Test Review: C. K. Conners. Conners Early Childhood Manual

General Description

The Conners Early Childhood (Conners EC; Conners, 2009) is a behavior and development rating scale intended to assess children in early childhood, specifically defined as ages 2 to 6 years. Using multiple informants across multiple settings, the Conners EC is administered for the purpose of early identification of disorders or developmental delays, screening, and planning and monitoring interventions for preschool-aged children. Although the assessment is designed to be administered individually, instructions are included to accommodate group screening. The assessment utilizes parent report (191 items) and teacher/caregiver-report (187 items) forms to facilitate a multi-informant approach that comprehensively captures a child’s behavior.

Scoring

The Conners EC can be scored using paper-and-pencil scoring, Conners EC Scoring Software, or Conners EC Online scoring. Both forms of the Conners EC consist of two domains: the Behavior scales and the Developmental Milestone scales.

The Conners EC Behavior scales are measured using a 4-point Likert-type scale from 0 to 3: “Not at all true (Never, Seldom),” “Just a little true (Occasionally),” “Pretty much true (Often, Quite a bit),” and “Very much true (Very often, Very frequently).” These responses are restricted to frequency of specific behaviors as observed within the past month. Similarly, the Conners EC Developmental Milestone scales are scored using a 3-point Likert-type scale (0-2) where responses are: “No, he/she never or rarely does this without help,” “He/she sometimes does this without help,” or “Yes, he/she always does this without any help.” In this instance, raters are asked to evaluate the child’s level of mastery in developmentally appropriate skills and thus are not restricted to the past month.

The Conners EC Behavior domain consists of eight subscales that target specific impairment clusters of interest in early identification. These subscales indicate impairment in the areas of inattention/hyperactivity, defiant/aggressive behaviors, social functioning/atypical behaviors, anxiety, mood and affect, and physical symptoms. The Behavior domain also produces a Global Index (ECGI), which indicates restlessness/impulsivity and emotional liability. Finally, the Behavior domain rating scale includes a measure of other clinical indicators, which includes clinical indicators of potential problem areas (e.g., tics, pica, and posttraumatic stress disorder) that may warrant further clinical attention.

The Conners EC Developmental Milestone domain yields five subscales that indicate specific areas of concern for development. These subscales are Adaptive Skills (dressing, eating/drinking, toileting, hygiene, and helping), Communication (expressive and receptive), Motor Skills (fine and gross), Play, and Pre-Academic/Cognitive skills.

Validity scales are also included, which measure Positive Impression, Negative Impression, and Inconsistency, and consist of items from the Conners EC. Positive and Negative Impression measure whether the rater is portraying an overly positive or negative bias in regards to the child’s behavior and each consist of 6 items. The Inconsistency scale consists of 6 item pairs in which disparate scores indicate potentially unmotivated or inattentive raters.

Standardized scores are provided in linear T-scores (M = 50, SD = 10). Linear T-scores were employed for their ease of interpretation due to several subscales producing skewed distributions in the normative sample. Scores are standardized based on gender (males and females) and chronological age, grouped in 6 month age intervals (i.e., 2.0-2.5 years, 2.6-2.11 years, etc.). Percentiles are provided in empirical percentiles based on the scores from the normative sample. Computer generated score reports include a T-score graph and a Table of Detailed Scores, which includes raw scores, T-scores, and a text interpretation guideline describing common characteristics of individuals who obtain high scores or high levels of developmental concern.

Test Materials

The Conners EC can be administered via paper-and-pencil or online formats and is designed to be very user friendly. Both forms take approximately 25 min to complete and all items were generated using the lowest reading level possible. Using the Flesch-Kincaid Grade Level Formula (Flesch, 1948; Kincaid, Fishburne, Rogers, & Chissom, 1975), items have a maximum reading level of 5.4 for the parent form and 6.1 for the teacher/caregiver form. The time constraints of the Conners EC Behavior items (i.e., occurring in the past month) ensure that respondents have known the child for an appropriate amount of time. The administration time, accessibility, and existence of multiple forms encourages the use of multiple raters, helping to ensure appropriateness of interpretation.

Aside from the technical manual (discussed in greater detail below), the publisher provides both software and online resources to assist in the interpretation of the scores. The Conners EC Scoring Software is provided on a USB drive, fostering easy transference from one computer to another. The Scoring Software generates score reports and highlights scores of interest that indicate areas of potential impairment. Scores must be manually entered if the test was taken via the paper-and-pencil format (i.e., no option to scan in scores). The tools accessed through Conners EC Online can be accessed from any computer with internet. Base scores and score distributions also are provided online to aid in interpretation of results. A specific benefit of the online scoring resources is that the raters have the option of completing the forms online, which allows for independent completion and eliminates the need to manually type in data.

