Upload
independent
View
0
Download
0
Embed Size (px)
Citation preview
Parent-Child Agreement in Prepubertal Depression:Findings with a Modified Assessment Method
NGA NGUYEN , M.D., SUZANNE WHIITLESEY, A.C.S.W., KATHY SCIMECA, R.N., M.S.,
DOLORES DIGIACOMO, R.N., BAO BUI, M.D., OSCAR PARSONS, PH.D.,
ALFRETIA SCARBOROUGH, M.P.H. , AN D DONNA PADDOCK, B.S.
ABSTRACT
Objective: Lack of or low parent-ehild (P-C) agreement is a well·documented problem in child psychopathology assess
ment. This study proposed to improve this agreement by using a modified assessment approach . Method: Ninety-three
depressed prepubertal children, aged 6 to 12 years, and their mothers underwent an assessment procedure that
combined multiple assessment measures given separately to child and mother (Schedule for Affective Disorders and
Schizophrenia for School-Age Children-Present Episode [K-SADS-PI , Children's Depression Inventory , and traditional
psychiatric interviews), confrontation of either and/or both informants with intra- and interinformant discrepanc ies, and
senior clinician's "best estimate" clinical judgment to solve discrepant ratings. Correlat ional statist ics (r, x, and z) were
used to compare child 's with mother's ratings on 20 K-SADS-P depress ive symptoms . Results: The major hypothes is,
that using our assessment procedure, P-C agreement would be significant and moderately high (rand" =.40 or higher),
was confirmed . The second hypothesis on dissociation of P-C agreement on behavioral versus ideational symptoms
was partially confirmed; the third hypothesis on adverse effects of maternal "depression" on P-C agreement was not
confirmed. Conclusion: Our assessment method has potential clinical application in enhancing diagnostic reliability of
childhood depress ion assessment. J. Am . Acad. ChildAdo/esc. Psychiatry. 1994, 33, 9:1275-1283. Key Words: parent
child agreement , diagnostic reliability, childhood depression assessment.
By reducing " criteria variance" and "information variance ," the use of DSM-II/ and DSM-II/-R (AmericanPsychiatric Association, 1980, 1987) and of standardized structured or semistructured interviews has enhanced diagnostic reliability in ad ult psych iatry, butnot in child psychiatry. The latter has been hampered
Accepted March 23. /994.Whm this study was conducted. Dr. Nguym was Associate Professor of
Child/Adolescent'Psychiatry, Department of Psychiatryand Behavioral Sciences,Oklahoma UniversityHealth Sciences Center (OUHSC). CollrgrofMedicine,Oklahoma City; shr is now at th« Uniuersity of Texas Medical Branch atGalveston. Dr. Parsons is Emeritus Profrssor, Ms. WhittlrsT] is ClinicalAssistantProjessor, Mrs. Scimeca is Clinical Instructor, Ms. Ditl iacomo is ResearchNurse, and Mrs. Scarborough and Mrs. Paddoct:are Research Staff. all in theDepartment of Psychiatryand BehavioralSciences, OUHSC Dr. Bui is ClinicalAssistant Proftssor ofPsychiatryand FamilyMedicine, DepartmentofPsychiatryand Behavioral Sciences and Family Medicine, OUHSC
This work was supported in part by grants to Dr. Nguym fto m the
Presbyterian Healt h Foundation and the Warrm Foundation.Reprint rrqursts to Dr. Nguyen, Department of Psychiatry and Behavioral
Sciences, University of Texas Medical Branch at Galveston, 30/ UniversityBlvd., Galveston, 7X 77555-0425.
0890-8567/94/3309-1275$03.00/0©1994 by the American Academyof Ch ild and Adolescent Psychiatry.
by an additional problem, "informant variance" (Kashani et al., 1985). Although the customary ch ildpsychiatry practice of using both the child and his orher parentis) to gather information about the child'ssymptoms has been shown to be helpful (G ershon etal., 1985; Verhulst et al., 1985), it inevitably introducesanother source of error, the parent-child (P-C) lackof, or low agreement, namely "i nformant variance."
There is now ample documentation of this issue oflow or lack of P-C agreement (Angold et al., 1987;Edelbrock er al., 1986; Herjanic and Reich, 1982; Ivens
and Rhem, 1988; Kazdin et al., 1983a-c; Weissman etal., 1980) . However, strategies aimed at solving thisdiagnostic dilemma have been mostly exploratory and
without any definitive results .The first type of strategy involves the identification
of specific sources of d iscrepancy that may adverselyaffect P-C concordance so that they can be readilyrecognized and taken into account; the following were
among the most commonly investigated and mostclinically relevant sources of P-C discrepancies: (1)
J. AM . AC AD. C HILD AD OL ESC . PSYCHI AT RY. 33 :9 . NOVEMBE RIDECEMBE R 199 4 1275
NGUYEN ET AL.
child's type of symptoms, i.e., behavioral versus ideational; (2) impact of parental psychopathology; and(3) dimension of P-C agreement, i.e., dichotomous(present/absent) versus continuous (degree of severity).The second type ofstrategy involves a variety ofempirical approaches suggested by investigators to solve P-Cagreement. These two strategies will be discussed inthe following sections.
Sources of P-C Discrepancies
Child's Type of Symptoms (Behavioral versus Ideational). Although Mokros er al. (1987) suggested that"it is overly simplistic to conclude that differences inratings between children and parents are largely afunction of the nature of the symptom being assessed"(p. 621), several studies showed significantly betterP-C agreement in behavioral (i.e., observable and factual) than in ideational (i.e., subjective and private)symptoms (Angold et al., 1987; Edelbrock et aI., 1986;Herjanic and Reich, 1982; Hodges et aI., 1990) .
