German Validation of the Conners 3U+00AE Rating Scales for Parents, Teachers, and Children
1. Introduction
Attention deficit/hyperactivity disorder (ADHD) is one of the most common psychological disorders of childhood and adolescence, with the three core symptoms inattention, hyperactivity, and impulsivity (Christiansen & Röhrle, 2012). Apart from onset in childhood, diagnostic guidelines require impairment in multiple settings, i.e., at home and at school/kindergarten. Thus, assessment of symptoms in those domains is recommended by different guidelines (e.g., NICE guidelines [CG072] (2008). URL: http://www.nice.org.uk/CG072fullguideline; Deutsche Gesellschaft für Kinder- und Jugendpsychiatrie und Psychotherapie, 2007). Although a diagnosis of ADHD should not be based solely on questionnaires, rating scales such as the Conners’ Rating Scales (CRS) are valuable adjuncts, since they offer parent, teacher, and self-ratings of children susceptible for ADHD (NICE guidelines [CG072] (2008). URL: http://www.nice.org.uk/CG072fullguideline). Accordingly, the CRS have been used widely for the assessment of ADHD in different studies (Charach, Chen, Hogg-Johnson, & Schachar, 2009; Deb, Dhaliwal, & Roy, 2008; Green, Wong, Atkins, Taylor, & Feinleib, 1999; Müller et al., 2011a, 2011b). They proved to be especially valuable for the use in large multicenter studies, since they are available in different languages (Müller et al., 2011a, 2011b), and because they assess not only the core ADHD symptoms based on DSM criteria, but also comorbid conditions such as conduct problems and social problems (Conners, 1997). Even though the scales are widely used internationally, cross-cultural comparability has seldom been verified, and culture and language invariance are only presumed (Huss, Iseler, & Lehmkuhl, 2001). The German version of the Conners’ Rating Scale-Revised (CRS-R; Conners, 1997) resulted in different factor models for both the parent (Huss et al., 2001) and the teacher version (Huss et al., 2002). Cross-cultural differences were largest for hyperactivity and impulsivity symptoms in parent ratings (Huss et al., 2001), whereas in teacher ratings those domains produced only marginal differences (Huss et al., 2002). The German inattention scale was almost identical to the American one in both studies (Huss et al., 2001; Huss et al., 2002). Large differences were found for conduct problems in the German CRS-R teacher version (Huss et al., 2002). Thus, cross-cultural differences need to be taken into account when different language versions of the American original are used.
The 3rd edition of the CRS (Conners 3U+00AE ;Conners, 2008) was the result of continuous development of the CRS (Conners, 1989) and their revision (CRS-R; Conners, 1997) for children aged 6 (parent and teacher forms) and 8 (self-report forms) to 18 years. According to the author, important refinements include validity assessment, refined T-score cut-off categories, and the assessment of defiance/aggression (Conners, 2008). All Conners 3U+00AE content scales (Hyperactivity/Impulsivity, Learning Problems, Executive Functioning, Defiance/Aggression, Peer-/Family-Relations) were deduced with exploratory factor analyses, except for the Inattention scale, which was constructed hypothetically. All symptom scales (ADHD Inattentive, ADHD Hyperactive/Impulsive, CD, ODD) were based theoretically on DSM-IV items. Though the Conners 3U+00AE have been in use for 8 years now, hardly any studies report psychometric properties apart from the results published in the test edition itself (Conners, 2008). To our knowledge there are no studies on the various adaptations of the Conners 3U+00AE in other languages.
As mentioned, cross-cultural generalizability of the scales is not self-evident (Huss et al., 2001). Thus, the present study aims to validate the German version of the Conners 3U+00AE (Lidzba, Christiansen, & Drechsler, 2013) by replicating the derived factor structure of the American original for the content scales; all results reported are more extensive than those found in the German Conners-3 (C). Since all symptom scales resemble DSM-IV items, they will not be subjected to factor analyses.
