; disagreements were
resolved by consensus. High interrater reliability for our research
program has been documented (King et al., 1997b).
RESULTS
Descriptive Statistics: Baseline Self-Report Measures
and Coded Suicidality
Means and standard deviations on baseline RADS,
BHS, SIQ-JR, and SPS for the total sample and for
boys and girls are listed in Table 1. t tests indicated
gender differences (girls greater than boys) on RADS
only. White and black adolescents did not differ
significantly on measures at baseline.
At baseline, adolescents_ most severe form of
suicidality in the past month (SSB) was as follows:
serious attempt (n = 79, 28%), mild attempt (n = 74,
26%), suicidal threat (n = 90, 31%), suicidal ideation
TABLE 1
Descriptive Statistics on Baseline Measures for Entire Sample
and by Gender
Total Mean Male Mean Female Mean t
(SD) (SD) (SD) Statistic
RADS 78.4 (17.5) 73.6 (17.3) 80.7 (17.2) 3.25*
BHS 8.8 (6.3) 7.7 (5.9) 9.3 (6.4) 1.70
SIQ-JR 40.9 (23.7) 37.3 (22.7) 42.6 (24.1) 1.63
SPS 78.2 (22.5) 74.2 (21.7) 79.9 (22.7) 1.79
Note: RADS = Reynolds Adolescent Depression Scale; BHS =
Beck Hopelessness Scale; SIQ-JR = Suicidal Ideation Questionnaire-
Junior; SPS = Suicide Probability Scale.
* p < .01.
390
Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
(n = 41, 14%), or insufficient information (n = 2, 1%).
SSB scores were collapsed to form two groups: suicide
attempters (mild or serious; n = 108 girls and n = 45
boys) and nonattempters (ideation or threat; n = 85
girls and n = 46 boys). Within the matched race
subsample, there were 30 white and 17 black attempters
and 20 white and 12 black nonattempters.
Relations Among Baseline Self-Report Measures
and Coded Suicidality
Associations among the baseline self-report mea-
sures were tested using Pearson correlations in the
total sample and within male, female, white, and black
subsamples. All of the measures, regardless of the
subsample considered, were significantly and positively
related (r = 0.56Y0.83, all p < .001). The relation
between RADS and BHS was significantly stronger
among girls (r [142] = 0.76, p < .001) than boys
(r [61] = 0.60, p < .001; Fisher z = 2.01, p < .05).
Otherwise, correlations did not differ significantly by
gender or race.
Relations between the self-report measures and the
baseline SSB were assessed using point-biserial correla-
tions in the total sample and within gender and race
subsamples. Self-reported ideation (SIQ-JR) was sig-
nificantly associated with nonattempt status in the total
sample (r = j0.12, p = .05), and higher general risk
(SPS) was significantly related to nonattempt status in
girls (r = j0.16, p < .05). These counterintuitive results
likely are due to the fact that unlike attempters,
nonattempters by definition had experienced some
baseline suicidal ideation. No other significant relations
were found.
Descriptive Statistics: Subsequent Suicidality
(6-Month Follow-up)
At follow-up, adolescents_ most severe form of
suicidality during the previous month (SSB) was as
follows: serious attempt (n = 6, 3%), mild attempt (n =
14, 6%), suicidal threat (n = 15, 6%), suicidal ideation
(n = 32, 14%), no suicidality (n = 134, 58%), or
insufficient information (n = 31, 13%). Again, two
groups were formed based on the presence (mild or
serious suicide attempt, threat, or ideation; n = 67) or
absence (n = 134) of suicidality.
According to codes assessing suicide attempts
between discharge and follow-up, 32 participants
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:3, MARCH 2007
 ASSESSMENT OF SUICIDALITY