Impact ofParent's Psychopathology. While impact ofparental depression on P-C agreement showed contradictory findings (Angold et al., 1987; Ivens and Rhem ,1988; Kashani et al., 1985; Moretti et al., 1985;Weissman et al., 1987), our preexperimental clinicalimpression was that maternal depression tended toadversely impact P-C agreement, because of the depressed mother's presumed low level of energy andoveridentificarion with her child 's symptoms.
Dimension of P-C Agreement (Dichotomous versusContinuous). Most studies investigated either the dichotomous dimension of P-C agreement (Angold etal., 1987; Chambers et al., 1985; Herjanic and Reich,1982; Ivens and Rhem , 1988; Reich et al., 1982;Weissman et al., 1987) or the continuous dimension(Hodges er al., 1990; Kazdin et al., 1983a-c; Knight etal., 1988) ; however, very few studied both dimensions(Mokros er al., 1987) . Because of their complementarynature, dichotomous and continuous dimensions ofP-C agreement will both be investigated for systematiccomparison in this study.
Empirical Approaches to Improve P-C Agreement
Empirical approaches include (1) administration ofmultiple measures for establishing a case (Beardsley etal., 1985 ; Sylvester et al., 1987); (2) "use of collateralinformation from other sources" (Kashani et al., 1985 ,p. 441); (3) submitting all available information to a
diagnostician for a " best estimate" diagnosis (Angoldet al., 1987); and (4) use of an expert clinician toreview parents' interviews and all available clinical data(Weissman et al., 1984).
Current Study
Given the above review, we used a modified approachto attempt to narrow the P-C rating gap in a sampleof depressed children. Our design combined (1) useof multiple assessment measures with child and mother;(2) confrontation of either and/or both informants notonly with interinforrnanr, but also with intrainformantdiscrepancies; and (3) use of clinical "best estimate"decision rules of a senior diagnostician, who monitoreddata gathering step-by-step via a one-way mirror toweight information and solve discrepant ratings. Thetwo latter approaches, i.e., the confrontation with intrainformant discrepancies and the senior diagnostician's first-hand observation of the entire data-gatheringprocess, constitute the unique factors of our modifiedassessment approach.
From all the above discussion and literature review,and the description of our assessment approach, threehypotheses were derived and tested . (1) Our assessmentapproach would result in an overall P-C agreementthat would be significant and at least moderately high,i.e., correlations of .40 or higher. (2) There would bea pattern of higher P-C agreement on the behavioralthan the ideational category of symptoms: P-Cagreement would be significant and at least moderately high in the behavioral category; P-C agreementwould be nonsignificant and low in the ideationalcategory. (3) Compared to "nondepressed" mothers,"depressed" mothers would have a significantlylower P-C agreement.
METHOD
Subjects
Subjects were recruited from consecutive admissions of prepubertal children to our outpatient Child and Adolescent DepressionClinic for diagnosis and treatment of affective disorders. Oursample consisted of 93 prepubertal subjects, 65 boys and 28 girls,aged 5 years 6 months to 11 years 11 months (mean age 9.37 ::!:1,56), and their mothers. Ethnic composition consisted of 75Caucasian, 15 African-American, 1 Hispan ic, and 2 mixed racialbackground. The sample was largely middle socioeconomic status(SES) according to Holl ingshead criteria (Hollingshead, 1975);73% were level II, III, and IV; 19% were level I or V; and 8%were undetermined.
1276 ]. AM. ACAD . C H I LD ADOLES C. PSYCHIATRY, 33 :9 , NOVEMBERIDECEMB ER 1994
Data gathered from clinical summary scores derived from ourmain assessment instrument, the Schedule for Affective Disordersand Schizophrenia for School-Age Children-Present Episode(K-SADS-P; Puig-Antich and Ryan, unpublished), were computedto assign DSM-III-R diagnoses to subjects. The diagnostic composition of our sample was as follows: 44 "pure" depressive disorders(major depressive disorder or dysthymia or minor depression); 45double depressions (major depressive disorder + dysthymia); 41depressive disorders comorbid with one nonaffective diagnosis(anxiety disorder or conduct disorder or oppositional defiant disorder or attention-deficit hyperactivity disorder); 10 depressive disorders comorbid with two nonaffecrive diagnoses; and two depressivedisorders comorbid with more than two nonaffective diagnoses.
Assessments, Interviewers, and Scoring Paradigm
Multiple assessment instruments were administered. TheK-SADS-P interview was used to make DSM-III-R diagnoses ofdepressive disorders in the child. The Children's Depression Inventory (CDI) (Kovacs, 1981) and the traditional clinical interviewsgiven separately to the child and to the mother were used asaccessory assessment tools to uncover inforrnanrfs)' intra- and/orinterinformant discrepancies. To test our secondary hypothesisinvolving the impact of maternal "depression" on P-C agreement,mothers were also given the Beck Depression Inventory (BDI)(Beck, 1967), a short, validated self-report scale (Beck er al., 1988)that gave a current (state) measure ofmothers' depressivesymptoms.
All above assessments concurrently took place in a video laboratory, where they were monitored step-by-step by the senior author(a child psychiatrist) via a one-way mirror. The K-SADS-Pinterviewers-a psychiatric social worker and two psychiatricregistered nurses, all seasoned child clinicians with a minimumof 8 years' experience-had at least 40 hours of intensive trainingin K-SADS-P interviews by the senior author, who, herself, hadreceived training by Drs. Puig-Anrich and Ryan's team. Interraterreliabilities between K-SADS-P interviewers were 85% to 90% asto DMS-III-R diagnoses of affective disorders before the study.Traditional clinical interviews were given separately to mother andchild by child fellows or senior psychiatry residents.
