Receiver Opwatkg Characteristics QROC) analysis of six depression scales extracted from the Child

Bek~viw Ckecklist (CBCL) and Youth Self-report IYSR) in a clinicahy referred sample of adolescents
1’41s
= 6671 sh owed that t?aeir performance in discriminating between depressed afid not depressed
patients was comparable to other specifically designed depression rating scales (area under the ROC
curves bemeen 0.75 and 0.82). Performance was better for boys th2.n for girls. These results suggest that
specific depression selfreport rating scales may be unnecessary in adolescents if CBCL or YSR data arz
available, and that self-reports are not more accurate than parent-reports in the identification of
dqxession in this age group.

Kq worrls= Depression rating scales; Child Behaviour Checklist; Adolescents; Receiver operating charac-
teristic analysis

able (Costello and AngoId, 1988; Mazdin, 1990),

with the result that clinician; and researchers
In recent years there has been a proliferation lack empiricai data on which to base decisions
of scales to measure childhood and adolescent about what instrument is best for what task. More
depression (I&din, 19901, and new scales con- often than not, this issue is resolved by making
tinue to bc developed (e.g. Joshi et al., 1990). arbitrary choices or by using several instruments
Yet, evaluation of the specific advantages of one simultaneousfy. The former creates difficulties in
instrument over others has received little atten- communication and integration of research re-
tion, and information to assess that is not avail- sults becomes difficult, while the latter produces
wastage of resources. Writing about a similar
problem in adult psychiatry, Endicott et al. (1981)
C~mqmndence w: J.M. Rey, Rivendell, I-iospkal Road, Cm- lamented that ‘the need for cumulative knowl-
cord Wesi, N.S.W. 2138, Australia. edpe in psychiatry is often frustrated by the use of
different procedures for measuring the same con- reports would be most accurate. while self-re-
structs. AS a result of the effort:; to generate ports were hypothesized as more accurate than
improved procedures. the findings from earlier parent reports.
proccdurcs are left behind in the literature and
cannot bc integrated with later results’.
It would also seem advantageous to be able to
extract a measure of depression from an inexpen-
sive. multipore, multi-informant instrument that
could be used in a variety of situations and that The sample consisted of A67 adolescents (SS%
could provide information not only about depres- boys, 42GJ girls) referred for assessment to the
sion but also about other areas of psychopathol- Rivendell Adolescent Wnit between 1983 and
ogy and psychosocial functioning. Such a measure 1986, for whom DS -III diagncses were avail-
would have adva,itages over more specialized able at thz time of analysis. This sample did not
scales. particularly in epidemiologica; or clinical include an:! of the subjects utilized by Hepperlin
contexts. Hcpperlin et al. (1990) tried to achieve et al. (19UOd.
this for adolescents by obtaining a m:asurc of The Unit is a specialized psychiatric facility
depression compar;tblc to the most commor,ly that accepts referrals form across the state of
used and better researched depression scale, the New SouIh Wales, Australia. About one third of
Children’s Depression Inventory (CDI; Kovacs, the referrals are from school counselors, 12%
1981). from the Child Behavior Checklist KBCL, from psychiatrists, the rest are from relatives or a
Achenbach and Edelbrock, 1983, 19871. Follow- variety of agencies. Disruptive behaviours, either
ing a different method Nurcombe et al. (1989) at home or at school is the most common reason
also obtained a depression factor from the CBCL for referr:ll (34%) followed by depression (12%),
which according to the authors. showed some peer problems (10%) and school refusal (7aI.
promis,: as a scale to measure depression in its All these youngsters and their parents had
own right. This paper extends the results of a completed the YSR, CBCL and other question-
previous study Gky and Morris-Yates, 1991) that naircs prirlr to the clinical assessment, following a
showed that a scale formed with items from the standard&d procedure described in detail else-
deprkon factor described by Nurcombe et al, where (Rcy st al., 1989). Twenty two per cent of
f 1989) did in t’sct discriminate between subjects the subjects were from the top three sociocco-
with a diagnosis of major depression and those comic lcv~ls (SESJ, 60% from the three middie
with other diagnoses. However. since this scale is levels and 18% from the two lower SES levels.
based on parent reports, it remains to be seen None of the patients was an inpatient at the time
whether it is more efficient than a measure based of assessment, subsequently about 20% were ad-
on self-reports, because depression is substan- mitted for inpatient treatment. Only children aged
tially a subjective phenomenon, and there is some 12 to 16 years were included in the study. The
suggestion that informants might underestimate average age was 13.7 years; 15.7% of cases aged
depressive feelings in children and adolescents 12 years; 29.4% 13 years; 27.3% 14 years; 27.3%
(Angold, 19881. 15 years; 17.5% 16 years.
The main aim of this study was to compare the
relative diagnostic effectiveness of several meas- Menstares
ures extracted from the CBCL and Youth Self- Six measures of depression were considered.
Report (YSR, Achenbach and Edelbrock, 2987) Two were extracted from the CBCL, three
in the identification of depressed and not de- from the YSR and two from both CBCL and
pressed subjects. As a secondary objective, it was YSR. The Child Behavior Checklist (CBCL)
intended to clarify whether there were advan- (Achenbach and Edelbrock, 1983) is a parent-
tages in using parent-reports, self-reports or a completed checklist designed to provide a stan-
combination of both. The expectation was that dardized description of behavior of children aged
scales based on a combination of parent and child 4-16 years. It consists of a Social Competence
 9

