Archives of Clinical Neuropsychology 18 (2003) 431 437 Book review Conners Adult ADHD Rating Scales (CAARS). By C.K. Conners, D. Erhardt, M.A. Sparrow. New York: Multihealth Systems, Inc., 1999 1. Test description In previous years, Attention Deficit Hyperactivity Disorder (ADHD) was thought to be a disorder that primarily affected children. Current research has shown that ADHD is a problem that also affects adults. Although the age of onset for this disorder is in childhood, the problems continue throughout adulthood and is not something an individual will overcome through the mere maturation process. Individuals with ADHD, in addition to having difficulties with sustaining attention, impulsivity, and hyperactivity, also are at risk for other difficulties throughout their lifetime. With the emphasis traditionally on assessment of ADHD in children, there has been a paucity of measures for use in the assessment of ADHD in adults. In addition, while assessment of children often relies on teacher(s) and parent(s) serving as informants, with adults the use of others as informants becomes more problematic. As a result, adult diagnosis is often based on self-report. The Conners Adult ADHD Rating Scales (CAARS) were designed to assess ADHD in adults. The CAARS utilizes short, long, and screening self-report and observer rating scale forms. The instrument is designed for individuals aged 18 through 50 years and older. The scales address ADHD symptoms as described in the Diagnostic and Statistical Manual Fourth Edition (American Psychiatric Association, 1994). Separate T-scores are provided for men and women. The various scales (short, long, screening) provide differing scores and vary in reliability and validity. The long version rating scales include measures of Inattention/Memory problems, Hyperactivity/Restlessness, Impulsivity/Emotional Lability, Problems with Self-Concept, DSM-IV Inattentive Symptoms, DSM-IV Hyperactive-Impulsive Symptoms, DSM-IV ADHD Symptoms Total, and ADHD Index. The short version rating scales only include DSM-IV Inattentive Symptoms, DSM-IV Hyperactive-Impulsive Symptoms, DSM-IV ADHD Symptoms Total, and ADHD Index. The CAARS also utilizes an Inconsistency Index. This score is made up of paired items that can be used to determine whether or not an individual completing the scale may be unmotivated, noncompliant, or may be answering in a haphazard fashion. It may also be used as an indicator of whether or not the respondent was able to understand wording of the test items. Test development, normative data, reliability, and validity of the scales will be discussed. 0887-6177/03/$ see front matter 2003 National Academy of Neuropsychology. doi:10.1016/s0887-6177(03)00021-0
432 Book review / Archives of Clinical Neuropsychology 18 (2003) 431 437 2. Test construction Using the DSM-IV symptom criteria, the Conners Rating Scales revised for Children and Adolescents (Conners, 1997), and current conceptualizations of adult ADHD (Wender, 1995) an item pool was developed for the scale, which resulted in the 4 factor-derived scales. The CAARS are available in three forms in both self-report and observer report. The long form (CAARS-S:L; CAARS-O:L) consists of 66 items with 9 subscales. The subscales include 4 factor-derived scales Inattention/Memory Problems (12 items), Hyperactivity/Restlessness (12 items), Impulsivity/Emotional Lability (12 items), and Problems with Self-Concept (6 items), three DSM-IV ADHD Symptoms subscales which correspond with the criteria established by the DSM-IV Inattentive Symptoms (9 items), Hyperactive-Impulsive Symptoms (9 items), Total ADHD Symptoms subscale, an ADHD Index which contains the set of items which will distinguish ADHD adults form nonclinical adults, and an Inconsistency Index which can be used to identify haphazard responses or difficulty in understanding the measure. These scales represent current conceptualizations of adult ADHD from the previously mentioned sources. From an item pool of 93 items, subjects were asked to respond to statements using a 4-point scale (0 for not at all, never and 3 for very much, frequently ). Factor analysis was used to determine which items should be retained. Items were included based on the following criteria. First, the item had to have conceptual coherence with the factor. Second, factors were determined if items loaded significantly on a given factor, greater than 0.30 and less than 0.30 on other factors. No other information was provided for how item loadings were determined. The ADHD Index provides a method of identifying those adults who are likely to be diagnosed with ADHD (Conners, Erhardt, & Sparrow, 1999, p. 51). The items for this scale were selected based on whether or not they appeared to differentiate between ADHD and non-adhd adults. The Inconsistency Index was designed to determine whether or not an individual was responding randomly to scales items. Items were paired and then correlated. Correlation