First, a K-SADS-P interviewer administered the K-SADS-P tothe mother, then observed the mother fill out a CDI-P on thechild and a BDI on herself in room A. Concurrently, a childfellow or psychiatry resident gave a traditional clinical interviewto the child, then observed the child fill out a CDI on himself orherself in room B; the senior author simulraneously monitoredthese two interviews via a one-way mirror. At the end of this firstperiod, both interviewers held a brief consultation with the seniorauthor (consultation I), with main focus on relevant and/or discordant data. Then these two interviewers switched rooms, theK-SADS-P interviewer administered the K-SADS-P interview tothe child in room B, while the child fellow or psychiatry residentgave the clinical interview to the mother in room A, again underthe monitoring of the senior author. At the end of this secondperiod, the two interviewers held another consultation with thesenior author (consultation II) regarding discrepant data. In caseof major inter- or intrainformant discrepancies, the K-SADS-Pinterviewer went back to probe and/or confront the child, themother, or both. A final meeting between the senior author andinterviewers was held to weight symptom scores, assign clinician'ssummary symptom scores, and formulate DSM-III-R diagnoses.
The following examples illustrate the scoring paradigm beingused and the process that determined the choice of confrontation:
PARENT-CHILD AGREEMENT
"solo" (child alone or mother alone) versus "joint" (child andmother together).
Example 1. In the traditional clinical interview, one child tearfullyadmitted to the child fellow that, after his grandfather's death ayear ago, he had been wishing "not to live anymore to join hisgrandfather in heaven"; he also endorsed the CDI item "I thinkabout killing myself, but I would not do it." However, he laterdenied having any suicidal ideation (score of 1) to the K-SADSinterviewer. As the latter had been debriefed by the senior diagnostician about the child's positive endorsement of suicidal idearion inthe CDI and the clinical interview, she gently confronted the childwith his discrepancies, resulring in the child's admission of frequentsuicidal idearions without suicidal plan (score of 3); this scorecoincided with the mother's K-SADS score of 3.
In this case of inrrainforrnanr discrepancy, a confrontation withthe child alone about his own inconsistencies was sufficient toyield a clinically satisfactory answer; thus, there was no need forjoint confrontation. The child's final score was changed from Ito 3, while the mother's score remained 3. Thus, the P-C agreementwas appropriately improved by this solo confrontation.
Example 2. A mother reported her child's absence of initialinsomnia (score of I), while the child subsequently endorsed 2-hourinitial insomnia (score of 4); when confronted alone with thediscrepancy, the child convincingly proved his point by statingthat, as he turned and tossed in his bed, he constantly checkedhis clock from his bedtime (l0:00 P.M.) until midnight before hecould finally fall asleep.
In this case of interinforrnant discrepancy, the child's answer tothe solo confrontation was convincingly factual, thus eliminatingthe need for joint confrontation. The mother, then, kept her scoreof I, while the child kept his score of 4. The P-C discrepancywas not improved by the confrontation process, which, nevertheless,helped increase the diagnostic accuracy. Indeed, without this confronrarion, there would have been no reliable clue as to whetheror not the clinician summary score (based on which the diagnosiswas made) should have been 1 or 4.
Example 3a. The mother endorsed lack of initial insomnia (scoreof 1) in the child, while the child subsequently and separatelyendorsed a 3-hour initial insomnia (score of 5) without being ableto back it up by factual data. Thus, a joint confrontation wasneeded to clarify these conflicting answers. It then became obviousthat the mother was a more objective judge: the child readilyagreed with her as she reminded him that, when she came to hisroom 1 hour and 30 minutes after his bedtime to wake him upfor bathroom break (because of the child's nocturnal enuresis), hewas usually found in deep sleep.
In this case of inter informant discrepancy, the joint confrontationappropriately helped to decrease the PC discrepancy from (mother =
I; child = 5) to (mother = 1; child = I).Example 3b. Suppose that the same child had remained adamant
about insomnia; his score would have remained 5 (to reflect hisconsistent answer), although the mother was more factual. Thus,the final P-C discrepancy would have remained high (mother'sscore = 1; child's score = 5). Nevertheless, this confrontation stillincreased the accuracy of the clinician's summary score (score of1 to reflect the mother's more factual answer).
In summary, these four examples showed that the "best estimate"clinical decision as to the most appropriate confrontarion type(solo or joint) depended on the clinical context in which the P-Cdiscrepancies arose: (l) inter- versus intrainformant discrepancies,and (2) presence or absence of data needed to determine whichanswer and/or which informant was more reliable.
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY. 33:9 NOVEMBERIDECEMBER 1994 1277
N G UYEN ET AL.
As above, the scoring paradigm was fairly straightforward: theclin ician scored the individual's final answer, as stated after theconfrontation , whereas the clinician 's summary score reflected thelatter's judgment as to wh ich answer or which informant wasmore believable.
In summary, this multiassessment procedure featured an operarionalizarion of data combina tion by (1) submitt ing all data assessments to a senior diagnostician 's first-hand mon itoring, (2) usingthe senior d iagnostician 's " best estimate" clin ical interpretation offindings-derived from the K-SADS-P int erviewer's confrontationof informants with inter- and in trainform ant discrepancies- tosolve discrepant data. Thus, we used " best estimate" clini cal decisionrules . rath er than "a prio ri" decision rules, to determine typeof con frontation , weight info rmation, and resolve discrepancies.Patterned after the " best estimate" method used by Leckman eral. (1982) on adult subjects, ours was, however, improved by thefact that ou r " best estimate" decision rules were derived from thesenior diagnostician's direct observation , rath er th an retrospectivereview, of all assessment data.