Method
Procedure and Subjects
This is a cross-sectional study with German-speaking children who are assessed by their parents, teachers, and themselves. Study centers were the Department of Pediatric Neurology and Developmental Medicine at the University Children’s Hospital Tübingen, the Department of Child and Adolescent Psychology at the University of Marburg, the Department of Child and Adolescent Psychiatry at the University of Zurich, the Department of Child and Adolescent Psychiatry, Psychotherapy and Psychosomatics, RWTH Aachen University, and the Department of Child and Adolescent Psychiatry of the Technical University of Dresden. Participants were mainly recruited in schools willing to contribute to the study. Teachers, parents, and children were provided with a short study description and asked to complete the German version of the Conners 3U+00AE rating scales (Lidzba et al., 2013) as well as questions on age, sex, and school. A study protocol in accordance with the criteria of the Declaration of Helsinki was reviewed and approved by each study center’s local institutional review board and by the school boards of the federal states of Baden Wuerttemberg and Saxony. Written informed consent was obtained from all participants, and their confidentiality was ensured. All subjects completed the questionnaires at home and sent them back to the study centers.
The total sample consisted of 741 teacher, 953 parent and 745 self-report child ratings. The majority of parent ratings were from mothers (68.4 %), and the majority of parent, teacher, and child ratings from elementary (parents: 35 %, teachers: 25.1 %, children: 14.5 %) and grammar schools (parents: 23.9 %, teachers: 24.4 %, children: 34.6 %). Slightly more males (51.3–55.1 %) than females participated in the study. According to parent reports, 7.2 % of the children had a former ADHD diagnosis. Table 1 shows the children’s age, sex, and school type separate for child self-report, teacher, and parent ratings.
Measures
Description of the Conners 3U+00AE Scales (Conners 3U+00A9 MHS, North Tonawanda, NY)
The Conners 3U+00AE assesses ADHD symptoms and related problematic behaviors in children and adolescents aged 6 to 18 years. Parent, teacher, and child-self-rating forms of the Conners 3U+00AE are available. In all three versions, symptoms are rated on a Likert scale with severity ratings from 0 (not at all/never) to 3 (very much/very frequently). The long form of the self-rating scales consists of 56 content scale items assessing not only ADHD (Inattention and Hyperactivity/Impulsivity), but also related problems in Executive Functioning, Learning Problems, Aggression, and Peer/Family Relations. Scales that relate directly to the DSM-IV-TR diagnostic criteria (symptom scales with 44 items) are included for ADHD (Inattention 10 items, and Hyperactivity/Impulsivity 11 items), as well as for the most commonly co-occurring group of disorders: Disruptive Behavior Disorders with CD (15 items) and ODD (8 items). The Conners 3U+00AE also features screener items for anxiety and depression. The parent, teacher, and self-report ratings differ primarily in their wording of questions, though there are also differences in lengths (teacher: 111 items, parent: 105 items, self-report 94 items). Only parent and teacher reports assess peer relations, whereas self-ratings require family relations.
The Conners 3U+00AE rating scales were translated into German, backtranslated, and norms for a German-speaking sample were established (Lidzba et al., 2013). All content scale items of this version were subjected to the following statistical analyses.
Statistical Analyses
All raw data were stored at a database in Tübingen, Germany (University Children’s Hospital Tübingen, Department of Pediatric Neurology). Data reduction and analyses were carried out using the statistical package SPSS 19.0 incorporating AMOS, a program for structural equation modeling.
According to Conners (2008), items were included in factor analyses if the following criteria were met: (a) Items loaded significantly (> .35) on a given factor and lower than .35 on all other factors; (b) Conceptual coherence was evident. The Scree-test and eigenvalues above 1.0 were used to select the number of factors for rotation.
