Does Psychiatric History Bias Mothers' Reports? An Application of a New Analytic Approach

Does Psychiatric History Bias Mothers' Reports? An Application of a New Analytic Approach HOWARD D. CHILCOAT, Sc.D., AND NAOMI BRESLAU, PH.D.

ABSTRACT ObJective: To evaluate whether mothers' psychiatric history biases reports of their children's behavior problems, mothers' and teachers' reports of children's behavior problems were compared using a recently developed statistical approach. Method: Child Behavior Checklists and Teacher's Report Forms were completed by mothers and teachers, respectively, about 801 six-year-old children. Mother's history of major depression, anxiety disorders, and substance use disorder was assessed by using the National Institute of Mental Health Diagnostic Interview Schedule. Generalized estimating equations were used for data analysis. Results: According to both teachers and mothers, matemal history of major depression was associated with more intemalizing problems: the association was significantly stronger when mothers were the informants. Mothers with history of any psychiatric disorder reported more externalizing problems in their children than expected, whereas teachers' reports of externalizing behaviors were unrelated to maternal psychiatric history. These findings could not be explained by variations in children's behaviors across settings. Conclusion: The generalized estimating equation models enabled simultaneous examination of whether children of depressed mothers have excess behavior problems and whether depressed mothers overreport behavior problems in their children. The results indicate that children of depressed mothers have more internalizing problems. In addition, depressed mothers overstate and overgeneralize their offspring's behavior problems. This study broadens the concerns with reporting bias beyond maternal depression to include other psychiatric problems. The results emphasize the potential for bias in family history studies that rely on informants. J. Am. Acad. Child Ado/esc. Psychiatry. 1997,36(7):971-979. Key Words: informants, validity, child behavior problems, generalized estimating equations.

Acupud Drumbn- 24. 1996. Drs. Chikoat and Bmlau arr with thr Drpartmrnt ofPsychiatry. Hrnry Ford Hralth Scirnm Crntrr, Drtroit. Support for this rrsrarch was provitkd by NIMH grant MH44586. Thr authors acltnowkdgr hrlpful suggrstionsftom Jim Anthony. John Richtm. and Nick Ialongo in thr prrparation ofthis manuscript. Rrprint rrqursts to Dr. Chilcoat. Drpartmrnt of Psychiatry. Hrnry Ford Hralth Scirnm Crnur. 1 Ford Piau. 3A. Drtroit. MI48202. 0890-8567/97/3607-09711$0.300/0© 1997 by the American Academy of Child and Adolescent Psychiatry.

and Moos, 1985; Conrad and Hammen, 1989; Ivens and Rehm, 1988; Richters and Pelligrini, 1989; Weissman et aI., 1987) have suggested that the excess in children's behavior problems reported by depressed versus nondepressed mothers reflects the effect of maternal depression on offspring (Richters, 1992). Maternal depression could influence child behavior problems through both genetic and environmental mechanisms. In addition, there might be other factors (e.g., poverty, marital discord) that influence both maternal depression and child problems, so that the observed relationship between them actually exists, even if it is not a causal one (Fergusson et aI., 1993). Furthermore, psychiatric distress in mothers might lower their threshold of tolerance for child misbehavior and might alter mother-ehild interactions, causing behavior problems that would not have existed otherwise (Richters, 1992). On the other hand, there are theoretical grounds for arguing that depressed mothers might overreport their children's behavior problems, because of the impairing

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY. 36:7. JULY 1997

971

Several empirical investigations have suggested that depressed mothers overreport problems in their children (Breslau et al., 1988; Estroff et al., 1984; Fergusson et al., 1985, 1993; Fergusson and Horwood, 1987; Forehand et aI., 1986; Friedlander et aI., 1986; Graham and Stevenson, 1985; Griest et al., 1979, 1980; Jensen et al., 1988; Moretti et aI., 1985; Panaccione and Wahler, 1986; Rickard et aI., 1981; Schaughency and Lahey, 1985; Webster-Stratton, 1988). Other studies (Billings