TABLE 2 increased the probability of suicide attempt between

Predictors of Any Suicidality at Follow-up: Results of Logistic hospital discharge and 6-month follow-up (Table 3).
Regression Analyses
Suicidal vs. Nonsuicidal Groups Incremental Predictive Utility of Baseline Measures
" SE " Wald Odds Ratio (95% CI) Hierarchical logistic regression analyses were con-
RADS model a ducted to examine whether additional baseline mea-
Gender .30 .33 0.82 1.35 (0.71Y2.56) sures would significantly improve prediction of
RADS .53 .17 10.23** 1.70 (01.23Y2.35) outcome groups beyond gender (first step) and the
BHS modelb two self-report measures assessing suicidality (SIQ-JR
Gender .52 .40 1.69 1.68 (0.77Y3.65)
BHS .62 .20 9.95* 1.87 (1.27Y2.75)
and SPS). In the first regressions, the outcome was any
SIQ-JR modelc versus no suicidality at follow-up (SSB), and SIQ-JR
Gender .23 .35 0.44 1.26 (0.64Y2.49) was entered on the second step (Wald = 13.23, p =
SIQ-JR .55 .17 10.49** 1.74 (1.25Y2.44) .000). The addition of SPS (third step) significantly
SPS modeld improved the model [$x2 (1) = 5.32, p = .021; Wald =
Gender .25 .38 0.45 1.29 (0.62Y2.68)
5.06, p = .024; overall model x2 (3) = 19.82, p = .000],
SPS .76 .19 15.97** 2.14 (1.47Y3.10)
and the effect of SIQ-JR was rendered nonsignificant
Note: Gender code: 1 = boy, 0 = girl. (Wald = 1.57, p = .210). Similarly, when the outcome
x (2) = 11.23, p = .004; Nagelkerke R 2 = 0.08; nonsuicidal
a 2
was suicide attempt between discharge and follow-up
and suicidal RADS mean (SD) = 73.55 (17.56) and 81.82 (15.96).
x (2) = 11.40, p = .003; Nagelkerke R 2 = 0.11; nonsuicidal
b 2 and SIQ-JR was entered on the second step (Wald =
and suicidal BHS mean (SD) = 6.91 (5.65) and 10.23 (6.02).
x (2) = 11.24, p = .004; Nagelkerke R 2 = 0.09; nonsuicidal
c 2
TABLE 3
and suicidal SIQ-JR mean (SD) = 35.48 (22.14) and 47.47 (23.33). Predictors of Suicide Attempts after Discharge: Results of Logistic
x (2) = 18.28, p = .000; Nagelkerke R 2 = 0.15; nonsuicidal
d 2
Regression Analyses
and suicidal SPS mean (SD) = 69.73 (21.23) and 84.73 (20.29).
Suicide Attempt vs. No Suicide Attempt Groups
* p < .01. ** p < .001. RADS = Reynolds Adolescent Depression
Scale; BHS = Beck Hopelessness Scale; SIQ-JR = Suicidal Ideation " SE " Wald Odds Ratio (95% CI)
Questionnaire-Junior; SPS = Suicide Probability Scale. RADS model a

Gender j.05 .44 0.02 0.95 (0.40Y2.23)

RADS .47 .21 4.95* 1.60 (1.06Y2.41)
(14%) had made a suicide attempt (mild or serious), BHS modelb
196 (84%) had not, and 4 (2%) had insufficient Gender .70 .56 1.58 2.01 (0.68Y6.00)
BHS .93 .27 10.62*** 2.54 (1.45Y4.45)
information. SIQ-JR modelc
Gender .10 .47 0.04 1.10 (0.44Y2.76)
Relations between Baseline Self-Report Measures SIQ-JR .61 .23 7.03** 1.84 (1.17Y2.88)
and Subsequent Suicidality SPS modeld
Logistic regressions were conducted separately by Gender .35 .50 0.50 1.42 (0.54Y3.75)
SPS .91 .25 13.02*** 2.49 (1.52Y4.09)
outcome and by baseline measure to predict the
presence/absence of suicidality in the past month at Note: Gender code: 1 = boy, 0 = girl.
follow-up and the presence/absence of suicide attempt x (2) = 5.71, p = .06; Nagelkerke R 2 = 0.04; nonattempter
a 2

between discharge and follow-up. Scores on baseline
measures were standardized to aid interpretation. All
regressions controlled for gender.
and attempter RADS mean (SD) = 76.24 (17.20) and 84.03 (18.35).
x (2) = 12.51, p = .002; Nagelkerke R 2 = 0.15; nonattempter
b 2
and attempter BHS mean (SD) = 7.61 (5.76) and 12.47 (5.94).
x (2) = 7.62, p = .02; Nagelkerke R 2 = 0.07; nonattempter
c 2