Data Analys is
T wen ty variables from the K-SAD S-P were selected for ana lysis.T he select ion of these variables was patterned after th at of Mo kroset al. (1987) because these variables rep resented the most easilyrecognized signs and symptoms of dep ression . These 20 variab leswere categori zed as 12 behavioral, i.e., observabl e and factual itemsand 8 ideational, i.e. , pr ivate and subjective items. (Refer to T able1 for categorization of these 20 iterns.)
We used lC and Pearson (r) correlation coefficient s to correlate, respectively, the dich otom ous (present versus absent) andcont inuous (degree of severity) dimensions of those 20 variables.K-SAD S-P symptom scores of 3 and above mean clinica lly signifi cant or presence of symptom, and scores of less th an 3 mea nabsence of sympto ms. Severity scores usually range from 1 to 6,with th e exceptio n of irrita bility, depressed mood , and suicida lideation, which range from I to 7, and self-pity, which rangesfrom I to 4.
Since a same "unblind" interviewer was used for moth er's andch ild's separa te K-SAD S-P intervi ews, we purposefully selectedcutoff scores that were high er th an the mean P-C cor relation r = .25calculated from 120 studies in a meta-analysis done by Achenbach eral. (1987). Amo ng the few invest igators who foun d relatively highr values, H odges and coworkers' (1990) conservative interpretationof the ir find ings (i.e., .40 $ r < .50 considered as mode rate) wasused for ou r r cutoff scores; th us, moderately high co rrelationswould be defined by .40 $ r < .50, high correlations by r::::: .50,low-moderate correlations by .30 $ r < .40, and low correl ationsby r < .30.
For lCvalues. we adopted H odges and cowo rkers' (1987) strategy.which followed the guidelines used in th e studies of chi ld int erviews:lC < .40 defined as low, .40 $ lC < .60 as modera te, lC ::::: .60 ashigh (Cos tello er al., 1984).
All statistical tests were two-tailed, and statistical significancefor r values was conservatively defined at p $ .0 1 to avoid a typeI error, du e to the large number of tests performed. In addition ,the p values of the individua l 20 items were adjusted to cont rolfor family-wise erro r, using the Bonferro ni correctio n.
Z statistics were used to evaluate differences between two independent correlatio n coefficient s. All statistical tests were two-tailed,with statistical significance for z values defined at p < .05. Basedo n the find in gs of Beck et a l. (1988). mat ern al BDI scoresof :::::13 were selected to indicate maternal " depression" ; alter-
TABLE 1Parent -C hild Correlations (r and x) Derived from 20 Mo ther
C hild K-SADS-P Symptom Sco res
x Pearson r
Behavioral categoryI. D epressed appearance .43 .49"*2. Irri tabi lity .20 .233. Aches/pains .40 .40"*4. Fatigue .40 .235. Dec reased con centrat ion .25 .38"*6. Psychom otor agitation .57 .45"*7. Psychomo tor retardatio n .35 .198. Social withdrawal .50 .32"*9. Insomnia .50 .33"*
10. H ypersomnia .54 .32"*II. Decreased appetite .47 .39"*12. Increased appetite .51 .52"*Summary score (items 1-12) Lr = .43*
Ideat ion al categoty13. Depressed mood .35 .45"*14. Nega tive self-image .28 .3 1"*I S. G uilt .30 .3Y*16. Hopelessness .40 .4 1"*17. Feeling " unloved" .23 .2518. Self-p ity .23 .29*19. Anhedo nia .38 .33"*20 . Suicidal ideation .37 .29*Summary score (items 13-20) Lr = .49*
Total: Behavioral and IdeationalSum mary score (items 1-20) Lr201T = .49*
Note: N = 93 mot her-child pairs. K-SADS -P = Sche dule forAffective Disorders and Schizo phrenia for Schoo l-Age C hild renPresent Episode.
, p adjusted with Bonferron i correction to control for fami lywise erro r.
* p :5 .01.
natively, maternal BDI scores of <13 were indicative of " nondepressed" mothers.
RESULTS
Sociodemographic Variables
Multivariate analyses of variance were computed toassess any systematic effects of child's age, gender, andSES in either child 's or parent's ratings on the Total(20 K-SADS-P items), Behavioral, and Ideational categories. Gender was dichotomized into male and female;dichotomization of age was based on median split(median age = 9.57 years); and that of SES was basedon median split (medial SES = 35). There were nosignificant effects.
To investigate possible differences due to erhnici ry,the sample was dichotomized into "white" (n = 75)
1278 J. AM, ACAD , CH IL D ADO LESC . PSYC H IATRY. 33 :9. NOV EM BER ID ECEMB ER 19 94
and "nonwhite" (n = 18), the latter consisting of black(n = 15), Hispanic (n = 1), and mixed races (n = 2).A 2 X 3 analysis of variance comparing "white" and"nonwhite" on the Behavioral, Ideational, and Totalcategories did not show any significant effect of ethnicity in child's ratings; in parent's ratings, there wasa significant effect of erhniciry on the Behavioral category (p= .03) due to one item, aches and pains (p= .01) ,and a significant effect of erhniciry on the Ideationalcategory (p = .008), due to two items, depressed mood(p = .03) and negative self-image (p = .003).
In summary, our analyses indicated that sociodernographic factors, in general, did not have significanteffects on child's and parent's ratings.
