Child Psychiatry Hum Dev DOI 10.1007/s10578-015-0616-1

ORIGINAL ARTICLE

Multi-domain Predictors of Attention Deficit/Hyperactivity Disorder Symptoms in Preschool Children: Cross-informant Differences John V. Lavigne1,2 • Karen R. Gouze1,2 • Joyce Hopkins3 • Fred B. Bryant4

Ó Springer Science+Business Media New York 2015

Abstract Numerous studies indicated that agreement between parent and teacher ratings of symptoms of attention-deficit/hyperactivity disorder in children of all ages is poor, but few studies have examined the factors that may be associated with rater differences. The present study examined the contextual, parent, parenting, and child factors associated with rater differences in a community sample of 4-year-old children. Parents and teachers of 344 4-year-olds recruited from preschools and pediatric practices completed the preschool versions of the Child Symptom Inventory. Measures of socioeconomic status, family stress and conflict, caretaker depression, parental hostility, support-engagement, and scaffolding skills, and child negative affect (NA), sensory regulation (SR), effortful control (EC), inhibitory control, and attachment security were obtained either by parental report or observational measures. v2 difference tests indicated that child factors of EC and SR, and contextual factor of stress and conflict, contributed more to parent-ratings of ADHD-I and ADHD-HI than to teacher-ratings of those same types of symptoms. Two factors contributed more to teacher-than to parent-rated ADHD-I, NA and caretaker depression.

& John V. Lavigne [email protected] 1

Department of Child and Adolescent Psychiatry (#10), Ann & Robert H. Lurie Children’s Hospital of Chicago, 225 East Chicago Avenue, Chicago, IL 60611, USA

2

Feinberg School of Medicine, Northwestern University, Chicago IL, USA

3

Department of Psychology, Illinois Institute of Technology, Chicago IL, USA

4

Department of Psychology, Loyola University Chicago, Chicago, IL, USA

Results indicate there are differences in factors associated with ADHD symptoms at home and school, and have implications for models of ADHD. Keywords ADHD
Parent–teacher agreement
Informant discrepancies
Preschoolers
Temperament

Introduction Attention deficit hyperactivity disorder (ADHD), by definition, is a disorder whose symptoms must be present in ‘‘two or more settings (e.g. at home, school, or work),’’ cause functional impairment, and emerge before age 12 [1]. To establish that ADHD symptoms are present in multiple settings, professional organizations such as the American Academy of Pediatrics [2, 3] and others [4] recommend obtaining information from both parents and teachers when evaluating and treating children with ADHD. Although cross-situational symptom manifestation is required to meet DSM diagnostic criteria, studies of inter-rater agreement for children of all ages show only low to moderate levels of agreement between teachers and parents whose locus of convenience for observing those symptoms are, respectively, school and home [5–8]. In a study of over 6000 children ages 4–17, for example, correlations between parent and teacher reports for both hyperactive-impulsive and inattentive symptoms ranged from 0.18 to 0.44. Agreement is slightly better for hyperactive-impulsive (r = 0.33–0.44) than inattentive (r = 0.18–0.32) symptoms [6], but at the highest level of agreement (r = 0.44) only 19 % of the variance in parent and teacher ratings is shared. When agreement between diagnostic categories is measured with kappa scores rather than correlations, agreement is also very low, ranging from -0.05 to 0.41 in two reports [7, 9].

123

Child Psychiatry Hum Dev

Some investigators have attributed these parent–teacher differences to methodological factors, e.g., lack of clarity in the wording of individual items contributing to differential interpretation, or need for more clearly interpretable anchors for symptom frequency [5]. Medication effects have also been identified as a possible explanation for the lack of consistency across raters, because some children are medicated at school but not at home. Nevertheless, other studies have found poor inter-rater agreement even when all children in the study are on medication and following a protocol that includes late afternoon—that is, home-based—medication doses along with school hours administration [7]. Others have suggested that differences between raters might be attributable to actual cross-situational differences in the children’s behavior rather than to measurement error [9]. Relatively few studies, however, have addressed which specific child or family factors might contribute to differences in inter-rater agreement on ADHD symptoms. Although Antrop et al. [5] speculated that underlying child cognitive or neuropsychological factors may contribute to differences between raters, and that a clearer understanding of such factors might improve the diagnostic process, they did not empirically test this possibility. Similarly, Narad et al. [6] recommended that research is needed to identify patient-related factors that contribute to parent–teacher agreement or disagreement. Still another possibility is that environmental factors, such as the degree of structure or the specific task demands, affect parent–teacher perceptions and subsequent symptom ratings. The low level of agreement between raters on ADHD symptoms is similar to that for another common externalizing childhood disorder, oppositional defiant disorder (ODD). Poor agreement between mothers and teachers for symptoms of ODD have been found for both preschool and school-age children in numerous studies [10–13]. Drabick et al. [14] found that source-specificity is stable from one school grade to another in young children. Based on such findings, investigators have suggested that ODD should be considered an informant- or source-specific disorder [10, 11, 13], and such findings have resulted in DSM-5 describing ODD as a source-specific disorder [1]. It is of interest that DSM-5 does not raise the possibility that ADHD is a source or situation-specific disorder, even though the level of inter-rater agreement is similar for ADHD and ODD symptoms. Origins of Source-Specific Differences in Rating of ADHD Symptoms Presently, greater attention is needed to the factors that may contribute to source-specificity of ADHD symptoms in children. Because so little attention has been given to

123

understanding the source-specificity of ADHD, there is no well-established conceptual framework to guide the study of specific factors that contribute to source-specific differences. There are, however, two sources of information that provide guidance in determining which factors may be associated with the development of ADHD, and these factors may also contribute to source-specific differences in parent–teacher ratings of ADHD symptoms. As a result, these factors may warrant further empirical investigation with regard to source specificity. First, models of the multi-domain factors associated with ADHD provide guidance in identifying factors that might also contribute to source-specific differences in ADHD symptoms and, hence, parent–teacher ratings. ADHD is recognized as a multi-factorial disorder [15]. A variety of different models attempt to identify the psychosocial factors and cognitive deficits that may be associated with symptoms of ADHD, but no single model fully incorporates all the deficits associated with the disorder [16]. Coghill et al. [17] put forth a general model that postulates causal factors within the biological, cognitive, behavioral, and environmental domains that contribute to ADHD. The biological basis of ADHD is often emphasized because the heritability of ADHD is high. Nigg et al. [18] suggest that there may be multiple pathways involving temperamental reactivity (e.g., negative affect; NA) and effortful control (EC) as causal factors for ADHD. One of the cognitive processes identified as possibly having a causal role in ADHD is inhibitory control (IC), i.e., the ability to inhibit a dominant response in favor of a less dominant but more desirable one, which is an important component of executive functioning [16]. At the environmental level, Johnston and Mash [19] emphasize that there may be a transactional process involving characteristics of the family and child that extends over time and contributes to the development and expression of ADHD symptoms. Among these factors, cross-sectional studies have found an association between ADHD symptoms and contextual variables of stress and family conflict, parental warmth or support, and parental coercion or hostility. In reviewing studies of parenting and psychosocial factors, Johnston and Mash [19] concluded that parenting practices affect ADHD symptoms and that contextual factors, such as life stress or marital conflict, are linked to ADHD, but not consistently. They concluded that developing models of family influences on ADHD symptoms is critical to understanding the genesis and maintenance of this disorder. Second, models of factors associated with other externalizing behaviors may be useful. A recent study examined many of the same factors noted above when studying crossinformant differences for another externalizing disorder. Lavigne et al. [13] examined a multi-level model of factors associated with the development of ODD that included