Each baseline measure significantly predicted suicid- and attempter SIQ-JR mean (SD) = 38.55 (22.77) and 51.92
ality (any versus none) at follow-up (Table 2). Higher (23.66).
scores significantly increased the probability that x (2) = 14.97, p = .001; Nagelkerke R 2 = 0.15; nonattempter
d 2

adolescents would show some form of suicidality
(ideation, threat, or attempt) over the previous
month, $6 months after discharge. Similarly, higher
scores on the four baseline measures significantly
and attempter SPS mean (SD) = 73.65 (21.28) and 91.24 (17.60).
* p < .05. ** p < .01. *** p < .001. RADS = Reynolds Adolescent
Depression Scale; BHS = Beck Hopelessness Scale; SIQ-JR =
Suicidal Ideation Questionnaire-Junior; SPS = Suicide Probability
Scale.

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:3, MARCH 20 07 391

Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
HUTH-BOCKS ET AL.

10.89, p = .001), results revealed that SPS (third step)
significantly improved the model [$ x2 (1) = 4.13, p =
.042; Wald = 3.92, p = .048; overall model x2 (3) =
16.64, p = .001], and the effect of SIQ-JR was reduced
[Wald = 1.71, p = .191]. Neither BHS nor RADS
significantly improved the prediction of either follow-
up outcome beyond the effects of SIQ-JR. Finally, no
measure improved the ability to predict either outcome
beyond the effects of SPS.
Clinical Utility of Baseline Measures as Predictors
of Suicidality
The clinical utility of the four baseline self-report
measures as predictors of follow-up outcomes was
examined using receiver-operating characteristic curves.
Receiver-operating characteristic curves reveal sensitiv-
ity (percent of true positives identified) and specificity
(percent of true negatives identified) rates for all
possible values of the predictor variable and are
relatively independent of event base rate. The area
under these curves (AUC) provides an index of overall
classification accuracy. The closer both sensitivity and
specificity are to 100%, the closer the AUC is to 1.00,
whereas chance prediction yields an AUC of 0.50.
When predicting any suicidality during the 1-month
period preceding the follow-up assessment, AUC values
for RADS, BHS, SIQ-JR, and SPS were 0.63, 0.66, 0.65,
and 0.69, respectively (all p < .01); when predicting suicide
attempt during the entire 6-month follow-up period,
AUC values were 0.63, 0.72, 0.67, and 0.73, respectively
(all p < .05). All AUC values were in the fair range.
As shown in Table 4, published cutoff scores on
RADS and BHS yielded unacceptably low sensitivity
for predicting suicidality over the month before follow-
up. Alternative cutoff scores that yielded moderate
($70%) and high ($90%) sensitivity also were
determined. A moderate level of sensitivity was
achieved with slightly lowered cutoff scores but with a
sacrifice in specificity. The published cutoff score on
SIQ-JR yielded moderate sensitivity but fair specificity.
Finally, the established SPS cutoff score yielded
sensitivity and specificity near the moderate range,
and sensitivity could be increased to moderate levels
with little cost to specificity.
When suicide attempts after hospital discharge were
predicted, results were improved, particularly for
sensitivity. Table 4 shows moderate levels of sensitivity
for all measures using published cutoff scores but with

