Assessing ADHD Through the Multi-Informant Approach: The Contribution of the Conners’ 3 Scales

Abstract

Objective: Symptoms of ADHD need to be present in at least two different settings to suggest a diagnosis, so multi-informant assessment approaches are usually adopted. The Conners’ Rating Scales offer three versions, completed by parents (Conners 3-P), teachers (Conners 3-T), and youth (Conners 3-SR). Nonetheless, there is a lack of studies examining the psychometric properties of the Parent and Teacher versions and the relation between each version of the Conners 3 scales. Method: The present work examined the psychometric properties of the Parent and Teacher scales in terms of the structural validity and reliability of their Content scales in an Italian sample. Moreover, to support the need to administer all Conners 3 scales together and to exclude their potential redundancy, cross-informant agreement between parents, teachers, and children was assessed. Results: Results supported the Conners 3 scales as reliable and valid tools to assess ADHD and provided evidence for their non-redundancy. Conclusion: The non-redundancy of the Conners 3 may promote clinicians to adopt a multi-informant perspective

Keywords

ADHD Conners 3 multi-informant assessment psychometric properties dimensionality

Introduction

ADHD is frequently diagnosed in childhood and adolescence, with a worldwide prevalence in school-aged children under 18 estimated to be 7.2% (Thomas, Sanders, Doust, Beller, & Glasziou, 2015). One of the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5; American Psychiatric Association, 2013) criteria for advancing a diagnosis of ADHD states that several symptoms need to be present in at least two different settings, such as at home and at school, and with both friends and relatives.

The Conners’ Rating Scales (Conners, 1989, 1997, 2008) are used worldwide to support the diagnosis of ADHD and the most common co-occurring difficulties, such as disruptive disorders and learning disabilities in children and youth. In fact, as ADHD is often associated with other comorbid impairments, such as disruptive behavior, learning difficulties, and negative family and peer relationships (Czamara et al., 2013; Jensen & Steinhausen, 2015; Spencer, Biederman, & Mick, 2007), the third edition (Conners 3; Conners, 2008) also examines those aspects. The Conners 3 scales consist of three versions completed by parents (Conners 3–Parent; Conners 3-P), teachers (Conners 3–Teacher; Conners 3-T), and youth (Conners 3–Self-Report; Conners 3-SR). To facilitate a comparison between the assessment of the three sources of information, the Conners 3 scales were developed so that content alignment across them was enhanced, so they dispose of similar structures, with almost overlapping subscales.

Multi-informant assessment is strongly suggested in clinical practice, as it allows to both obtain a global, exhaustive evaluation of the child, and to detect contextual variations in mental health, thus, facilitating clinicians in forming an accurate judgment (De Los Reyes et al., 2015; Gizer et al., 2008; Hughes & Gullone, 2010; Hunsley & Mash, 2007; Power, Dombrowski, Watkins, Mautone, & Eagle, 2007). Collecting information from different points of view has significant implications for clinical practice. In fact, knowing the child’s most relevant difficulties and in which context(s) they occur more strongly and more frequently is fundamental for intervention purposes and to support children’s strengths (De Los Reyes et al., 2015; Hawley & Weisz, 2003). For instance, if hyperactivity symptoms were particularly prevalent at home rather than at school, the intervention program should focus on strengthening parents’ strategies. In line with this, the Conners 3 scales’ multi-informant nature represents one of their main strengths, as they make it possible to obtain information on how the child behaves both at home and school, and to merge those hetero-evaluations with a self-assessment that provides an insight into personal feelings and thoughts that others might not be aware of.

Given that the Conners 3 Rating Scales are widely used in both clinical and research contexts, the present study aimed at assessing the psychometric properties of the Italian version of the Conners 3 scales in their parent and teacher versions to test whether they represented a reliable and valid instrument to assess ADHD in children and youth. In fact, as the psychometric properties of the Conners 3-SR scale were already investigated before (see Izzo, Donati, & Primi, 2018), and that hetero-assessments make it possible to gather information that people filling a self-evaluation may be unaware of, this research examined the psychometric properties of the Conners 3-P and the Conners 3-T scales in terms of the dimensionality of these scales, their reliability, and cross-informant agreement.