scale and a Behavior Problem scale. The Behav- Gort WSR) (Achenbach and Edelbrock, 1987) is
ior Problem scale consists of 1118items (e.g. ‘Acts essentially identical to the CBCL with the excep-
too young for his/her age’) which are endorsed tions that it is a self-rating scale for children or
according to three response options: not true adolescents aged 11 to 18 years, it is worded in
(rating = O), somewhat/ sometimes true (rating = the first person (e.g. ‘I act too young for my age’)
1) and very true (rating = 2). The Youth Self-Re- and it contains only 102 items of the 118 in the

TABLE 1
Item content of the various depression scales. I: CBCL-NUR; 2: YSR-CDT; 3: YSR-DEPB; 4: YSR-DEPG; 5: Anxious/depressed
(all items from parent and child reports except items 18 and 91 from YSR on!;). Composite scale equals CBCL-NUR plus
YSR-CDL divided by two.

Item Item content Scale

number
1
8 Can’t concentrate +
9 Obsessions +
12 Lonely + +
13 Confused + f
14 Cries a lot + + +
17 Day dreams +
lti Harms self or attempts suicide + -l- + +
24 Doesn’t eat well +
27 Easily jealous +
30 Fears going to school + + + +
31 Fears doing something bad + -I” + +
32 Feels has to be perfect + +
33 Feels unloved + + +
34 Feells persecuted + -I-
Xi Feels worthless + + + + +
42 Likes to be alone f +
4s Nervous, tense + f t +
46 Nervous movements + +
47 Nightmares +
48 Not liked -I-
50 Too fearful or anxious + + +

51 Feels dizzy +

52 Feels too guilty + “I- +

54 Overtired + -t +

56B Headaches +
62 Poorly coordinated +

69 Secretive +
71 Self-conscious + + +

75 Shy or timid + + +

77 Sleeps more +
tsn Stares blankly +
86 Stubborn +
x7 Sudden changes in mood + -k -t

x9 Suspicious -t -t -t

91 Talks about killing self -I- + + +

100 Trouble sleeping + + +

IO2 Underactive, lack energy + + +

103 Unhappy, sad or depressed f + + + t

111 Withdrawn +
Worrying “I- + + + +
112
10

CBCL due to thz exclusion of 16 items (replaced Composite.

