Preliminary Studies of the Reliability and Validity of the Childrenʼs Depression Rating Scale

Preliminary Studies of the Reliability and Validity of the Children'sDepression Rating Scale

ELVA O. POZNANSKI, M.D., JANET A. GROSSMAN, R.N., M.S.N., YAEL BUCHSBAUM, PH.D.,MARTA BANEGAS, M.D., LINDA FREEMAN, M.D., AND ROBERT GIBBONS, PH.D.

The Children's Depression Rating Scale, revised version (CDRS-R), is a reliable, clini­cian-rated scale which differentiates the depressed from the nondepressed child. The sumscore of the CDRS-R appears to provide a better estimate of depressive symptomatologythan does clinical impression. The relationship of the sum of the CDRS-R with globalclinical ratings of depression indicates that the scale measures the severity of depressionwhich is its primary purpose. The scale is not affected by the age of the child in our clinicalsample, and the content of the items grouped as mood, somatic, subjective, and behavior allshow good correlations with depression. The CDRS-R has been shown to be useful in avariety of settings, suggesting it is useful in both primary and secondary depressions.

Journal of theAmerican Academy of Child Psychiatry, 23,2:191-197, 1984.

The Children's Depression Rating Scale, revisedversion (CDRS-R), is a clinician-rated instrument forthe assessment of the severity of depression in chil­dren ages 6-12 years. The majority of depressionrating scales for children which have been publishedare self-report scales (Birleson, 1980; Kovacs, 1981).A self-report scale reliably quantifies subjective dys­phoria but is less useful when the individual does notperceive his/her own affective state or rates himselfor herself disproportionately high or low as comparedto others in a given clinical state.

The general need for depression rating scales forchildren has become more evident as cross-sectionalstudies of the incidence of childhood depression showthis is a common disorder in psychiatric populations.In addition, as treatment strategies emerge, the needto measure change both in an individual child and ingroups of children is apparent.

The senior author's early stance that depression inchildren could be directly observed rather than in­ferred formed the basis from which the scale wasdeveloped (Poznanski and Zrull, 1970). The Children's

Dr. Poznanski is Professor, Ms. Grossman is Instructor, Dr.Buchsbaum is Research Associate, and Dr. Banegas is AssistantClinical Professor, Department of Psychiatry, University of Illinoisat Chicago. Dr. Freeman is Assistant Professor of Psychiatry, De­partment of Psychiatry, University of Illinois at Chicago, and StaffPsychiatrist, Institute for Juvenile Research, University of Illinois atChicago. Dr. Gibbons is Assistant Professor of Biostatistics, Depart­ment of Psychiatry, Illinois State Psychiatric Institute, Chicago.

Requests for reprints should be addressed to Dr. Poznanski, De­partment of Psychiatry, University of Illinois at Chicago, 912 S.Wood St., Chicago, IL 60612.

The authors wish to thank Lee Robins for his help in datacollection.

This study was supported in part by the National Institute ofMental Health, Grant PHS MH 34196-02.

0002-7138/84/2302-0191 $02.00/0 © 1984 by the American Acad­emy of Child Psychiatry.

Depression Rating Scale emphasizes the central im­portance of the child's ability to report verbally symp­toms to their affective state and the direct observationof the child's behavior. Like other depression ratingscales, the CDRS-R relates to the diagnosis of depres­sion but cannot in itself make that diagnosis as infor­mation relating to other psychiatric diagnoses is lack­ing. The reliability of the child in reporting affectivesymptoms has been confirmed in several studies(Brumback and Weinberg, 1977; Carlson and Can­twell, 1979; Cytryn et aI., 1980).