Child Psychiatry Hum Dev

contextual, parent, parenting, and child factors that might be associated with source-specific parent and teacher ratings of ODD symptoms in preschool children. Results indicated that one contextual factor, high levels of family conflict, was associated with teacher as well as parent ratings of ODD symptoms. A parental factor, higher levels of parenting hostility/coercion, was associated with higher levels of parent-rated but not teacher-rated ODD symptoms. This pattern of results suggested that the effects of family conflict, but not necessarily parenting hostility per se, may have generalized across situations to influence manifestation of ODD symptoms at school where teachers also observe them. Lower levels of parent-rated child EC were also associated with parent-rated but not teacher-rated ODD symptoms. While that pattern of results raises the possibility that the temperamental characteristic of child EC is more strongly associated with ODD symptoms at home more than at school, it is also possible that common method-variance inflated the strength of the relationship between parent-rated EC and ODD symptoms, in comparison to parent-rated EC and teacher-rated symptoms. The Present Study The present study examined the associations between psychosocial factors across multiple levels (contextual, parent, parenting, and child) and differences in parent and teacher ratings of ADHD symptoms in a group of 4-yearolds. Sonuga-Barke et al. [20] noted the importance of studying the multiple factors that may be associated with ADHD in a preschool population because environmental factors contributing to ADHD symptoms, such as negative parenting, might have particularly strong effects in the preschool years when they could influence executive functioning and hyperactivity. Furthermore, parental negativity and harsh parenting may increase risk particularly during this developmental period because school entry is a major life transition that, if managed poorly, might produce increased hyperactivity and, if managed well, might ameliorate these symptoms. Other studies also emphasize the importance of examining ADHD symptoms in preschoolers, with these studies indicating that the symptoms of ADHD in preschoolers are similar to those reported in school-age children [21–23], and that stability of the ADHD diagnosis is comparable for preschoolers and older children. Stability is relatively high over a period of 6 years, with 89 % retaining that diagnosis [24]. We hypothesized that the contextual, parent, and parenting factors related to parent–teacher discrepancies in symptom reports of ADHD would follow a pattern similar to that found with another externalizing problem, ODD. That is, we anticipated that many of these contextual, parent, and parenting factors would be associated with

ADHD symptoms at home but not necessarily at school. We anticipated that temperamental factors reported by parents based on home observations would also show a stronger association with behavior at home than at school. Conversely, we hypothesized that the cognitive factor of IC would show a stronger relationship with teacher reports of school behavior of ADHD symptoms, because the specific demands of the school setting require a greater need to exhibit strong control over executive functioning than do the specific demands of the home setting.

Method Participants Participants in this study included 344 children from a larger longitudinal study of factors associated with behavior problems in preschoolers (reference withheld for blind review) [13]. The larger study involved 796 participants who were recruited from 23 pediatric practices and 13 urban public schools in a Midwestern metropolitan area. This study sample comprised 43.2 % of the original study sample for whom both parent and teacher data were available at age 4 years (M = 4.42, SD = 0.33). Difficulties in collecting teacher data were related to the large number (350) of preschools which study children attended. Exclusion criteria for the sub-sample were the same as for the larger study: all children were 4 years of age, spoke English or Spanish in their homes and did not meet exclusion criteria for intellectual disability or autism. Of this sample, (a) 178 (51.7 %) were boys, and (b) parentreported racial/ethnic identity of the children was 217 (63.1 %) White non-Hispanic, 46 (13.4 %) African American, 54 (15.7 %) Hispanic, 8 (2.3 %) Asian, 14 (4.1 %) two or more race/ethnic groups or other, and 5 (1.4 %) not reporting; (c) the five Hollingshead social classes [54] were represented [Class I and II (high), 279, 81.1 %; Class III–V, 65, 18.9 %]. There were 10 fathers who were the primary child-care providers who completed the study questionnaires; the children’s mothers were the respondents for the others. Children whose teachers completed the questionnaire were more likely to be White, v2(4, n = 339) = 19.70, p \ .001, Cohen’s w = 0.24, and to come from the higher two social classes, v2(1, n = 344) = 13.92, p \ .001, w = 0.20. The children whose teachers completed the forms did not differ from those whose teachers did not complete forms on child’s age, t(342) = 1.69, p = .09, Cohen’s d = 0.18, or gender, v2(1, N = 344) = 1.67, p = .20, w = 0.07. Missing data and item frequencies were examined, and less than 5 % of the data were missing from the sample for whom teacher data were available. Because imputation is generally less

123

Child Psychiatry Hum Dev

biased than listwise deletion [25], missing data were imputed using SPSS V15.0 Expectation Maximization methodology using maximum likelihood procedures. Measures For the parent study, a multi-informant approach was used that included direct observation of child behavior and parent–child interaction, as well as parent- and teacherreports of various measures as described below. In addition, multiple measures were used to assess some of the constructs (e.g., caregiver depression); for the present study, these were converted to z scores and summed to create composite measures. For each of the measures, a index of internal consistency is reported. Coefficient alpha is commonly reported as the index of internal consistency in psychological studies [26, 27]. Alpha is commonly used to measure internal consistency with scales that have multiple items ostensibly measuring the same construct, and the magnitude of coefficient alpha is affected by the number of items on the scale [26]. When scores for latent factor are estimated based on manifest indicators taken from the summary scores from a small number of different measures, coefficient alpha is likely to be small even if the cumulative number of items across those scales is large. In those instances, an index of composite reliability [28, 29] is preferable to coefficient alpha. We reported coefficient alpha when internal consistency was measured for a single scale, and composite reliabilities when more than one scale was used to measure the latent factor. There is no minimum, fixed level of coefficient alpha that is acceptable or satisfactory for internal consistency [26, 27]. Nunnally’s [30] recommendation of a minimum acceptable level of 0.7 is often cited, but Peterson [27] noted that Nunnally had changed that recommendation from 0.5 or 0.6 to 0.7 without explanation, and Peterson notes other recommendations ranging from 0.5 to 0.95 depending upon whether the research was preliminary, basic, or applied. Schmitt [26] notes that the acceptable level of coefficient alpha actually depends upon the test use, and that the main concern with the use of a test with low internal consistency is that it might attenuate the validity of the relationship between a predictor and outcome measure, i.e., the strength of the relationship may be underestimated or appear statistically non-significant when it is in fact significant. Schmitt notes that a measure with a reliability as low as 0.49 may still be useful. As will be seen, in the present study, four measures showed internal consistencies indices between 0.70 and 0.75, while one (NA) was as low as 0.62, and others were [0.75. Each of these measures showed significant relationships with other psychological variables included in the study that would

123

not have been expected if low internal consistency prevented identifying significant predictor-outcome relationships, and they showed psychologically meaningful results in other reports using these data [13]. For these reasons these variables were retained for analyses. Contextual Measures Socioeconomic Status (SES) Parent reports of employment and education were coded for SES as a continuous measure using the Hollingshead Four-Factor Index of Social Status [31]. Parents also provided demographic information on child age, gender, ethnic/racial background, etc. Life Stress A composite measure of life stress was created from the (a) Parenting Stress Index-Short Form (PSI-SF) [32]; (b) McCubbin Family Changes and Strains Scale [33]; and (c) Perceived Stress Scale [34]. The PSI-SF is a 36-item self-report measure of parental distress. The Perceived Stress Scale is a 14-item measure of the degree to which adults appraised life as stressful, unpredictable, and uncontrollable. The McCubbin Family Changes and Strains Scale measures life stress. The composite reliability of this measure was 0.80; alpha coefficients and composite reliabilities reported herein are for the present study. Higher scores reflect greater stress. Family Conflict Scores on the following three parent-rated measures were combined to create a ‘‘family conflict’’ composite score; (a) the conflict scale of the Family Environment Scale (FES) [35]; (b) the McCubbin Family Distress Index [36]; and (c) the McCubbin Family Problem-Solving/Communication Scales [37]. The composite reliability for the conflict measure was 0.71. Parent and Parenting Measures Parental Depression Parents completed the Beck Depression Inventory (BDI) [38] and the Center for Epidemiological Studies- Depression Scale (CES-D) [39]. The BDI is a self-report measure of clinical depression with high concurrent validity [40]. The CES-D is a self-report measure of the frequency of depressive symptoms, also correlates highly with other depression scales [39]. Coefficient alpha for the parental depression composite score was 0.70.