by filler items) considered to be inappropriate or This scale was formed by adding up depression
unlikely to be self-reported in this age group. scores on parent reports obtained with CBCL-
The six measures of depression were* NUW and those on self-reports obtained with
YSR-CD1 and dividing them by two.
CBCL-NUR.
Nurcombe et al. (1489) identified a depression Diagmsis
factor in CBCLs obtained from a series of 216 Two senior clinicians made independent DSM-
hospitalized adolescents which, they suggested, III diagnoses using all the information available
could be used as a scale to measure depression. in the file, including reports of psychiatric inter-
This scale consists of 22 CBCL items (Table 1). views with the child and the family, school re-
ports, reports of psychological and educational
Y,SR-CDL testing and parents’ and children’s item responses
Using a referred sample of 207 adolescents in the questionnaires. once the independent
(different from the sample used in this study) DSM-III diagnoses were made, the cases in which
Mepperiii,l et al. (1990) showed that scores on a there was disagreement were reviewed by the
scale made up of 15 items (Table Ii from ihe same two clinicians who, after further considera-
YSK had a correlation of 0.74 with depression tion of the available data, made a joint decision
scores obtained for the same subjects with the about which diagnostic criteria were met by the
CD!. This correlation is similar to the test-retest child and the consequent DSIWII diagnosis
reliability of child self-reports (Achenbach et al., (consensus diagnosis). This procedure was carried
1987). out with the first 367 cases, which were the object
of a reliability study (Rey et al., 1989), while the
Y.SR-DEFB. other 300 casts only had diagnoses made by con-
Factor analysis of the YSR (Achenbach and sensus by the same two clinicians, following the
Edelbrock, 1987) extracted a ‘depression’ factor, same procedure but without prior independent
but different items loaded on this factor for boys diagnosis.
(20 items) and for girls (32 items). Factor analysis Chance corrected agreement (K) for the inde-
of the CBCL (Achenbach and Edelbrock, 1983) pendent diagnoses were: overall, 0.59; major de-
found no clearly identifiable ‘depression’ factor pression, 0.57; dysthymic disorder, 0.3fi; separa-
for boys and a ‘depressed withdrawal’ factor for tion anxiety disorder, 0.80 (Rey et al., 1989).
girls. YS is formed with the 20 items of Estimated reliability of consensus diagnoses (K
the depression factor found in the YSR of boys with Spearman-Brown correction for two rates;
(Table 1). Guilford, 1954) were: overall, 0.74; major depres-
sion, 0.73; dysthymic disorder; 0.53; separation
YSR-DEFG. anxiety disorder, 0.89.
This scale contains the 32 items (Table 1) in Seventy eight percent of patients received a
the depression factor extracted in the YSR from single diagnosis, 21% two diagnoses and 1% had
girls (Achenbach and Edelbrock, 1987). three concurrent diagnoses. The most common
combination was conduct disorder and attention
Anxious / depressed. deficit disorder with hyperactivity. Major depres-
This measure corresponds to the ‘anxious/ sion was diagnosed in 3.6% of patients, while
depressed’ factor extracted by Achenbach et al. dysthymia was diagnosed in 9% of casts, a fur-
(1984) where combining information from multi- ther 1.3% had a diagnosis of bipolar disorder.
ple sources (parent, child, teacher). It contains Overall, 13.9% had a mood disorder diagnosis.
the items presented in Table 1, column 5, from Conduct disorder was the most common diagno-
the CBCL and YSR except for items 18 and 19 sis (28.3%) followed by attention deficit disorder
that are taken oniy from the YSR. This scale is with hyperactivity (13.5%). Anxiety disorders were
common to boys and girls. diagnosed in 17.3% of patients, the most: frequent
 11

diagnosis in this group was separation anxiety of these scales (Costello and Angold, 1988; Hsiao
(9.3%). et al., 1989; Mossman and Somoza, 1989). ROC
analysis was carried out using Dorfman and Alfs
Statistical analysis maximum likelihood estimation program
Analysis of tests, scales or diagnostic systems RSCORE-2 (Dor.?nan and Aff, 1969; Swets itnd
in terms of the receiver operating characteristic Pickett, 1982). QnIy data about the areas under
(RQC) of signal detection theory provides a mea- the ROC curves (AUC) are presented, other
sure of diagnostic accuracy not influenced by statistics of the ROC curves are available on
decision biases and prior probabilities, and it request.
places the performances of different systems on a Cases were not included in analyses if more
common, easy to interpret scale (Swets, 19881. than 10% of item responses in the scale being
This technique has been recommended as an studied were missing; this may result in slightly
appropriate means of testing the relative accuracy different number of cases in the various groups.

TABLE 2
Comparison of depression scores according to six scales between diagnostic grocps.