In the original work (Conners, 2008), Content scales were empirically deduced by exploratory factor analyses (EFA), except for the Inattention scale, which was not supported because items belonging to this aspect cross-loaded with the Learning Problems factor. Consequently, this scale was theoretically defined based on the opinions of clinical experts—so the items were rationally selected—and confirmatory factor analyses (CFA) were conducted only on the other empirically found scales. Therefore, the composition of the Inattention scale and the dimensionality of the Content scales including the Inattention scale were not empirically tested (Conners, 2008). However, in a recent study examining the German version of the Conners 3 scales (Christiansen et al., 2016), CFAs were conducted on all Content scales, including the Inattention scale, with results revealing a good fit and supporting the Inattention scale with psychometric evidence. Moreover, empirical data from a recent Italian study (Izzo et al., 2018) confirmed this structure with the Inattention scale as a separate factor, though analyses were performed only for the Conners 3-SR scale. Thus, the first aim of the present work (Study 1) was to examine the dimensionality of the Conners 3-P and Conners 3-T scales to test whether the original structure could be confirmed and whether there is empirical support for the Inattention scale as a stand-alone factor. Moreover, we aimed at assessing the accuracy of the Content scales in terms of their internal consistency.

Furthermore, the relations between the Conners 3-P, Conners 3-T, and Conners 3-SR scales need to be assessed as well, to test the level of correspondence between different informants. In fact, although the multi-informant procedure is widely used, it is not unusual that informant discrepancies occur, that is, the report of an informant does not agree with that of another informant (Achenbach, 2006; Achenbach, McConaughy, & Howell, 1987; De Los Reyes et al., 2015; Goodman, De Los Reyes, & Bradshaw, 2010; Lapouse & Monk, 1958). However, discrepancies do not necessarily indicate poor validity of the examined instrument, because several factors may explain their occurrence, such as (a) belonging to different versus the same context of observation—that is, pairs of parents obtain greater correspondence than parent-teacher pairs; (b) the nature of the mental health concern, as levels of correspondence are greater when referring to externalizing rather than to internalizing problems; (c) the nature of the measurement method, because correspondence is greater when reports are on a dimensional scale than when they are on a categorical scale (Achenbach et al., 1987; De Los Reyes et al., 2015). Examining the relations between the different versions of the Conners 3 scales may help in identifying the extent to which each version contributes to the understanding and the description of the child profile. Thus, an additional study (Study 2) was conducted to assess the validity of the Conners 3 scales in terms of the cross-informant correlations between parents, teachers, and the children/youth themselves, expecting to find moderate correlations, in line with the original work (Conners, 2008).

In sum, Study 1 aimed at testing the psychometric properties of the Conners 3-P and Conners 3-T scales in terms of their dimensionality and reliability, whereas Study 2 examined the validity of the Conners 3 scales in terms of the cross-informant correlations between the parent, teacher, and self-report assessments.

Study 1

Although the Conners 3 is one of the most commonly used scales to help clinicians in assessing ADHD, it is important to further investigate its factorial structure. In fact, EFAs in the original work (Conners, 2008) did not distinguish between the Inattention and the Learning Problems scales, as items cross-loaded on the same factor. In line with that, EFAs in the German study were unable to separately extract factors for the Inattention, Learning Problems, and Executive Functioning scales, as items loaded on a broader factor—although CFAs supported them as stand-alone scales (Christiansen et al., 2016). Moreover, the Italian study performing CFAs on the Conners 3-SR scale found an extremely high correlation between the Inattention and the Learning Problems scales—though the model considering them as a unitary factor showed worse fit than the model that identified them as distinct scales (Izzo et al., 2018). Building on these findings, several models testing the structure of the Conners 3-P and Conners 3-T were tested in the present study. In detail, a solution that considered the Inattention, the Learning Problems, and the Executive Functioning scales as separate factors, and some models that combined them into broader factors were investigated. Furthermore, the reliability of the scales in terms of their internal consistency to examine their accuracy was assessed.

Method

Participants

Participants were 876 parents of 6- to 18-year-old non-clinical children and adolescents (mean age: 10.99, SD: 2.92; 49% male), attending primary (51%), middle (37%), and high schools (12%) in the North-Center (72%) and South (28%) of Italy. A total of 84% of the participating parents were mothers.

Furthermore, 832 teachers of 6- to 18-year-old non-clinical children and adolescents were recruited. However, many participants were excluded because of a high percentage of missing answers, so the final sample included 487 teachers of children and adolescents (mean age: 11.23, SD: 3.4; 50% male) attending primary (52%), middle (29%), and high schools (19%) in the North-Center (67%) and South (33%) of Italy. A total of 98% of the participating teachers were women.