values range from r =.65 to r =.42 for the eight pairs included on this subscale. Scores of 8 or higher may be considered an indication of an inconsistent response pattern. The cutoff of 8 or higher on the Inconsistency Index was established based how successful the score is at differentiating 100 respondents from 100 sets of random numbers. The DSM-IV Symptoms scale is derived from the Conners (1997) self-report measure for adolescents that adapts the ADHD criteria of the DSM-IV. Wording was modified for use with adult clients. The short and screening forms were designed with fewer items than the long version using confirmatory factor analysis testing a one-dimensional model to select five items for each subscale (Conners et al., 1999). Items were selected only if they had high loadings with the latent variable (Conners et al., 1999). The short form (CAARS-S:S; CAARS-O:S) contains 26 items and 6 subscales. An abbreviated version of the 4 factor-derived scales included Inattention/Memory Problems (5 items), Hyperactivity/Restlessness (5 items), Impulsivity/Emotional Lability (5 items), and Problems with Self-Concept (5 items) which are subsets from the long form. The short form also includes the 12-item ADHD Index and the Inconsistency Index. The screening form (CAARS-S:SV; CAARS-O:SV) contains 30 items and the 3 DSM-IV ADHD symptoms measures. The DSM-IV ADHD symptom subscales include Inattentive
Book review / Archives of Clinical Neuropsychology 18 (2003) 431 437 433 Symptoms (9 items), Hyperactive-Impulsive Symptoms (9 items), Total ADHD Symptoms subscale, and the ADHD Index (12 items). 3. Standardization The CAARS self-report forms norms were developed on 1,026 adults aged 18 through 50 years and higher. The observer norms were developed on 943 individuals. Individuals who participated were from the United States and Canada. No other specific information is provided for the sample. Age and gender differences were found in the data analysis resulting in separate norms for age and gender. The manual provides detailed tables containing means and standard deviations for both self-report and observer forms as well as results of ANOVA for each subscale. 4. Administration Administration of the CAARS requires a protocol and a pencil or a ballpoint pen. The long form can be completed in approximately 30 min while the short and screening versions can be completed in approximately 10 min. Nine administration steps are provided. Examiners should tell the individual that the scale will provide information about feelings and behaviors. The 4-level Likert scale should also be explained 0 for not at all or never, 1 for just a little, once in a while, 2 for pretty much often, and 3 for very much, very frequently. Individuals are to be instructed to select the best answer if they are unsure how to respond. A post administration debriefing also is recommended in the manual. According to the Dale Chall formula (1948), that establishes reading level of an instrument, the CAARS, overall, requires a fourth grade reading level in order to be completed. The manual provides a complete readability analysis of all of the forms. The CAARS manual indicates that the rating scales can be used for routine screening of individuals in clinics, treatment centers, private practice, prisons, and psychiatric hospitals. When determining which form to use with individuals, the manual briefly contrasts each measure. The long form should be administered whenever possible since it encompasses all symptoms of ADHD and will provide the most information. The short form may be more suited when individuals repeatedly will complete the scale over a period of time (e.g., treatment or research study). In selecting the short form over the screening version, practitioners or researchers should consider their own needs. These two versions are approximately the same in length. The short version addresses core symptoms of ADHD through the ADHD Index, related problem areas, and contains the Inconsistency Index. The screening version only included the DSM-IV subscales and the ADHD Index. 5. Scoring and interpretation The CAARS manual provides detailed steps to score protocols, which includes procedures specific to the form being scored. Hand scoring requires a minimal amount of time, which