CHILCOAT AND BRESLAU

effect of depression on cognItIOn and perception (Clifford and Hemsley, 1987; Gotlib et aI., 1988; Kowalik and Gotlib, 1987; McCabe and Gotlib, 1995). In this study, we have adapted a regression strategy, put forth recently by Fitzmaurice et al. (1995), to examine the question of bias, using mothers' and teachers' reports. The method permits us to test simultaneously the hypotheses that (1) children of depressed mothers have excess behavior problems and that (2) depressed mothers overreport behavior problems in their children. Although previous research posed the question of bias versus "real" effect as contending alternatives, the hypotheses are mutually compatible. The first hypothesis is tested using teachers' and mothers' reports in combination; the second, by comparisons between mothers' and teachers' reports, that is, by testing interactions. Evidence that teachers report more behavior problems in children of depressed than of nondepressed mothers would support an effect of maternal depression on behavior problems in offspring, if informants who are relatively free of bias due to maternal depression are used. Evidence that based on mothers' ratings, maternal depression is associated with a significant increment in children's problems, above the level observed in teachers' data, would suggest that depressed mothers overreport child behavior problems. The alternative explanation that children of depressed mothers display behavior problems at home but not at school cannot be addressed directly in the model. However, the plausibility of that explanation would be considerably dampened if there was no support in the data that children in general display more behavior problems at home than in school. Previous studies have focused on maternal depression. Although a range of psychiatric disorders in parents is associated with psychopathology in young offspring (Loeber et aI., 1995; Lynskey et al., 1994; Rice et aI., 1987; Rosenbaum et aI., 1992; Vanyukov et aI., 1993), the extent to which these associations might be biased, when based on mothers' reports, is unknown. Furthermore, because of the comorbidity of depression with anxiety or other disorders (Eaton and Chilcoat, 1992; Regier et aI., 1990; Robins et aI., 1991), a more critical investigation of the question must consider the potential influence of comorbid disorders on depressed mothers' reports. In this study, we examine also mothers' history of anxiety and substance use disorders.

METHOD

972

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY. 36:7. JULY 1997

Sample Data come from an ongoing study of the neuropsychiatric consequences oflow birth weight in children (Breslau et aI., 1994). The methods of this study have been described in detail elsewhere (Breslau et aI., 1994, 1996; Chilcoat et aI., 1996). In brief, the study sample consisted of 823 children who were participants in a study of the effects of birth weight on neuropsychiatric functioning. The children were selected from rhe 1983 to 1985 lists of newborn discharges from two major hospitals in southeast Michigan, one in the city of Detroit and the other in a nearby suburban middle-class community. A stratified sampling procedure was used, in which random samples of 140 low birth weight (:52.500 g) and 93 normal birth weight newborns were drawn from each hospital for each year from 1983 to 1985. Children with severe neurological impairment were excluded. The target sample was reduced also by those who had died, who were placed in foster care, or whose families had moved out of the metropolitan area. This resulted in a target sample of 618 low birth weight and 477 normal birth weight children eligible for participation in the study. A total of 473 low birth weight and 350 notmal birth weight children (75% of those eligible) completed a 3to 4-hour session of neuropsychological testing between 1990 and 1992, when the childten were 6 or 7 years old. Characteristics of the entire sample, as well as by level of variables used in the sampling design, are presented in Table 1. The assessment included a neuropsychological battery, measures of children's behavior ptoblems, and history of psychiatric disorders in the mothers.

Measures Mothers and teachers rated children's behavior problems, using the Child Behavior Checklist (CBCL) and the parallel Teacher's Report Form (TRF), respectively. The TRF was mailed to each child's teacher; 801 TRFs (97%) were returned. All but two mothers completed the CBCL (99%). In this repott, we have focused on the Internalizing and Externalizing "broad-band" scales of the CBCL and TRF (Achenbach, 1991). The Internalizing scale is a sum of items from three subscales: Withdrawn, Somatic Complaints, and Anxious/Depressed. The Externalizing scale is a sum of three subscales: Hyperactive, Delinquent, and Aggressive. Raw scores were converted to T scores, based on age/sex-defined percentiles from normative samples. Only those scale items that were common to both the CBCL and TRF were used. The National Institute of Mental Health Diagnostic Interview Schedule (DIS) was used to assess DSM-III-R psychiatric disorders in the mothers. Data were gathered in personal interviews with mothers by trained interviewers, who were blind to the birth weight status of the child. The interviewers who conducted the interview with the mother were different from those who examined the children, to preclude contamination. Lifetime history of major depression, anxiety disorders, and abuse of or dependence on alcohol or illicit psychoactive drugs was determined by computer algorithms. History of anxiety disorder was defined as the presence of one or more of the following disorders in lifetime: simple phobia, social phobia, agoraphobia, generalized anxiety disorder. posttraumatic stress disorder, panic disorder, and obsessive-compulsive disorder. The DIS was completed by 801 (97%) mothers. There were 140 (17%) mothers with lifetime history of major depression, 291 (36%) with anxiety disorder, and 105 (13%) with substance abuse or dependence (ND). Ninety mothers (11 %) had a history of major depression as well as anxiery disorder, 50 (6%) had a history of major

MOTHERS AS INFORMANTS

TABLE 1 Sample Characteristics Detroit Utban

Suburban

LBW (n = 238)

NBW (n = 176)

LBW (n = 235)

Black (%)

80.2

73.9

11.5

3.0

42.9

Mother's education (%)
29.5 23.5 37.0 9.7

21.6 26.1 38.6 13.6

7.2 30.2 39.2 23.4

7.5 30.5 33.9 28.2

16.8 27.5 37.3 18.4

31.8 (6.4)

31.7 (5.5)

34.3 (5.0)

34.7 (4.8)

33.1 (5.6)

37.8

32.9

14.0

8.6

23.8

Characteristic

Maternal age: mean (SD) Single mother (%) Note: LBW

(n

NBW = 174)

(n

Total = 823)

= low birth weight; NBW = normal birth weight.

depression without anxiety disorder, and 201 (25%) mothers had anxiety without major depression.