First, a series of exploratory factor analyses with all content scale items was conducted to determine factor structure and item loadings for the German parent, teacher, and child samples (German models). Subsequently, Cronbach’s U+03B1 was calculated for the emerging scales. Second, all content scale items were used for confirmatory factor analyses restricted to the factors as published by Conners (2008). Again, Cronbach’s U+03B1 was calculated for the emerging scales. Third, goodness of fit for the replicated American model was calculated using structural equation modeling. We calculated several model fit indices to evaluate the results of our analyses. The root mean square residual (RMR) measures the mean absolute value of the covariance residuals (Kline, 2005). Values less than .05 indicate a good model fit (Blunch, 2008), though other authors state that a value of less than .10 signals an acceptable model fit (Arbuckle, 2008; Weiber & Mühlhaus, 2009). The standardized root mean square residual (SRMR) eliminates scaling effects of the RMR. Values .10 indicate a good model fit (Weiber & Mühlhaus, 2009). The global fit index (GFI) can be considered as a measure of the proportion of variance and covariance a given model is able to explain. A GFI equal or higher .90 can be considered as reflecting a good model fit (Raykov & Marcoulides, 2006). The adjusted global fit index (AGFI) takes the number of parameters used in computing the GFI into account. An AGFI equal to or higher than .90 can be considered as showing a good model fit (Weiber & Mühlhaus, 2009).
With Cronbach’s U+03B1 we investigated the internal consistency of the postulated factors in the different questionnaires. Values ≤ .10 are considered to be acceptable (George & Mallery, 2010).
Results
Exploratory Factor Analyses of the Conners 3U+00AE Parent, Teacher, and Child Rating Scales
Our exploratory factor analyses for the three German Conners 3U+00AE versions resulted in overall different factor loadings from the American original (for comparison purposes, the factor structure of the original scales can be found for parent, teacher, and child ratings in Tables 8–10 in the supplemental material, which also report the results of the confirmatory factor analyses). Different from Conners’ study (2008), the hypothetical subscale Inattention could be extracted for all three scales (parent, teacher, child ratings). This subscale also explained most variance in all three exploratory factor analyses. Different from the American original, we were unable to separately extract factors for Learning Problems/Executive Functioning in any version; items loaded instead on a broader factor named Inattention/Learning Problems. In contrast, the American factor Defiance/Aggression is perceived more differentially by German parents with a factor related to Aggression in the sense of CD and a factor more related to Defiance in the sense of ODD. Further, child ratings resulted in a factor Impulsive Communication, separating impulsivity items related to blurting out and disrupting communication from impulsive behavior. See Table 2 for details on the EFA analyses as well as Tables 5, 6, 7 in the supplementary material.
Confirmatory Factor Analyses
Only content scale items of the parent, teacher, and child ratings were used for this analysis. For the parent (n = 953), teacher (n = 741), and child ratings (n = 745), a maximum of 4.2 %, 2.8 %, and 4.3 % single items were missing completely at random, respectively. These were replaced with the expectation maximization (EM) algorithm (Enders, 2010). Under the assumption of conceptual coherence between factors, they were allowed to correlate and oblique rotations were used. Unweighted least squares (ULS) were used for estimation, since this procedure does not make assumptions about the distribution (Blunch, 2008). The more degrees of freedom the resulting model has, the higher the probability of rejecting the model. Thus, models with high degrees of freedom are favorable since they are assumed to be very robust (Blunch, 2008; Raykov & Marcoulides, 2006). All models (parent, teacher, child) show very high degrees of freedom and overall very satisfying model fits. Table 3 shows the degrees of freedom and model fit indices with their cut-off values for the three confirmatory models (parent, teacher, child). Details from confirmatory analyses are presented in the supplemental material (Tables 8 to 10).
Since we did not assume the independence of subscales, but rather correlations between the content scales, intercorrelations between scales were analyzed. The content scales of the parent, teacher, and child ratings correlated between .32 and .93, as shown in Table 4. Especially the assumed dependence between Inattention and Learning Problems/Executive Functions was demonstrated for all three versions of the Conners 3U+00AE scales.
Cronbach’s U+03B1 was acceptable for all scales of all versions. In the parent version, values were .92 for Inattention, .92 for Hyperactivity, .80 for Learning Problems, .87 for Executive Functions, .85 for Aggression, and .74 for Peer Relations. In the teacher version, values were .94 for Inattention, .96 for Hyperactivity, .94 for Learning Problems/Executive Functions, .94 for Aggression, and .88 for Peer Relations. In the child version, values were .89 for Inattention, .86 for Hyperactivity, .72 for Learning Problems, .86 for Family Relations, and .76 for Aggression.