N Mean Mean SD r-test il P

z-score
CBCL-NUR
1. Major depression 23 22.5 i.03 7.7
2. No major depression 634 14.0 - 0.04 8.4 11-2) 5.12 0.000
3. Dyst hymia 62 17.5 0.40 8.3 (l-3) 2.44 0.016
4. Separation anxiety 57 14.7 0.05 7.9 (l-4) 3.94 O.OGa

YSR-CDI
I. Major depression 24 19.5 1.13 7.0
2. No major depression 635 11.4 - 0.04 6.8 (l-2) 5.76 0.000
3. Dysthymia 60 16.8 0.74 7,2 (l-3) 1.58 0.118
4. Separation anxiety 61 10.9 -0.11 6.3 (l-415.52 0.000

Y SR-DEPB
I. Major depression 14 24.51 1,25 9.4
2. No major depression 368 14.5 - 0.05 7.7 (l-2) 4.89 0.000
3. Dyst hymia 27 18.5 0.46 7.7 (1-3) 2.31 0.026
4. Separation anxiety 30 14.7 - 0.02 6.2 if -4) 4.25 0.000

YSR-DEPG
1. Major depression 10 37.3 0.79 11.4
2. No major depression Xl 26.4 - 0.03 13.2 (l-2) 257 0.011
3. Dyst hymia 33 36.2 0.71 12.6 (l-3) 0.24 0.814
4. Separation anxiety 30 23.5 - 0.25 14.2 (l-4) 2.79 0.008

Composite
1. Major depression 23 21.0 1.24 6.1
2. No major depression 627 12.9 - 0.05 6.3 (l-2) 6.25 0.000
3. Dysthymia 62 17.4 0.66 6.0 (l-3) 2.45 0.016
4. Separation anxiety 57 13.0 - 0.03 6.1) (I-4) 5.36 0.000

Anxious/depressed
1. Major depression 24 35.0 1.10 10.0
2. No major dspiessicul 623 22.3 - 0.04 10.9 (l-2) 5.59 0.000
3. Dysthymia 60 29.a 0.64 11.2 (l-3) 1.9s 0.055
4. Separation anxiety 59 21.6 -0.11 10.5 (l-4) 5.35 n.000

” two-tailed I-test.
12

If lessthan one in ten items were missing, those ROC analysis

items were scored as having the average item R0C analyses were carried out for the com-
score for that scale. parisons described above. Areas under the KQC
curves (AI-K) are shown in Table 3. An AUC of
0.5 means that diagnostic accuracy is no better
W.lk§ than random allocation of subjects to the various
groups, while an AUC of 1.0 means perfect dis-
crimination between diagnostic groups.
Three comparisons were carried out: betweer. All scales provided better discrimination than
patients with major depression and patients with random allocation of diagnoses. Discriminating
other diagnoses; between subjects suffering from ability by scales for major depression vs. no major
major depression and from dysthymia; and be- depression is similar to that for major depression
tween adolescents with major depression and with vs. separation anxiety. Overall, performance of
separation anxiety disorder. The last two cornpar- the scales parallels the findings obtained in the
isons were chosen because of the diagnostic and analysis of means. Taking into consideration the
phenomenological overlap among these disor- AUCs and confidence intervals, all scales show
ders. In cases with a diagnosis of major depres- broadly similar results. However, the best overa
sion and other concurrent diagnosis, the patient discrimination was obtained with the composite
was included in the major depression group. The scale. YSR-DEPB performed better than YSR-
group with other diagnoses was largely composed DEW. This could be due to the scales them-
of cases with disruptive disorders. selves or to the sex of the patient. To control for
Means tests for the five scales in the above the latter an estimate of the diagnostic accuracy
groups are presented in Table 2. Mean scores in was also carried out for the Composite scale
all scales are significantly different between pa- separately for boys (AUC = 0.86) and girls (AUC
tients with and without major depression, and = 0.79, suggesting that differences in accuracy
between subjects with major depression and those are largely accounted for by the sex of the pa-
with separation anxiety. To facilitate compar- tient. Depression scores on the Composite scale
isons, the means of the standardized (z) scores in boys with major depression were higher
are also given in Table 2. KL33SD) than those for girls, but not significantly