434 Book review / Archives of Clinical Neuropsychology 18 (2003) 431 437 includes transferring scores to columns, adding up each column, and completing the Inconsistency Index. If items were not completed a 0 can be transferred to the column for that item. One should note that if two or more items are missing for a subscale, the subscale should be considered invalid. If three or more items are missing from the short or screening form or five or more items are missing from the long form, the entire protocol is considered invalid. When profiling results, raw scores are converted to T-scores. Profiles are specific to the gender and age group of the individual. Separate norms are available for men and women. When scoring, the correct gender must be selected first, and then the correct age category column is selected. Raw scores from the scoring sheet are circled in the columns. The circles are then connected by a line to obtain the shape of the profile (Conners et al., 1999, p. 9). In addition to the detailed scoring and profiling instructions, the manual also includes examples of this process with copies of protocols to guide users. The manual recommends interpreting the CAARS by examining item responses, subscale scores, and profile patterns. Symptoms indicated at the item level provide information for making recommendations. Subscale scores can be interpreted using T-scores or percentile. Scores depending on range fall into qualitative categories, though the manual cautions against using these categories as absolutes. In profile interpretation, the examiner should look at the number of subscales that fall above a T-score of 65. The number of scales with clinically relevant elevations the greater the chance that scores indicate a significant problem. The manual also provides a step-by-step approach to interpreting the CAARS. First, Does the CAARS provide valid information about ADHD symptoms? The examiner should examine the Inconsistency Index to determine if responses are consistent and a valid representation of the individual s symptoms. Second, Which item responses are elevated? Third, examine the subscale scores and the overall level of symptomatology. The manual points out that the ADHD Index represents a measure of the overall level of ADHD-related symptoms. This is the best screen for identifying those at-risk for ADHD (Conners et al., 1999, p. 23). The T-scores above 65 are clinically significant. Fourth, the examiner must integrate information from the self-report and observer report forms. Fifth, integrate information from other sources. Sixth, consider the diagnosis and make recommendations. The manual also provides six case studies and CAARS results to provide concrete examples when the scale can be used and they present problems that commonly are presented in clinical practice. 6. Technical adequacy 6.1. Reliability Four types of reliability tests were completed on the CAARS. First, internal consistency reliability was calculated. Internal consistency reliability indicates whether or not items on a scale consistently measure the same construct. Cronbach s alpha across age, subscales and forms men ranged from 0.64 to 0.91 and for women ranged from 0.49 to 0.90. Second, mean inter-item correlations were calculated. This reliability measure indicates the degree to which items on a particular scale consistently measure the same construct. According to Hogan and Nicholson (1988) the higher the mean inter-item correlations the greater chance
Book review / Archives of Clinical Neuropsychology 18 (2003) 431 437 435 the scale is one-dimensional. Across age, subscales and forms Cronbach s alpha for men ranged from 0.31 to 0.68 and for women ranged from 0.26 to 0.63. Third, test retest reliability measures the stability of the scale given multiple administrations. Test retest reliability was calculated based on 33 men and 28 women who were seen at an adult ADHD clinic. The CAARS-S:L was completed twice, 1 month apart. Correlation values range from.88 to.91. The CAARS-O:L was examined in a sample of nonclinical individuals (24 men and 26 women) who were rated by their spouses. Correlation values ranged from.85 to.95. Finally, standard error of measurement (SEM) and standard error of prediction (SEP) were calculated. The SEM was calculated for all age groups and for self-report and observer forms. The manual provides a complete table with all of the SEM values. The SEP was calculated based on the test retest reliability estimates, which may provide evidence of the degree to which scores will vary between original scores (Conners et al., 1999). The SEP values are calculated for all age groups and forms and the manual provides complete tables with all of the values. 6.2. Validity Confirmatory factor analysis was conducted on the short and long self-report and observer forms. Goodness of Fit Index (GFI), Adjusted Goodness of Fit Index (AGFI), Non-normed Fit Index (NNFI), and Confirmatory Fit Index were conducted. Guidelines for acceptable values follow the recommendations of Cole (1987) and Marsh, Balla, and McDonald (1988): GFI > 0.850, AGFI > 0.800, NNFI > 0.900, CFI > 0.900. The four factor structure of the CAARS-S:L, CAARS-S:S, CAARS-O:L, and CAARS-O:S (Inattention/Memory Problems, Hyperactivity, Impulsivity/Emotional Lability, and Problems with Self-Concept) met the criteria for good fit. Intercorrelations of the self-report and observer scales (short and long forms) were completed. The manual provides four tables that detail the information. Discriminant validity studies, which examines whether or not an instrument is able to differentiate