Statistical Analysis We used a regression analytic strategy using generalized estimating equations (GEE) (Diggle et aI., 1994; Liang and Zeger, 1986; Zeger and Liang, 1986) to estimate the effect of maternal depression on children's behavior problems, using mothers and teachers as raters. The use of this approach was motivated by a recent report, which demonstrated a regression method for combining data on children's disorders from multiple informants (Fitzmaurice et aI., 1995). The GEE approach has several distinct advantages over methods previously applied to this question: (I) it is generalized, enabling the analysis of a variety of types of outcomes, including continuous (gaussian), count, and binary data; (2) responses from multiple informants can be examined in the same model, allowing simultaneous esrimation of the association between risk factors of interest and outcomes according to each informant; and (3) differences in the magnitude of associations between risk factors of interest (e.g., maternal depression) and informants (i.e., mothers versus teachers) can be tested through the use of interaction terms (Fitzmaurice et aI., 1995). It is the second and third features together that permit us to estimate the extent to which depressed mothers might overreport behavior problems in their offspring. In this analysis, two data records were included for each child: one based on the mother's report and one based on the teacher's. Each record contained child behavior ratings for the specified rater, an indicator for whether the rater was the mother or the teacher, and information on other risk factors, namely mother's history of specific psychiatric disorder, birth weight status, and demographic characteristics, which was identical in each of the child's two data records. The following basic model was used to test our hypotheses (for the purpose of a clearer illustration, terms for birth weight status and demographic characteristics included in the models are not displayed):

and 0 if the informant is the teacher; MOMDEP = 1 if the mother has a lifetime history of major depression and 0 if not; ex, the intercept in the model, is the expected T score for children with no maternal history of depression, as rated by teachers; III is the difference in T scores between mothers and reachers for children whose mothers are not depressed; 112 is the difference in teachers' T scores for children whose mothers are depressed compared with those whose mothers are not depressed. If 112 l' 0, then there is evidence that maternal depression is associated with children's behavior problems, based on teachers' reports. (The meaning of III and 112 is in part defined by the inclusion of the interaction term in the modeL) IlJ is the interaction coefficient, which can be interpreted as the difference in the magnitude of the association of maternal depression with children's behavior problems when mothers are the informants compared with that obtained using teachers' ratings (1l2)' If 13\ > 0, the magnitude of the association of maternal depression with child behavior problems as reported by mothers is greater than that reported by teachers. GEE estimates regression coefficients and their standard errors, taking the correlation between responses into account. Consequently, the potential bias due to the fact that each child had two records is reduced and estimates are more efficient (in terms of standard errors) (Diggle et aI., 1994; Liang and Zeger, 1986; Zeger and Liang, 1986). We examined two child outcomes: internalizing and externalizing problems. For each outcome, two models were constructed. The first model included indicators for mother's lifetime history of major depression and maternal depression-informant interaction. The second model expands the inquiry of maternal psychiatric disorder by adding terms to represent anxiety and substance use disorder. Each of the models also included terms for urban versus suburban site of birth, birth weight status, and sex of child.

RESULTS

where child's outcome (Y) is measured by the T score from the CBCL or TRF; INFORMANT = 1 if the informant is the mother

We present first descriptive data on children's behavior problems and statistical tests of group differences. Mean mother (CBCL) and teacher (TRF) T scores for child internalizing and externalizing problems by mother's lifetime psychiatric disorder are presented in

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY. 36:7, JULY 1997

973

+ III (INFORMANT) + 112 (MOMDEP) + IlJ (INFORMANT X MOMDEP)