TABLE 4
Clinical Utility of Measures for Predicting Suicidality and Suicide Attempts
Suicidal vs. Nonsuicidala Suicide Attempt vs. No Attemptb
Raw Score Sensitivity, % Specificity, % Raw Score Sensitivity, % Specificity, %
RADS
Published cutoff 77 60 54 77 66 49
High sensitivity 62 90 26 66 91 25
Moderate sensitivity 71 70 38 76 69 47
BHS
Published cutoff 9 60 64 9 74 59
High sensitivity 3 89 29 6 90 44
Moderate sensitivity 7 70 54 11 68 69
SIQ-JR
Published cutoff 31 74 47 31 77 41
High sensitivity 12 89 15 12 89 15
Moderate sensitivity 34 69 50 46 69 61
SPS
Published cutoff 78 65 64 78 80 57
High sensitivity 61 90 38 72 92 50
Moderate sensitivity 74 72 60 79 72 59

Note: Alternative cutoff scores for moderate sensitivity are provided even when published cutoff scores yielded moderate sensitivity, with the
intent of evaluating possible improvements in specificity. RADS = Reynolds Adolescent Depression Scale; BHS = Beck Hopelessness Scale;
SIQ-JR = Suicidal Ideation Questionnaire-Junior; SPS = Suicide Probability Scale.
a
Suicidality in the last month was assessed at 6-month follow-up.
b
Suicide attempt was assessed for the entire period between hospital discharge and 6-month follow-up.

392 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:3, MARCH 2007

Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.