Instruments

The Parent and Teacher Conners 3 Rating Scales (Italian version: Primi & Maschietto, 2017) include items aiming to measure youth’s ADHD symptoms and their most frequent associated complications, that is, learning problems, defiance/aggression, and difficulties in social relationships. They can be administered to parents and teachers of 6- to 18-year-old children and adolescents. Although the parent and teacher scales are similar, the Content scales vary in number according to the specific version: The Conners 3-P Scale contains 6 Content scales (i.e., Inattention, Hyperactivity/Impulsivity, Learning Problems, Executive Functioning, Defiance/Aggression, and Peer Relations), with a total of 62 items. The Conners 3-T Scale consists of 5 Content scales (i.e., Inattention, Hyperactivity/Impulsivity, Learning Problems/Executive Functioning, Defiance/Aggression, and Peer Relations), totaling 69 items. Items are rated on a Likert-type scale ranging from 0 (not true at all/never) to 3 (very much true/very frequently).

Procedure

A study protocol in accordance with the criteria of the Declaration of Helsinki was reviewed and approved by each head teacher and school board of different primary, middle, and high schools in North-Center and South Italy. Parents and teachers were given a short description of the study. Written informed consent was obtained from all participants, and their confidentiality was ensured. All participants completed their questionnaire through a self-administered procedure, though parents completed it at home, whereas teachers completed it in class, during schooltime. The time needed to complete the scales was approximately 20 min.

Statistical analyses

First, a preliminary missing value analysis of data was performed. Cases were excluded when missing data exceeded the allowable number of omitted responses as specified by the original manual (Conners, 2008). Analyses revealed that, for the Conners 3-P Scale, all cases presented an acceptable number of missing answers, whereas for the Conners 3-T Scale, 345 protocols needed to be excluded. When missing data were acceptable, values of omitted answers were estimated through the application of a formula calculating a prorated score, as described in the original manual (Conners, 2008).¹

To examine the dimensionality of the Conners 3-P and Conners 3-T scales, CFAs were conducted on the Content scales. In line with the original work (Conners, 2008), items belonging to the same factor that showed high intercorrelations were combined into parcels, and those resulting combined items were used as observed variables. Parcels were then loaded onto their respective scales, and the scales correlated with each other. Adequate fit for the model was defined according to the Comparative Fit Index (CFI; Bentler, 1990), and the Steiger-Lind Root Mean Square Error of Approximation Index (RMSEA; Steiger & Lind, 1980); in details, in line with the original manual (Conners, 2008), adequacy of the model was defined as CFI values equal to .90 or greater, and RMSEA values of .10 or below.

To assess reliability of the Conners 3-P and T scales, internal consistency of the Content scales using Cronbach’s alphas was obtained.

Results

CFA

Conners 3-P Scale

As a first step, the six-factor structure of the Conners 3-P was tested. Goodness-of-fit indicators suggested that it had adequate fit to the data (Table 1). Each item loaded strongly and significantly on its hypothesized factor, and the correlations between the five factors were all significant, ranging from .26 to .94.

Table 1.

Goodness-of-Fit Statistics for the Confirmatory Factor Analysis Models for the Conners 3 Scales.

Scale	Model	χ² (df)	χ²/df	CFI	RMSEA (Confidence Interval)
Conners 3-P	Six-factor Model	1,090.90 (284)	3.84	.93	.06 (.05-.06)
	Five-factor Model (IN + EF)	1,143.55 (289)	3.96	.92	.06 (.06-.06)
	Five-factor Model (IN + LP)	1,221.70 (289)	4.23	.92	.06 (.06-.06)
	Four-factor Model (IN + EF + LP)	1,282.23 (293)	4.38	.91	.06 (.06-.07)
Conners 3-T	Five-factor Model	1,489.07 (220)	6.77	.90	.10 (.10-.11)
	Five-factor Model with covariance errors	1,147.49 (216)	5.31	.93	.09 (.08-.09)
	Four-factor Model (IN + LE)	1,610.62 (224)	7.19	.90	.11 (.11-.12)

Note. χ² = chi squared; df = degrees of freedom; CFI = Comparative Fit Index; RMSEA = Root Mean Square Error of Approximation; IN = Inattention; EF = Executive Functioning; LP = Learning Problems; LE = Learning Problems/Executive Functioning; Conners 3-P = Conners 3-Parent scale; Conners 3-T = Conners 3-Teacher scale.

Nonetheless, as the correlations between the Inattention scale and both the Executive Functioning and the Learning Problems scales were high (i.e., equal to .94 and to .91, respectively), a five-factor model considering Inattention and Executive Functioning as a unique factor, and a five-factor model considering Inattention and Learning Problems as a unique factor were tested. However, in both cases, the results showed a poorer fit than the original six-factor model (Table 1).

