Introduction

Children and parents are both important informants of children’s behavioral and emotional problems. It is well established that discrepancies exist between children’s reports and reports by parents [3, 8, 21]. In a recent review on informant discrepancies, De Los Reyes and Kazdin [21] concluded that there is less agreement between informant’s ratings of internalizing problems compared to externalizing problems [3, 24].

Since there is familial aggregation for internalizing problems, partly resulting from disorder specific genetic influences [32, 36, 48, 50], more parents from children with internalizing problems experience internalizing problems themselves, compared to other parents. The distorting effects of maternal internalizing problems on ratings of child problems have been extensively studied. There is empirical evidence suggesting that maternal depression and anxiety are associated with parent–child informant discrepancies. Inconsistencies however exist as to the source of the distortion (depression or anxiety) and the gender and age of the child in relation to the maternal informant discrepancy [21]. Most studies on informant discrepancies regarding the child’s internalizing problems have included only mothers, although some studies have included both parents [27, 38, 46].

Results of epidemiological studies in the general population [9, 10, 51] as opposed to clinical studies [12, 26, 30, 34, 35, 37, 43, 45] show that youths tend to report more internalizing problems than parents do about them. Prevalence rates of child internalizing problems in the general population derived from parental ratings are lower compared to those derived from the child itself [18, 21]. However, when parents experience internalizing problems, they tend to report more symptoms of internalizing problems about their child, compared to the child itself [41]. Richters [44] concluded that previous studies on maternal depression distortion were not empirically validated. Recent studies with clinical samples of youths [6, 38] and general population samples [11, 15, 27, 41, 46, 53] explored the association between maternal depression and anxiety and reporting discrepancies on child internalizing problems, taking previous methodological flaws into account by comparing their ratings to teacher ratings [11, 15, 53] or using different methodology [46].

In the present study, we try to elucidate some of the inconsistencies mentioned by De Los Reyes and Kazdin [21] by analyzing the effect of maternal depressive and anxious symptoms separately on informant discrepancies for boys and girls. Furthermore statistical analyses are done, using standardized difference scores, since these scores are proven to correlate equally with both informants’ ratings [22] and in this way adjust for methodological problems, mentioned by Richters [44]. We address the following research question: to what extent are maternal depression and anxiety symptoms predictors of positive mother–child reporting discrepancies on affective and anxiety problems in children? We make use of a unique large cohort from the general population, in which the effects of current and lifetime maternal internalizing problems on reporting of child internalizing problems are assessed.

Materials and methods

Instruments

Predictor variables: maternal depressive, anxiety and stress symptoms

The Depression Anxiety Stress Scale (DASS-21) [39] is a self-report questionnaire to assess anxiety, depression, and feelings of stress in adults. It contains 21 items covering the past week, rated on a four-point scale, ranging from ‘not at all’ (=0), to ‘very much, or most of the time’ (=3). By summing item scores, three syndrome scores of each seven items can be derived: Depression, Anxiety, and Stress. The good psychometric properties of the DASS-21 [5, 16, 20, 31] were replicated for the Dutch version [7]. For the sample of respondents used in this study, Cronbach’s alpha coefficients for the subscales of the DASS were: .83 for the DASS depression scale, .78 for the DASS anxiety scale and .86 for the DASS stress scale, demonstrating good internal consistency. Inter correlations between the subscales of the DASS for this sample were: .59 between DASS depression and DASS anxiety, .58 between DASS anxiety and DASS stress and .71 between DASS depression and DASS stress, using Pearson’s correlations, suggesting shared method variance. Therefore the DASS stress scale was left out of the analyses.

In order to asses the number of respondents who score in a range suggestive of clinically significant depression and anxiety, DASS scores were dichotomized based on normative data for a non-clinical adult population using the percentiles indicating ‘severe’ and ‘extremely severe’ as a cut-off score. These cut off scores correspond with a DASS-21 depression score of 12 and a DASS-21 anxiety score of 8 [20].