Review of the Literature

Rating scales can be divided into two types, namely,self-report and clinician-rated scales. Typically, a self­report protocol, with its obvious advantage of usingless professional time, has not proven as reliable asclinician-rated scales. Carroll et a1. (1973), in a reviewof the literature of depression rating scales, found amoderate correlation of self-reported symptoms ofdepression, with a clinician's rating of depression (r =0.42) in a sample of adults. Prusoff et a1. (1972)reported correlations between these two types of scalesas ranging from 0.11 to 0.63. The majority of thesestudies have used different scales for the self-reportedrating and clinician's rating. This presents a majormethodological problem. The Hamilton Scale (Ham­ilton, 1960), which is clinician rated, is extensivelyused in clinical and research work with adults. TheCarroll Scale (Carroll et al., 1981), which is a self­inventory, was specially designed to correspond withthe Hamilton. As such, a comparison of the Carrolland Hamilton Scales overcomes the problem of usingtwo unrelated instruments, and the correlation be­tween these scales is higher than in the comparison ofunrelated scales (r = 0.67) (Carroll et al., 1981).

191

192 POZNANSKI ET AL.

Self-rating scales with young children need to be"interviewer-assisted" in that children can understandoral language before they can read it. Thus, mostchildren under 9 years of age must have a scale readto them. The latter process introduces a potential biasin that the child's relationship to the adult readingthe scale may influence the response. The influenceand direction of this type of bias has not been studiedin either adult or child populations.

The Children's Depression Inventory (CDI) (Ko­vacs, 1981) was the first and is the most widely usedself-report measure used in childhood depression re­search. It was developed by Kovacs from the BeckDepression Inventory (Beck, 1969) for adults. TheCDI has 27 items with a 3-point scale for severity.The CDI correlates moderately with global clinicaldepression rating (r = 0.55). Its main disadvantage isits reliability (Carlson and Cantwell, 1979), which hasbeen a difficulty in any self-report measure.

In addition to the CDI, the only other self-reportmeasure of depression in children that has had re­ported statistical analysis is a scale devised by Birle­son (1980). His scale has 18 items which are simplyand concretely worded and pose a forced choice situ­ation, that is, a yes or no answer.

At least three clinical rating scales for depressionin children are used in research. McKnew et al. (1979)developed and used the Children's Affective RatingScale (CARS). The major drawback to this scale isthat each item has a 10-point scale for severity withoutsubcategory definitions, making it more difficult toobtain good interreliability. In addition, its subscaleon fantasy is difficult to elicit and is subject to inter­pretation. Many mildly depressed children, as well asschizophrenic children with depressive affect, can pro­vide fantasy material. However, reports by moderatelyto severely depressed children are characterized byslow, short answers. Therefore, eliciting fantasy ma­terial can be difficult to obtain.

Two diagnostic structured interviews, the BellevueIndex of Depression (Petti, 1978) and the KiddiesSchedule for Affective Disorders and Schizophrenia(K-SADS) (Puig-Antich et al., 1978) are used for boththe diagnosis of depression in children and as a mea­sure of severity. Both of these structured interviewsfor depression in childhood have been used diagnos­tically in clinical studies of depressed children. Nostatistical analyses of their data have been publishedrelative to their use as severity rating measures ofdepression.

Review of the CDRS

A Depression Rating Scale for Children (CDRS)was devised and subsequently first used in a formalstudy of a random sample of children in an inpatient

pediatric unit in a medical hospital (Poznanski et al.,1979). This population was selected primarily for easeof entry. The initial study of CDRS reliability andvalidity was done by two child psychiatrists, inde­pendently rating each child on a global rating ofdepression and the CDRS. The two sum scores of theCDRS had an inter-rater reliability of r = 0.96. As anindex of convergent validity, the correlation of onerater's total CDRS score with a global clinical ratingof depression from the other interviewer was r = 0.89.

Since the above study was conducted in a pediatricunit, a high occurrence of secondary depression wasexpected. The next step was to rate consecutive ad­missions to a Children's Unit of a Regional MentalInstitution in the same manner as described above(Poznanski et al., 1983). In this study, four psychia­trists were used, two experienced and two inexperi­enced. The between-rater correlation was r = 0.80.Similarly, across-rater correlations (i.e, correlation be­tween experienced and inexperienced raters) were r =0.75 and r = 0.74. In this study, as well as in the studyin the pediatric ward, a high correlation was foundbetween global ratings of depression and the sum ofthe CDRS-R, indicating that the scale was indeedmeasuring severity.