Discussion
We explored the factor structure of the German translation of the Conners 3U+00AE (Lidzba et al., 2013) in a large sample of German-speaking children. Exploratory factor analyses of the parent, teacher, and child ratings resulted in overall different structures from the American original, though matching ADHD symptoms very well. As opposed to the American model, we were able to extract an Inattention/Learning Problem scale for all three versions of the Conners 3U+00AE. Thus, we provide psychometric evidence for the Inattention scale, which was only hypothetically derived in the American model. But, different from the American original, we were unable to extract separate subscales for Inattention, Learning Problems, and Executive Functions for all three versions, though items related to those scales loaded on one factor that could be termed Inattention/Learning Problems. Since inattentive behavior is obviously very closely related to learning problems and executive functions, it might be the case that parents, teachers, and children in Germany do not differentiate between those different domains, but rather consider learning problems and executive functions to be part of attention problems. When asked, teachers, parents, and children report that school and homework situations requiring skills from those three domains are the most problematic situations (Breuer & Döpfner, 1997; Östberg & Rydell, 2012; Salbach et al., 2005). Thus, this may be perceived as one big problematic domain, fine differences not being distinguished.
The EFAs of teacher ratings resulted in the highest amount of variance explained with almost 60 %, followed by parent ratings with 53 %, and the child ratings with only 41 %. Internal consistency was highly satisfactory for the majority of derived subscales.
Of the teacher ratings the subscale Peer Relations only resulted in a Cronbach’s U+03B1 of .65. Taking into account that teachers only rate relations observed at school which might considerably differ from out-of-school relations might explain the relatively lower internal consistency.
On the child form, the subscale Aggression resulted in an U+03B1 of .62. It might be harder for children to judge their own conduct problems and family relations, though they did not seem to have difficulties rating core ADHD symptoms. Interestingly, a factor we termed Impulsive Communication Style emerged in the exploratory analysis of the child ratings. Different from the American original, children in this sample seem to differentiate between impulsive actions closely linked to hyperactive behavior, and impulsivity related to communication, such as blurting out and disrupting communication of others.
The confirmatory factor analyses with only the content scale items included in the original American study showed such good model-fits in our German speaking sample for all three Conners 3U+00AE versions that ratings according to those structures are very well justified. The Cronbach’s U+03B1s for all scales of all versions were acceptable to high. This is very important with respect to clinical practice and research, since they provide clinicians and researchers with carefully constructed and psychometrically sound scales for the assessment of juvenile ADHD. Considering that many studies today are international multicenter studies, it is highly important to be able to rely on psychometrically sound instruments with cross-cultural comparable factor structure which meet diagnostic requirements as formulated in the existing diagnostic manuals (DSM-IV/ICD-10), and guidelines (NICE guidelines [CG072] (2008). URL: http://www.nice.org.uk/CG072 fullguideline; Deutsche Gesellschaft für Kinder- und Jugendpsychiatrie und Psychotherapie, 2007).
The high dependence of Inattention and Learning Problems/Executive Functions, demonstrated for the exploratory factor analyses and explaining the largest amount of variance there, is reflected in the very high intercorrelations between those factors in the confirmatory analyses, highlighting the importance of this impairment domain.
Limitations
Some limitations have to be taken into account. When doing psychometrics on a scale, one should generally choose the sample of subjects chosen to complete the scale which is similar to the population the scale was written for. In this case, the intended population is children with ADHD, though this study is based on a convenience sample in which only a minority of 7.2 % had an ADHD diagnosis (according to parent information). Thus, any psychometric statistics generated are biased by properties of the sample and might not be attributable solely to the scale. Even though we assessed a large German-speaking sample, this is not normative for the German population because of the convenience sampling. The majority of participants came from higher educational schools or elementary school, and only a small minority from special education contexts, where more children with ADHD might have been found. Children diagnosed with ADHD were assessed as well, however, and results of this sample as well as data on convergent and discriminant validity and reliability are currently in preparation, as are the cross-validation data of the three Conners 3U+00AE forms. Finally, longitudinal studies would greatly enhance predictive validity of this measure.