P-C Correlations on Ratings of Symptoms
Table 1 shows that overall P-C agreement on severityof symptoms was moderate and highly significant asreflected by "Total summary score" (i.e., the sum ofthe 20 individual K-SADS items): L r OlT = .49, P =.0001. Of the 20 items, 14 (70%) showed significantcorrelations after the probability values were adjustedwith the Bonferroni correction to control for familywise error; of those 14 correlations, 1 was high (increased appetite: r = .52); 5 were moderate, i.e., .40:s: r < .50 (depressed appearance, aches/pains, psychomotor agitation, depressed mood , and hopelessness),and 8 were low-moderate, i.e., .30 :s: r < .40 (decreasedconcentration, social withdrawal, insomnia, hypersomnia, decreased appetite, negative self-image, guilt,and anhedonia).
Although somewhat weaker than r findings, K
findings in Table 1 indicated an overall acceptable P-Cagreement on presence or absence of symptoms: of the20 items, 10 items or 50% had K values in the moderaterange (.40 :s: K < .60). However , as a striking exceptionto this trend for mostly moderate K and r values,irritability showed low K and rvalues (K = .20; r= .23).
In general, our first hypothesis , that overall P-Cagreement would be significant and moderate,was supported.
In Table 1, K and r statistics are also reportedaccording to nature ofsymptoms, i.e., Behavioral versusIdeational. P-C agreement regarding severity of symptoms in the Behavioral category was significant andmoderate: the Behavioral summary score was moderateand highly significant (Lr = .43, P = .0001); 9 of the
PARE NT- CHILD AGREEMEN T
12 behavioral items (or 75%) had r values in the lowmoderate (.30 :s: r < .40) to high (r ~ . 50) range,with adjusted p values s .01. The Ideational categoryshowed similar results to the Behavioral category: theIdeational summary score was moderate and highlysignificant: (Lr = .49, P = .0001); 5 of the 8 ideationalitems (or 62.5%) had r values in the low-moderate tomoderate range, with adjusted p values s .01. Thus,our second hypothesis (that there was a significantlygreater P-C agreement in Behavioral than Ideationalcategory) was not supported by the r statistics. Incontrast to the r statistics, K statistics showed that P-Cagreement regarding presence or absence of symptomswas significantly greater in the Behavioral than Ideational category: 9 of 12 (or 75%) behavioral itemsas opposed to 1 of 8 (or 12.5%) ideational items hada K ~ .40; p < .02 (Fisher's Exact Test, two-tailed).
In summary, our second hypothesis was only partiallysupported: a pattern of significantly greater P-Cagreement on the Behavioral than Ideational categorywas supported for presence or absence of symptoms,but not for severity of symptoms.
Comparison of P-C agreement between "depressed"mothers (n = 22) and "nondepressed" mothers (n =64) is shown in Table 2. There was no significantdifference in P-C agreement between the two groupsof mothers for Toral summary score of 20 items;"depressed" mothers: L l(d)201T = .54; "nondepressed"
mothers: Ll(nd/OIT = .42; z = .59, P = .55. Similarly,
there were no significant between-group differences oneither the Behavioral summary score or the Ideationalsummary score: Behavioral summary score for "depressed" mothers, L/(d) = .41, and of "nondepressed"mothers, Ll(nd) = .40; z = .05, P = .96; Ideationalsummary score of "depressed" mothers, Ll(d) = .43,and of "nondepressed" mothers, L/(nd) = .44; z = .05 ,P = .96. In addition, none of the 20 items showedsignificant between-group differences, with only onestriking exception, suicidal ideation. With respect tosuicidal ideation, "depressed" mothers unexpectedlyshowed a significantly higher P-C correlation than"nondepressed" mothers: "depressed" mothers, r= .69;" nondepressed" mothers, r = .009; z = 3.23 , P = .001.
Thus, our third hypothesis , that compared to nondepressed mothers, depressed mothers would have a significantly lower P-C agreement, was not supported.
J. AM . ACAD. CHILD ADOLES C. PSYCHIATRY. 33:9 NOVEMBERIDECEMBER 1994 1279
N GUY EN ET AL.
TABLE 2" Depressed" Mo thers (n = 22) versus " Nondepressed" Mothers (n = 64): Parent-Child Pearson Correlation (r)
Derived from Mother-Child K-SADS-P Scores
Behavioral categoryI. Depressed appearance2. Irritabili ty3. Aches/pains4. Fatigue5. Decreased concentration6. Psychomotor agitation7. Psychomotor retardation8. Social withdrawal9. Insomnia
10. Hypersomnia11. Decreased appet ite12. Increased appetiteSummary score (items 1-12)
Ideational category13. Depressed mood14. Negative self-image15. Guilt16. Hop elessness17. Feeling " unloved"18. Self-pity19. Anhedonia20. Suicidal ideationSummary score (items 13-20)
Total: Behavioral and IdeationalSummary score (items 1-20)
" Depressed"Mothers: 'id)
.58*.11.65"'*.39.17.48.27.25.27.09.44.54*
L,'idl = .41
.20
.07
.19
.13
.08
.48
.30
.694••
L,'id) = .43
"Nondepressed"Mothers: 'ind)
.41a.*
.23
.374.*
.14
.424 .e
.38"'*
.18
.31.36*.42"'*.30.47"'*
L,'ind) = .40·
.444.*
.40a.*.36·.45"'*.31.23.29.009
L,'indl = .44·
Comparison'id) vs. 'ind): Z
z = .05; NS
z = 3.23· *z = .05; NS
z = .59; NS
Note: K-SADS-P = Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present Episode: NS =
not significam.a p adjusted with Bonferroni correction to control for family-wise error.* p ~ .01; *. P < .00 1.