Although the CDRS scale was performing well, sinceits inception 5 years of clinical research in childhooddepression resulted in the recognition that severalmodifications would improve the clinical utility of thescale.

Scale Description and Modification

The CDRS-R is a clinician-rated scale for severityof depression of children ages 6-12 years. It usuallytakes about 20-30 min to interview a child in order tomake a rating. Depressed children generally takelonger to interview than nondepressed children. Allpossible sources of information can be used, i.e., ad­ditional information from the parent, child careworker, teacher or other sources. However, the em­phasis is placed on information obtained from thechild.

The initial items for the CDRS were selected on thebasis of clinical experience, as there are not manywell-developed objective methods for validating thediagnosis of depression in children. Drug response hasbeen used in part with adults. Other measures, suchas the Dexamethasone Suppression Test, are currentlybeing developed for use in children, but are, as yet,less reliable than clinical diagnosis (Poznanski et al.,1983).Hence, the cardinal manifestations of childhooddepression for the CDRS were by necessity derivedfrom clinical experience.

The original CDRS had the following items: de­pressed mood, weeping, self-esteem and pathological

CHILDREN'S DEPRESSION RATING SCALE 193

guilt, morbid ideation, suicidal ideation, schoolwork,social withdrawal, irritability, anhedonia, tempo ofspeech, appetite, sleep, hypoactivity, physical com­plaints and fatigue. In the CDRS-R, the items dealingwith self-esteem and guilt have been split into separatecategories and a verbal item for feelings of depressionby the child was added. The original item of depressedmood in the CDRS is retained in the CDRS-R and israted on the basis of nonverbal behavior. Thesechanges were made for the following reasons. A verbalitem of depressed mood is useful in order to distinguishclinically those children who relate feelings of unhap­piness without showing depressive affect from chil­dren who deny feelings of unhappiness, despite man­ifesting depressive affect. Clinically, children in thefirst group, i.e., those who have only verbal dysphoria,are less depressed than children who show persistentdepressive affect. A report of dysphoria by the parentis sometimes difficult to interpret. The parents' re­ports of depression in their child's behavior are fre­quently contaminated by feelings that the parents mayhave themselves, particularly if the parents are de­pressed. Guilt and self-esteem are separated so thatthe characteristics of each behavior could be analyzedseparately. Hence the CDRS-R has 17 items, 14 ofwhich are scored on the basis of verbal observationand three on nonverbal items: tempo of language,hypoactivity, and nonverbal expression of depressedaffect.

The original CDRS had 15 items with a total scoreof 61 points. Since a rating of 1 = normal, the baselineis the total number of items, i.e., 15 for CDRS and 17for the CDRS-R, rather than O. The new CDRS-R has17 items with a total of 113 points.

The range of possible points in each subcategorywas 1-5 in the CDRS and has been increased from 1to 7 points in the CDRS-R. The additional 2 pointswere added as clinicians were rating between numberson the old scale, thus creating fractions. In addition,lengthening the pathological end of the scale increasedthe chances that slight lessening or worsening ofsymptomatology would be recorded. The lengtheningof each subscale is shown in Table 1.

If the subcategory description does not seem todescribe the child's behavior, the 1-7 scale can be usedfree of description with the following guidelines, 1 =normal,2 = doubtful pathology, 3 = mild symptoma­tology,5 = moderate symptomatology and 7 = severesymptomatology for the items.

All items have a description of normalcy and mild,moderate, and severe psychopathology. Wheneverpossible, the subcategory descriptions are meant toreflect increasing severity, both in terms of the fre­quency of the behavior and the intensity of such abehavior. Most of the time, the frequency and inten-

sity of behavior increase together. However, occasion­ally one might encounter a child who shows a veryintense behavior which occurs infrequently. The im­portance given to the frequency versus the intensityof behavior is ultimately dependent on clinical judg­ment. An example of subcategory description is shownin Table 2.

