The Conners EC Global Index is a 10-item scale that assesses the presence of general psychopathology over the last month in children between 2 and 6 years. This scale is part of the Conners Early Childhood (Conners, 2009), but it also can be administered independently. Items are scored on a four-point Likert scale and divided into two subscales that assess symptoms related to ADHD: restless-impulsive and emotional lability. The scale has a cut-off for elevated scores (T 65-69) and for very elevated scores (T ≥70), indicating more and many more concerns, respectively, than typically reported in children. The parents’ and teachers’ forms have shown a high internal reliability (Cronbach's alpha), with values of .89 and .92 respectively. The restless-impulsive’ subscale has an internal reliability of .87 in the parents’ form and .91 in the teachers’ form, while the emotional lability subscale ranged between .75 and .86, respectively. Inter-rater agreement is .75 (p=.001) for the total scale, .75 (p=.001) for the restless-impulsive subscale and .62 (p=.001) for the emotional lability subscale.

The Conners 3 ADHD Index (Conners’ Hyperactivity Index) is a 10-item scale that assesses the presence of the most prominent symptoms of ADHD over the last month in children between 6 and 18 years. This scale is part of the Conners Rating Scale 3rd Edition (Conners, 2008), but it also can be administered independently. The authors recommend the use of this index as a research and clinical screening tool in large samples. Items are based on DSM-5 criteria and scored on a four-point Likert scale. The scale has a cut-off for elevated scores (T 65-69) and for very elevated scores (T ≥70). Parents’ and teacher’ forms have shown a high internal reliability, with values of .90 and .93 respectively. Inter-rater agreement is.56 (p=.001).

The Schedule for Affective Disorders and Schizophrenia for School Aged Children (K-SADS-PL; Kaufman et al., 1997) is a semi-structured psychiatric interview that yields DSM diagnoses and has been widely used in studies of child psychopathology. We used the K-SADS-PL to collect information on child ADHD symptomatology from parents. The Spanish version has shown a high inter-rater reliability (K = .91) for the ASD scale (Ulloa et al., 2006).

The Child Behavior Checklist is a parent report to assess psychopathological symptoms among preschoolers (CBCL 1½-5; Achenbach & Rescorla, 2000) and school-aged children (CBCL 6-18, Achenbach & Rescorla, 2001). CBCL 1½-5 has a list of 110 items of behavioural problems that provided scores for seven scales; emotionally reactive, anxious/depressed, somatic complaints, withdrawn, sleep problems, attention problems and aggressive behavior. CBCL 6-18 has a list of 113 items of behavioral problems that provided scores for eight scales; anxious/depressed, withdrawn/depressed, somatic complaints, social problems, thought problems, attention problems, rule-breaking behavior and aggressive behavior. Both questionnaires provide DSM scales. In the preschool form, De la Osa, Granero, Trepat, Domenech, and Ezpeleta (2016) reported an internal reliability between .65 and .86 for the syndrome scales and .50 and .78 for the DSM-5 scales. The internal reliability of the Spanish school-aged version ranged from .71 to .87 (Sardinero, Pedreira, & Muñiz, 1997).

The academic achievement was recruited by latest school marks and used to obtain teacher qualitative and quantitative reports. Among preschoolers, school marks were coded as low (1), medium (2) and high (3) achievement in self-care, social interaction, motor skills, science, art, mathematics and language. Among school-aged children, school ratings included motor skills, science, art, mathematics, language and foreign language. School marks were coded as insufficient, sufficient, good, very good and excellent and ranged from 0 to 10. The mean score of all the school marks was calculated to obtain the child global achievement.

A sociodemographic questionnaire addressed to parents and teachers was developed ad-hoc in order to gather information during the first phase from the children (age, gender, place of birth and previous diagnosis) and their families (place of birth, educational level and parents’ work and marital status). Socioeconomic level was estimated according to Hollingshead.

If any items were missing from any of the questionnaires administered and they did not amount to more than 10% of the total, we imputed data on the basis of responses to similar items.

According to the classification of Montero and León (2007), this was a descriptive survey study. The EPINED is a double phase epidemiologic cross-sectional initiated in 2013 in Tarragona (Spain). Before beginning the study, we obtained permission from the Catalan Department of Education and the Ethics Committee at the Sant Joan University Hospital (Reus). All the school boards contacted agreed to participate in the study.