somewhat low specificity for RADS and SIQ-JR. All of
the measures were more sensitive at predicting suicide
attempts compared with any suicidality, perhaps
because of the smaller timeframe for measuring
suicidality. The published SPS cutoff score yielded an
especially strong sensitivity rate with an adequate
corresponding specificity.
DISCUSSION
Participants in the present study showed mean
baseline scores on self-reported depression, hope-
lessness, suicidal ideation, and general suicide risk
comparable to or higher than those in previous
studies with psychiatrically hospitalized youth (e.g.,
King et al., 1997a; Reinecke et al., 2001; Steer et al.,
1993) and substantially higher than in nonclinical
samples and community-based norming samples (e.g.,
Mazza and Reynolds, 1998; Tatman et al., 1993).
Given that all of the participants were hospitalized
acutely for suicidality, it also is not surprising that
mean scores on baseline measures met or exceeded
published cutoff scores for Bclinical significance.[
As expected, self-reported depression, hopelessness,
suicidal ideation, and general risk were significantly
interrelated within the total sample and within male,
female, white, and black subsamples. Intercorrelations
were equally strong by gender with one exception:
Depression and hopelessness were more strongly related
for girls than for boys. These results are consistent with
several studies of associations between depression and
related constructs (e.g., King et al., 1997b; Mazza and
Reynolds, 1998; Shain et al., 1990). Notably, few prior
studies (e.g., Hovey and King, 1996) have documented
relations among suicide risk factors within minority
samples. Thus, the present finding that correlations
among self-report measures were equally strong within
black and white subsamples addresses this gap in the
literature, and supports the convergent validity of these
instruments for suicidal black adolescents.
All 4 measures significantly predicted whether youth
showed suicidal thoughts or behaviors across a 1-month
period $6 months after hospital discharge and whether
youth attempted suicide between discharge and this
follow-up assessment. These findings are especially im-
pressive given the high levels of distress uniformly reported
by suicidal participants at baseline and the challenge of
assessing future risk of suicidality in such a group.
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:3, MARCH 20 07
Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
ASSESSMENT OF SUICIDALITY
Clinical Implications
Overall, these prospective findings support the
usefulness of self-report measures in predicting future
suicidality within clinical populations of suicidal youth.
The incremental value of considering more than one of
these self-report measures for assessing suicide risk was
not supported. Results suggest that the SPS may be
especially clinically useful for predicting a broad range
of suicidality. Specifically, when SPS was considered a
predictor of either suicide risk outcome, none of the
other measures contributed significantly to the models.
These findings may be particularly meaningful, given
that the SPS has been used less frequently with
adolescents. The SPS may perform well because the
items are less face valid, include more subtlety with
respect to suicidality, and assess a broad combination of
constructs theoretically linked to suicide (e.g., negative
self-evaluation, hostility). These measurement charac-
teristics may be especially important for assessing risk
among hospitalized adolescents who often become
aware of the relationship between self-reported suicidal
ideation and length of hospital stay.
Finally, sensitivity and specificity for established and
possible alternative cutoff scores on the four baseline
measures have implications for everyday clinical
decision making. In some cases, published cutoff scores
yielded unacceptable levels of sensitivity for predicting
suicide-related outcomes in this acutely suicidal sample,
and specificity was too low to be clinically useful. This is
not surprising given that suggested cutoff scores were
based on different populations (i.e., community or
outpatient samples), outcomes (e.g., depression), and
follow-up time frames. However, sensitivity rates using
BHS and SPS cutoff scores for predicting suicide
attempt and SIQ-JR cutoff scores for predicting both
suicidality and suicide attempt were acceptable. Specif-
ically, 74% to 80% of individuals were correctly
identified at follow-up. Given the importance of
accurately identifying suicidal individuals, sensitivity is
generally thought to be more critical than specificity
when screening for suicide risk (Goldston, 2003).
However, because increasing levels of sensitivity usually
come at a price of decreased specificity, clinicians must
balance the cost of overidentifying individuals.
When published cutoffs were unacceptable, moderate
sensitivity was achieved for the present sample with
alternative cutoff scores, albeit with sacrifice in specificity.
393
HUTH-BOCKS ET AL.
Still, rates of sensitivity and specificity in these instances
were similar to or higher than the few other existing studies
reporting such rates in clinical samples (Goldston et al.,
2001; Larzelere et al., 1996; Osman et al., 1998). High
sensitivity ($90%) for prediction of suicide attempts was
accompanied by unacceptably low specificity in all cases
except for the SPS. Overall, the SPS appeared to be
particularly strong because it predicted both suicide
attempts and a broader range of suicidality.
Limitations
The present study had several limitations. First,
generalizability is limited to acutely suicidal, psychiat-
rically hospitalized adolescents. Findings also do not gen-
eralize to youth who subsequently die by suicide because
we did not have adequate statistical power to focus on this
outcome. Because of the relatively low recruitment rate
(perhaps related to the acute state of eligible adolescents
and recruitment from a private, nonuniversity hospital),
unknown differences (e.g., severity of psychopathology)
mayhaveexistedbetweenparticipantsandnonparticipants.
The current sample also comprised primarily middle-class
families in one geographic region, which may limit
generalizability further. In addition, statistical power
limited our ability to examine relationships between
instrument scores and follow-up suicidality within sub-
samples defined by gender and race. Conclusions about the
prediction of suicide attempts at follow-up also must be
made with caution given the relatively small number of
attempts during that time. Finally, although some research
suggests that multiple attempters may differ from other
suicidal adolescents (Goldston et al., 1999), too few
multiple attempters participated in the present study to
statistically account for this variable in analyses.
In conclusion, clinicians working with at-risk youth
must consider the nature of their population, outcomes
of interest, and clinical resources when choosing
screening measures and cutoff scores. Clinicians
assessing future suicide risk with hospitalized and/or
suicidal youth may consider the use of a brief self-
report measure, such as BHS, SIQ-JR, or SPS, and the
published or alternative cutoff scores that yielded
moderate sensitivity in the present study. If there is a
particular concern about future suicide attempts, then
clinicians may find the SPS to be a useful screening
measure, with attention given to the published cutoff
score or, for greater sensitivity, a slightly lowered
cutoff score. In the present study, both options yielded
394
Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
acceptable specificity. Although the results of this
study shed light on the usefulness of specific measures,
suicidality continues to be a complicated and serious
phenomenon that is challenging to predict. Whenever
possible, suicide risk assessment should not be based
on one source of information (e.g., self-report). Future
research is needed to replicate these findings and to
clarify the potential role of different instruments in
identifying at-risk youth.
Disclosure: The authors have no financial relationships to disclose.
REFERENCES
Beck A, Steer R (1988), Manual for Beck Hopelessness Scale. New York: The
Psychological Corporation
Cole D (1989), Psychopathology of adolescent suicide: hopelessness, coping
beliefs, and depression. J Abnorm Psychol 98:248Y255
Cull J, Gill W (1988), Suicide Probability Scale (SPS) Manual. Los Angeles:
Western Psychological Services
Goldston D (2003), Measuring Suicidal Behavior and Risk in Children and
Adolescents. Washington, DC: American Psychological Association
Goldston D, Sergent D, Reboussin B, Reboussin D, Frazier P, Harris A
(2001), Cognitive risk factors and suicide attempts among formerly
hospitalized adolescents: a prospective naturalistic study. J Am Acad
Child Adolesc Psychiatry 40:91Y99
Goldston D, Sergent D, Reboussin D, Reboussin B, Frazier P, Kelley A
(1999), Suicide attempts among formerly hospitalized adolescents: a
prospective naturalistic study of risk during the first 5 years after
discharge. J Am Acad Child Adolesc Psychiatry 38:660Y671
Hodges K (1996), Child and Adolescent Functional Assessment Scale. Ypsilanti,
MI: Department of Psychology: Eastern Michigan University
Hovey J, King C (1996), Acculturative stress, depression, and suicidal
ideation among immigrant and second-generation Latino adolescents.
J Am Acad Child Adolesc Psychiatry 35:1183Y1192
Keane E, Dick R, Bechtold D, Manson S (1996), Predictive and concurrent
validity of the Suicidal Ideation Questionnaire among American Indian
adolescents. J Abnorm Child Psychol 24:735Y747
King C, Hovey J, Brand E, Ghaziuddin N (1997a), Prediction of positive
outcomes for adolescent psychiatry inpatients. J Am Acad Child Adolesc
Psychiatry 36:1434Y1442
King C, Katz S, Ghaziuddin N, Brand E, Hill E, McGovern L (1997b),
Diagnosis and assessment of depression and suicidality using the NIMH
Diagnostic Interview Schedule for Children (DISC 2.3). J Abnorm Child
Psychol 25:173Y181
King C, Kramer A, Preuss L, Kerr D, Weisse L, Venkataraman S (2006),
Youth-nominated support team for Suicidal adolescents (version 1): a
randomized controlled trial. J Consult Clin Psychol 74:199Y206
King C, Segal H, Kaminski K, Naylor M, Ghaziuddin N, Radpour L (1995),
A prospective study of adolescent suicidal behavior following hospital-
ization. Suicide Life Threat Behav 25:327Y338
Kumar G, Steer R (1995), Psychosocial correlates of suicidal ideation in
adolescent psychiatric inpatients. Suicide Life Threat Behav 25:339Y346
Larzelere R, Smith G, Batenhorst L, Kelly D (1996), Predictive validity of
the suicide probability scale among adolescents in group home
treatment. J Am Acad Child Adolesc Psychiatry 35:166Y174
Mazza J, Reynolds W (1998), A longitudinal investigation of depression,
hopelessness, social support, and major and minor life events and their
relation to suicidal ideation in adolescents. Suicide Life Threat Behav
28:358Y374
National Center for Health Statistics, Centers for Disease Control (2005),
Health, United States, Tables 46 and 62. Available at: http://www.cdc.
gov/nchs/fastats/suicide.htm. Accessed June 6, 2006
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:3, MARCH 2007
 ASSESSMENT OF SUICIDALITY