between clinical and nonclinical groups, also are reported in the manual. One study (Erhardt, Conners, Epstein, Parker, & Sitarenios, 1999) found that the CAARS produced an overall correct classification rate of 85%. The ADHD Index also was cross-validated on a sample (N = 192) of ADHD and non-adhd adults (Conners et al., 1999). Sensitivity was 71%, specificity was 75%, positive predictive power was 74%, negative predictive value was 72%, false positive rate was 25%, false negative rate was 29%, kappa coefficient was 0.458, and overall classification rate was 73%. These results indicate the ADHD Index may be used to identify adults who would benefit from a full assessment (Conners et al., 1999). Construct validity is reported regarding current ADHD symptoms and retrospective reports of symptoms from childhood or adolescence and the relationship between self-report and observer ratings of symptoms (Conners et al., 1999). Erhardt et al. (1999) studied the relationship between adult and childhood ADHD symptoms. The CAARS-S:L and Wender Utah Rating Scale (WURS; Ward, Wender, & Reimherr, 1993) were completed by 101 clinical individuals. The two scales correlate as follows: Inattention/Memory Problems, r =.37; Hyperactivity, r =.48; Impulsivity/Emotional Lability, r =.67; and Problems with Self-Concept, r =.37
436 Book review / Archives of Clinical Neuropsychology 18 (2003) 431 437 (P <.01) (Conners, 1997). The relationship between self-report and observer forms were examined based on a sample of 188 adults who completed the CAARS-S:L and also were rated by a significant other. Moderate to high correlations were found (.12.61). 7. Summary Research provides evidence that ADHD is not only a disorder of childhood and that symptoms and the disorder can continue during adolescence and into adulthood. A need for a rating scale measure for adults exists and Conners et al. (1999) have developed such a measure. This rating scale system provides measures in three versions (long, short, and screening) for two informants (self and observer reports). In terms of normative data, the manual provides very limited information on the sample of individuals. Information about ethnicity, region of the United States, and Canada, and socioeconomic status is important for generalization of test scores and to determine for which individuals the test is appropriate. Reliability of the CAARS ranged from moderate to high for internal consistency and low to moderate for mean inter-item correlations. The CAARS was found to correlate moderately with one other measure of adult ADHD and discriminate between clinical and nonclinical groups. Further study of the CAARS is needed as part of the standardization process. Since no information is provided on the ethnic representation of the normative sample, and no studies address issues of ethnic differences or bias, caution should be exercised when using the CAARS with individuals of minority status. Research is needed in this area. Acknowledgments The author would like to acknowledge and express appreciation to Dr. Cynthia A. Riccio, Department of Educational Psychology, Texas A & M University for all of her time, encouragement, and assistance with this manuscript. References American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author. Cole, D. A. (1987). Utility of confirmatory factor analysis in test validation research. Journal of Consulting and Clinical Psychology, 55, 584 594. Conners, C. K. (1997). Conners Rating Scales Revised. New York: Multihealth Systems. Conners, C. K., Erhardt, D., & Sparrow, M. A. (1999) Conners Adult ADHD Rating Scales (CAARS). New York: Multihealth Systems, Inc. Dale, E., & Chall, J. S. (1948). A formula for predicting readability. Coumbus, OH: Ohio State University Bureau of Educational Research. (Reprinted from Educational Research Bulletin, 27, 11 20, 37 54). Erhardt, D., Conners, C. K., Epstein, J. N., Parker, J. D. A., & Sitarenios, G. (1999). Self-ratings of ADHD symptoms in adults II: Reliability, validity, and diagnostic sensitivity. Manuscripts submitted for publication. Hogan, R., & Nicholson, R. A. (1988). The meaning of personality test scores. American Psychologist, 43, 621 626.
Book review / Archives of Clinical Neuropsychology 18 (2003) 431 437 437 Marsh, H. W., Balla, J. R., & McDonald, R. P. (1988). Goodness-of-Fit Indexes in confirmatory factor analysis: The effect of sample size. Psychological Bulletin, 103, 391 410. Ward, M. F., Wender, P. H., & Reimherr, F. W. (1993). The Wender Utah Rating Scale: An aid in the retrospective diagnosis of child hood attention deficit hyperactivity disorder. American Journal of Psychiatry, 150, 885 890. Wender, P. H. (1995). Attention-Deficit Hyperactivity Disorder in adults. Oxford, UK: Oxford University Press. Katherine DeGeorge Macey Department of Educational Psychology 4225 Texas A & M University, College of Education College Station, TX 77843-4225, USA E-mail address: kldegeorge@tamu.edu