Child Psychiatry Hum Dev

Parent Support and Hostility The parent-completed Parent Behavior Inventory (PBI) [41] includes two factor-analytically derived subscales, Support/Engagement (a = 0.85) and Hostility/Coercion (a = 0.73). Parental Sensitivity/Scaffolding A semi-structured videotaped parent–child interaction paradigm, The Three Boxes task [42], was used to assess parental sensitivity/scaffolding. This measure includes three tasks, two of which are designed to be somewhat challenging for the child to complete without help, and a third that involves freeplay. Trained research assistants rated parental behaviors on a 7-point Likert scale (1 = very low, 2 = very high) for each of the following dimensions—supportive presence, respect for autonomy, quality of assistance, cognitive stimulation, confidence, and hostility (reverse scored). Factor analysis of these dimensions revealed a one-factor solution that was used as a composite measure of sensitivity/scaffolding (a = 0.81). Coders were trained to a criterion of 80 % reliability. Interrater reliabilities ranged from 0.80 for quality of assistance to 0.69 for maternal hostility, with the lower rating for the latter likely to have been affected by a low base-rate (mean reliability = 0.74). Child Measures Child Temperament Child temperament factors examined included NA, EC, IC, and sensory regulation (SR). Child NA Child NA was measured using a modified version of the parent-completed Children’s Behavior Questionnaire (CBQ) [43]. In studies examining the relationship between temperament and child behavior problems, particular attention must be made to reducing the effects of item contamination which could inflate the magnitude of the associations between measures of temperament and behavior problems [44, 45]. To reduce contamination, we followed procedures developed by Lengua et al. [46] using expert opinion and confirmatory factor analysis (CFA) to refine the temperament measures. CFA shows that the factor structure is the same for the full set of items and those retained after reducing item contamination [13]. For this study, the modified NA scale was a composite measure calculated by summing the 7 items from the CBQ NA subscales that the experts felt were measures of NA rather than psychopathology (a = 62.) As noted above, while a cutoff

of 0.70 is often used for assessing the adequacy of alpha, lower scores may be acceptable when the measure has other advantages [46], i.e., in this study, the expert-rated NA items were distinct from other temperament variables, and showed a good fit in the overall measurement model of risk factors for psychopathology [13]. Child EC The CBQ was also used to measure EC after reducing item contamination with measures of behavior problems and other temperament variables (e.g., sensory regulation). After eliminating contamination items from the CBQ, attentional focusing and IC scales were used to create a composite EC measure (a = 0.72). Child IC Whereas child EC is often defined as a ‘‘hot’’ measure of emotion regulation because it assesses the child’s ability to control emotions when emotionally aroused (e.g. frustrated), IC is typically considered a ‘‘cool’’ measure because it involves the inhibition of a prepotent response in an emotionally neutral problem solving task [47–49]. In this study, the Statue subtest from the Attention/Executive Function Domain of the Developmental Neuropsychological Assessment (NEPSY) [50] was used to assess IC. This measure assesses motor inhibition by asking the child to maintain a body position for 75-seconds while inhibiting responses to sound distracters (a = 0.91). Child SR Sensory regulation (SR), a child temperament variable which assesses the child’s reactivity to and processing of sensory input, was measured with the 38-item Short Sensory Profile (SSP; [51]. This parent questionnaire assesses threshold of response and reactivity across all sensory domains in children ages 3–10. Scores on the SSP differ for groups of children with and without sensory processing difficulties and are correlated with physiological measures of sensory processing [51]. Item reduction for this scale followed the same procedure as that used for the CBQ. The final expert-rated items on the index of SR in the present study measured the tactile, movement, visual/auditory sensitivity, taste/smell, and low energy/weakness scales of the SSP (a = 0.82). Attachment The Attachment Q-Sort (AQS) [52] is a continuous measure of attachment security. The AQS shows good

123

Child Psychiatry Hum Dev

convergent validity with the Strange Situation Paradigm [53]. The AQS was completed by trained research assistants; inter-rater reliability in the present study was 0.77 in a 20 % random-sample. Child Receptive Vocabulary The Peabody Picture Vocabulary Test (PPVT-L) [54] is an individually-administered measure of single-word receptive language skills. The PPVT correlates 0.88 with the Wechsler Intelligence Scale for Children verbal index [55]. Internal consistency is 0.94 and test–retest reliability is 0.89 [56]. ADHD Parents rated symptoms of ADHD on the Early Childhood Inventory (ECI), the preschool version of the Child Symptom Inventory (CSI): Parent Checklist for young children [57], while teachers completed the ECI-4: Teacher Checklist. In both scales, parents and teachers rate symptom occurrence on a 4-point scale (from ‘‘never’’ to ‘‘very often’’). Items on these checklists were derived from DSMIV diagnostic criteria. In this sample, coefficient alpha for parent-rated ADHD-Inattentive type (ADHD-I) was 0.88 and for teacher rated ADHD-I, 0.90. For parent-rated ADHD hyperactive-impulsive type (ADHD-HI), coefficient alpha was 0.88 and for teacher rated ADHD-HI, 0.90. Multiple studies have demonstrated that ADHD consists of a twofactor structure, with both ADHD-I and ADHD-HI present in young children [58, 59], so those factors were examined. Procedure In the pediatric practices, parents were approached individually in the waiting rooms by study staff. In the public schools, staff members approached parents at morning dropoff and school events that occurred during the enrollment period. Research assistants then scheduled a home visit, and questionnaires were mailed to interested families. At the home visit, research assistants administered study tasks (e.g., PPVT, Statue subtest, Q-Sort, Three Boxes Task) and additional questionnaires (including demographic questionnaire and ECI). After the visit, the ECI-Teacher form was mailed to the child’s teacher. Institutional Review Boards at the authors’ institutions approved the procedures; informed consent was obtained from a parent. Data Analysis In the preliminary analyses, Pearson correlations between predictors and source-specific outcomes were calculated using IBM SPSS (version 23). While these correlations

123

allow for examination of the relationships between each of the predictors and outcomes, the magnitude of the correlations may be increased by variance shared between predictors. In contrast, with hierarchical linear regression, the contributions of specific predictors to ADHD symptoms can be assessed. Because prior studies of teacher–parent rating differences for ADHD have not examined a wide variety of contextual factors simultaneously, the specific associations of various factors with ADHD-HI and ADHDI symptoms at school or at home are not entirely clear. For this reason, separate hierarchical linear regression analyses of teacher and parent-reported ADHD symptoms were examined using IBM SPSS (version 23). Because models of factors associated with ADHD have emphasized child characteristics, child variables (EC, IC, SR, PPVT, NA, and attachment security) were entered first, followed by parenting (support-engagement, hostile parenting, scaffolding), parent (caretaker depression), and contextual and demographic variables (stress, conflict, SES, race). Race as well as SES was included because the subsample including teacher reports differed from the total sample on those measures. The hierarchical approach and order of entry of blocks of variables was chosen for several reasons: (a) causal models of the neuroscience of ADHD stress the importance of multiple levels or domains of analysis influencing the behavioral manifestations of ADHD; (b) within such models [17, 18], child biological (including temperamental and self-regulatory factors) and cognitive (including verbal skills that might aid in self-regulation) factors are thought to be primary or core causal factors in manifestations of ADHD, with other environmental factors (e.g., family climate, broader social factors such as socioeconomic factors) being secondary; (c) a similar multi-domain model of symptoms of another externalizing disorder, oppositional defiant disorder (ODD), goes beyond the existing ADHD models by specifying a sequence of potential causal factors within the environmental domain described by Nigg and colleagues. This model specifies, in order, a sequence of proximal to distal factors as child domain, parenting domain, parent domain, and contextual domain factors associated with externalizing symptoms. Support for associations between and across this sequence of domains has been reported in both a cross-sectional and longitudinal cascade model for ODD [13]. Entry following the same proximal–distal sequence extends this multiple level of analysis to the study of ADHD symptoms in both school and home settings. Betas reported in the tables are those obtained at final entry. A multi-group approach comparing predictors of parent and teacher reports using a structural equation modeling approach to data analyses was considered but rejected because the sample size was not sufficient for testing appropriate models for teacher-reported symptoms.