TABLE 3
Areas under the ROC curves for six depression scales when discriminating between patients with major depression and those with
other diagnoses (95% confidence intervals between parentheses)

Scale Discrimination between major depression and:

No major Dysthymia Separation
depression anxiety
CBCL-NU R 0.78 0.68 0.76
(0.69-0.87) IO.55-0.80) I0.65-0.87)
YSR-CD1 Q.81 0.62 0.82
(0.72-0.89) (0.49-0.75) (0.72--11.92)
YSR-DEPB 1).7x O.68 (1.80
(0.64-0.92) (OSM1.86) (O.63-0,97)
YSR-DEPG t-l.73 0.4’) 0.79
W.bI-0.85) (0.3 I -0.68) Ul.66-0.93)
Composite 0.82 0.70 0.82
(0.73-O.YO) (0.57-0.82) CO.71-0.93)
Anxiety/depression
!I;79 M.7 0.83
(0.72-0.87) u~.50-0.76~ W.73-0.92)
 13

TABLE 4 siparatim anxiety disorder are valid diagnostic

Sensitivity (SE) and spxificity (Sk’1 of the Compositor scak at entities, and on the understanding that reliability
selected cut-off prlints for the contrasb between major dc- of diagnoses sets a limit on the performance of
pressian and 0thrr di,gnostiC groups
any diagnostic test.
Reliability of the diagnoses considered here
Cut-off TOtal Boys Girls

SlXl-G
according to DSM-III criteria is only moderate,
SE SP SE SP SE SP
but comparable to that obtained in other reliabil-
Ih 0.X3 O.bX 0.7’1 0.71 (1.90 0.49 ity studies in children and adolescents. For exam-
1X 0.78 0.76 0.79 0.80 0.W MI
ple, Rey et al. (1989) reported an average reliabil-
20 O.hS 0.x3 0.7 1 O.Xh 0.X1 I Oh8
33
__ OX 0.89 0.71 II.91 Wio (I.76 ity across published studies of K = 0.56 for major
-- depression, k = 0.36 for dysthymia, and k = 0.74
for separation anxiety. A detailed study of the
Diagnostic Interview Schedule for Children-Re-
vised (DISC-R) reported poor agreement be-
so. At the contrary, depression scores for girls tween diagnoses obtained from the DISC-R and
with dysthymia were significantly higher than for those obtained from clinicians using a semi-stcuc-
boys 0 = XL_, T3* df = 56, 1’ < 0.(102) and the size of tured interview covering the same information
the difference was substantial IN74SD). That is, base as the DISC-R and employing a similar
boys with major depression had scores as high or (computerized) procedure to assign diagnoses.
higher than girls, while dysthymic boys had dc- Agreement for major depression was K = 0.36 for
pression scores significantly lower than dysthymic parent interview and K = 0.39 for child interview.
girls. Boys with other disorders also had signifi- The parallel results for separation anxiety were
cantly lower scores (0.5SD) than girls (I = 6.78, 0.30 and 0.27 (Schaffer et al., 1988). The overall
dC = 573. I’ = O.U~NN diagnostic performance of the various scales
Perform;ince of the different depression scales within this clinical sample is only moderate. Be-
according to age (younger = 12, 13 years: older = cause reliability of clinical diagnosis sets a ceiling,
14- 56 years) could not be estimated because there the performance of the scales would probably
were not enough cases of major depression (N = have been higher if diagnoses were more reliable.
3) in the younger age groups. However, scores on The results reported here might underestimate
the Composite scale were significantly higher (t the ability of these scales to discriminate between
= 4.29, df = 595, P < O.(?OObin the older than in depressed and not depressed patients.
the younger age group, but the size of the differ- Mood disorders in this clinical sample were
ence was not large (0.35SD). Similar results were not diagnosed often (13.9%) and major depres-
obtained with the other scales. sion, in particular, was diagnosed in only 3.6% of
Table 4 shows senstivity GE) and specificity cases. This may be reflection of the patterns of
(SP) at various cut-off points for the Composite referral to the unit, predominantly disruptive dis-
scale, when comparing diagnostic accuracy be- orders, and that clinicians were conservative when
tween major depression and other disorders. SEs assigning major depression diagnosis.
and SPs are modest. For example, a cut-off of 20 Confidence intervals of the AlJCs for the five
identified 65% of patients with major depression scales overlap (this may be due largely to the
in this referred sample: a false negative rate of small number of subjects in the major depression
15% and a rate of 17% for false positives. It also
-_ group). Therefore, we can only conclude that
shows that SE and SP vary according to the sex of there are no significant differences in accuracy
the patient. between them. Yet, some scales appear to Per-
form better than others, but this can only be
Discussion considered suggestive at this stage. Overall, there
is no support for the assumption that self-reports
These results need to be interpreted on the are more accurate in identifying depressed ado-
assumption that major depression, dysthymia and lescents than parent reports Mngold, 1988).
1-l