DISCUSSION
Our findings of an overall highly significant andmoderate P-C agreement on both presence/absenceand severity of symptoms suggest that our main goalof increasing P-C agreement was attained in this outpatient sample of depressed prepubertal children. Indeed,our findings compare favorably with those of similarstudies: Edelbrock et al. (1986), using the DiagnosticInterview Schedule for Children, found a negative P-Ccorrelation on the depressive symptom summary score(r = - .08) and a low overall P-C correlation (r = .10)for children aged 6 to 9 years; Weissman et al. (1980)found r = .04 on the Center for Epidemiologic StudiesDepression Scale; Kazdin et al. (1983a) found r = .10on the COl and r = .27 on the Bellevue Index ofDepression (BID) ; Kazdin et al. (1983b) found r =
.01 on the COl, r = .17 on the Depression Symptom
Checklist, and r = .32 on the BID; Kazdin et al.(1983c) found r = - .03 on the COl, r = .21 on theBID, and r = .07 on the Depression Symptom Checklist; Herjanic and Reich (1982) found that 73% of a168-item structured interview had J( values < .30; Ivensand Rhem (1988) found that 100% of the 9 K-SADSdepressive symptom clusters had J( values < .40; Angoldet al. (1987) found that of 21 K-SADS-EpidemiologicVersion depressiveitems, 17 items or 81% had J( values< .35; of these 17 items, 5 items or 30% had negativecorrelations ranging from J( = - .02 to - .07.
In spite of our improved P-C correlations, they werestill only modest overall. As described in the "Method"section, the confrontation process (either solo or joint)mayor may not help the informant process moreobjectively his or her response. The cognitive immaturity of the prepubertal children in this study may
1280 J. AM. ACAD . C H ILD ADO LESC . PSYCHIATRY, 33 :9, N OV EM BERIDECEMBER 19 94
explain, in part, the relatively modest overall improvement in P-C agreement.
The finding of low correlations (K = .20, r = .23)for irritability is consistent with the hypothesis ofdifferences in parents' and children's thresholds fortolerance of symptoms: Parents and children are "proneto report the symptoms which bother them the most"(Herjanic, 1984, p. 125). Thus, one can reasonablypresume that the child's irritability likely "bothered"the mother more than the child. A closer look at ourdata showed that, compared to their children, mothersreported significantly greater frequency and severity ofirritability in their children.
The pattern of greater P-C correlations in behavioralthan ideational types of symptoms found specificallyfor the K statistics (presence or absence of symptoms)and not for the r statistics (severity of symptoms) needsexplanation. Our approach (confronting the motherand/or child with intra- and/or interinformant discrepancies) may be more effective with the factual, observable behavioral (versus private, subjective ideational)type of symptoms, and the more straightforward typeof yes/no answer (versus complex answer involvingseverity grading), and thus be responsible for the specificpattern of behavioral versus ideational dissociation inP-C agreement on presence or absence of symptoms.
Our other finding of consistently significant andmoderate r values across the Behavioral and Ideationalcategories is congruent with that of Knight et al.(1988), who also found consistently significant andmoderate to high r values between children and parentsacross different depressive measure scales of the COLKnight and coworkers' explanation for their findingswas that patterns of P-C agreement might vaty withsample composition (i.e., subject's age and inpatientversus outpatient status). While they used a homogeneous sample of prepubertal children, many otherstudies used heterogeneous samples of prepubertal children mixed with adolescents, and even young adults(Angold et al., 1987; Herjanic and Reich, 1982; Stavrakaki et al., 1987). In addition, Knight et al. suggestedthat P-C agreement might also be affected by theseverity of the child's depression, which could be inturn reflected by the child's inpatient versus outpatientstatus. This might explain, they added, the low P-Cagreement found in inpatient samples (Kazdin et al.,1983a-c) as opposed to the higher P-C agreementfound in their outpatient sample. Thus, the similarity
PARENT-CHILD AGREEMENT
in sample composition (i.e., homogeneous outpatientsample of depressed prepubertal children) betweenKnight and coworkers' study and ours might accountfor the similar results in the two studies.
The absence of significant differences in P-C correlation between our "depressed" and "nondepressed"subgroups of mothers is not consistent with either thefindings of Moretti et al. (1985) or of Kashani et al.(1985), who reported adverse impact (overreportingand underreporting, respectively) ofparental depressionon the reporting of children's depressive symptoms,or those of Weissman et al. (1987), who documentedpositive impact of parental depression (i.e., improvement in P-C agreement on the children's psychopathology). Our findings were, on the other hand, congruentwith those ofIvens and Rehm (1988), who found thatmother-child discrepancies were not a function ofmaternal depression. The use of different methodologies to label "maternal depression" may contribute tothe above discrepancies in findings. Kashani et al.(1985) and Weissman et al. (1987) used mothers withformal diagnoses of unipolar or bipolar depression,while Moretti et al. (1985) and Ivens and Rehm (1988)used criteria similar to ours (e.g., maternal BDI scores)to define maternal depression.
Furthermore, the discrepant findings in impact ofmaternal depression on P-C agreement in the twoabove-cited studies that used the BDI to label maternal"depression" deserve additional explanation. Morettiet al. (1985) used similar self-report scales to define"depression" both in the child (COl and Children'sDepression Scale [Lang and Tisher, 1978]) and inthe mother (BDI) and found an impact of maternal"depression" on P-C agreement. On the contrary,the lack of impact of maternal "depression" on P-Cagreement found in Ivens and Rhem's study and inours might be attributed to the fact that both studiessimilarly used different assessment tools (K-SADS interview versus BDI self-report scale) to define differentconditions (depressive diagnoses versus current/statedepressive symptoms, respectively, in the child andthe mother).