Child Psychiatry Hum Dev

To determine if possible suppression effects could be operating, we examined the Pearson correlations to determine if any of the predictor variables were highly correlated [60, 61]. If any pair of variables was highly correlated, separate analyses were conducted that included only one variable from each pair. With 14 variables and a sample size of 344, power to detect a significant R2 of 0.40 was 0.90. The regression-based approach used included the use of composite variables. There is clear precedent for creating composite variables by converting individual variables into standard scores and combining them. This approach has been used in reports from the fields of medicine, epidemiology, and management [62–64], as well as a prior report by us [13]. This approach is warranted in constructing composite scales when the constituent items have different measurement response-scales and one wishes to weight the multiple items equally by converting raw scores to standardized scores before being summed [30]. Because of the large number of comparisons, corrections for multiple comparisons were made using a sequentiallyrejective Sidak procedure [65]. There are both advantages and disadvantages to using such corrections. On the one hand, their use has confirmatory value, because it reduces the likelihood that findings significant after multiple corrections are not due to chance (i.e., familywise Type I error rate). On the other hand, relying solely on findings that are significant after multiple corrections reduces statistical power and may thus lead to inappropriately rejecting new findings that may be important when a new area of investigation is being explored. In the present study, to be significant after multiple corrections, p values needed to be \.0001, a rather strict standard that increases Type II error. Because so little attention has been paid to factors that might contribute to differences in parent–teacher ratings of ADHD symptoms, we chose to consider factors that were not significant after correcting for multiple comparisons to be important trends and did discuss their implications. In addition to determining whether there were differences in the specific variables significantly associated with parent versus teacher reports, we also assessed whether the sizes of these differences were significant using structural equation modeling (SEM) with LISREL version 8.8 [66]. The goodness-of-fit statistics of two models were compared for each variable: (a) a baseline model in which the paths from the predictor to the parent and teacher report of ADHD were freely estimated; and (b) a nested, comparison model in which the paths from the predictor to the parent and teacher outcome was constrained to be equal to one another. Because the second model is nested within the first, a Chi square difference test can be used to assess the statistical significance of the difference in the fit of the two models [67]. A significant difference in Chi square value

(with difference in df) indicates that the two unstandardized path coefficients are significantly different from one another in magnitude [66, 68]. To reduce the number of comparisons of parent and teacher reports, we present the results for ADHD-I and ADHD-HI, but not ADHD-C. Studies have confirmed the utility of a two-factor model of ADHD with both young children [58] and older children [6].

Results Descriptive Statistics and Correlations Means and standard deviations for the ADHD symptom reports were: parent-rated ADHD-I, M = 5.61, SD = 3.92, skewness index (SI) = 1.33, kurtosis index (KI) = 3.66; parent-rated ADHD-HI, M = 5.86, SD = 3.97, SI = 1.25, KI = 3.01; teacher-rated ADHD-I, M = 4.60, SD = 4.37 SI = 1.26, KI = 1.62; teacher-rated ADHD-HI, M = 4.93, SD = 4.52, SI = 1.23, KI = 1.50. Table 1 includes the correlations among predictor and outcomes variables. The correlations between parent and teacher ratings for each type of ADHD symptom were very low (ADHD-I, r = 0.22. p \ .001; ADHD-HI, r = 0.26. p \ .001) but, because of the large sample size, statistically significant. Shared variance between parent and teacher report did not exceed 7.3 % for any type of ADHD symptoms. For parent-rated ADHD symptoms, each of the predictor variables was significantly associated with ADHD-I and ADHD-HI symptoms, although the magnitude of these correlations was low. Fewer of the predictor variables were significantly correlated with ADHD symptoms for teacher ratings; for all three types of ADHD symptoms, poorer EC, IC, verbal skills, more parental hostility, poorer parental scaffolding skills, poorer child attachment security, more stress, and lower SES were associated with more ADHD symptoms of each type, while SR, NA, SE, caretaker depression, and conflict were not. Directions of effects were the same for parent and teacher ratings of ADHD. Similarly, correlations among predictor variables were low to moderate. Given that the EC measure included items assessing IC that were parent-observed at home, it is interesting that the correlation between EC and IC was so low (r = 0.16), indicating that the two forms of self-regulation were relatively independent of one another and best construed as correlated but independent constructs. Similarly, since EC included an attention component but the overlap with ADHD symptoms was eliminated based on expert ratings, the correlations between EC and both parent and teacher rating of ADHD-I and ADHD-HI were only low to moderate. Although variable inflation factors were

123

123 0.28 0.18

3. Sensory regulation (SR)

4. Verbal (PPVT)

-0.39 -0.2 -0.2

15. Parent ADHD hyperactive (ADHD-HI-P)

16. Teacher ADHD inattentive (ADHD-I-T)

17. Teacher ADHD hyperactive (ADHD-HI-T)

-0.26

-0.24

-0.2

-0.18

-0.03

-0.04

-0.3

-0.3

0.14

-0.3

-0.3

-0.34

0.19

0.1

-0.2

0.28

-0.25

0.17

1

3. SR

-0.18

-0.2

-0.16

-0.15

0.5

-0.12

-0.24

-0.25

0.29

0.38

-0.13

0.23

0.04

1

4. PPVT

-0.2

-0.02

0.25

0.25

-0.03

0.24

0.21

0.21

-0.2

-0.08

0.12

-0.22

1

5. NA

-0.06

-0.05

-0.17

-0.19

0.26

-0.04

-0.27

-0.3

0.18

0.23

-0.09

1

6. SE

0.17

0.16

0.23

0.21

-0.04

0.28

0.23

0.29

-0.19

-0.12

1

7. Host

-0.24

-0.26

-0.21

-0.22

0.4

-0.05

-0.26

-0.26

0.25

1

8. Scaff

-0.14

-0.14

-0.28

-0.27

0.16

-0.29

-0.23

-0.27

1

9. Att

0.05

0.05

0.3

0.28

-0.32

0.33

0.66

1

10. C-dep

0.1

0.1

0.31

0.31

-0.26

0.32

1

11. Stress

0.06

0.06

0.33

0.32

-0.11

1

12. Conf

-0.12

-0.13

-0.13

-0.12

1

13. SES

0.23

0.22

0.98

1

14. ADHD-IP

0.26

0.25

1

15. ADHDHI-P

N = 796. 0.09 B |r| \ 0.12, is significant at two-tailed p \ .05; 0.13 B |r| B 0.15 is significant at two-tailed p \ .01; 0.16 B |r| is significant at two-tailed p \ .001

-0.39

0.05

-0.05

-0.33 0.13

-0.04

-0.13

0.27

0.23

-0.07

0.01

-0.06

0.13

0.11

1

2. IC

-0.22

14. Parent ADHD inattentive (ADHD-I-P)

ADHD

13. Socioeconomic status (SES)

12. Conflict (Conf)

11. Stress

Contextual/demographic

-0.2

0.31

9. Attachment (Att) Parent

10. Caretaker depression (C-dep)

0.22

8. Scaffolding skills (Scaff)

0.32 -0.19

7. Caretaker hostility (Host)

6. Supportive-engagement (SE)

Parenting

-0.48

0.16

2. Inhibitory control (IC)

5. Negative affect (NA)

1

1. Effortful control (EC)

Child

1. EC

Table 1 Correlations between risk and outcome variables

0.99

1

16. ADHD-IT

Child Psychiatry Hum Dev

Child Psychiatry Hum Dev

within acceptable limits (\10.0), indicating that multicollinearity was not excessive, one correlation, stress with caretaker depression was high (r = 0.66), which raised the possibility that this strong correlation could lead to suppression effects when both variables were included in the same regression analyses even if multicollinearity was not excessive.

predictors that is associated with a decrease in symptoms of ADHD-I [61]. After correcting for multiple comparisons, none of these factors were confirmed as significant predictors of teacher-rated ADHD-I symptoms.

Analyses of Inattention Symptoms

Parent-Rated ADHD-HI Symptoms

Parent-Rated ADHD-I Symptoms

Table 3 presents the regression analyses for parent and teacher ratings of ADHD-HI. For parent-rated ADHD-HI symptoms, the overall R2 of 0.28 was significant, F = 9.12, p \ .001, and the change in R2 was significant for each block of predictor variables. Before correcting for multiple comparisons, poorer child EC and IC, more difficulties with SR, and greater family stress and conflict were significant predictors of parent-rated ADHD-HI symptoms. There were no indications of suppression effects involving stress and caretaker depression. After correction for multiple comparisons, only parent-rated EC was confirmed as a significant predictor of parent-rated ADHD-HI symptoms.