O’Carroll P, Berman A, Maris R, Moscicki E, Tanney B, Silverman M
(1996), Beyond the tower of Babel: a nomenclature for suicidology.
Suicide Life Threat Behav 26:237Y252
Osman A, Downs W, Kopper B et al. (1998), The Reasons for Living
Inventory for Adolescents (RFL-A): development and psychometric
properties. J Clin Psychol 54:1063Y1078
Pfeffer C (1986), The Suicidal Child. New York: The Guilford Press
Prinstein M, Nock M, Spirito A, Grapentine W (2001), Multimethod
assessment of suicidality in adolescent psychiatric inpatients: preliminary
results. J Am Acad Child Adolesc Psychiatry 40:1053Y1061
Reinecke M, DuBois D, Schultz T (2001), Social problem solving, mood, and
suicidality among inpatient adolescents. Cognit Ther Res 25:743Y756
Reynolds W (1987), Reynolds Adolescent Depression Scale: Professional
Manual. Odessa, FL: Psychological Assessment Resources
Reynolds W (1988), Suicide Ideation Questionnaire: Professional Manual.
Odessa, FL: Psychological Assessment Resources
Reynolds W, Mazza J (1998), Reliability and validity of the Reynolds
Adolescent Depression Scale with young adolescents. J Sch Psychol
36:295Y312
Reynolds W, Mazza J (1999), Assessment of suicidal ideation in inner-city
children and young adolescents: reliability and validity of the Suicidal
Ideation Questionnaire-Jr. Sch Psych Rev 28:17Y30
Shain B, Naylor M, Alessi N (1990), Comparison of self-rated and clinician-
rated measures of depression in adolescents. Am J Psychiatry
147:793Y795
Spirito A, Valeri S, Boergers J, Donaldson D (2003), Predictors of continued
suicidal behavior in adolescents following a suicide attempt. J Clin Child
Adolesc Psychol 32:284Y289
Steer R, Kumar G, Beck A (1993), Self-reported suicide ideation in
adolescent psychiatric inpatients. J Consult Clin Psychol 61:1096Y1099
Tatman S, Greene A, Karr L (1993), Use of the Suicide Probability Scale
(SPS) with adolescents. Suicide Life Threat Behav 23:188Y203