It is of interest that scores in subjects with depression may be easier to identif) in adolescent
majL>r depression are significantly different from boys than in girls. Dysthymic girls had depression
those without major depression (and those with scores almost one SD higher than dysthymic boys.
separation anxiety disorder) on all five SC&S. whiic that was not the case for those with major
The size of the difference I z-SCOITS, Table 1) is depression. Girls with other disorders also had
largest for YSR-DEPB Il.30SDI followed by the depression scores significantly higher than boys
Composite scale (1.29SDL Y SR-CD] t ].17SD), (0.5 SD). Thi s may reflect the higher preva!;;:nce
Anxious/ depressed ! 1.1LcSDJ, CBCLNU R ( 1 .W of depressive symptoms in girls and this, in turn,
SD) and YSR-DEPG (0.82 DI. Similar analyses may make it more difficult to discriminate be-
with other scales have produced conflicting rc- tween major depression and other disorders, par-
suits (Costello and Angold, 1988). For example ticularly dysthymia, in girls than in boys. Sex
Nelson et al. (1987) reported no difference in differences in the prevalence and pattern of de-
CDI scores between depressed and conduct dis- pressive symptoms in adolescents needs further
ordered children and adolescent inpatients, while study.
Rotundo and Hcnsley ( 198% found significant The AUCs for the comparison between ado-
diffcrcnccs bctwecn dcprcsscd and non-dc- lescents with and without major depression (Ta-
pressed subjects. hle 3) art: comparable tu the AUC found by
The depression factor found in the YSR of Mossman and Somoza (1989) for the composite
girls (YSR-DEPG) is substantially less accurate of seven studies using the Dexamethasone Sup-
than the one found in boys (YSR-DEPB). The pression Test (DST; AUC = 0.79). This suggests
difference is particularly marked when comparing that the scales considcrcd here arc as accurate as
major depression with dysthymia. However, this is the DST in discriminating patients with major
largely accounted for by the sex of the patient depression from those with other disorders. How-
rather than the scale itself. These results provide cvcr, pcrformancc of depression scales may have
some evtdcnce supporting the validity of these been enhanced by the fact that scales and diag-
YSR scales. The anxious/deprcsscd factor re- nosticians use the same sources of information
ccntly extracted by Achenbach et al. (19896 that (reports by parents and children), and diagnosti-
uses parent and child reports, discriminates be- cians were not blind to actual item responses in
tween major depression and separation anxiety as the questionnaires, while the DST is a diagnostic
accurately as bctwccn major depression and other test based on information independent from that
disorders. This suggests that this factor taps used to make a clinical diagnosis. This may have
mainly a depression construct, despite the fact cnhanccd the performance of these scales.
that, at face value, item content corresponds to a SE and SP of these scales in this referred
mixture of anxiety and depressive symptoms. sample are only moderate, but comparable to
In a recent report, Roberts et al. (19911 com- those reported for other depression scales in clin-
pared the performance of two depression scales, ical samples. For example, Asarrsow and Carlson
the Center for Epidemiological Studies Depres- (198% found a SE = 88% and SP = 88% in a
sion Scaie KES-D) and the Beck Depression group of 82 inpatients assessed with the K-SADS-
Inventory (BDI) in a community sample (N = E for a cut-off score of 13 on the Depression
1,704) diagnosed with the K-SADS administered Self-Rating Scale. Administering the same scale,
by trained interviewers. They reported an AUC 3irlcson et al. (1987) reported an SE = 07% and
for males of 0.93 (BDI) and 0.87 (CES-D). For SP = 77% in a group of 155 outpatients with
females the AUCs were 0.83 for both scales. clinical ICD-9 diagnosis, when a cutpoint of 15
Results for girls are comparabIe to those found was used. Kashani et al. (1990) described a SE =
with the Composite scale in our study, while 48% and SP = 87% for a cutpoint of 16 on the
those for boys are higher. The pattern of higher Beck Depression Inventory administer-cd to a
accuracy for boys than girls is similar to that group of 100 adolescent inpatients diagnosed with
found in this study. This suggests that major the DICA.
 AL’hcnhack. T.M. and Edelbro~k. C.S. (1987) Manual fr,r the
Yuuth Self-Report rtnd f’rtrfile. University of Vermont,
It can be concluded that scales extracted from kpartmenr uf Psychiatry. Burlington. VT.
Achcnhtich, T.M.. McConaughy. S.H. and Howei]. C.T. f 19x7)
the CBCL and YSR can indeed be used to idcn-
Chibd/ad~~escenF behavioral and emotional problems: im-
ti@ depressed subjects. In clinical samples, they plicntionr of cross-informant f3melations for situational
seem to be as accurate discriminating Individuals specificiQ.PsychoI. BuEl. IfIl, 213-232.
with major depression as other specifically de- .kkIhCh. -t-hf.. ~OIICWFS, C.K., Quay, H.C., Verhulst, F.C.