Y = ex

CHILCOAT AND BRESLAU

Table 2. Mothers with a history of a psychiatric disorder reported higher levels of internalizing and externalizing problems than mothers without each specified disorder (p < .0001 for all mothers' ratings). For teachers' ratings, differences in mean T scores by mother's psychiatric disorder were considerably smaller and reached statistical significance (p < .05) only for differences in internalizing behavior by history of major depression. Results from the basic GEE analysis for internalizing problems, adjusted for site of birth, birth weight status, and sex of child, indicate that teachers rate children of depressed mothers as having more internalizing problems than children of nondepressed mothers (difference = 3.47,95% confidence interval [CI] = 1.39 to 5.55). The interaction of informant by maternal depression was statistically significant (p = .020), and the difference in internalizing problems between children of depressed and nondepressed mothers was nearly twice as large when mothers were the raters (difference = 6.54, 95% CI = 4.73 to 8.35) than when teachers were

raters. In summary, when the mother of the target child was depressed, the teacher reported a higher level of internalizing problems in the child than would otherwise be predicted. According to the depressed mother, the excess in child's internalizing problems was even greater. In the next phase of the analysis of children's internalizing problems (Table 3), we examined mothers' history of major depression, taking into account comorbidity with anxiety disorder, as well as with maternal substance NO. Based on teachers' ratings, maternal depression signaled significantly more internalizing problems in the children when it occurred in combination with a history of anxiety disorder (difference = 4.23,95% CI = 1.56 to 6.90). The informant by maternal disorder (depression plus anxiety) interaction indicated that this difference was significantly larger when mothers were informants (p = .006), resulting in a 9-point difference in internalizing problems for children of depressed and anxious mothers, compared with no maternal history of either disorder (difference = 8.98, 95% CI = 6.77 to

TABLE 2 Comparison of Mean CBCL (Mother) and TRF (Teacher) TScores for Internalizing and Externalizing Behaviors by Mother's Lifetime History of Specified Psychiatric Disorders Mother's Psychiatric History: Major Depression Present

Not Present Mean

SO

tTest

p Value

54.77 10.09 54.08 10.70

48.17 48.32

9.32 9.93

7.50 6.15

<.0001 <.0001

52.43 11.73 52.43 9.58

48.92 50.82

9.63 9.40

3.26 1.81

.0013 .0711

Rater: Scale

Mean

Mother: Internalizing Mother: Externalizing Teacher: Internalizing Teacher: Externalizing

SO

Anxiery Disorder Present

Not Present Mean

SO

t Test

p Value

52.87 9.66 52.81 10.35

47.31 47.34

9.28 9.73

8.04 7.46

<.0001 <.0001

50.38 10.35 51.61 9.58

49.04 50.81

9.94 9.36

1.77 1.14

.0766 .2567

SO

t Test

p Value

Rater: Scale

Mean

Mother: Internalizing Mother: Externalizing Teacher: Internalizing Teacher: Externalizing

SD

Substance Abuse or Dependence Present Mean

Mother: Internalizing Mother: Externalizing

54.82 9.33 54.53 10.10

48.50 9.56 48.54 10.10

6.32 5.66

<.0001 <.0001

Teacher: Internalizing Teacher: Externalizing

49.96 11.05 52.96 11.08

49.46 50.82

0.46 1.87

.6425 .0636

Nou: CBCL

974

Not Present Mean

Rater: Scale

SO

9.96 9.14

= Child Behavior Checklist; TRF = Teacher's Report Form. J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY. 36:7. JULY 1997

MOTHERS AS INFORMANTS

11.19). According to both teachers and mothers, children whose mothers were depressed but not anxious had slightly higher levels of internalizing problems than children with neither maternal disorder. The magnitude of this difference was the similar regardless of informant (p = .81), although this estimate was more precise for mothers' reports, as indicated by the narrower confidence interval (Table 3). Although teachers did not rate children of anxious, nondepressed mothers as having more internalizing problems than children with no maternal disorder (difference = 0.37, 95% CI = -1.23 to 1.98), there was a difference when mothers were informants (difference = 3.62, 95% CI = 2.09 to 5.14). In summary, the teacher observed more internalizing problems than expected when the mother was both depressed and anxious. These problems were even greater according to the depressed and anxious mother. Mothers who were depressed but not anxious concurred with teachers that their children had a relatively small increase in internalizing problems. However, anxious mothers without history of major depression rated their children as having higher than expected levels of internalizing problems, whereas teachers did not. Teachers' reports indicate that there was no increase in internalizing problems when comparing offspring of mothers with a history of substance AID with those with no maternal substance ND. In contrast, a considerable increase was found when mothers were informants (difference = 4.26, 95% CI = 2.28 to 6.22). We carried out a parallel GEE analysis in which children's externalizing problems was the outcome. In the basic GEE model, which compared level of externaliz-

ing problems for children whose mothers were depressed with that of those with no maternal history of depression, we found only a slight and nonsignificant difference in teachers' ratings of externalizing problems comparing children of depressed and nondepressed mothers (difference = 1.45, 95% CI = - 0.27 to 3.17). In contrast, the magnitude of this difference was significantly greater when mothers were the informants, as indicated by the coefficient for the informant-maternal depression interaction in the GEE model (interaction, p < .001). On average, depressed mothers rated their children 5.51 points higher (95% CI = 3.64 to 7.38) on externalizing problems than mothers with no history of depression. Thus, although teachers did not link children's externalizing problems to maternal depression, depressed mothers attributed more externalizing problems to their children than did nondepressed mothers. Table 4 presents the GEE results for externalizing problems for maternal history of major depression stratified by comorbidity with anxiety disorder. History of maternal substance ND is included as well. As viewed by teachers, children's level of externalizing problems was not significantly related to mothers' history of psychiatric disorder, as can be seen from the confidence intervals, which include the null value. In contrast, mothers with history of major depression and anxiety, as well as those with anxiety alone, reported more externalizing problems in their children, compared with mothers with neither disorder, but those with depression alone did not. Teachers rated children of substance-abusing mothers as having similar levels of externalizing problems as