Conversion of the CDRS to the CDRS-R

The conversion formula for converting a CDRSscore to a CDRS-R score was determined by linearregression. The following equation is satisfactory ifthe CDRS score is above 20:

CDRS-R = -12.1 + 1.6 x CDRS

Method

Fifty-three children who were referred for possibledepression were evaluated in two outpatient clinical­research units, first at the University of Michigan andlater at the University of Illinois. The demographic

TABLE 1

Comparison of Scaling: CDRS and CDRS-R

CDRS CDRS-R

0 Unable to rate 01 Normal 12 Doubtful 23 Mild 3

44 Moderate 5

65 Severe 7

TABLE 2

Description for CDRS-R Item of Anhedonia

CAPACITY TO HAVE FUN (0-7)o Unable to rate1 Interest and activities realistically appropriate for age, person­

ality, and social environment. Shows no appreciable change withpresent illness. Any feelings of boredom transient

2 Doubtful3 Mild. Describes some activities realistically available several

times a week but not on a daily basis. Shows interest but notenthusiasm. May express some episodes of boredom more thanonce a week

4 Mild to moderate5 Moderate. Is easily bored. Complains of "nothing to do." Partic­

ipates in structured activities with a "going through the motions"attitude

6 Moderate to severe. Shows no enthusiasm or real interest. Hasdifficulty naming activities. May express interest primarily inactivities that are (realistically) uruiuailable on a daily or weeklybasis

7 Severe. Has no initiative to become involved in any activities.Primarily passive. Watches others play or watches television butshows little interest in program. Requires coaxing and/or push­ing to get involved in activity

194 POZNANSKI ET AL.

characteristics of these two populations suggest it isreasonable to combine these two groups as shown inTable 3.

The age and socioeconomic status (SES) are similarin the two populations. The sex ratios show a predom­inance of males in both samples as reported by otherresearchers. The major difference in these two popu­lations was an ethnically mixed population in theChicago sample as compared to an entirely whitepopulation in Ann Arbor.

Children were referred for an outpatient psychiatricevaluation by a diverse group of professionals, agen­cies and clinics, both private and public. Children wereaccepted for evaluation in our clinic if they met ourinclusion-exclusion criteria. For inclusion in ourstudy, the child must have been between the ages of 6and 12 years, not have a major physical illness, havean IQ over 70, and be off all mood altering drugs for2 weeks prior to the evaluation. The child must alsohave had a reliable adult informant to give the child'spast history. Legal consent to participate in our studywas required.

In both the University of Michigan and the Univer­sity of Illinois clinics, the diagnostic evaluation wascarried out over 2 days, 2 weeks apart. During theentire evaluation, the child and parent(s) were inter­viewed independently by different clinicians. The re­search protocol included an unstructured interview ofthe child, the K-SADS, a clinical global rating ofdepression, and a Global Assessment Score (Endicottet al., 1976). The parent or parents were interviewedabout the child simultaneously while the child wasinterviewed by a different clinician, using the sameinstruments in order to ensure independent data col­lection.

The majority of the children in the two sampleswere interviewed for the K-SADS and the CDRS-Rby two different child psychiatrists. Also, in order toreduce rater bias effects, the CDRS-R obtained on thesecond visit was administered by the clinician who didnot previously rate the child.

The results of the above interviews were not shared

TABLE 3Demographics of Two Clinical Populations

Chicago Ann Arbor p

until a clinical conference was held following thesecond day of evaluation. Thus each child had twoCDRS-R scores obtained by different raters 2 weeksapart. The purpose of the clinical conference is toshare information, establish the diagnosis, and for­mulate the recommendations for treatment. A follow­up visit to the clinic was scheduled 6 weeks after thefirst visit, both to obtain a third CDRS-R and to givethe parents an interpretation of our findings.