Table 2 presents the results of regression analyses predicting parent and teacher ratings of ADHD-I. For parentrated ADHD-I symptoms, the overall R2 of 0.27 was significant, F = 8.50, p \ .001. R2 change was significant for each block of variables—child, parent, parenting, and contextual. Before correcting for multiple comparisons, poorer child EC, more difficulties with SR, and greater family stress and conflict were significant predictors of parent-rated ADHD-I symptoms. Subsequently, separate regression analyses were conducted that included stress but not caretaker depression as a predictor and vice versa, because of the possibility of suppression effects. In these analyses, stress continued to be associated with parentrated ADHD-I symptoms and caretaker depression did not. After correction for multiple comparisons, only child EC remained a significant predictor of parent-rated ADHD-I symptoms, while the other previously significant relationships are best considered trends. Teacher-Rated ADHD-I Symptoms For teacher-rated ADHD-I, the overall R2 of 0.16 was modest but significant, F = 4.34, p \ .001. The blocks of child and parenting variables were significantly associated with teacher-rated ADHD-I, but the inclusion of parent and contextual factors did not improve the prediction of teacher-rated ADHD-I. Before correcting for multiple comparisons, poorer child EC and IC, lower levels of NA, greater parental hostility, and poorer scaffolding skills were significant predictors of teacher-rated ADHD symptoms. None of the parent or contextual factors was associated with teacher-rated ADHD. Indications of suppression effects between stress and caretaker depression were not found. Because the correlations of NA with each of the ADHD outcome measures were not significant, the increased magnitude of the relationship of NA with ADHD-I when multiple predictors were present indicates a positive suppression effect (with an amplified magnitude of the beta coefficient but not a change in direction of the relationship between NA and ADHD-I). That is, there is some aspect of NA that is not associated with other

Analyses of Hyperactivity Symptoms

Teacher-Rated ADHD-HI Symptoms For teacher-rated ADHD-HI, the overall R2 of 0.16 was modest but significant, F = 4.53, p \ .001. The blocks of child and parenting variables were significantly associated with teacher-rated ADHD-I, but the inclusion of parent and contextual factors did not improve the prediction of teacher-rated ADHD-HI. Before correcting for multiple comparisons, poorer child EC and IC, greater parental hostility and poorer scaffolding skills were significant predictors of teacher-rated ADHD-HI symptoms. None of the parent or contextual factors was associated with teacher-rated ADHD-HI. There were no indications of suppression effects involving stress and caretaker depression. After correcting for multiple comparisons, IC was confirmed as a predictor of teacher-rated ADHD-HI. Comparison of Parent and Teacher Ratings for ADHD-I: v2 Difference Tests Results of v2 difference test were similar for ADHD-I and ADHD-HI. That is, for both types of symptoms, child factors of EC and SR, and contextual factor of stress and conflict, contributed more to parent-ratings of ADHD-I and ADHD-HI than to teacher-ratings of those same types of symptoms. Only two factors contributed more to teacherthan to parent-rated ADHD-I, NA and caretaker depression. Suppression effects likely affected the results for NA and, while the effect for caretaker depression was greater

123

Child Psychiatry Hum Dev Table 2 Regression analyses and comparisons of magnitudes of path coefficients: Inattention symptoms Block

Variable

Parent ADHD-I R2/R2 change Child

Teacher ADHD-I Beta

0.21***

R2/R2 change

Beta

0.12***

EC

-0.39***

-0.34**

3.88*

IC SR

-0.32 -0.15*

-0.77*** 0.03

1.07 7.30**

NA

0.03

-0.21*

8.14**

PPVT

0.01

-0.02

0.55

Parenting

0.02* SE

0.03* -0.02

Hostile

0.07

0.06

1.91

0.24*

0.11

Scaffold

-0.12

-0.27**

0.59

Attachment

-0.24

-0.01

1.36

-0.28

5.77**

Parent

0.01* Caretaker depression (with stress)

0.00 0.00

Caretaker depression (without stress) Contextual

0.17 0.02*

-0.17 0.00

Stress (with depression)

0.24*

0.15

Stress (without depression)

0.24*

0.04

Conflict SES

0.29* 0.12

-0.02 -0.20

7.16** 0.06

-0.24

0.11

Race (White versus minority) Total

v2 difference v2 value (df = 1)

Regression Analyses

0.37 0.27***

6.61**

0.16***

* p \ .05; ** p \ .01; *** p \ .001. Italicized betas are not significant after correcting for multiple comparisons. The v2 difference test compares the magnitudes of the unstandardized path coefficients

for teachers, it was not significant for either parent- or teacher-rated ADHD in the regression analyses.

Discussion Although DSM-IV and, now, DSM-5, state that ADHD can only be diagnosed when symptoms are present in more than one situation, DSM-5 provides no justification for requiring ADHD to be a cross-situational disorder. A large number of previous studies show that there is poor agreement between teacher and parent reports of all types of ADHD symptoms [5–8]. The results of the present study are consistent with those of these prior studies in finding only low correlations between parent and teacher reports of symptoms of ADHD-I, and ADHD-HI; in the present study, the shared variance between parent and teacher ratings did not exceed 7.3 % for any type of ADHD. Interestingly, shared variance between parent and teacher ratings of ODD in this same sample [13] was only slightly

123

less (2.9 % for ODD). Results from that study, and others [10] have been consistent in showing that teachers and parents differ in ratings of ODD. Such findings led the authors of DSM-5 to note that ODD symptoms may be source-specific. Yet, ADHD, which has similarly low agreement across source-specific raters, has been viewed as a disorder that is cross-situational and DSM demands cross-situational consistency to justify the diagnosis. The rationale for considering ADHD, with low inter-rater agreement for parents and teachers, to be a cross-situational disorder and another, ODD, with low parent–teacher agreement, to be source-specific is unclear. If low correlations between parents and teachers justify considering ODD to be source-specific, it seems that there is similar justification for considering ADHD to be source-specific. As such, greater attention needs to be paid to the factors that might contribute to differences in the manifestations of such symptoms across sources or situations. Presently, models exist that suggest multiple factors may contribute to the manifestation of ADHD symptoms, but no model has

Child Psychiatry Hum Dev Table 3 Regression analyses and comparisons of magnitudes of path coefficients: hyperactive-impulsive symptoms Block

Variable

v2 difference

Regression analyses Parent ADHD-HII R2/R2 change

Child

Beta

0.22***

R2/R2 change

Beta

v2 value (df = 1)

0.12***

EC

-0.39***

-0.34**

4.87*

IC SR

-0.40* -0.14*

-0.92*** 0.06

1.17 7.28**

NA

0.04

-0.21

7.26**

PPVT

0.01

-0.02

0.21

Parenting

0.03*

0.03*

SE

0.01

0.03

1.21

Hostile

0.11

0.28**

0.15

Scaffold

-0.09

-0.24*

0.53

Attachment

-0.29

0.00

1.64

Parent

0.01* Caretaker depression (with stress)

0.00 0.03

Caretaker depression (without stress) Contextual

Total

Teacher ADHD-HII

-0.32

0.20 0.02*

5.97*

-0.18 0.01

Stress (with depression)

0.23*

0.19

Conflict

0.32*

-0.02

7.53

SES Race (White v minority)

0.11 0.37

-0.21 -0.22

0.02 0.18

0.28***

4.34*

0.16***

* p \ .05; ** p \ 0.01; *** p \ .001. Italicized betas are not significant after correcting for multiple comparisons. The v2 difference test compares the magnitudes of the unstandardized path coefficients

been presented that attempted to explain the factors contributing to differential ratings between parents and teachers. For that reason, we extended a multi-domain model of factors associated with another externalizing disorder, ODD, which had been used to study sourcespecific differences for symptoms of that disorder, to examine source-specificity for ADHD symptoms. Extending this model to studying ADHD source-specificity seemed justified for two reasons: (a) both ODD and ADHD are similar in being externalizing disorders and comorbidity between the two types of symptoms often occurs [69]; and (b) recent studies have called for further research into the role that a range of factors across domains—child, parenting, parent, and contextual—might play in the expression of symptoms of ADHD. The present study used three different, complementary approaches to (or levels of) data analysis. The first approach examined bivariate, Pearson correlations between predictors and teacher- and parent-rated ADHD. This approach assesses the relationship between each predictor and outcome, aids in comparing the results of the present study with prior studies, and helps determine whether or not suppression effects are present in the regression

analysis. Because the predictor variables tend to be intercorrelated, Pearson correlations may inflate the magnitude of the relationships between predictors and outcomes and should therefore be interpreted with caution. At the second level, hierarchical regression analyses were used to determine which predictors had a specific, significant association with ADHD that was independent of other predictors for teacher- as well as parent-rated ADHD. These regression analyses also assess the amount of variance the predictors explain the outcome measures. A limitation of the regression approach is that it does not enable direct comparison of the differences each predictor might make to the teacher and parent ratings of ADHD. Comparing the predictors of parent- and teacher-rated ADHD required the third level of analysis using SEM to contrast the magnitude of the unstandardized regression coefficients for each predictor with parent- and teacher-rated ADHD symptoms simultaneously. There were both similarities and differences in which predictors were associated with parent versus teacher ratings of ADHD-I. For both parent and teacher ratings, poorer EC was associated with more ADHD-I symptoms. Poorer SR was associated with more parent-rated but not