As a final step, as the correlation between the Learning Problems and the Executive Functioning scales in the six-factor model was high (i.e., equal to .86), a four-factor model that merged the Inattention, the Learning Problems, and the Executive Functioning scales into a unique factor was tested. Nonetheless, results showed a poorer overall fit (Table 1), so the original six-factor model was empirically supported as it showed a better fit to the data than the other proposed solutions (Figure 1).

Figure 1.

Confirmatory Factor Analysis for Conners 3-Parent scale—final model.

Conners 3-T Scale

Using CFA performed on the Conners 3-T scale, a five-factor model in line with the original structure was tested. Since goodness-of-fit indicators of the model were not adequate (Table 1), Modification Indices (M.I.) were examined. They suggested the addition of four covariance errors between parcels 5 and 6 of the Hyperactivity/Impulsivity factor, between parcels 2 and 3 of the Learning Problems/Executive Functioning factor, between parcels 1 and 2 of the Defiance/Aggression factor, and between parcels 2 and 3 of the Peer Relations scale. These links were theoretically justified, given that (a) the two parcels of the Hyperactivity/Impulsivity factor examined verbal hyperactivity (namely, talking too much, blurting out answers, and interrupting others); (b) the two parcels of the Learning Problems/Executive Functioning factor dealt with difficulties in executive functions, such as forgetting or not following instructions; (c) the two parcels of the Defiance/Aggression factor both referred to harming others or other people’s belongings; and (d) the two parcels of the Peer Relations factor pertained to poor social skills and being isolated. The modified model showed a better fit than the first model (Table 1). Each parcel loaded strongly and significantly on its hypothesized factor and the correlations between the five factors were all significant (p < .001), ranging from .46 to .97.

Because the correlation between the Inattention and the Learning Problems/Executive Functioning scales was .97, a four-factor model that combined the Inattention and the Learning Problems/Executive Functioning scales into a single factor was tested. However, results showed a poorer overall fit (Table 1), so the original five-factor model with the four covariance errors was empirically supported (Figure 2).

Figure 2.

Confirmatory Factor Analysis for Conners 3-Teacher scale—final model.

Reliability

Conners 3-P Scale

Concerning reliability, in line with the original study (Conners, 2008), we aimed to measure the internal consistency of the Content scales. Cronbach’s alphas for the Conners 3-P scale showed an acceptable internal consistency with alpha’s values ranging from .71 to .90 (Table 2).

Table 2.

Cronbach’s Alpha Values for the Conners 3-Parent and Teacher Content Scales.

Scale	Conners 3-Parent	Conners 3-Teacher
Inattention	.90	.96
Hyperactivity/Impulsivity	.89	.96
Learning Problems	.84	—
Executive Functioning	.83	—
Learning Problems/Executive Functioning	—	.96
Defiance/Aggression	.74	.94
Peer Relations	.71	.88

Conners 3-T Scale

Referring to Cronbach’s alphas for the Conners 3-T scale, the Content scales showed acceptable to high levels of reliability, ranging from .88 to .96 (Table 2).

Discussion

Despite the widespread use of the Conners 3 scales, only one study (Christiansen et al., 2016) examined the psychometric properties of the Parent and Teacher scales to test whether the original structure (Conners, 2008) could be confirmed. Moreover, in the original work aiming to develop the Conners 3 scales, EFAs did not support the Inattention scale, as the items belonging to this sphere cross-loaded with the Learning Problems factor (Conners, 2008). Therefore, the Inattention scale was only rationally defined, and CFAs did not include it. The present study has made several contributions to the analyses of the psychometric properties of the Conners 3-P and Conners 3-T scales. First, when examining the dimensionality of the Content scales to test for the presence of the original structure, a six-factor model for the Conners 3-P scale and a five-factor model for the Conners 3-T scale were identified, thus, providing empirical evidence for the Inattention scale.

Furthermore, as correlations between the Inattention scale and both the Learning Problems and Executive Functioning scales were extremely high for the Conners 3-P scale, thus, suggesting a potential overlap between them in line with previous studies (Christiansen et al., 2016; Conners, 2008; Izzo et al., 2018), a five-factor model merging Inattention and Learning Problems scales, and a five-factor model merging Inattention and Executive Functioning scales were tested. In addition, as the correlations between the Learning Problems and Executive Functioning scales were also extremely high, a four-factor model merging those two scales with the Inattention scale was tested. However, results showed a poorer fit than the six-factor solution, so this study not only confirmed the originally proposed six-factor model (Conners, 2008), thus, supporting its generalizability to a different cultural context, but it also provided psychometric, empirical support to the Inattention scale. The same results were found for the Conners 3-T scale, because the five-factor structure in line with the original one showed a better fit than a four-factor solution that combined the Inattention and the Learning Problems/Executive Functioning scales into a single factor. Finally, Cronbach’s alphas for the Content scales of both the Conners 3-P and the Conners 3-T scales were high, thus, supporting their reliability.