signed instruments and laboratory tests such as and HOWL% C.T. t 19x9) Replication nf empiricatty derived
syndromes as a brtsis fur taxonomy of child/adolescent
the DST. If both CBCL and YSR arc available,
psychopaFhalogy. .I Abnorm. Child Psychicrl. 17. 29%32.3.
the composite of CBCL-NUR and YSR-CDI may A~dd, A. t I9XKI Childhood 2nd adolescent depression, II:
be the best measure to utilize. If only the YSR is research in clinical populatirrns. 6i. J. Ps&iatry 15, 476
available. any of the three scales described here 492.

may be used, although YSR-DEPG appears to Asarntrw. J.R. and Carlcon. G.A. (IYXh) Depression self-rat-
ing scale: utility with child psychiatric inpatients. J. Con-
perform wf)rst, particularly when discriminating
sult. C‘lin. Psychology 53. 491-499.
hetwcern major depression and dysthymia. There Birlcson. P.. I ludwn. I.., Buchanan, D.G. and Wolf. S. i 19X7)
may be some advltntagc in using different cut off Clinical cvsluation of self-rating scale for depressive disor-
scores for buys and girls hecausc of the differ- der in childhood. J. Child Psychol. Psychiatry 2% Xi-1%
C’ostello. E.J. and Artgold. A. t 15)X8) Scaies to assess child and
ences in :i+.:uracy found in this study and by
adolcscenl depression: Checklists. screens and nets. J. Am.
Roberts et al. WI91 I. This, and the pattern of
Acad. Child Adtrlesc. Psychiatry 27, 72f1-737.
depressive symptoms according to age requires Dorfmart. D.D. and Alf, E. (1YhY) Maximum likelihood esti-
further study. In cases in which only parent rc- mation ot parameters of signal detection theory and deter-
ports are available, CBCL-NUR may be used. mination of confidence intervals: rating method data. J.
l-bwever, more rcscarcR w the efficiency of multi Mathem& Psychology h, 4X7-496.
Endicott, J., Cohen, .I.. Ncc, J., Fteiss. J. and Sarantakos. S.
vs. single purpose instrunlents needs to be carried ( IWI ) Hrtmilton Dcprcssinn Rating Scale extracted from
out. The findings reported hcrc will allow the regular and change,7 =A t:ersions of the Schcdulc f-a- Affective
a/cry large body of normal and clinical CBCL and Disorders and Schiznphrcnia. Arch. Gen. Psychiatry 3K.
YSR data already available to he used in the 9% I(i.7.
Guildford. J.P. I l%4) Psychometric Methods. New York. Mc-
study of adolescent depression. This study also
Graw Hill.
illustrates the usefulness of Receiver Operating licppcrlin. C.M., Stewart. G.W. and Rey, J.M. (IYYO) Extrac-
Characteristic (ROC) analysis in this type of rc- tion of depression scores in ndolesccnts from a general
search. purpose behaviour checklist. J. Affect. Dis. 18 105-I 12.
tlsiao, J.K.. Qrtko. J.J. ;:~.d Potter. W.Z. (‘;;83i i;iagnosing
Acknowledgments diagnoses:Receiver Operating Characteristic methods and
psychiatry. Arch. Gen. Psy<hi;,try 46, 664-667.