Results

Relationship of CDRS-R to Diagnosis of Depression

Thirty-four ofthe 53 children qualified for a clinicaldepression by RDC criteria, combining the "Definite"and "Probable" groups (the "Definite" group by DSM­III criteria and by Poznanski criteria (Poznanski etal., 1979». The mean CDRS-R of the RDC, DSM-IIIand Poznanski diagnostic groups of depressed childrenas previously described varied from 50 to 52, suggest­ing that group differences between depressed childrenusing these diagnostic criteria were minor. Thirty-twochildren met all three sets of diagnostic criteria, whilethe additional two children varied with each set ofdiagnostic criteria. Poznanski criteria differ from RDCand DSM-III primarily in having a nonverbal as wellas a verbal definition of dysphoria. Since there wereminor changes in group composition with the differentdiagnostic criteria, Poznanski criteria have been usedfor consistency within our clinical research unit andhave been used by our group since the beginning ofsystematic research data collection in childhooddepression. Table 4 shows the relationship between aclinical diagnosis of depression using Poznanski's cri­teria and the mean CDRS-R.

The difference between the depressed and nonde­pressed CDRS-R means is probably greater in thegeneral population than shown in the Table 4 becausechildren are referred to our clinic for suspected dys­phoria. Hence, the comparison sample probably hasmore borderline depressed children than a more typi­cal outpatient psychiatric population.

Our clinical experience has been that a child withan initial CDRS-R score of 40 or more ultimatelyobtains a diagnosis of a clinical depression. The meanCDRS-R of 52 and 49 with a standard deviation of 10

MeanDiagnosis N CDRS-R S.D.

TABLE 4

Comparison of CDRS-R with Clinical Diagnosis of Depression

Age X = 9.2 X = 9.8 NSS.D. = 2.2 S.D. = 1.10

Sex M = (60%) M = (81%)F = (40%) F = (19%)

SES X = 4.00 X = 3.38 NSS.D. = 1.24 S.D. = 1.24

Race B 43% W 100%W40%H 17%

Major depressive disorder:Definite group"Definite" and "probable" groups

Nondepressed psychiatric disorder

283419

524929

10114

CHILDREN'S DEPRESSION RATING SCALE 195

and 11 fits with this clinical experience. Two childrendiagnosed as clinically depressed had scores below 40and their scores were outside the range of one standarddeviation of the remaining group of "Definite" and"Probable" depressed groups. Although it needs to betested, it appears that some children with clinicaldepressions and low CDRS-R scores may deny de­pressive symptomatology if there is a stigma of de­pressive illness in their family.

Five children had sum scores over 60 and thesechildren clinically appear to be severely depressed.Thus the group of children with CDRS-R scores be­tween 40 and 60 points contain the majority of chil­dren with mild to moderate depression and a divisioninto the two clinical subgroups of mild and moderateis purely arbitrary. In our experience, the duration ofthe depressive episode appears to be clinically moreimportant than the exact CDRS-R score at the timeof evaluation; however, this problem merits furtherstudy.

Correlation between a Global Rating of Depressionand the CDRS-R

Prior to scoring the CDRS-R, the clinician gave thechild a Global Rating of Depression based on anoverall clinical impression. This rating,which is a 1­7-point scale, was then correlated with the totalCDRS-R score. The correlation of the CDRS -R withthe global rating was r = 0.87. Hence , an improvementof the CDRS-R over the CDRS is that it increases thecorrelation with Global Rating of Depression.

One way to study the correlation ofthe mean GlobalRating of Depression with the mean sum of the CDRS­R scores is to assign somewhat arbitrary points to theglobal rating based on clinical experience. Table 5shows the association between groups based on globalratings of depression and average CDRS-R scores.

The majority of children felt to be nondepressedclinically were given a global rating under 2.5 on a 7­point scale. Global ratings over 4 were rare for thenondepressed group, partially due to hesitancies onthe part of clinicians to rate a child as moderately orseverely depressed on clinical appearance alone.

The correlation of the initial global rating with thesum of the CDRS-R indicates that a wide range ofscores are encountered, particularly with global rat-

TABLE 5

Association between Global Ratings of Depression and CDRS -R

Mean Global N CDRS-R S.D.

<2.4 31 33 9.12.5-3.5 10 45 6.53.5- 4.5 4 60 6.5>4.5 7 68 11

ings of 1 and 2. However, the sum of the CDRS-R atthe first interview was found to be more predictive ofthe final clinical diagnosis of depression or a nonde­pressed psychiatric diagnosis after the total evaluationthan the initial global rating. Hence, the value of adepres sion rating scale systematically to assess de­pressive symptomatology has some diagnostic value.