123

Child Psychiatry Hum Dev

teacher-rated ADHD-I symptoms, while the opposite pattern was found for NA—that is, NA was more strongly related to teacher-rated than to parent-rated ADHD-I. For EC, SR, and NA, differences in the magnitude of the relationships with parent- and teacher-rated ADHD symptoms were great enough to be statistically significant. In contrast, poorer IC was significantly associated with more teacher-rated symptoms but not significantly associated with parent-rated symptoms of ADHD, but the difference between the associations was not large enough to be statistically significant on v2 difference tests. This pattern of results suggests two conclusions about situational demands that might affect the expression of ADHD-I differentially at home and school: (a) the home environment may contain more situations than school in which sensory demands on the ability to focus or redirect attention are placed on the child and are difficult to manage. Such demands likely include home-based tasks such as dressing, eating, and hygiene, and might also be a function of greater noise and chaos at home than at school. Thus, these behaviors may be more likely to be observed by parents than teachers; (b) the type of IC that the Statue task involves has been described as ‘‘cool’’ rather than ‘‘hot’’ IC, i.e. the child must inhibit a prepotent response in a situation in which emotional control is not involved [47]. The association of IC as measured by statue with ADHD-I symptoms at school rather than home suggests that, in contrast to home, the school situation places more demands on the child for cool IC. Such demands might include refraining from raising your hand when it is not time to do so or stopping to think and look at the addition or subtraction sign on a math problem before completing it. If the child with symptoms of ADHD-I has difficulty with these types of tasks, then they may be more likely to manifest symptoms at school, and hence in teacher’s reports, than at home. Overall, however, the difference in the demands across situations is small as reflected in the v2 difference test. It is noteworthy that the pattern of association for parenting and contextual/parent behaviors differed with ADHD-I symptoms observed by teachers and parents in the regression analyses. That is, none of the parenting factors were independently associated with parent-rated ADHD-I symptoms, while higher levels of parent hostility and poorer parental scaffolding skills were associated with teacher-rated ADHD-I symptoms in the classroom. Overall, the differences in effects were small as indicated by nonsignificant v2 difference tests. In contrast, stress and conflict were associated with higher levels of parent-reported ADHD-I symptoms, but these contextual factors were not associated with ADHD-I symptoms reported by teachers. These results suggest the following: (a) the

123

significant block effect for parenting, and the correlations of each parenting variable with parent-rated ADHD-I symptoms indicates there is a small but significant cumulative statistical association between parenting and ADHDI at home, but no single aspect of parenting had an independent association with ADHD-I as observed by parents; (b) parental hostility seems to have an independent association with on ADHD-I as observed by teachers, with higher parental hostility associated with higher levels of ADHD-I behavior observed at school; (c) scaffolding skills exhibited by parents are associated with lower levels of ADHD-I observed by teachers. Although causal affects cannot be determined in an essentially correlational, crosssectional design, this raises the possibility that parental scaffolding skills may generalize to the classroom and be associated with better child functioning on tasks requiring cool IC skills.; (d) contextual factors have a stronger association with ADHD-I symptoms exhibited at home, and little to no association with such symptoms observed at school. That is, the effects of family stress, conflict, and SES seem to be unrelated to ADHD-I exhibited in school, after taking parenting and child factors into consideration. The results for ADHD-HI were similar to ADHD-I for parent-rated symptoms, except that IC was significantly associated with parent-rated ADHD-HI but not ADHD-I. The parent and teacher results differed for ADHD-HI in that NA was unrelated to teacher ratings of ADHD-HI. This suggests that child NA is either more freely expressed at home or, perhaps, that ADHD-HI symptoms cause enough parent–child conflict that higher levels of child NA develop at home over time. It may be that as children with ADHD-HI get older, the demands of school increase, and the school environment is less tolerant of their behavior, such that a similar trend would be seen in teacher ratings with stronger associations between ADHD-HI and child NA at older ages. Because a prior study [13] examined the same set of predictors for source-specificity for ODD, it is possible to compare the factors associated with teacher and parent ratings of ODD with the factors associated with teacher and parent ratings of ADHD-I and ADHD-HI. Multi-domain factors associated with parent-rated ODD and ADHD-I symptoms were very similar: among the child factors, for both parent-rated ODD and ADHD-I symptoms, poorer EC and SR were associated with more symptoms and made an independent contribution to those symptoms in the regression analyses. For contextual factors, conflict was associated with increased ODD and ADHD parent-rated symptoms. For parent factors, caretaker depression did not have a direct effect on either ADHD or ODD symptoms. In addition, some differences emerged with regard to parenting factors associated with ADHD and ODD. For

Child Psychiatry Hum Dev

parenting, none of the parenting factors made an independent contribution to ADHD symptoms, but hostility made an independent contribution to greater ODD symptoms. For teacher-rated symptoms, there were important differences in factors associated with ADHD-I and ODD. For child factors, only EC was associated with teacher-rated ODD symptoms; in contrast, along with EC, IC and NA were also associated with teacher-rated ADHD-I. Thus, both cool and hot IC were associated with teacher ratings in the academic setting, but only ‘‘hot’’ EC was associated with ODD at home. It is also of interest that more parenting factors made an independent contribution to ADHD at school than to ODD. That is, poorer scaffolding skill was associated with more teacher-rated ODD, but greater hostility as well as scaffolding skills had an independent association with ADHD as rated by teachers. Contextual factors did not make an independent contribution to either teacher-rated ODD or ADHD. The present findings differ somewhat from those which had been hypothesized. We hypothesized that the relationship between contextual, parent, and parenting factors would follow a similar pattern for ADHD as it did with ODD. That is, we anticipated that these factors would be associated with ADHD symptoms at home but not necessarily in school. This was true for child and contextual factors and parent-rated ADHD, but not for parenting factors. We also hypothesized that the cognitive factor of IC would show a stronger relationship with teacher-rated ADHD symptoms than with parent-rated symptoms because of a greater need for children to exhibit ‘‘cool’’ executive functioning control at school. Contrary to expectations, while IC was significantly related to ADHD-HI symptoms at school and not at home, the difference was not significant. These findings also have implications for models of factors associated with ADHD. Coghill et al. [17] argued that models of ADHD need to include within-child biological, cognitive, and behavioral, domains as well as environmental domains external to the child that are associated with ADHD symptoms. The present study’s findings suggest that the domains associated with ADHD symptoms may vary by the contextual environment, with factors differing on whether the ADHD-symptoms are displayed at home or at school, where they are likely to be observed by parents versus teachers, respectively. Furthermore, Coghill et al. suggest that any model must account for heterogeneity among individuals in terms of the degree that cognitive and other risk factors contributed to ADHD symptoms. The results of this study suggest that the context in which the ADHD symptoms are displayed— teacher- versus parent-observed—contributes to that heterogeneity, and that the role of factors within each of the domains will vary by observed situation.

The present study has several important limitations. First, this study employed a cross-sectional design, so significant findings reflect associations among variables that may suggest the possibility of causal relationships to be explored in future research, but conclusions about causality associated with significant findings are not warranted from these findings per se. Second, the study design, which recruited preschoolers from over 350 different preschool settings to achieve the needed sample size and participant diversity, made it impossible to examine differences in the classrooms and teachers. These contextual variables may contribute to differences in child behavior across settings and warrant greater attention from researchers. Third, because of the large number of statistical analyses in this study, the results need replication. It is important that additional multi-domain studies be conducted that include a wide range of possible predictors, in order to reduce omitted variable effects and estimate the independent contributions of contextual and child factors to ADHD symptoms and their source-specific manifestations. Fourth, further studies are needed to determine the degree to which common method variance (CMV) may have affected the findings. CMV effects may have been present when parents provided ratings of some predictors as well as the ADHD symptom outcome variables. Clearly, CMV effects cannot account for all study findings in which parents rated ADHD symptoms because the parent-rated predictors showed similar results when teachers rated outcomes for a number of parent-rated predictors, and in some instances the results for parent-rated parenting variables showed stronger associations with teacher-reported ADHD symptoms. Nonetheless, CMV effects may have influences some of the findings. Finally, as noted earlier, informant discrepancies are sometimes attributed to measurement error rather than actual cross-situational differences in the child’s behavior [70]. It is also possible that other characteristics of the raters could affect their perceptions—and, hence, their ratings—of the child’s behavior. Perhaps the best studied and most often cited rater characteristic that might affect perceptions of the child’s behavior is maternal depression. The depression-distortion hypothesis suggests that maternal depression biases mother’s reports and inflates ratings of child behavior problems [71] While the belief that maternal depression inflates reports of child problem behaviors has been based on comparisons of ratings of depressed mothers and teachers or other informants [72, 73], there is little evidence that rater differences result from distortion rather than genuine source- or situation- differences in behavior [71]. In the present study, there is little evidence that parental depression distorted results because (a) parental depression was not significantly associated with ADHD-I or ADHD-HI in the regression analyses; (b) because parental depression was included in those analyses, the effects of other variables on ADHD-I or