Evaluation of Resident Communication Skills and Professionalism: A Matter of Perspective? William B. Brinkman, MD,
Sheela R. Geraghty, MD, Bruce P. Lanphear, MD, MPH, Jane C. Khoury, MS, Javier A. Gonzalez del Rey, MD, Thomas G.
DeWitt, MD, Maria T. Britto, MD, MPH
Objective: Evaluation procedures that rely solely on attending physician ratings may not identify residents who display poor
communication skills or unprofessional behavior. Inclusion of non-physician evaluators should capture a more complete
account of resident competency. No published reports have examined the relationship between resident evaluations obtained
from different sources in pediatric settings. The objective of this study was to determine whether parent and nurse ratings
of specific resident behaviors significantly differ from those of attending physicians. Methods: Thirty-six pediatric residents
were evaluated by parents, nurses, and attending physicians during their first year of training. For analysis, the percentage of
responses in the highest response category was calculated for each resident on each item. Differences between attending physician
ratings and those of parents and nurses were compared using the signed rank test. Results: Parent and attending physician ratings
were similar on most items, but attending physicians indicated that they frequently were unable to observe the behaviors of
interest. Nurses rated residents lower than did attending physicians on items that related to respecting staff (69% vs 97%),
accepting suggestions (56% vs 82%), teamwork (63% vs 88%), being sensitive and empathetic (62% vs 85%), respecting
confidentiality (73% vs 97%), demonstrating integrity (75% vs 92%), and demonstrating accountability (67% vs 83%). Nurse
responses were higher than attending physicians on anticipating postdischarge needs (46% vs 25%) and effectively planning care
(52% vs 33%). Conclusions: Expanding resident evaluation procedures to include parents and nurses does enhance information
that is gathered on resident communication skills and professionalism and may help to target specific behaviors for improvement.
Additional research is needed to determine whether receiving feedback on parent and nurse evaluations will have a positive impact
on resident competency. Pediatrics 2006;118:1371Y1379.

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:3, MARCH 20 07 395

Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.