Predictor variables: maternal lifetime depression and anxiety

Maternal lifetime depression and anxiety were assessed with the TRAILS Family History Interview (FHI) [42]. A research assistant conducted this interview at home with the mother. A vignette, describing the main characteristics, followed by a question about lifetime occurrence was used to assess both depression and anxiety disorders (generalized anxiety disorder, social phobia, specific phobia, panic disorder or obsessive compulsive disorder). The response format was ‘no’ (0), ‘probably not’ (1), ‘probably so’ (2), ‘yes ‘(3) and ‘I do not know’ (9). For the present study the response format was dichotomized; aggregating 0–1 to ‘disorder absent’ and 2–3 to ‘disorder present’. The Pearson’s correlation coefficient between the FHI depression and anxiety disorders scale for this sample was .35, showing diversity between the scales. Since the interview was constructed specifically for TRAILS, to obtain an indication about the absence or presence of psychiatric disorders in the mothers of participants, no further information on psychometric properties is available.

Outcome variables: CBCL–YSR standardized difference scores on affective and anxiety problems

The Child Behavior Checklist (CBCL) [4] is a parent questionnaire to assess behavioral and emotional problems in 4- to 18-year-old-children. In this study the subscales according to the Diagnostic and Statistical Manual of mental Disorders-IV, text revised [1] of the CBCL were used, reflecting: affective problems, anxiety problems, somatic problems, attention deficit hyperactivity problems, oppositional defiant problems and conduct problems. High concurrent validity was found with DSM-IV clinical diagnoses and other standardized rating scales [4]. In the sample of Achenbach et al. [2] the mean Cronbachs’ alpha of the DSM-oriented scales was .82 and the test–retest reliability was .85. The cross-informant agreement between parents and youths was .45.

In this study, the DSM-IV subscales affective problems and anxiety problems of the CBCL were used. The affective problems scale consists of 13 items that are consistent with the DSM-IV diagnoses of major depressive disorder and dysthymia. The subscale anxiety problems, consisting of six items, represents three DSM-IV anxiety disorders: generalized anxiety disorder (three items), separation anxiety disorder (two items) and simple phobia (one item). The scoring format is 0 = not true, 1 = somewhat or sometimes true, 2 = very true or often true. In order to select borderline and clinical cases of affective and anxiety problems, the normative sample of Achenbach and Rescorla [4] was used. For boys the clinical cut-off score correspondents with 5 (CBCL affective problems), 4 (CBCL anxiety problems), 8 (YSR affective problems) and 6 (YSR anxiety problems) or higher. For girls the clinical cut-off score correspondents with 5 (CBCL affective problems), 5 (CBCL anxiety problems), 10 (YSR affective problems) and 7 (YSR anxiety problems or higher. For the present sample the Cronbach’s alpha coefficients were .68 for the DSM-IV scale affective problems and .66 for the DSM-IV scale anxiety problems.

The Youth Self-Report (YSR) [4] was used to obtain adolescents’ self-reports. The YSR, developed for 11–18-year-olds, has roughly the same format as the CBCL, except that items are worded in the first person. In the sample of Achenbach et al. [2] the mean Cronbach’s alpha coefficient of the DSM-oriented scales of the YSR was .76 and the test–retest reliability .79. In the present sample the Cronbach’s alpha coefficients of the scales we used were .72 for the DSM-IV scale affective problems and .62 for the DSM-IV scale anxiety problems.

Two standardized difference scores were constructed by subtracting the YSR-scale scores from the CBCL-scale scores on affective problems and anxiety problems. Taking into account the recommendations of De Los Reyes and Kazdin [22] on measuring informant discrepancies, the raw difference scores were converted into standardized Z-scores. The Z-score has a mean of 0 and a standard deviation of 1. In this way the difference scores correlate equally with the informant’s ratings used, which is in line with the comments of Richters [44] that no informant can be considered a ‘gold standard’ by which to interpret another informant’s rating (see the Appendix).