123

Child Psychiatry Hum Dev

ADHD-HI symptoms was measured independently of parental depression, i.e., the effects of parental depression was ‘‘controlled,’’ so the significant relationships between other variables and ADHD symptoms could not be attributed to parental depression; (c) when the differences between teacher and parent ratings was examined, parental depression was associated with higher teacher ratings of ADHD symptoms. If depression-distortion were operating, one would have expected the opposite pattern, with parental depression leading to higher estimates of ADHD symptoms by parents. That does not mean, of course, that other parent or teacher characteristics may not affect informant differences in their observations, and this would be important area for future studies to address. Nevertheless, this study has important implications both for our theoretical understanding of ADHD and our practices regarding diagnosis of this disorder. From a theoretical standpoint, this study highlighted the extent to which ADHD, often considered a ‘‘biological’’ disorder, actually manifests differently depending upon a set of psychosocial factors and the nature of specific task demands. As such, it is critical to identify these factors over time so that interventions to address these symptoms in children can be enhanced by appropriate psychosocial interventions along with the current ‘‘best practice’’ of pharmaceutical intervention. Also, this study has practical implications for the diagnosis of ADHD in children. If, as this study and others have suggested, there is little agreement between parents and teachers in reporting ADHD symptoms, then perhaps it is time to re-consider whether the mandate for report of symptoms in two or more situations should continue to be required as the gold standard for ADHD diagnosis. Summary While DSM-5 indicates that ADHD, by definition, is a disorder whose symptoms must be present in two or more settings, such as home and school, many studies indicate that agreement between parent and teacher ratings of symptoms of ADHD is poor. Before such differences are attributed simply to measurement error, greater attention is needed to understanding the factors that might contribute to such differences. Using a community sample of 4-year-old children, the present study examined the contextual, parent, parenting, and child factors associated with ratings of ADHD symptoms provided by parents and teachers. Predictors of parent and teacher ADHD symptoms from those four domains were obtained via parent reports as well as observations of child behavior and parent–child interaction. A large number of factors were significantly associated with parent and/or teacher ratings. After correcting for multiple comparisons, however, EC was a significant predictor of parent-rated symptoms of inattentive symptoms of

123

ADHD, while none of the multi-domain factors predicted teacher-rated symptoms of ADHD. For parent-rated ADHD-HI symptoms, only parent-rated EC was a significant predictor after correcting for multiple comparisons; in contrast, IC was confirmed as a predictor of teacher-rated ADHD-HI. When comparing the relative differences in the contributions of the multi-domain factors to parent and teacher ratings, v2 difference tests indicated that child factors of EC and SR, and contextual factor of stress and conflict, contributed more to parent-ratings of ADHD-I and ADHD-HI than to teacher-ratings of those same types of symptoms. Two factors contributed more to teacher- than to parent-rated ADHD-I, NA and caretaker depression. Overall, the results indicate there are differences in factors associated with ADHD symptoms at home and school, and have implications for models of ADHD. Acknowledgments This research was supported by National Institute of Mental Health Grant MH 063665.

References 1. American Psychiatric Association (2013) Diagnostic and statistical manual of mental disorders, 5th edn. Arlington, VA 2. American Academy of Pediatrics (2000) Clinical practice guideline: diagnosis and evaluation of the child with attentiondeficit/hyperactivity disorder. Pediatrics 2000(105):1158–1170 3. Subcommittee on Attention-Deficit/Hyperactivity Disorder of the Steering Committee on Quality Improvement and Management (2011) ADHD: clinical practice guideline for the diagnosis, evaluation, and treatment of attention-deficit/hyperactivity disorder in children and adolescents. Pediatrics 128:1007–1022 4. American Academy of Child and Adolescent Psychiatry (2007) Practice parameter for the assessment and treatment of children and adolescents with attention-deficit hyperactivity disorder. J Am Acad Child Adolesc Psychiatry 46:894–921 5. Antrop I, Roeyers H, Oosterlaan J, Oost Van (2002) Agreement between parent and teacher rating of disruptve behavior disorders in children with clinically diagnosed ADHD. J Psychopathol Behav Assess 24:67–73 6. Narad ME, Garner AA, Peugh JL, Tamm L, Antonini TN, Kingery KM et al (2015) Parent–teacher agreement on ADHD symptoms across development. Psychol Assess 27:239–248 (Epub ahead of print) 7. Lavigne JV, Dulcan MK, LeBailly SA, Binns HJ (2012) Can parent reports serve as a proxy for teacher ratings in medication management of attention-deficit hyperactivity disorder? J Dev Behav Pediatr 33:336–342 8. Wolraich ML, Lambert EW, Bickman L, Simmons T, Doffing MA, Worley KA (2004) Assessing the impact of parent and teacher agreement on diagnosing attention-deficit hyperactivity disorder. J Dev Behav Pediatr 25:41–47 9. Mitsis EM, McKay KE, Schulz K, Newcorn JH, Halperin JM (2000) Parent–teacher concordance for DSM-IV attention-deficit/ hyperactivity disorder in a clinic-referred sample. J Am Acad Child Adolesc Psychiatry 39:308–313 10. Drabick DAG, Gadow KD, Loney J (2007) Source-specific oppositional defiant disorder: comorbidity and risk factors in referred elementary schoolboys. J Am Acad Child Adolesc Psychiatry 46:92–101

Child Psychiatry Hum Dev 11. Munkvold L, Lundervold A, Lie SA, Manger T (2009) Should there be separate parent and teacher-based categories of ODD? evidence from a general population. J Child Psychol Psychiatry 50:1264–1272 12. Korsch F, Petermann F (2015) Agreement between parents and teachers on preschool children’s behavior in a clinical sample with externalizing behavior problems. Child Psychiatry Hum Dev 45:617–627 13. Lavigne JV, Dahl KP, Gouze KR, LeBailly SA, Hopkins J (2015) Multi-domain predictors of oppositional defiant disorder symptoms in preschool children: cross-informant differences. Child Psychiatry Hum Dev 46:308–319 14. Drabick DAG, Bubier JL, Chen D, Price J, Lanza H (2011) Source-specific oppositional defiant disorder among inner-city children: prospective prediction and moderation. J Clin Child Adolesc Psychol 40:23–35 15. Deault LC (2010) A systematic review of parenting in relation to the development of comorbidities and functional impairments in children with attention-deficit/hyperactivitiy disorder (ADHD). Child Psychiatry Hum Dev 41:168–192 16. Sergeant JA, Geurts H, Huijbregts SC, Scheres A, Oosterlaan J (2003) The top and the bottom of ADHD: a neuropsychological perspective. Neurosci Biobehav Rev 27:583–592 17. Coghill D, Nigg JT, Rothenerger A, Sonuga-Barke E, Tannock R (2005) Whither causal models in the neuroscience of ADHD? Dev Sci 8:105–114 18. Nigg JT, Goldsmith HH, Sachek J (2004) Temperament and attention deficit hyperactivity disorder: the development of a multiple pathway model. J Clin Child Adolesc Psychol 33:42–53 19. Johnston C, Mash EJ (2001) Families of children with attentiondeficit/hyperactivity disorder: review and recommendations for future research. Clin Child Fam Psychol Rev 4:183–207 20. Sonuga-Barke EJS, Auerbach J, Campbell SB, Daley D, Thompson M (2005) Varieties of preschool hyperactivity: multiple pathways from risk to disorder. Dev Sci 8:141–150 21. Wilens TE, Biederman J, Brown SS, Tanguay S, Monuteaux MC, Blake C et al (2002) Psychiatric comorbidity and functioning in clinically referred preschool children and school-age youths with ADHD. J Am Acad Child Adolesc Psychiatry 41:262–268 22. Egger HE, Kondo D, Angold A (2006) The epidemiology and diagnostic issues in preschool attention-deficit/hyperactivity disorder. Infants Young Child 19:109–122 23. Gadow KD, Sprafkin J, Nolan EE (2001) DSM-IV symptoms in community and clinic preschool children. J Am Acad Child Adolesc Psychiatry 40:1383–1392 24. Riddle MA, Yershova K, Lazzareto D, Paykina N, Yenokyan G, Greenhill L et al (2013) The preschool attention-deficit/hyperactivity disorder treatment study (PATS) 6-year follow-up. J Am Acad Child Adolesc Psychiatry 52:264–278 25. Graham JW (2009) Missing data analysis: making it work in the real world. Ann Rev Psychol 60:549–576 26. Schmitt N (1996) Uses and abuses of coefficient alpha. Psychol Assess 8:350–353 27. Peterson RA (1994) A meta-analysis of Cronbach’s coefficient alpha. J Consum Res 21:381–391 28. Mosier CI (1943) On the reliability of a weighted composite. Psychometrika 8:161–168 29. Evans LD (1996) A two-score composite program for combining standard scores. Behav Res Methods Instrum Comput 28:209–213 30. Nunnally J, Bernstein I (1994) Psychometric theory, 3rd edn. McGraw-Hill, New York 31. Hollingshead AB (1975) Four-factor index of social position. Yale University Department of Sociology, New Haven 32. Abidin RR (1995) Manual for the parenting stress index. Psychological Assessment Resources, Odessa