Finally, in our study, "depressed" mothers showeda significantly higher P-C correlation than "nondepressed" mothers with respect to the child's suicidalideation. Clinical experience suggests that suicidal ideation in a young child may be especially painful for"average" parents to admit, and, therefore, is more
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 33:9 NOVEMBERIDECEMBER 1994 1281
NGUYEN ET AL.
likely to be dismissed or downplayed by them. Onthe other hand, by virtue of their own experientialinsight with suicidal ideation, "depressed" mothersmay be more attuned to their young children's suicidal ideation.
As our findings were achieved by the implementationof a modified assessment procedure, the pros and consof each component of our method will be addressedin the following discussion.
First, we will address the use of the same K-SADSinterviewer for the parent and child and the "unblind"nature of our method. All assessment findings (CDI,traditional clinical interview, K-SADS) from differentinformants (parent and child) and/or from the sameinformant were made available to the K-SADS interviewer (via consultations I and II with the seniordiagnostician) for use in confronting the child, themother, and/or both with intra- and interinformantdiscrepancies. The use of the same unblind interviewercould be criticized as being methodologically weakerthan that of two independent interviewers, blind to oneanother's findings, thus being responsible for possiblyyielding a more inflated level of P-C agreement. Whilethere is definite methodological advantage in usingblind and independent interviewers, this approach creates its own set of problems. As Angold et al, (1987)pointed out, "disparate sources of information eitherhave to be combined to produce a diagnosis or analyzedseparately, with the result in the latter case, that thesame subject may contribute to the 'sick' group forone set of analyses and the 'well' group for another"(p. 912). On the other hand, with the combinationof sources of information, one is confronted with thevery same problem of "operationalization of the processof data combination" (Angold et al., 1987, p. 913) asthat encountered in the unblind interviewing of parentand child.
Second, the method of confronting the child, themother, and/or both with interinformant discrepancieswas criticized by Angold et al. (1987) as "inappropriate"because "such an approach clearly involves placingpressure on one informant to agree with the other"(p. 912). Consequently, they argued, "it will probablyincrease agreement, but may not produce any more'truth'" (p. 912). While in agreement with the firstpart of their argument, we will take issue with thesecond by examining the four examples provided inthe "Method" section.
As the scoring paradigm allows for each informantto keep his or her stated final score, and as the clinicianhas no control over the inforrnantls)' "idiosyncratic"response to confrontation, the confrontation proceduremay increase P-C agreement as in examples 1 and 3a,or may not affect at all P-C agreement as in examples2 and 3b. However, diagnostic accuracy or "truth"was consistently improved in all four examples, becausethe clinician's summary scores-which formed the basisfor computing DSM-III-R diagnoses-were designedto reflect the most factual and reliable answers thatemerged from the confrontation process.
In conclusion, the above points provide reasonablesupport for the use of our modified assessment methodto decrease informant variance. The results are encouraging, providing that customary cautions be taken intheir generalization because of the earlier describedlimitation of our assessment method.
For the clinician, our findings make more salientthree important yet underrecognized problems: (1) Theclinician should pay attention not only to interinformant but also intrainformant discrepancies that wouldultimately affect diagnostic accuracy. (2) Bias factorsmay exert not only individual but also interactionaleffects on P-C agreement, such as our finding ofselective interaction between the behavioral versus ideational nature of symptoms with the dichotomousdimension ofsymptoms. (3) Another often underrecognized bias factor is parents' and children's differencesin thresholds for tolerance of symptoms.
The following principles of our modified assessmentmethod may be adapted by the clinician to increasediagnostic accuracy: (1) use of different assessmentinstruments (self-report scales, traditional clinical interview, standardized structures interviews) to maximizethe chance of uncovering not only overt but also covertintra- and/or interinforrnant discrepancies, and (2)adaptation of the "best estimate" clinical judgment tothe clinical context in which discrepancies arise, inorder to select the most appropriate confrontationstrategies and information-weighing techniques to clarify and solve these discrepancies.