The standardized CBCL–YSR difference scores on (1) affective problems and (2) anxiety problems are the two outcome variables of this study. A positive difference score means that the mother reports more problems than the child. A negative difference score indicates that the child reports more problems than the mother.

Participant recruitment

This study is part of the Tracking Adolescents’ Individual Lives Survey (TRAILS), which is a large longitudinal epidemiological study of Dutch pre-adolescents. The present study used data from the first assessment wave of TRAILS, which ran from March 2001 to July 2002. The target sample consisted of 10-to 12-year-olds from five municipalities in the North of the Netherlands, including both urban and rural areas. The response rate was 76.0%, resulting in 2,230 participants. Although significant differences between responders and non responders were found (56.9% of participants were boys in the non-responders vs. 49.2% of boys in responders; 44.2% of non responders had a parent with lower secondary education or less vs. 32.6% in responders; 28.4% of non responders needed additional help because of learning difficulties vs. 21.1% of responders), there were no indications of differences in the prevalence of psychopathology based on teacher ratings between responders and non responders [23].

Of the 2,230 participants, only data of mothers and children for whom complete data on both the DASS, CBCL and YSR were available, were included in the statistical analyses. Respectively 191 mothers, 186 mothers and 55 children were excluded from analyses because they did not complete the DASS, CBCL, or YSR. Due to overlap in non-response, the total number of incomplete mother–child pairs was 244. The final sample of this study consisted therefore of 1,986 participants (89%).

To test for differences between responders and non-responders on demographic characteristics, socioeconomic status (SES), Total Intelligence Quotient (TIQ) and ethnicity, χ 2- and ANOVA tests were used. SES was based on the highest educational level of mother and father, the type of occupation of the father and mother and the family income. SES was divided into lower class (1), middle class (2) and high class (3) by means of the frequency distribution (low SES = lowest 25%, middle SES = middle 50% and high SES = highest 25%). TIQ was based on two subtests of the WISC-R; Vocabulary and Block Design [49]. Demographic characteristics of this study sample seem to be roughly comparable to those of the Dutch population in 2001 based on data of the Dutch Central Bureau of Statistics [25].

Statistical analyses

First, mean, standard deviations and the number and percentage of respondents who scored in the range, suggestive of clinical significant problems were calculated for all variables. Second, Spearman’s correlation coefficients were calculated between the current and lifetime maternal internalizing problems, informants’ characteristics and the standardized difference scores. Third, in order to analyze the association between the predictor variables and difference scores by the mother, linear regression analyses were conducted. Standardized affective and anxiety CBCL–YSR difference scores were used as outcome variables and the DASS depression and anxiety subscales and FHI lifetime depression and lifetime anxiety scores were used as predictors: First, every independent variable (DASS depression, DASS anxiety, FHI lifetime depression, FHI lifetime anxiety) was entered separately in the analysis with the standardized difference scores on affective and anxiety problems as outcome variables; next, DASS depression and DASS anxiety combined were entered and FHI lifetime depression and anxiety combined were entered on both outcomes. Finally, all independent variables were entered simultaneously on both outcomes. SES and TIQ were included in every model as covariates. In this way the unique and combined predictive value of the current and lifetime maternal depression and anxiety problems on both difference scores could be assessed. All statistical analyses were conducted for boys and girls separately.

Results

Demographics

The final study sample consisted of 1,986 participants: 971 boys (49%; mean age 11.1 years, SD 0.5), 1,015 girls (51%; mean age 11.1 years, SD 0.5) and 1,986 mothers (mean age 40.7 years, SD 4.6). Eighty-nine percent of the children was of Dutch origin, 11% was originally from other countries, including Turkey, Morocco, Surinam, Indonesia, Dutch Antilles or other (Dutch population: 78.3% of Dutch origin and 21.7% originally from other countries[25]). Fourteen percent of the children belonged to a one-parent family and 86% to a two-parent family (Dutch population: 16.9% single parent families and 83.1% two-parent families [25]). Fourteen percent of the families had a monthly income of less than €1134,-, 65% had an income between €1134,- and €2950,- and 14% had an income above €2950,- (7% did not know their income or did not answer the question) (Dutch population: mean income pro family = €2358.33 [25]).