Content of the CDRS Items Relative to the Child'sAge Function and Severity of Depression

The items on the CDRS-R were broken down intofour groups based on clinical experience. These fourgroups were the following:

1. MoodDepressed feelings (verbal)Depressed feelings (nonverbal)IrritabilityWeeping

2. SomaticAppetite-either increase or decreaseSleep-initial, middle and/or terminal insomniaExcessive fatiguePsychomotor retardation-includes tempo ofspeech and hypoactivityPhysical complaints (nonorganic).

3. SubjectiveSelf-esteemGuiltMorbid ideationSuicide ideation

4. BehaviorAnhedoniaSocial withdrawalSchoolwork

Correlation analyses (using Pearson product-mo­ment correlation coefficient) were then carried out , toassess the strength of the relationship between eachof these groups and the variables of age, Global As­sessment Score (GAS), Global Rating Depression(Global) and the sum score of the CDRS-R. Thecorrelations are shown in Table 6.

The highest correlations were obtained for each ofthe subgroups to items (i.e., mood, somatic, subjectiveand behavior) and both the global rating of depression

TABLE 6

Correlations between Groups of Items and Associated Variables·

Age GAS Global CDRS-R

Mood -0.02 -0.30 0.86** 0.88**Somatic 0.004 -0.10 0.75** 0.86**Subjective 0.11 -0.42* 0.75** 0.84**Behavior 0.17 -0.25 0.53** 0.68**

. *p < 0.02, "» < O.OOL

196 POZNANSKI ET AL.

Chicago sample as it is done by two different raters: r= 0.86 and N = 32.

The correlations between the 2-week and 6-weekratings (r = 0.81, N = 36) were also significant, indi­cating the stability of both the syndrome and themeasure.

Discussion

The CDRS and the CDRS-R have been used inthree different clinical populations: a pediatric unit ina medical hospital, a psychiatric inpatient unit, andpsychiatric outpatient clinics. Our increasing knowl­edge of the clinical phenomenology of childhooddepression undoubtedly influenced the clinicians' per­ception of depressive symptomatology and may thusaffect the rating of the instrument. Nevertheless, themean score of the CDRS-R in these populations goesin the expected direction. For example, the meanCDRS-R was lowest in a random sample of medicallyill children, higher in consecutive admissions to apsychiatric inpatient unit, and highest in the groupsof children specifically referred for possible depres­sions. Table 7 summarizes these findings.

Thus the CDRS-R has been shown to be useful ina variety of settings, and in diagnosing both primaryand secondary depressions.

ReferencesBECK, A. T. (1969), Depression: Clinical, Experimental and Theo­

retical Aspects. London: Staples Press, pp. 200-201.BIRLESON, P. (1981), the validity of depressive disorder in childhood

and the development of a self-rating scale: a research report. J.Child Psychol. Psychiat., 22:73-88.

BRUMBACK, R. & WEINBERG, W. (1977), Childhood depression: anexplanation of a behavior disorder of children. Percept. Mot.Skills,44:911-916.

CARLSON, G. & CANTWELL, D. (1979), A survey of depressivesymptoms in a child and adolescent psychiatric population. ThisJournal, 18:587-599.

CARROLL, B. J., FIELDING, J. M. & BLASHKI, T. G. (1973), Depres­sion rating scales: a critical review. Arch. Gen. Psychiat., 28:361­366.

- FEINBERG, M., SMOUSE, P. E., RAWSON, S. G. & GREDEN, J.P. (1981), The Carroll rating scale for depression; I. Development,reliability and validation. Brit. J. Psychiat., 138:194-200.

CYTRYN, L., McKNEW, D. & BUNNEY, W. (1980), Diagnosis ofdepression: a reassessment. Amer. J. Psychiat., 137:22-25.

ENDICOTT, J., SPITZER, R. L., FLEISS, J. & COHEN, J. (1976), The

TABLE 7

Mean CDRS-R Scores in Three Samplesand the sum score of the CDRS-R. The correlationsbetween each of the four groups and the CDRS-Rscore account for more of the variance than the cor­relations to the Global Rating, although both groupsof scores are highly significant. These findings suggestthat the scale is unidimensional and is tapping severalsymptom clusters which make up a depressive syn­drome.