Study 2

The multi-informant assessment, which is one of the main strengths of the Conners 3 scales, is strongly suggested to obtain a global, more precise assessment of the child’s impairments. However, it is important to examine the amount and specificity of information provided by each version of the scale about the assessed child, to establish the extent to which each version contributes to the understanding and the description of the child’s profile. In fact, if accordance between different informants’ assessments was particularly high, this could suggest redundancy between the different points of view, so clinicians may opt to administer only one version of the scales. In contrast, extremely low correlations may raise doubts regarding the validity of the scales. To answer these questions, the second study aimed at measuring the validity of the Conners 3 scales in terms of cross-informant correspondence between the different sources of information, so the correlation coefficients among the three versions of the Conners 3 scales were computed. We expected to find moderate correlations, which are not as high as to suggest redundancy in conducting all three assessments.

Method

Participants

A sample of 293 8- to 18-year-old non-clinical children and adolescents (mean age: 11.82, SD: 2.42; 52% was male) was recruited. They attended primary (47%), middle (43%), and high schools (10%) in the North-Center (77%) and South (23%) of Italy. Moreover, their parents and teachers were recruited, resulting in a total of 879 participants (293 children, 293 parents, and 293 teachers). Thus, for each child, we collected all three versions of the Conners 3 scales.

Instruments

Along with the previously described Conners 3-P and Conners 3-T scales, the Italian version of the Conners 3-SR scale (Primi & Maschietto, 2017) was administered. It includes 56 items that compose the five Content scales, which assess the core symptoms of ADHD and their associated impairments in children and youth. The Content scales are the Inattention scale, the Hyperactivity/Impulsivity scale, the Learning Problems scale, which examines school difficulties, the Defiance/Aggression scale, assessing physical and/or verbal aggressiveness toward others, and the Family Relations scale, which evaluates negative family interactions. The symptoms are rated on a Likert-type scale with severity ratings from 0 (not true at all/never) to 3 (very much true/very frequently). A recent study (Izzo et al., 2018) found that the scale showed good psychometric properties. Cronbach alphas for this sample were .86 for the Inattention scale, .83 for the Hyperactivity/Impulsivity scale, .77 for the Learning Problems scale, .81 for the Defiance/Aggression scale, and .71 for the Family Relations scale.

Procedure

A study protocol similar to that of Study 1 was implemented. However, parents were also asked to provide their informed consent to allow their children to participate in the research. Written informed consent was obtained for all participants, and their confidentiality was ensured. All participants completed the questionnaire through a self-administered procedure, though parents completed them at home, whereas students and teachers completed them in class, during schooltime. The time needed for parents and teachers was approximately 20 min, whereas students completed both the Conners 3-SR and the Gambling Behavior Scale for Adolescents (GBS-A) scales in approximately 45 min.

Statistical analyses

First, means and standard deviations for each Conners 3 Content scale were calculated separately for each version—parent, teacher, and self-report—to measure differences in assessments from different points of view. Moreover, to assess cross-informant correspondence, correlation coefficients (Pearson’s r) were calculated between each pair of informants: parent and teacher ratings, parent and self-report ratings, and teacher and self-report ratings.

Results

Means and standard deviations for each Conners 3 Content scale in the parent, teacher, and self-report versions are reported in Table 3. As the results show, self-reported mean scores were higher on every Content scale when comparing them with hetero-evaluations from both parents and teachers. Furthermore, when comparing parents’ and teachers’ assessments, educators reported higher scores than parents on the scale evaluating negative peer relations, whereas there were no manifest differences on the other Content scales.

Table 3.

Means and Standard Deviations for Each Conners 3 Content Scale.

Scale	Conners 3-PM (SD)	Conners 3-TM (SD)	Conners 3-SRM (SD)
Inattention	7.17 (5.79)	7.64 (7.79)	7.95 (5.95)
Hyperactivity/Impulsivity	7.18 (6.67)	6.27 (9.77)	11.11 (7.17)
Learning Problems	4.45 (4.16)	—	5.30 (4.20)
Executive Functioning	6.62 (4.85)	—	—
Learning Problems/Executive Functioning	—	11.38 (10.67)	—
Defiance/Aggression	3.00 (3.58)	3.18 (6.00)	3.33 (4.31)
Peer Relations	1.98 (2.49)	3.28 (4.22)	—
Family Relations	—	—	4.42 (3.45)

Note. P = parent; T = teacher; SR = self-report.