Within this study sample no differences were found between responders (N = 1,986) and non-responders (N = 244) regarding age (χ 2 = 4.3, P = ns) and sex (χ 2 = 0.86, P = ns) of the children. Significant differences between responders and non-responders were found for SES (responders: 23% low SES, 50.6% middle SES, 26.4% high SES, vs. non-responders: 47.3% low SES, 39.5% middle SES, 13.2% high SES; χ 2 = 60.976, P < 0.0005). Significant differences between responders and non- responders were also found for ethnicity (responders: 88.7% Dutch and 11.3% other ethnicity vs. non- responders: 65.5% Dutch and 34.5% other ethnicity, χ 2 = 85.352, P < 0.0005). Finally, significant differences between responders and non-responders were found regarding one- versus two-parent families (responders: two parent family 85.9%, one parent family 14.1% vs. non-responders: two parent family 75.6%, one parent family 24.4%, χ 2 = 15.273, P < 0.0005) and IQ (responder: mean IQ = 97.9 vs. non responders: mean IQ = 91.3; ES = .019, P < 0.000).

Descriptive statistics

Mean and standard deviations of all variables and the number and percentages of respondents with borderline or clinical significant affective and anxiety problems are presented in Table 1.

Table 1 Mean (M), standard deviations (SD) and number (N) and percentages (%) of respondents with clinical significant affective and anxiety problems for boys, girls and mothers

Correlations between maternal internalizing problems and the standardized difference scores

Tables 2 and 3 shows small (ranging between .06 and .15) but significant positive correlations between both current and lifetime maternal depression and anxiety scores and the standardized difference scores. Small but significant negative correlations were found between the covariate SES and all outcome variables, except the affective difference score for girls. Significant negative correlations were also found between the covariate TIQ and the anxiety difference scores for boys and girls. SES and TIQ were added as covariates in the regression analyses.

Table 2 Correlations between maternal current (DASS) and life time (FHI) depression and anxiety, informant characteristics and the standardized difference scores on affective and anxiety problems for boys (N = 971) and girls (N = 1,015) separately
Table 3 Linear regression analyses of the current (DASS) and lifetime (FHI) maternal internalizing problems and the standardized CBCL–YSR affective and anxiety difference scores, for boys (N = 971) and girls (N = 1,015) separately, adjusted for SES and TIQ

Predicting CBCL–YSR difference scores by the DASS depression and anxiety subscales and FHI lifetime anxiety and lifetime depression

All unique predictor variables were significant in the models on both difference scores for boys and girls, explaining 0.8–4.1% of the variance. Combining DASS depression and anxiety in four predictive models on both outcome scores for boys and girls separately, only DASS depression remained a significant predictor in every model. The combined scales explained 1.6 tot 4.4% of the variance. Combining FHI lifetime depression and anxiety problems in four models predicting both outcome scores for boys and girls, again only FHI lifetime depression significantly predicted all outcome scores. FHI lifetime anxiety reached significance predicting the difference score on anxiety problems in boys as well. The combination of both predictors explained 1.9–3.2% of the variance. Finally, combining all predictive variables in four models on the standardized affective and anxiety difference scores for boys and girls separately, DASS depression remained significant in three models. The DASS depression score did not significantly explain the difference score on anxiety problems in boys. FHI-lifetime depression also remained significant in three out of four models. It did not significantly predict the anxiety difference score in girls. DASS anxiety was not a significant predictor in all models. FHI lifetime anxiety remained significant in the model predicting the anxiety difference score in boys. The variance explained by the combined predictor models varied between 2.7 and 4.8%.