The three groups which had the highest correlationto sum scores on the CDRS-R and accounted for thegreatest proportion of the variance, were the groupscontaining mood, somatic and subjective items. Thehighly significant correlation with mood items wouldbe expected, since depression in children is both ob­servable by trained clinicians and is verbalized bychildren when queried. The significant correlationbetween the somatic items and CDRS score suggeststhat the "vegetative" signs such as sleep and appetitedisturbances, are as centrally important in childhooddepression as they are in adult depression. The sub­jective items of self-esteem, guilt, morbid and suicidalideation also had a significant correlation, suggestingthat these are symptoms which can be assessed inchildren and which contribute to overall depressivesymptomatology.

The group of behavioral items, i.e., anhedonia, socialwithdrawal and problems with schoolwork, are oftenitems which depend on reports from multiple sources(e.g., parents, school 1 report) and by direct observa­tion. While the correlation for these behavior items tosum score or the CDRS-R was still significant, it didnot account for as much of the variance as the otherthree groups. It may be that items in this behaviorgrouping are not unique to depression, and this ac­counts for the slightly lower correlation.

The lack of correlation with age suggests that theratings on the CDRS-R are not affected by age of thechild in our psychiatric population. Therefore, chil­dren such as we see in the age range of 6-8 years canbe evaluated using the CDRS-R, although it is unclearif more reliance is placed on clinical judgment whenrating younger and less verbal children.

A modest significant correlation was obtained forthe subjective group of items and the Global Assess­ment Score. It may be that symptoms such as loweredself-esteem, feelings of guilt and morbid or suicidalideation impinge upon the child's overall level of func­tioning more so than mood, somatic, and behavioralitems. However, further analyses which will allow usto assess the contribution of each of these items, aswell as to investigate the factor structure of the CDRS­R, are clearly needed.

Test-Retest Reliability of the CDRS-R

The inter-rater reliability was determined by thecorrelation between the 0 and 2-week score in the

Sample

Children's Medical Inpatient-randomsample

Children's Psychiatric Inpatient-con­secutive admissions

Children's Outpatient Population­referred to dysphoria:

Ann ArborChicago

N

30

30

2231

MeanCDRS-R

30

36

3844

CHILDREN'S DEPRESSION RATING SCALE 197

global assessment scale: a procedure for measuring overall sever­ity of psychiatric disturbance. Arch. Gen. Psychiat., 33:766-771.

HAMILTON, M. (1960), A rating scale for depression. J. Neural.Neurosurg. Psychiat. , 23:56-62.

KOVACS, M. (1981), Rating scales to assess depression in school­aged children. Acta Paedopsychiat., 46:305-315.

McKNEW, D., CYTRYN, L., EFRON, A., GERSHON, E. & BUNNEY,W. (1979), Offspring of patients with affective disorders. Brit. J .Psychiat., 134:148-152 .

PETTI, T. (1978) , Depression in hospitalized child psychiatry pa­tients: approaches to measuring depression. This Journal, 17:49­59.

POZNANSKI, E. & ZRULL, J. (1970), Childhood depression: clinical

characteristics of overtly depressed children. Arch. Gen. Psychiat. ,23:8-15 .

-- COOK, S. C. & CARROLL, B. J. (1979), A depression ratingscale for children. Pediatrics, 64:442-450.

---- & CORZO. (1983), Children's depression rating scale : itsperformance in a residential psychiatric population. J. Clin. Psy­chiat., 44:200-203.

PRUSOFF, B. A., KERMAN, G. L. & PAYKEL, M. D. (1972), Concord ­ance between clinical assessments and patient's self-report indepression. Arch. Gen. Psychiat., 26:546-552.

PUIG-ANTICH, J. , BLAU, S., MARX, N. GREENHILL, L. & CHAM­BERS, W. (1978), Prepubertal major depressive disorder. ThisJournal, 17:695-707.