Conclusion
The results of the study show that the German version of all three forms of the Conners 3U+00AE have good factorial validity. Even though the exploratory factor analyses yielded different factors than the American original study, confirmatory factor analyses showed excellent model-fits, so that, with respect to international studies, the use of the American factor structure is justified and to be recommended to facilitate international research on ADHD.
Conflicts of interest: Karen Lidzba received financial support by the Hogrefe Publishing Group for validation of the Conners 3U+00AE, but this has not influenced results. All other parties have not received funding for the study, except reimbursement of mailing costs. There are no conflicts of interest.
References
Literature
(2008). Amos 17.0 user’s guide. Chicago: SPSS.
(2008). Introduction to structural equation modeling using SPSS and AMOS. London: Sage.
(1997). Die Erfassung von problematischen Situationen in der Familie. Assessment of difficult situations at home. Praxis der Kinderpsychologie und Kinderpsychiatrie, 46, 583–596.
(2009). Using the Conners’ Teacher Rating Scale-Revised in school children referred for assessment. Canadian Journal of Psychiatry/La Revue canadienne de psychiatrie, 54, 232–241.
(2012). Psychische Störungen des Kindes- und Jugendalters. Psychological disorders of childhood and adolescence. In W. Rief & M. Berking (Eds.), Klinische Psychologie und Psychotherapie: Band II (pp. 213–224). Berlin: Springer.
(1997). Conners’ Rating Scales-Revised. North Tonawanda, NY: Multi-Health Systems.
(1989). Conners› Rating Scales. Manual. Toronto, Ontario, Canada: Multi-Health Systems.
(2008). Conners 3 Edition. Manual. Toronto, Ontario, Canada: Multi-Health Systems.
(2008), The usefulness of Conners› Rating Scales-Revised in screening for attention deficit hyperactivity disorder in children with intellectual disabilities and borderline intelligence. Journal of Intellectual Disability Research, 52, 950–965.
u. a.). (2007). Leitlinien zur Diagnostik und Therapie von pschischen Störungen im Säuglings-, Kindes- und Jugendalter. Hyperkinetische Störungen. Köln: Deutscher Ärzte Verlag.
(2010). Applied missing data analysis. New York: Guilford.
(2010). SPSS for Windows step by step: A simple guide and reference (18.0 update). Boston: Allyn & Bacon.
(1999). Diagnosis of attention-deficit/hyperactivity disorder. Technical Reviews, No. 3. Rockville, MD: Agency for Health Care Policy and Research.
(2001). Interkultureller Vergleich der Conners-Skalen: Lässt sich die US-amerikanische Faktorenstruktur an einer deutschen Stichprobe replizieren? [translation please] Zeitschrift für Kinder- und Jugendpsychiatrie und Psychotherapie, 29, 16–24.
(2002). ADHS im Lehrerurteil: Ein Vergleich von Klinik- und Normstichprobe anhand der Conners-Skalen. Kindheit und Entwicklung, 11, 90–97.
(2005). Principles and practice of structural equation modeling. New York: Guilford.
(2013). Conners 3 Skalen zu Aufmerksamkeit und Verhalten: Deutschsprachige Adaptation der Conners 3 Editon (Conners 3) von C. Keith Conners. Bern: Huber.
(2011a). The impact of study design and diagnostic approach in a large multi-centre ADHD study. Part 1: ADHD symptom patterns. BMC Psychiatry, 11, 54.
(2011b). The impact of study design and diagnostic approach in a large multi-centre ADHD study: Part 2: Dimensional measures of psychopathology and intelligence. BMC Psychiatry, 11, 55.
(2012). An efficacy study of a combined parent and teacher management training programme for children with ADHD. Nordic Journal of Psychiatry, 66, 123–130.
(2006). A first course in structural equation modelling. Mahwah, NJ: Erlbaum.
(2005). Die Wirksamkeit eines Gruppentrainings für Eltern hyperkinetischer Kinder. Zeitschrift für Kinder- und Jugendpsychiatrie und Psychotherapie, 32, 59–68.
(2009). Strukturgleichungsmodellierung: Eine anwendungsorientierte Einführung in die Kausalanalyse. Berlin: Springer.
www.nice.org.uk/CG072fullguideline [please provide full bibliographical information]