Discussion

The present study investigated to what extent current and lifetime maternal depression and anxiety are related to positive reporting discrepancies between mother and child regarding the child’s internalizing problems in a large general population sample. Regression analyses showed positive but small relations between current and lifetime maternal internalizing problems and the mother–child reporting difference score on affective and anxiety problems in children. This indicates that a higher level of maternal internalizing problems is associated with a larger difference score. These findings suggest that maternal internalizing problems contribute to the maternal perception of their child’s problems tending to be higher in relation to their child’s own appraisal. This finding is supported by most previous studies, showing that mothers with internalizing problems in the general population report more internalizing problems than youths themselves and/or another observer [11, 15, 27, 41, 53]. Mothers without internalizing problems report fewer problems compared to their child [41].

Combining maternal depression and anxiety scores as predictors of the difference score in this study, current and/or lifetime maternal depression reached significance in every model and predicted both reporting differences for boys and girls, whereas anxiety did not reach significance in all but one model. The association between maternal depression and a positive reporting discrepancy was confirmed by previous general population studies on mother–child reporting discrepancies associated with maternal internalizing problems. However, previous studies that differentiated between maternal depression and anxiety, reported an association between maternal anxiety and reporting discrepancies as well [11, 15, 27, 53]. Results of the present study show that both predictors seem to be associated with a positive reporting discrepancy but when they are adjusted for each other, maternal current and lifetime depression appears the main predictor of the mother–child reporting discrepancies.

Although significant associations were found between maternal current and lifetime depressive symptoms and positive reporting discrepancies, these associations explained three to five percent of the variance. According to Cohen [17], effects accounting for 1.0–5.9% of variance are considered small. Maternal depressive symptoms were only slightly elevated, which could explain the fact that maternal depression only accounted for a small part of the variance of the reporting discrepancy. Najman et al. [41] demonstrated that an increase in maternal internalizing symptoms coincides with mothers reporting increasingly more symptoms compared to their child. Therefore it seems that although the association between maternal depressive symptoms and a positive mother–child reporting discrepancy has been demonstrated, in the general population this association is small and does not bias research, using maternal reports on child problems.

Moving beyond studies using questionnaires, the association between maternal depressive and anxiety symptoms and maternal reporting behavior has also been demonstrated by experimental studies. Conrad and Hammen [19] showed that depressed mothers made more negative comments in interactions with their symptomatic child (experiencing either internalizing or externalizing symptoms) during a discussion compared to non-depressed mothers. More recently, a maternal dysphoria-related bias was demonstrated [52]; mothers who experienced dysphoria (anxious and depressive symptoms) rated the negative emotions and behaviors of their own child and a control child more negative compared to independent raters. Considering children with ADHD, Chi and Hinshaw [14] demonstrated more negative ratings by mothers experiencing depression, on the problematic behaviors of their child and on their own parenting style, compared to teachers’ and child’s ratings. Consequently the maternal judgment distortions (and not the depressive symptoms) predicted future mother–child interaction problems like showing inadequate levels of maternal warmth.

Previous research on adult depression indicates that depressogenic cognitive affective structures form a stable underlying feature of depression-vulnerable individuals. Furthermore these cognitive structures are responsible for the occurrence of a depressive episode [33, 47]. In this way, former or current depressive symptoms may be associated with a bias in current reporting behavior. The number of mothers with a current clinical significant depression score in this study is quite small but more mothers reported a life time depression. Although depressive disorders are also associated with cognitive deficits in executive functioning (e.g. decision making), memory and attention deficits [13], these deficits are correlated with depression severity. Therefore they do not account for the reporting behavior of mothers with depressive symptoms in this study since maternal depressive symptoms were only slightly elevated and only few mothers scored in the range for clinical depression.

In this study a difference between boys and girls appeared in predicting discrepancy scores on child’s anxiety symptoms. In boys, maternal lifetime depression and anxiety were associated with this positive difference score. In girls, current maternal depression and lifetime maternal depression (showing a non-significant trend) were associated. In a previous study Briggs-Gowan [11] found opposite results, demonstrating a correlation between maternal anxiety and mother–teacher difference scores in girls and maternal depression and mother–son difference scores in boys. Krain and Kendall [38], studying a clinical population of children with anxiety disorders, found no effect of parental anxiety on positive parent–child reporting differences. Based on the outcomes of this study and previous studies, specific effects of maternal depressive or anxious symptoms on differences between reporting-discrepancies regarding boys and girls in the general population are inconsistent. This finding reflects the conclusions of De Los Reyes and Kazdin [21] based on their review study that in sum, child gender may not be related to informant discrepancies.