33. McCubbin HI, Patterson J (1996) Family member well-being. In: McCubbin HI, Thompson A, McCubbin MA (eds) Resiliency, coping and adaptation—inventories for research and practice. University of Wisconsin, Madison p, pp 735–782 34. Cohen ST, Kamarck T, Mermelstein R (1983) A global measure of perceived stress. J Health Soc Behav 24:385–396 35. Moos RH, Moos BS (1981) Family environment scale manual. Consulting Psychologists Press, Palo Alto 36. McCubbin HI, Thompson A, Elver KM (1996) Family distress index. In: McCubbin HI, Thompson A, McCubbin MA (eds) Resiliency, coping and adaptation—inventories for research and practice. University of Wisconsin, Madison, pp 783–788 37. McCubbin MA, McCubbin HI, Thompson A (1996) Family problem solving and communication. In: McCubbin HI, Thompson A, McCubbin MA (eds) Resiliency, coping and adaptation—inventories for research and practice. University of Wisconsin, Madison, pp 639–686 38. Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J (1961) An inventory for measuring depression. Arch Gen Psychiatry 4:561–571 39. Radloff LA (1977) The CES-D Scale: a self-report depression scale for research in the general population. Appl Psychol Meas 1:385–401 40. Beck AT, Steer RA, Garbin MG (1988) Psychometric properties of the Beck Depression Inventory: twenty-five years of evaluation. Clin Psychol Rev 8:77–100 41. Lovejoy MC, Weis R, O’Hare E, Rubin EC (1999) Development and initial validation of the Parent Behavior Inventory. Psychol Assess 11:1–12 42. NICHD Early Childhood Research Network (1999) Child care and mother–child interaction in the first three years of life. Dev Psychol 35:1399–1413 43. Rothbart MK, Ahadi SA, Hershey KL, Fisher P (2001) Investigations of temperament at three to seven years: the Children’s Behavior Questionnaire. Child Dev 72:1394–1408 44. Eisenberg N, Valiente C, Spinrad TL, Cumberland A, Liew J, Reiser M et al (2009) Longitudinal relations of children’s effortful control, impulsivity, and negative emotionality to their externalizing, internalizing, and co-occurring behavior problems. Dev Psychol 45:988–1008 45. Lemery KS, Essex MJ, Smider MA (2002) Revealing the relation between temperament and behavior problem symptoms by eliminating measurement confounding: expert ratings and factor analyses. Child Dev 73:867–882 46. Lengua LJ, West SG, Sandler IN (1998) Temperament as a predictor of symptomatology in children: addressing contamination of measures. Child Dev 69:164–181 47. Nigg JT (2006) Temperament and developmental psychopathology. J Child Psychol Psychiatry 47:395–422 48. Hongwansihkul D, Happaney KR, Lee WSC, Zelazo PD (2005) Assessment of hot and cool executive fuction in young children: age-related changes and individual differences. Dev Neuropsychol 28:617–644 49. Zelazo PD, Carlson SM (2012) Hot and cool executive function in childhood and adolescence: development and plasticity. Child Dev Perspect 6:354–360 50. Korkman M, Kirk U, Kemp S (2007) NEPSY-II: clinical and interpretive manual. Psychological Corporation, San Antonio 51. McIntosh D, Miller L, Shyu V, Dunne W (1999) Overview of the short sensory profile (SSP). In: Dunn W (ed) The sensory profile: Examiner’s manual, pp 59–73 52. Waters E Attachment Q-set (version 3.0) (1987) State University of New York at Stony Brook. Stony Brook, NY 53. van IJzendoorn M, Vereijken C, Bakersmans-Kranenburg M, Riksen-Walraven J (2004) Assessing attachment security with the attachment Q-sort: meta-analytic evidence for the validity of the observer AQS. Child Dev 75:1188–1213

123

Child Psychiatry Hum Dev 54. Dunn LM, Service A (1997) Examiner’s manual for the PPVTIII. Peabody Picture Vocabulary Test, AGS, Los Angeles 55. Hodapp A, Gerken K (1999) Correlations between scores for Peabody picture vocabulary test-III and the Wechsler intelligence scale for children-III. Psychol Rep 84:1139–1142 56. Dunn L, Dunn L (1997) The peabody picture vocabulary test, 3rd edn. American Guidance Service, Circle Pines 57. Gadow KD, Sprafkin J (2000) Early childhood symptom inventory-4. Screening manual. Checkmate Plus, Stonybrook 58. Strickland J, Keller J, Lavigne JV, Gouze KR, Hopkins J, LeBailly SA (2011) The structure of psychopathology in a community sample of preschoolers. J Abnorm Child Psychol 39:601–610 59. Sterba SK, Egger HE, Angold A (2007) Diagnostic specificity and nonspecificity in the dimensions of preschool psychopathology. J Child Psychol Psychiatry 48:1005–1013 60. Kline RB (2011) Principles and practice of structural equation modeling, 3rd edn. Guilford, New York 61. Gaylord-Harden NK, Cunningham JA, Grant KE, Holmbeck GN (2010) Suppressor effects in coping research with African American adolescents from low-income communities. J Consult Clin Psychol 78:843–855 62. Andres PL, Finison LF, Conlon T, Thibodeu LM, Murstat TL (1988) Use of composite scores (megascores) to measure deficit in amyotrophic lateral sclerosis. Neurology 38:405–408 63. Save the Children, State of the World’s Mothers (2004) Children having children. Save the Children’s Foundation, Westport 64. Deshpande SP, Fiorito J (1989) Specific and general beliefs in union voting models. Acad Manage 32:883–897 65. Yarnold PR, Soltysik RC (2005) Optimal data analysis: a guidebook with software for windows. American Psychological Association, Washington

123

66. Joreskog KG, Sorbom D (2006) LISREL for Windows. Scientific Software International, Chicago 67. Bollen KA (1989) Structural equations with latent variables. Wiley, New York 68. Bryant FB, King SP, Smart CM (2006) Multivariate statistical strategies for construct validation in positive psychology. In: Ong AG, Van Dulmen MHM (eds) Oxford handbook of methods in positive psychology. Oxford University Press, New York, pp 61–82 69. Jensen PS, Hinshaw SP, Kraemer HC, Lenora N, Newcorn JH, Abikoff HB et al (2001) ADHD comorbid findings from the MTA study: comparing comorbid subgroups. J Am Acad Child Adolesc Psychiatry 40:147–158 70. De Los Reyes A, Kazdin AE (2005) Informant discrepancies in the assessment of childhood psychopathology: a critical review, theoretical framework, and recommendations for further study. Psychol Bull 131:483–509 71. Richters JE (1992) Depressed mothers as informants about their children: a critical review of the evidence for distortion. Psychol Bull 112:485–499 72. Youngstrom EA, Loeber R, Stouthamer-Loeber M (2000) Patterns and correlates of agreement between parent, teacher, and male adolescent ratings of externalizing and internalizing problems. J Consult Clin Psychol 68:1038–1050 73. Hinshaw SP, Han SS, Erhardt D, Huber A (1992) Internalizing and externalizing behavior problems in preschool children: correspondence among parent and teacher ratings and behavior observations. J Clin Child Adolesc Psychol 21:143–150