Through the teachers, we sent to all the families a letter informing them about the study and asking for their written informed consent. The participation was voluntary and disinterested. The assessment procedure was conducted by a research group from the Rovira i Virgili University consisting of psychology and medicine professionals specialized in child and adolescent psychopathology. Individual assessments of the child and family interviews were conducted in the schools in order to facilitate participation.

In the first phase, socio-demographic data from all 1,796 children were collated and Conners EC GI and Conners 3 AI were completed by parents and teachers in order to assess ADHD symptoms and identify children at risk of the disorder. An agreement with the Catalan Department of Education allowed us to collect anonymous data from teachers about the children of non-participating families, obtaining data from the whole sample. A total of 49.1% of the families agreed to participate in the study and signed the written informed consent. Thus, information from both informants (teacher and family) was obtained of 882 children (see figure 1).

We considered children to be at risk of ADHD if they obtained a T-score ≥ 65 from both informants. This cut-off score was chosen to include all the subjects with elevated and very elevated scores (Conners, 2008, 2009). In the second phase, parents of subjects at risk of ADHD and a subsample without risk (controls) answered the CBCL and were assessed using K-SADS-PL ADHD. Academic achievement was also recruited. Diagnosis was performed in accordance with DSM-5 criteria and by means of an individual neuropsychological and psychopathological assessment of the child and its family. All the families agreed to participate in this phase and received a complete report of the results.

Descriptive statistics were used to summarize the demographic characteristics of the sample and the average score for the different age groups and forms. Factorial analysis (FA) was used to obtain the factor structure of the Conners 3 AI and Conners EC GI. An Exploratory Factor Analysis (EFA) was conducted on a randomly selected half of the sample. On the second half, a Semi-Confirmatory Factor Analysis (S-CFA) was performed using a semi-specified target matrix. Finally, the same S-CFA was replicated with the whole sample. For all the FA, we used an optimal implementation of Parallel Analysis (Timmerman & Lorenzo-Seva, 2011), an unweighted least squares’ extraction method and, to obtain an oblique rotated solution, we used the direct oblimin procedure (Browne, 1972). Polychoric correlation dispersion matrices were used because of the ordinal nature of the data and because several items showed skewness or kurtosis greater than one in absolute values. For the S-CFA solution, three commonly recommended fit indices were used: the Goodness of Fit Index (GFI), the root mean square of the residuals (RMSR) and the congruence index. As Lorenzo-Seva and ten Berge (2006) suggest, congruence values in the range .85-.94 were interpreted as correspondent with a fair similarity and values higher than .95 were interpreted as implies that the two factors (or components) compared can be considered equal. The evidence of convergent and external validity was computed by Pearson correlations and evidence of the internal reliability of the overall scales and factors was assessed using the reliability estimates of the factor scores, which take into account the ordinal nature of the item responses. To assess the reliability of the overall scale, the first canonical factor was extracted (Ferrando & Lorenzo-Seva, 2013) and its reliability estimate was obtained.

Statistical analyses were performed by SPSS 22.0 and FAs and internal reliabilities were conducted using FACTOR 9.2 (Lorenzo-Seva & Ferrando, 2013).

All the socio-demographic and psychopathological data are shown in table 1. In the first phase there were no significant differences in age, gender, ethnicity or SES between participating and non-participating families. In the second phase, there were significantly more boys than girls in Nursery (z = 2.79, p = .005) and Primary Education (z = 2.09, p = .004).

Conners 3 AI and EC GI have shown good psychometric properties in both parents’ and teacher’ forms. To our knowledge, no studies have been conducted to validate the psychometric properties of the new Conners indexes in different cultures. Some studies have been conducted on Spanish populations with previous versions (Amador et al., 2003; Amador et al., 2002; Farre-Riba and Narbona, 1997), but data for the latest version were not available. Due to their briefness and efficacy in ADHD detection, these scales seem to be appropriate for screening procedures in primary health care, school and research settings.