Regarding multi-informant correspondence, the correlations between the three informants were found to be all significant and moderate in size, indicating that there was a good deal of consistency between different informants’ ratings of the same youth across the Conners 3 scales. The correlations were not high enough, however, to suggest redundancy. In detail, the mean parent to teacher correlation was .37 (ranging from .32 to .43), the mean parent to self-report correlation was .39 (ranging from .30 to .45), and the mean teacher to self-report correlation was .34 (ranging from .30 to .41; Table 4).

Table 4.

Conners 3 Across-Informant Correlations.

Scale	Parent to Teacher	Parent to Self-Report	Teacher to Self-Report
Inattention	.39***	.43***	.30***
Hyperactivity/Impulsivity	.36***	.36***	.31***
Learning Problems^a	.43***	.45***	.41***
Executive Functioning^b	.32***	—	—
Defiance/Aggression	.37***	.30***	.35***
Peer Relations	.32***	—	—

***

p < .001.

Correlation between the Conners 3–Teacher (T) Learning Problems/Executive Functioning scale and the Conners 3–Parent (P) and Conners 3–Self-Report Learning Problems scale.

Correlation between the Conners 3-T Learning Problems/Executive Functioning scale and the Conners 3-P Executive Functioning scale.

Discussion

One of the main strengths of the Conners 3 scales is their multi-informant assessment, which allows a gathering of information from different points of view, thus, obtaining a more detailed profile of the youth’s impairments. In fact, the different sources of information assess the child from different perspectives—that is, the parent provides a profile based on how the child behaves at home, the teacher describes the student according to what can be observed in class, and the children respond on the basis of their personal perceptions and insights.

The Conners 3-P, T, and SR scales present similar structures with similar item content, so multi-informant concordance in scores is expected to be found. Nonetheless, previous meta-analyses (Achenbach et al., 1987; De Los Reyes et al., 2015) found that multi-informant correspondence tended to be low. Study 2 aimed at measuring the correlation coefficients among the three versions of the Conners 3 scales in our Italian sample. In line with the original work (Conners, 2008), correlations between all sources of information—parent to teacher, parent to youth, and teacher to youth—were moderate, though not high enough to suggest redundancy. In our study, we confirmed previous results (Achenbach et al., 1987; De Los Reyes et al., 2015), which found that, rather than indicating poor psychometric quality of the instrument, low to moderate levels of agreement between informants could be due to several factors: (a) the nature of the mental health concern, as there were greater levels of correspondence when referring to externalizing problems (with correlations between informants around .36), than in the case of internalizing difficulties (with rs around .28); (b) the nature of the measurement method, because correspondence was greater when reports were on a dimensional scale than when they were on a categorical scale; and (c) different observational contexts, because correspondence tended to be greater between pairs of informants belonging to the same context, such as pairs of parents or pairs of teachers, rather than between pairs of informants belonging to different contexts, such as parent and teacher. Considering that (a) the Conners 3 scales aim at assessing externalizing, directly observable concerns such as ADHD, (b) ratings are on a 4-point Likert-type scale, and (c) to facilitate the comparison between the assessment of the three sources of information, the scales have very similar structures, subscales, item content, and item formulation, it was not surprising that moderate cross-informant correspondence was found.

It is also noteworthy that children and adolescents rated themselves as more problematic on every Content scale when comparing their scores with the evaluations from both parents and teachers. This finding is in line with previous studies and is not limited to ADHD (e.g., Begovac, Rudan, Skočić, Filipović, & Szirovicza, 2004; Grigorenko, Geiser, Slobodskaya, & Francis, 2010; van der Ende & Verhulst, 2005). Moreover, educators reported higher scores than parents on the scale evaluating negative peer relations, whereas there were not differences on the other Content scales. Possible explanations for these results may include the fact that parents and teachers may be less aware of their sons/students’ problems than children/adolescents themselves. In line with this, parents may be less aware of the child’s social relationships than teachers, as educators may have more time to observe the student’s interactions with peers. Otherwise, children and adolescents may be too critical of themselves, thus, reporting more difficulties than parents and teachers. Overall, teachers and children may be more objective than parents in reporting their difficulties, as parents may be susceptible to biases in social desirability and may tend to describe the child as less problematic. These findings may confirm the importance of collecting information from all three sources of information. However, it is important to remember that a global clinical evaluation should also include observations and interviews, which may shed light on discrepancies among the three versions of the Conners 3 scales.