TABLE 3 Expected Difference in Children's Internalizing Problems by Presence of Maternal Disorder, Based on Teachers' and Mothers' Reports Expected Difference (95% Confidence Interval) Maternal Disorder MDD and anxiery MDDonly Anxiery only Substance ND

Mother

Teacher 4.23 (1.56-6.90) 2.62 (-0.69-5.93) 0.37 (- 1.23-1.98) -0.43 (-2.66-1.80)

8.98 3. I 1 3.62 4.26

(6.77-11.19) (0.39-5.83) (2.09-5.14) (2.28-6.22)

Comparison of Differences Based on Teachers' vs. Mothers' Reports (p Value)· .006 .81 .003 .002

Note: Results from generalized estimating equation models for maternal history of major depression, anxiery disorder, and substance use disorder (coefficients adjusted for child's sex, birth weight, and site of birth). MDD = major depressive disorder; ND = abuse or dependence. ·Statistical test comparing differences based on teachers' versus mothers' reports, as indicated by informant by maternal disorder interaction term in generalized estimating equation models.

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 36:7, JUl.Y 1997

975

CHILCOAT AND BRESLAU

TABLE 4 Expected Difference in Children's Externalizing Problems by Presence of Maternal Disorder, Based on Teachers' and Mothers' Reports

Mother

Comparison of Differences Based on Teachers' vs. Mothers' Reports (p Value)"

7.88 ( 5.04-10.72) 2.19 (-0.45-4.83) 3.48 (1.83-5.13) 4.12 (2.03-6.69)

<.001 .82 <.001 .08

Expected Difference (95% Confidence Interval) Teacher

Maternal Disorder MOD and anxiety MOD only Anxiety only Substance NO

0.90 1.79 0.21 1.82

(-1.33-3.13) (-0.84-4.42) (-1.38-1.80) (-0.41-4.05)

Note: Results from generalized estimating equation models for maternal history of major depression, anxiety disorder, and substance use disorder (coefficients adjusted for child's sex, birth weight, and site of birth). MOD = major depressive disorder; AID = abuse or dependence. • Statistical test comparing differences based on teachers' versus mothers' reports, as indicated by informant by maternal disorder interaction term in generalized estimating equation models.

Our findings provide general support for both hypotheses. They suggest significant effects of maternal depression on children's behavior and significant overreporting by depressed mothers. First, evidence of a "real" effect of maternal depression was supported by our finding that teachers reported more internalizing

problems in children of depressed versus nondepressed mothers. Second, evidence for distortion was provided in results indicating that the excess in children's internalizing and externalizing problems was significantly larger when we compared ratings from depressed versus nondepressed mothers. Teachers are unlikely to be aware of mothers' history of depression and should provide reports of a child's behavior problem that are relatively unbiased with respect to mothers' depression. Therefore, the finding that maternal depression signaled more internalizing problems when teachers were informants suggests a "real" effect of maternal depression on internalizing problems. When comorbidity of depression with anxiety disorders was taken into account, teachers reported a difference only for children of depressed and anxious mothers compared with children with neither maternal disorder. No significant differences were found for maternal depression or anxiety disorder alone, based on teachers' reports. Regardless of maternal disorder, teachers reported no significant differences in externalizing problems. The findings in this report also support the hypothesis that depressed mothers overreport behavior problems in their offspring. Relative to teachers, depressed/anxious mothers appeared to overstate their children's internalizing problems and overgeneralized children's problems by reponing an excess in externalizing problems. In addition, there was evidence that mothers with a history of substance NO might overstate their children's behavior problems, especially internalizing problems, even when analyses were adjusted for maternal depression and anxiety. Although these findings support