REFERENCES
Achenbach TM, McConaughy S, Howell C (1987), Child/adolescentbehavioral and emorional problems: implicarions of cross informantcorrelarions for siruarional specificity. Psychol Buill 01:213-232
1282 ]. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 33:9, NOVEMBERIDECEMBER 1994
American Psychiarric Associarion (1980) . Diagnostic and Statistical ManualofMental Disorders. 3rd edition (DSM -l/I). Washingron . DC: AmericanPsychiarric Associarion
American Psychiatric Association (1987). Diagnostic and Statistical Ma nualojM ental Disorders,3rd edit ion-revised (DSM-Il/-R). Washington , DC :American Psychiarric Association
Angold A, Weissman MM , John K er al. (1987), Parent and child reporrsof depressive symptoms in children at low and high risk of depression.J Child Psycbol Psychiatry 28:901-915
Beardsley WR. Klerman GL. Keller MB. Lavori PW. Podorefsky DL(1985). But are they cases? Validiry of DSM -l/I major diagnoses inchildren identified in a family srudy. Am J Psychiatry 142:687-69 1
Beck AT (1967) , D epression: Clinical, Experimental and Theoretical Aspects.New York: Hoeber (repub lished as Depression: Causes and Treatment.Philadelphia: Universiry of Pennsylvania Press, 1972)
Beck AT, Sreer RA. Sarbin MG (1988). Psychometric properti es of theBeck Depression Inventory: rwenty-five years of evaluation. ClinicalPsychology Review 8:77-100
Chambers WJ. Puig-Anrich J, H irsch er al. (1985). The assessment ofaffective disorders in children and adolescents by semi-structu red interview: the test-retest reliabiliry of the Schedule for Affective Disordersand Schizoph renia for School-Age Children-P resent Episode. Arch GenPsychiatry 42 :696-702
Costello AJ. Edelbrock LS, Dulcan MK. Kalas R, Klaric SH (1984), Reporton the NIMH Diagnostic Interview Schedule fOr Children (DIS-C).Washington . DC: Natio nal Institute of Mental Health
Edelbrock C. Costello AJ. Dulcan MK. Conover NC . Kalas R (1986).Parent-child agreement on child psychiarric symptom s via structuredinterview. J Child Psychol Psychiatry 27 :18 1-190
Gershon ES, McKnew 0 , Cyrryn L et al.(l985), Diagnoses in school-agechildren of bipolar affective disorder patients and normal cont rols. JAffict Disord 8:283- 29 1
Herjanic B (1984), Systematic diagnostic interviewing of children: presentstate and future possibilities. Psychiatric Developments 2:115-130
Herjanic B, Reich W (1982). Development of a structured psychiatricinterview for children. J Abnorm Child PsychollO:307-324
Hodges K. Gordon Y. Lennon MP (1990). Parent-child agreement onsymptoms assessed via a clinical research interview for children: theChild Assessment Schedule (CAS).J ChildPsycbol Psychiatry 31:427-436
Hodges K. McKnew D. Burback OJ. Roebuck L (1987) . Diagnosticconcordance between the Chi ld Assessment Schedule (CAS) and theSchedule for Affective Disorders and Schizophrenia for School-AgeChi ldren (K-SADS) in an outpa tient sample using lay interviewers. JAm A cad Child Adolesc Psychiatry 26:654-66 1
Hollingshead AB (1975), Four Factor Index of Social Status. New Haven.CT: Yale Universiry Depart ment of Sociology
Ivens C. Rehm L (1988). Assessment of childhood depression: correspondence between reports by child. mothe r and father. JAm Acad ChildAdolesc Psychiatry 27:738--741
PA RENT- CHILD AG REEMENT
Kashani JH . Orvaschel H. Burk JP. Reid JC (1985). Informant variance:the issue of parent-child disagreement. J Am Acad Child Psychiatry24:437-4 41
Kazdin AE, Esveldt-Dawson K, Unis AC, Rancurello MD (l983a), Childand parent evaluations of depression and aggression in psychiarricinpatient children. J Abnorm Child Psycho! I I :40 I-4 I4
Kazdin AE. French NH , Unis AS (l 983b), Child. mother and fatherevaluations of depression in psychiatric inpatient children. J AbnormChild Psychol 11:167-180
Kazdin AE. French NH . Unis AS. Esvelot-Dawson, K (l983c). Assessmentof childhood depression: correspon dence of child and parent ratings.JAm Acad Child Psychiatry 22 :157- 164
Knight 0 , Hensley YR. Warers B (1988), Validation of the Children'sDepression Inventory in a prepubertal sample. J Child Psycho! Psychiatry 29:853-863
KovacsM (198 I). Rating scales to assessdepression in school-aged children.Acta Paedopsychiatr 46:305-3 15
Lang M, T isher M (1978), Children's Depression Scale. Victoria, Australia:Australian Council for Educational Research
Leckman JF. Scholomskas D. Thompson WD, Belanger A, WeissmanMM (1982), Best estimate of life time psychiatric discrepancies: amethodological study. Arch Gen Psychiatry 39 :879-883
Mokros HB. Poznanski E. Grossman JA. Freeman LN (1987) . A comparisonof child and parent rarings of depression for normal and clinicallyreferred children . J Child Psycho! Psychiatry 28 :613-627
Moretr i MM , Fine S. Haley G, Marr iage K (1985). Childhood andadolescent depression: child-report versus parent-report information. JAm Acad Child Psychiatry 24 :298-302
Reich W, Herjanic B. Welner Z. Gandhy PR (1982) . Development of astructu red psychiatric interview for children : agreement on depressioncomparing chil d and parenr inte rviews . J Abnorm Child Psycho]10:325- 336
Stavrakaki C. Vargo B, Roberts N, Boodoosingh L (1987). Con cordanceamong sources of information for ratings of anxiery and depression inchildren. JAm Acad Child Adolesc Psychiatry 26:733-737
Sylvester CEo Hyde TS, Reichler R (1987). The diagnostic interview forchildren and personaliry inventory for children in stud ies of childrenat risk for anxiery disorders and depression. JAm Acad Child AdolescPsychiatry 26:668-675
Verhulst FC. Berden GF, Sanders-Wou dstra JAR (1985) . Mental healthin Dutch children: II. T he prevalence of psychiarric disorder andrelationship between measures. Acta Psychiatr Scand 72( supp1324):1-45
Weissman MM. Leckman JF, Merikangas KR. Gammon GO. PrusofTBA(1984), Depression and anxiery disorders in parents and children: resultsfrom the Yale family study. Arch Gen Psychiatry 4 1:845-852
Weissman MM. Orvaschel H , Padian N (1980), Children's symptomsand social functioning self-report scales: compa rison of moth ers' andchildren's reports. J N ero Me nt Dis 168:736 -740
Weissman MM . Wickramararne P. Warne r V (1987), Assessingpsychiarricdisorders in children; discrepancies between mothers' and children'sreports. Arch Gen Psychiatry 44 :747-753
] . AM. ACAD. C H I LD AD O LESC. PSYC H IATRY , 33:9 NO VEM BERID ECEMB E R 1994 1283