General Discussion

The Conners’ Rating Scales, which are currently in their third edition, are among the most commonly used measures to assess ADHD. The third edition (Conners, 2008) includes three versions, completed by parents (Conners 3-P), teachers (Conners 3-T), and youth themselves (Conners 3-SR). While the Conners 3-SR version was validated for the Italian context elsewhere (see Izzo et al., 2018), the Italian version of the parent and teacher scales had not yet been validated. Because one of the most important strengths of the Conners 3 scales is its multi-informant nature, and the psychometric properties of the self-report form have already been tested, the present work aimed at examining the measurement properties of the Conners 3-P and the Conners 3-T scales—so that clinicians may select instruments with proven reliability and validity.

The present study has made several original contributions to the analyses of the psychometric properties of the Conners 3-P and T scales, and regarding the usefulness of the multi-informant version of the Conners 3 scales. First, CFAs for both the Conners 3-P and the Conners 3-T scales revealed that, even when compared with alternative structural solutions, the original six-factor model for the Conners 3-P, and the original five-factor model for the Conners 3-T, were detected. These results are noteworthy, as they provided empirical evidence and psychometric support for the Inattention scale, which in the EFAs of the original work (Conners, 2008) could not be distinguished from the Learning Problems scale. Furthermore, reliability of each factor of both the Conners 3-P and the Conners 3-T scales in terms of their internal consistency was assessed, showing high Cronbach’s alpha values.

Moreover, the current study found that, though being lower than in the original work (Conners, 2008), correspondence between the three informants—parents, teachers, and youth—was moderate, thus, indicating that, though examining similar constructs, the three versions were not redundant, as they provided information from different standpoints. The modest multi-informant correspondence—which is in line with previous studies (see Achenbach et al., 1987; De Los Reyes et al., 2015)—may reflect the fact that the assessed behaviors and symptoms tend to be contextual-specific or to show variations among different contexts. Thus, discrepancies among different Conners 3 versions represent precious information, as each informant provided data that complemented those of other informants, rather than being overlapping and redundant, thus promoting the adoption of a multi-informant perspective to obtain a comprehensive assessment.

Some limitations should be considered for future studies. First, this study focused on the general population rather than on a clinical sample, so generalization of results to the clinical population should be made with caution, and future research should recruit clinical samples so as to repeat these analyses. Second, future studies may examine correlations between similar informant types, such as parent-parent or teacher-teacher. Eventually, different methods of evaluation such as observations and clinical interviews should be integrated.

Despite the above-mentioned limitations, the current findings have important practical implications in the clinical field. First, clinicians can now be confident that all versions of the Conners 3 scales—the parent, the teacher, and the self-report versions—are valid and reliable measures. Furthermore, the three versions of the Conners 3 scales were not redundant when examining cross-informant agreement. Therefore, the lack of redundancy strongly supports the multi-informant nature of the Conners 3 scales and advocates their administration. Consequently, clinicians aiming to use the Conners 3 scales should collect information from all three sources of information—parent, teacher, and the youth themselves—to take full advantage of the benefits that multi-informant instruments offer by obtaining complementary information.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

ORCID iD

Viola Angela Izzo

Author Biographies

Viola Angela Izzo is a PhD student of “Dottorato Toscano di Neuroscienze,” University of Florence, and tutor of psychometrics at the University of Florence.

Maria Anna Donati is a lecturer of “Teoria e Tecniche dei Test” in the course “Scienze e Tecniche di Psicologia Clinica e della Salute,” and “Metodi Avanzati di Ricerca e Valutazione Psicologica” in the course “Psicologia Clinica e della Salute” at the University of Pisa.

Caterina Primi is a lecturer of Psicometria and Test Psicologici at the School of Psychology of the University of Florence, is responsible for the “Laboratorio di Psicometria,” and is director of the “Corso di Perfezionamento: I modelli dell’Item Response Theory (IRT).”

References

Achenbach

T. M.

(2006). As others see us: Clinical and research implications of cross-informant correlations for psychopathology. Current Directions in Psychological Science, 15, 94-98.

Achenbach

T. M.

McConaughy

S. H.

Howell

C. T.

(1987). Child/adolescent behavioral and emotional problems: implications of cross-informant correlations for situational specificity. Psychological bulletin, 101, 213.

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Arlington, VA: American Psychiatric Publishing.

Begovac

Rudan

Skočić

Filipović

Szirovicza

(2004). Comparison of self-reported and parent-reported emotional and behavioral problems in adolescents from Croatia. Collegium Antropologicum, 28, 393-401.

Bentler

P. M.

(1990). Comparative fit indexes in structural models. Psychological Bulletin, 107, 238-246.