976

J. AM. ACAD. CHIl.D ADOLESC. PSYCHIATRY. 36:7. JULY 1997

those with no maternal substance A/O (adjusted for maternal history of major depression or anxiety disorders, difference = 1.82, 95% CI = -0.41 to 4.05). There was an increase in this difference that was of marginal statistical significance when mothers were informants (interaction, p = .08). Of the 105 mothers with a history of substance NO, 68% had either an anxiety disorder or major depression, leaving only a very small subset with "pure" substance use disorder. Further analysis that did not adjust for comorbid disorders suggested that externalizing problems were higher in children of substance-abusing mothers, regardless of informant (not shown). We found no support for the notion that children are more likely to display behavior problems at home than in school. On the contrary, in each of the GEE models, the coefficient for informant, which represents the difference between mother and teacher repons in the absence of maternal disorder, was negative and, for all but one model, statistically significant (p < .00l). These results indicate that teachers rated children in the subset with no maternal history of psychiatric disorder as having significantly more problems than mothers rated them. DISCUSSION

MOTHERS AS INFORMANTS

the distortion hypothesis, alternative explanations should be considered. Previous studies have framed this research issue as an either/or question: Either depressed mothers distort repons of their children's problems or children of depressed mothers actually do have more problems. This report indicates that it may be misleading to force a choice between distortion versus real effects of maternal psychiatric disorder, as both explanations found support in the data. For example, teachers' and mothers' agreement that children of depressed and anxious mothers have increased internalizing problems provides further evidence that these children actually have more internalizing problems. At the same time, depressed and anxious mothers report, on average, more of these problems than teachers, indicating that these mothers overreport internalizing problems in their offspring. For externalizing problems, there only is evidence of overreporting by mothers with psychiatric disorders. The greatest challenge to the distortion explanation is the possibility that the discrepancies between mothers and teachers might be due to variations in children's behaviors at home and at school (Achenbach et al., 1987). The results of this study enable us to examine this challenge. We found that in the absence ofhistory ofpsychiatric disorders, mothers reported fewer behavior problems in their children than teachers, suggesting that children generally display more behavior problems at school than at home. A related potential explanation is that children of depressed mothers specifically, in contrast to children in general, manifest an excess of behavior problems at home but not in school. Using a family interactional perspective (Patterson and Dishion, 1988), one might hypothesize that depressed mothers are more irritable, resulting in disruptions in discipline and, possibly, behavior problems in the child. If teachers are able to maintain high levels of control in the classroom, then disturbed offspring of depressed mothers might not display problems in school at the same level as in the home. There is little empirical support for this argument. Furthermore, Patterson and Dishion (I988) found that neither maternal depression nor maternal drug use was associated with irritable discipline on the part of the mother. Nonetheless, this alternative explanation cannot be ruled out in our data. Additional support for the plausibility of the overreporting explanation can be found in family history studies of psychiatric disorders, in which informants'

psychiatric disorder appears to be a source of bias (Chapman et aI., 1994; Roy et aI., 1994, 1996). In these studies, informants' psychiatric history tends to increase the sensitivity of family history information, while simultaneously decreasing specificity. Mothers' overreporting was largest when major depression was comorbid with anxiety disorder, whereas no evidence of overreporting was found for maternal depression without anxiety. It should be noted that previous research has consistently demonstrated high lifetime comorbidity of major depression with anxiety disorders (Eaton and Chilcoat, 1992; Regier et aI., 1990; Robins et al., 1991). In our sample, more than two thirds of mothers with a history of major depression also had a history of an anxiety disorder. Consequently, the observed effect of maternal depression comorbid with anxiety is likely to apply to the majority of depressed mothers. The findings from this study have important implications for psychiatric research, especially for family studies of psychiatric disorders, which compare the rates of disorder in relatives of affected versus unaffected probands. Evidence of overreporting by depressed mothers in this study underscores the concerns expressed previously about the psychiatric status of informants as a potential source of bias. This is a particular hazard in studies of children, in which mothers often are used as the sole informant about child behaviors. The use of multiple informants offers no guarantees, because there are no well-defined rules for combining discordant information that guard against such a bias (Klein, 1991; Piacentini et aI., 1992). In this report, we focused specifically on the interaction between maternal psychiatric history and type of informant (mother versus teacher) in relation to children's behavior problems. However, the analytic strategy can be used more generally to determine whether ratings from multiple sources can be combined, for example, when analyzing associations between risk factors and child outcomes, as described by Fitzmaurice et al. (1995). If there is no risk factor-informant interaction, data from both informants can be combined to produce a single and more precise estimate of the association between the risk factor and child outcome. Furthermore, this analytic strategy can be used when there are more than two informants by extending the model to include additional terms for each additional informant and corresponding interaction terms.

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY. 36:7. JULY 1997

977

CHILCOAT AND BRESLAU

Findings from this study have implications for both diagnosis and treatment in clinical practice. Clinicians, who often must rely on mothers as informants about their young patients, should consider the mothers' psychiatric history in the course of diagnosing and treating children's disorders. In addition, our findings underscore the importance of collecting information about children's behavior problems from multiple sources, especially when the mother is depressed and anxious. This study extends the inquiry of discrepancies between mothers' reports and those of other informants a step forward. The analytic strategy enabled a more critical evaluation of alternative interpretations. Findings from this study indicate that mothers, in general, report fewer problems than teachers. Further research is needed to determine whether children of depressed mothers have a unique susceptibility to displaying more behavior problems at home than at school.