Factor analysis produced clearly defined and easily labelled factors in accordance with the clinical consistency of the items. Given the statistical indicators, we found high item-factor correlations and a good explained variance of the models, ranging from .62 to .83. As has been previously described in the original North American validated version, in the Conners 3 AI forms, we found two underlying factors corresponding with inattentive and hyperactive-impulsive symptoms. Arias Martínez et al. (2013) obtained similar results for the parents’ form in a Spanish population. Although they consider the scale as “unidimensional enough”, we contemplate the two factor structure based on statistical and clinical criteria. Our results also confirm the two factor structure regarding emotional lability and restless-impulsive symptoms of the Conners EC GI parents’ forms. However, in the teachers’ form we also found a third factor that computed those items that assessed inattentive symptoms, which suggests that inattention could be identified by teachers as an early indicator of ADHD.

The Conners indexes have shown good consistency in our sample. The internal reliabilities were excellent in all the overall scales and factors (ordinal alpha values ranging from .84 to .98). Of all the responses, the teachers’ were the most reliable. These results are similar to those obtained in the original validation (.75-.93), where higher values were also obtained in the teachers’ forms (Conners, 2008, 2009). Previous versions of the scales have also demonstrated a similar and high internal reliability in a Spanish community (.92) and clinical (.87) samples (Amador et al., 2003). Parent-teacher agreement (Cohen's kappa) values were moderate; results were close to the original values in the Conners 3 AI, but were much lower in the Conners EC GI. However, inter-rater agreement was within the .09–.43 range compiled by Narad et al. (2015) from studies on ADHD symptoms across development.

Regarding the evidence of external validity, our data showed that ADHD symptoms interfere with the academic achievement in all age group. The inattentive factor was most associated with academic achievement. Similar relationships have been found in previous studies (González-Castro, Rodríguez, Cueli, García, & Álvarez, 2015; Papaioannou et al., 2016; Scholtens, Rydell, & Yang-Wallentin, 2013) supporting the idea that teachers are good at detecting symptoms of inattention in its early stages. In fact, early intervention on ADHD core symptoms has shown to improve school performance in children (Arnold, Hodgins, Kahle, Madhoo & Kewley, 2015; Moreno-García, Delgado-Pardo, de Rey, Meneres-Sancho, & Servera-Barceló, 2015).

The Spanish forms of the Conners indexes showed moderate to high correlations with K-SADS-PL and CBCL scores, although our results were lower than the original versions. In regard to Conners EC GI, correlations between parents’ form and CBCL ADHD scale, externalizing scale and total score ranged from .43 to .72 in comparison to the original values that ranged from .56 to .93 (Conners, 2009). In relation to Conners 3, the original version only compared the whole scale with CBCL scores and this could explain that their correlations were higher (from .63 to .89) than those obtained in the present study. On the other hand, we didn’t administer the Teacher Report Form and this reason may explain the lowest correlations in teacher’ forms.

In keeping with previous research, mean scores of the Conners indexes were significantly higher among boys (with a high effect size ranging from 0.22 to 0.58) and on parents’ forms (Català-López et al., 2012; Conners, 2008, 2009, Ezpeleta et al., 2014; Narad et al., 2015; O’Neill, Schneiderman, Rajendran, Marks, & Halperin, 2014; Wichstrøm et al., 2013). With regard to percentiles, our scores were higher than the values described in the original version. Raw scores obtained at Pc 98 (T-score ≥ 70) in Conners 3 AI were 16 (TF) and 18 (PF), while the original version describes values of 13 and 6, respectively. Minor differences were observed in the Conners EC GI original version, which describes values of 22 and 20, respectively whereas we obtained 21 (TF) and 26 (PF). This may be due to cultural differences in children's behavior and the informants’ perceptions.

In summary, our results indicate that Conners 3 AI and EC GI are valid questionnaires for the detection of inattentive and hyperactive-impulsive symptoms in children from a non-clinical population. In this sense, the use of these short indexes would improve ADHD screening procedures both in clinical and school settings. However, the present study does have some limitations. Firstly, we consider that replications in larger samples along with an additional statistical analysis (such as ROC curves) should be carried out in order to establish the accuracy, sensitivity and specificity of the Conners indexes and to determine the optimal cut-off points. Also, it would be interesting to carry out validation studies in clinical population.

This study was funded by the Ministry of Economy and Competitiveness of Spain and the European Regional Development Fund (ERDF; ref. PSI2015-64837-P). The sponsor of the study had no role in the study design, data gathering, data analysis, and data interpretation, or writing of the report. It was also supported by the Ministry of Education of Spain (grant FPU2013-01245).