Christiansen

Hirsch

Drechsler

Wanderer

Knospe

E. L.

Günther

Lidzba

(2016). German validation of the Conners 3U+ 00AE rating scales for parents, teachers, and children. Zeitschrift für Kinder-und Jugendpsychiatrie und Psychotherapie, 44, 139-147. doi: 10.1024/1422-4917/a000408

Conners

C. K.

(1989). Manual for Conners’ Rating Scales. North Tonawanda, NY: Multi-Health Systems.

Conners

C. K.

(1997). Conners’ Rating Scales—Revised technical manual. North Tonawanda, NY: Multi-Health Systems.

Conners

C. K.

(2008). Conners 3rd edition: Manual. Toronto, Ontario, Canada: Multi-Health Systems.

10.

Czamara

Tiesler

C. M.

Kohlböck

Berdel

Hoffmann

Bauer

C. P.

. . . Heinrich

(2013). Children with ADHD symptoms have a higher risk for reading, spelling and math difficulties in the GINIplus and LISAplus cohort studies. PLoS ONE, 8(5), e63859.

11.

De Los Reyes

Augenstein

T. M.

Wang

Thomas

S. A.

Drabick

D. A.

Burgers

D. E.

Rabinowitz

. (2015). The validity of the multi-informant approach to assessing child and adolescent mental health. Psychological Bulletin, 141, 858-900.

12.

Gizer

I. R.

Waldman

I. D.

Abramowitz

Barr

C. L.

Feng

Wigg

K. G.

. . . Rowe

D. C.

(2008). Relations between multi-informant assessments of ADHD symptoms, DAT1, and DRD4. Journal of Abnormal Psychology, 117, 869-880.

13.

Goodman

K. L.

De Los Reyes

Bradshaw

C. P.

(2010). Understanding and using informants’ reporting discrepancies of youth victimization: A conceptual model and recommendations for research. Clinical child and family psychology review,13, 366-383.

14.

Grigorenko

E. L.

Geiser

Slobodskaya

H. R.

Francis

D. J.

(2010). Cross-informant symptoms from CBCL, TRF, and YSR: Trait and method variance in a normative sample of Russian youths. Psychological Assessment, 22, 893-911.

15.

Hawley

K. M.

Weisz

J. R.

(2003). Child, parent and therapist (dis)agreement on target problems in outpatient therapy: The therapist’s dilemma and its implications. Journal of Consulting and Clinical Psychology, 71, 62-70.

16.

Hughes

E. K.

Gullone

(2010). Discrepancies between adolescent, mother, and father reports of adolescent internalizing symptom levels and their association with parent symptoms. Journal of Clinical Psychology, 66, 978-995.

17.

Hunsley

Mash

E. J.

(2007). Evidence-based assessment. Annual Review of Clinical Psychology, 3, 29-51.

18.

Izzo

V. A.

Donati

M. A.

Primi

(2018). Conners 3–Self-Report Scale: An empirical support to the dimensionality of the content scales. Clinical Child Psychology and Psychiatry, 23, 556-566. doi:10.1177/1359104518757289

19.

Jensen

C. M.

Steinhausen

H. C.

(2015). Comorbid mental disorders in children and adolescents with attention-deficit/hyperactivity disorder in a large nationwide study. Attention Deficit and Hyperactivity Disorders, 7, 27-38.

20.

Power

T. J.

Dombrowski

S. C.

Watkins

M. W.

Mautone

J. A.

Eagle

J. W.

(2007). Assessing children’s homework performance: Development of multi-dimensional, multi-informant rating scales. Journal of School Psychology, 45, 333-348.

21.

Primi

Maschietto

(2017). Conners 3TM: Standardizzazione italiana [Conners 3TM: Italian Standardization]. Firenze, Italy: Giunti O.S. Psychometrics.

22.

Spencer

T. J.

Biederman

Mick

(2007). Attention-deficit/hyperactivity disorder: diagnosis, lifespan, comorbidities, and neurobiology. Journal of Pediatric Psychology, 32, 631-642.

23.

Steiger

J. H.

Lind

J. C.

(1980, May). Statistically based tests for the number of common factors. Paper presented at the annual meeting of the Psychometric Society, Iowa City.

24.

Thomas

Sanders

Doust

Beller

Glasziou

(2015). Prevalence of attention-deficit/hyperactivity disorder: A systematic review and meta-analysis. Pediatrics, 135(4), e994-e1001.

25.

van der Ende

Verhulst

F. C

. (2005). Informant, gender and age differences in ratings of adolescent problem behaviour. European Child & Adolescent Psychiatry, 14, 117-126.