Studies considering the long-term implications of informant discrepancies in the general population are scarce. One prognostic study on parent–adolescent disagreement regarding behavioral and emotional problems as a risk factor for adverse outcome 4 years later showed an association between negative mother–child discrepancies on the subscales ‘withdrawn’ and ‘anxious/depressed’ of the CBCL and YSR and adverse outcome 4 years later. However, these associations did not remain significant when adjusted for CBCL and YSR scores as independent predictors [28]. Longitudinal studies on parent–child informant discrepancies regarding clinical samples of children with internalizing problems showed mixed results: a longitudinal study into parent–child diagnostic agreement on anxiety disorders in a clinical sample of children, aged 9–13 years old, showed that although parent–child agreement was poor to moderate at pre-treatment, post-treatment and 7.4 year follow-up, treatment outcome was good and improvement was maintained at follow-up [45]. Ferdinand et al. [29] demonstrated an increasing risk for future police/judicial contacts at 4 year follow up for referred adolescents who received much higher anxiety/depression scores by their parents compared to their own scores. The possible clinical implications on the long term functioning of the child/adolescent of these parent–child reporting discrepancies on child internalizing problems should be explored in further studies.

Strengths and limitations of the study

This study was conducted in a large population sample, consisting of 1,986 mothers and children. Our large population based cohort can be considered representative for the Dutch general population on demographic variables. Although this study sample consisted of 89% of the initial TRAILS sample and some significant differences on demographic variables between this research sample and the respondents who were left out because of incomplete data were found, we assume that the prevalence of internalizing problems in children is comparable to the initial sample. Research on the initial sample of 2,230 respondents showed that there were no differences in the prevalence of psychopathology between responders and non responders despite of differences in demographic characteristics [23]. Furthermore demographic characteristics of this study sample are generally comparable to those of the Dutch population [25]. Assessment of both current and lifetime maternal internalizing problems, and of anxiety and depression in both mothers and children, further supported the strength of the study. The number of mothers reporting a life time period of depression is comparable to the original TRAILS sample and to the CIDI-DSM-IV life time rates obtained by direct interviewing in the Netherlands Mental Health Survey and Incidence Study (NEMESIS) [42]. Standardized difference scores were computed, proven to be the most consistent estimates among informant discrepancies and informant characteristics by De Los Reyes and Kazdin [22].

In this study multiple tests have been performed, which increases the risk on chance findings. However, since results of this study were all consistently heading in the same direction and several P-values are below .001, outcomes can be considered as representative of ‘true’ associations.

A limitation of the study could be that fathers’ reports on their children were not obtained. However, the few studies on reporting bias due to parental (internalizing) problems in the general population that included fathers [27, 46], found little differences between mothers’ and fathers’ reports of children’s internalizing problems. Considering the familial aggregation of anxiety and depression, partly explained by genetic factors [32, 50], and the co-occurrence of mental health problems in marital partners [40], the effect of paternal internalizing problems on reporting behavior should be included in future research. Furthermore lifetime anxiety and depression were not assessed with a clinically validated interview.

Conclusions

Considering inconsistencies of previous studies, regarding the relative effects of maternal depressive and maternal anxiety symptoms [21], this study clearly shows that maternal depressive symptoms are related to an increase of the reporting discrepancy on child’s internalizing problems, whereas maternal anxiety almost was not. However, since only a small percent of variance was explained it seems that maternal depressive symptoms do not bias general population research, using maternal report of child’s internalizing problems. No clear pattern regarding the gender of the child related to reporting discrepancies emerged from this study.

More research is needed on long term consequences of parent–child reporting discrepancies, regarding child’s internalizing symptoms as a risk factor for future adverse functioning of the child, since these studies are few and show mixed results.