Fergusson DM. Horwood LJ, Gretten ME. Shannon FT (1985). Family life events, maternal depression. and maternal and teacher descriptions of child behavior. P-diatrics 75:30-35 Fergusson DM. Lynskey MT. Horwood LJ (1993). The effect of maternal depression on maternal ratings of child behavior.] Abnorm Child Psychol 21:245-269 Fitzmaurice GM. Laird NN. Zahner GEP, Daskalakis C (1995), Bivariate logistic regression analysis of childhood psychopathology ratings using multiple informants. Am] Epid,miol 142:1194-1203 Forehand R, Lautenschlager GJ. Faust J. Graziano WG (1986), Parent perceptions and parent-ehild interactions in clinic-referred children: a preliminary investigation of the effects of maternal depressed moods. &hav

R'5 Tha 24:73-75 Friedlander S, Weiss DS, Traylor J (1986), Assessing the influence of maternal depression on the validity of the Child Behavior Checklist.] Abnorm

Child Psycho/14: 123-133 Gotlib IH. Mount JY. Cordy NI. Whiffen VE (1988), Depression and perceptions of early parenting: a longitudinal investigation. Br] Psychiatry 152:24-27 Graham P, Stevenson J (1985). A twin study of genetic influences on behavioral deviance.] Am Acad Child Psychiatry 24:33-41 Griest DL. Forehand R, Wells KC, McMahon RJ (1980), An examination of differences between nonclinic and behavior problem clinic-referred children.] Abnorm PsychoI89:497-500 Griest DL, Wells KC. Forehand R (1979), An examination of predictors of maternal perceptions of maladjustment in clinic-referred children. ]

Abnorm PsychoI88:277-281

REFERENCES Achenbach TM (1991). Manualftr th, Child &havior Ch,cklist/4-J8 and J99 J Profll,. Burlington: University of Vermont Department of Psychiatry Achenbach TM. McConaughy SH. Howell CT (1987). Child/adolescent behavioral and emotional problems: implications of cross-informant correlations for situational specificity. Psychol Bull 10 1:213-232 Billings AG, Moos RH (1985). Children of parents with unipolar depression: a controlled I-year follow-up.] Abnorm Child Psycho/14: 149-166 Breslau N. Chilcoat H. DelDotto J. Andreski P, Brown G (1996). Low birth weight and neurocognitive status at six years of age. Bioi Psychiatry 40:389-397 Breslau N. Davis Gc. Prabucki K (1988), Depressed mothers as informants in family history research: are they accurate? Psychiatry R'5 24:345-359 Breslau N, Del Dotto JE, Brown GG et at. (1994). A gradient relationship between low birthweight and IQ at age 6 years. Arch P-diatr Adol'5c M-d 148:377-383 Chapman TF. Manuzza S. Klein DF. Fryer AJ (1994), Effects of informant mental disorder on psychiatric family history data. Am] Psychiatry IS I :574-579 Chilcoat HD. Breslau N, Anthony JC (1996). Potential barriers to parent monitoring: social disadvantage. marital status. and maternal psychiatric disorder.] Am Acad Child Adol'5c Psychiatry 35: 1673-1682 Clifford PI, Hemsley DR (1987), The influence of depression on the procession of personal attributes. Br] Psychiatry 150:98-103 Conrad M, Hammen C (1989). Role of maternal depression in perceptions of child maladjustment.] Consult Clin PsychoI57:663-667 Diggle PJ, Liang K-Y. Zeger SL (1994). Analysis ofLongitudinal Data. New York: Oxford University Press Eaton WW, Chilcoat HD (1992), The latent class structure of anxiety and depression in treated and untreated samples of two US populations. In: Primary H,alth Car< and Psychiatric Epid,miology. Cooper B, Eastwood R. eds. London: Routledge. pp 307-318 EstroffTW. Herrera C. Gaines R. Shaffer D, Gould M, Green AH (1984), Maternal psychopathology and perception of child behavior in psychiatrically referred and child maltreatment families.] Am Acad Child Adolrsc

Ivens C. Rehm LP (1988), Assessment of childhood depression: correspondence between reports by child. mother, and father. ] Am Acad Child