Joshi. P.T.. Capozolli, J.A. and Coyle. J.. . (ic390) The 2&ns
TM. Achenbach. V.R. Hensley, J.M. Plapp
Iiopkins Depression S&t!: normative: data and validation
and G.W. Stewart made comments to earlier in child psychiatry patients. J. Am. Acad. Child Adoksc.
drafts. Some of the dater included here were Psychiatry 29, 2XLXX.
presented at the 12th Congress of the Kashani. J.H.. Sherman. D.D., Parker, D.R. and Reid. J.C.
(IW!) Utility of the Beck Depression inventory with clinic
I.A.C.A.P.A.P., Kyoto (Japan), July lh-2CI, 1990.
rt:fcrred adolcsccnts. J. Am. Acad. Child Adolesc. Psychia-
This work was carried out with support from the
try 29. 27X-282.
National Health and Medical Research Council Kazdin, A.E. ( IWO) Childhabd depression. 3. Child PsychoI.
of Australia. Psychiatry 31. 121-160.
K(lvacs, M. ( 1981) Rating scales to assess depression in school
aged children. AcFa Paedopsychia;%r; 46, 305-315.
Mossman, D. and Scmoza, E. (1989) Mayimizing diagnostic
Achenhach, T.M.. Edelhrock and C.S. (19X3) Manual for thr information from the: dexamttthasone suppression tat.
Child Behavior Checklist and Revised Child Behavior Pro- Arch. Gen. Psychiatry. 36, 653-660.
file, University of Vermont, Department of Psychiatry. Nelson. W.M.. Politzno, P.M.. Finch. A.J.. Wendell. N. and
Burlington, VT. Mayh;dl, C. (1987) Children’s Depression Inventory: nor-
 mative data and utility with emotionally disturbed chil- pression scales. J. Am. Acad. Child Adolesc. Psychiatry 30,
dren. J. Am. Acad. Child Adolesc. Psychiatry 26, 43-48. 5%hh.
Nurcombe, B., Seifer, R., Scioli. A., Tramontana, M.G.. Rotundo, N. and Hensley, V.R. (19851 The Children’s De-
Crapentine. W.G. and Beauchesne. 14-C. (1989) Is major pression Scale. A study of its validity. J. Child Psychol.
depressive disorder in adolescence : distinct diagnostic Psychiatry 26. 917-927.
entity?. J. Am. Acad. Child Adolesc. Psychiatry 28, 333- Shaffer, D.. Schwab-Stene, M., Fisher, P., Davies, M., Piacen-
342. tini, J. and Gioia, P. (198,) A revelsed version of the
Rey, J.M., Plapp, J-M. and Stewart, G.W. (1989) Reliability of Diagnostic Interview Schedule for children (DISC-R)
psychiatric diagnosis in referred adolescents. J. Child Psy- Washington, DC: National Institute of Mental Health.
chol. Psychiatry 30, 879-888. Swets, J.E. (19%) Measucin~ the accuracy of diagnostic sys-
Rey, J.M. and Morris-Yates, A. (1991) Ad~llescent depression tems. Science 240, 1285-1293.
and the Child Behavior Chechlist (CBCL). J. Am. Acad. Swets. J.E. and Picket& R.M. 11982) Evaluation of diagnostic
Child Adolesc. Psychiatry 30, -123-427. systems: Methods from signal detection theory, New York,
Roberts, R-E.. Lewinsohn. P.M. and ipeltey, J.R. (1991) Academic Press,
Screening for adolescent depression: .r comparison of de-