Test Manual

The test manual is very well organized and uniquely user friendly for the various professionals who might use this test. At the same time, the manual succeeds in retaining its technical accuracy. Development and rationale of the test are clearly defined, and all expected technical data are provided. The manual is designed in a way that a trained paraprofessional or staff member can administer and score the assessment by following the protocols, and the results are interpreted by a qualified assessor.

Test Construction and Item Analysis

All items were generated based on clinical experience, Diagnostic and Statistical Manual of Mental Disorders 4th Edition-Text Revision (DSM IV-TR; APA, 2000) constructs, and research literature. Items generated were designed to address the previously determined domains of interest, and exploratory factor analyses were run on the initial item pool. Items that did not load greater than .35 on any factor, or cross-loaded on multiple factors, were excluded from the final form. Final items were then administered to a normative sample to generate validity and reliability data.

The Positive and Negative Impression Validity Scales were created using items likely to detect potential patterns of positive or negative bias. More items were created than necessary so that the strongest items could be identified and retained for these scales. The Inconsistency index was constructed in a similar fashion, except pairs of intercorrelated items were tested for strength in detecting inconsistent responses.

Bias for sex and race by specific subscales was analyzed and effects were determined to be small for both. Sex accounted for 1.9% of the variance for the parent forms and 1.1%-4.1% of the variance for the teacher/caregiver forms. Race accounted for 1.5%-5.4% of the variance for the parent forms and 1.3%-4.0% for the teacher/caregiver forms. Age was found to account for a large portion of the variance for the Conners EC Developmental Milestone domain for both the parent and teacher/caregiver forms (27.3%-44.5% and 30.0%-53.4%, respectively), as would be expected from a measure of development.

Standardization Sample

A total of 3,281 Conners EC assessments were collected nationwide. The normative sample was extracted from 2,567 ratings of children from the general population, including 40 boys and 40 girls in each age group. The racial/ethnic distribution of the sample resembled the U.S. population. In addition, a normative sample of 714 ratings for children with clinical diagnoses was obtained for the purpose of ensuring the accuracy of the Conners EC indicators.

The sample was then stratified by demographic variables (e.g., sex, ethnicity, age, etc.) in order to match the U.S. population by proportion (based on the 2000 U.S. Census). Once this was done, 40 boys and 40 girls from each age group were randomly extracted from each rating group to create the normative sample, with the only limitation to the randomization being that the racial/ethnic distribution of the sample had to resemble the U.S. population.

Reliability

Results of the Conners EC normative sample data demonstrated excellent levels of consistency on both the parent and teacher forms of the Behavior (B) and Developmental Milestone (DM) scales. Measures of internal consistency (Cronbach’s alpha) were high for both the parent (M_B = .86, M_DM=.93) and teacher/caregiver (M_B = .89, M_DM = .93) forms. The only subscale that did not meet the criteria for good internal consistency (defined as .70 or above by the test developers) was the Sleep Problems subscale (M = .64), which is only found on the parent form.

The assessment also demonstrated statistically significant test–rest reliability. A sample of 78 parents and 68 teachers filled out their respective forms on two separate occasions over a 2 to 4 week interval. Correlation coefficients were found to be statistically significant (p < .001); however, the coefficients for the Atypical Behaviors (r = .71) and Sleep Problems subscale (r = .67) were noticeably low in comparison to all others.

Finally, the manual reports excellent inter-rater reliability. Inter-rater reliability correlation coefficients were calculated using the parent form, with statistically significant agreement (p < .001) found for scores on all scales. Inter-rater reliability correlations were not conducted using the teacher/caregiver form because many children in the standardization sample were primarily cared for by a single teacher or childcare provider. Instead, across-informant validity correlations were conducted and are discussed below.

Validity

Results with the normative sample also demonstrated adequate validity through different approaches. Using data from a clinical group within the normative sample, a series of Analyses of Covariance (ANCOVA) and discriminant function analyses (DFA) were conducted to demonstrate discriminative validity. ANCOVA analyses indicated that clinical classification had medium to large effect size (p < .01) on scores for every scale on both the parent forms (partial η² = .122-.535) and teacher/caregiver forms (partial η² = .099-.491). DFA also indicated a high degree of correct classification for both parents (86.5%) and teachers/caregivers (86.5%).

Convergent and divergent validity were also demonstrated through use of correlation coefficients (r). All scales that theoretically measured the same constructs were moderately to strongly related to one another (p < .001). For example, there was a notably high correlation for results of both the parent (r = .91) and teacher/caregiver (r = .86) forms of the Conners EC Defiant/Aggressive Behaviors scale and the Behavioral Assessment System for Children, 2nd Edition (BASC-2; Reynolds & Kamphaus, 2006) Aggression scales. Conversely, scores for scales that theoretically do not measure the same constructs were not significantly related to one another.

Across-informant validity was also demonstrated using correlation coefficients (r) to measure similarity in scores for all scales between the parent forms and the teacher/caregiver forms. Correlations for all scales were found to be moderate to large (r = .46-.87, p < .001), indicating consistency between both parent and teacher/caregiver ratings of the same child.