Adolrsc Psychiatry 27:442-450 Jensen PS, Traylor J. Xenakis SN, Davis H (1988), Child psychopathology rating scales and interrater agreement. I: parents' gender and psychiatric status.] Am Acad Child Adol'5c Psychiatry 27:451-461 Klein RG (1991), Parent-ehild agreement in clinical assessment of anxiety and other psychopathology: a review.] Anxi,ty Disord 5: 187-198 Kowalik SB. Gotlib IH (1987). Depression and marital interaction: concordance between intent and perception of communication.] Abnorm Psycho! 96:127-134 Liang K-Y. Zeger SL (1986), Longitudinal data analysis using generalized linear models. Biom'trika 73: 13-22 Loeber R. Green SM, Keenan K, Lahey BB (1995), Which boys will fare worse? Early predictors of the onset of conduct disorder in a six-year longitudinal study.] Am Acad Child Adolrsc Psychiatry 34:499-509 Lynskey MT, Fergusson DM. Horwood LJ (1994), The effect of parental alcohol problems on rates of adolescent psychiatric disorders. Addiction 89: 1277-1286 McCabe SB. Gotlib IH (1995). Selective attention and clinical depression: performance on a deployment-of-attention task. ] Abnorm Psycho! 104:241-245 Moretri MM. Fine S, Haley G. Marriage K (1985), Childhood and adolescent depression: child-report versus parent-report information. ] Am

Acad Child Psychiatry 24:298-302

Fergusson DM. Horwood LJ (1987), The trait and method components of ratings of conduct disorder, part II: factors related to the trait component of conduct disorder scores.] Child Psychol Psychiatry 28:261-272

Panaccione VF. Wahler RG (1986), Child behavior, maternal depression. and social coercion as factors in the quality of child care. ] Abnorm Child Psycho! 14:263-278 Patterson GR, Dishion TJ (1988), Multilevel family process models: traits. interactions. and relationships. In: R-lationships Within Family. Hinde RA, Stevenson-Hinde J. eds. Oxford. England: Clarendon Press. pp 283-310 Piacentini JC. Cohen P, Cohen J (1992), Combining discrepant diagnostic information from multiple sources: are complex algorithms better than simple ones? ] Abnorm Child Psycho! 20:5 1-63 Regier DA. Burke JD. Burke KC (1990), Comorbidity of affective and anxiety disorders in the NIMH Epidemiologic Catchment Area Program. In: Comorbidity ofMood and Anxi,ty Disordm, Maser JD, Cloninger CR. eds. Washington. DC: American Psychiatric Press. pp 113-122 Rice J, Reich T. Andreasen NC et at. (1987). The familial transmission ofbipolar illness. Arch Gm Psychiatry 44:441-447

978

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 36:7, JULY 1997

Psychiatry 23:649-652

MOTHERS AS INFORMANTS

Richters J, Pelligrini D (1989), Depressed mothers' judgments about theit children: an examination of the depression-distortion hypothesis. Child

Dro 36:884-893 Richters JE (1992), Depressed mothers a5 informants about their children: a critical review of the evidence for distortion. Psycho/ Bull 112:485-499 Rickard KM, Forehand R, Wells KC, Griest DL. McMahon RJ (1981), Factors in the referral of children for behavioral treatment: a comparison of mothers of clinic-referred deviant, c1inic-teferred non-deviant. and non-clinic children. &hav &s Thtr 19:201-205 Robins LN, Locke BZ, Regier DA (1991), An overview of psychiatric disorders in America. In: PsychiAtric DisortUrs in Ammca, Robins LN, Regier DA, eds. New York: Free Press, pp 328-386 Rosenbaum JF, Biederman J, Bolduc EA, Hirschfield DR, Faraone SV, Kagan J (1992), Comorbidity of parental anxiety disorders a5 risk for childhoodonset anxiety in inhibited children. Am] PsychiAtry 149:475-481 Roy MA, Walsh D, Kendler KS (1996), Accuracies and inaccuracies of the family history method: a multivariate approach. Acta Psychiatr Scand 93:224-234

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 36:7, JULY 1997

Roy MA, Walsh D, Prescott CA, KemUer KS (1994), Bia5es in the diagnosis of alcoholism by the family history method. Akohol Clin Exp Ru 18:845-851 Schaughency EA, Lahey BB (1985), Mothers' and fathers' perceptions of child deviance: roles of child behavior. parental depression, and marital satisfaction.] Consult Clin PsychoI53:718-723 Vanyukov MM, Moss HB, Plail JA. Blackson T. Mezzich AC, Tarter RE (1993). Antisocial symptoms in preadolescent boys and in their parents: associations with cortisol. Psychiatry Res 9-17 Webster-Stratton C (1988), Mothers' and fathers' perceptions of child deviance: roles of parent and child behaviors and parent adjustment. ]

Consult Clin PsychoI56:909-915 Weissman M. Wickramarante P, Warner U et aI. (1987). Assessing psychiatric disorders in children: discrepancies between mothers' and children's reports. Arch Gm Psychiatry 44:747-753 Zeger SL. Liang K-Y (1986), Longitudinal data analysis for discrete and continuous outcomes. Biometrics 42: 121-130

979