The Diagnostic Accuracy of Parent Rating Scales in Discriminating Children with and Without Attention-Deficit/Hyperactivity Disorder

Attention–Deficit Hyperactivity/Disorder (ADHD) remains a highly controversial psychological condition due to its complex nature and co-occurrence with several comorbid disorders; thus, it makes differential diagnosis especially difficult (if not impossible). The overdiagnosis of ADHD and subsequently the over-prescription of psychostimulants to these children are the most essential parts of ADHD controversy. There exists no accurate diagnostic method to help professionals to establish a correct diagnosis of ADHD; therefore, they have been recommended to use multiple methods and information sources, such as interviews with parents and teachers, behavioral rating scales, direct observation techniques, and neuropsychological tests to ensure the correct diagnosis. This study aimed to investigate the diagnostic power of behavioral rating scales responded by parents in diagnosing ADHD as well as their other psychometric properties.

This was a descriptive and methodological study. The statistical population included all male students in Mashhad City, Iran, in the 2014–2015 academic year. The study sample consisted of 40 students selected by convenience sampling method. To gather the quantitative data, we observed students in classroom settings and interviewed their teachers and parents. Of these boys, 20 were healthy and 20 had received ADHD diagnosis by a physician before participating in our study. We administered a short three–subscale form of the Wechsler Intelligence Scale for Children–Revised (Wechsler, 1974) to all these 40 children to ensure they have no intellectual disability. Besides, the study groups were matched concerning intelligence quotient. Then, we trained blind interviewers to administer semi–structured interviews with the children’s parents to verify their initial ADHD diagnoses; they also requested the parents to rate their children’s classroom behaviors using the Conners Parent Rating Scale–Revised (CPRS–R) (Conners, 1997) and the Swanson, Nolan, and Pelham–4 Rating Scale (SNAP–IV) (Swanson et al., 2001) two times with a one–month interval. We also used the Receiver Operating Characteristic (ROC) analysis to calculate the sensitivity and specificity as well as False Positive Rate (FPR) and False Negative Rate (FNR) for these scales. Additionally, we used Cronbach’s alpha coefficient and Pearson correlation coefficient methods to evaluate the internal consistency and test-retest reliability of both rating scales in SPSS. The significance level of the tests was set at 0.05.

Multivariate Analysis of Variance (MANOVA) followed by one–way Analysis of Variance (ANOVA) and posthoc tests data indicated that all subscales of CPRS–R and SNAP–IV could significantly distinguish between ADHD and non–ADHD children. Cronbach’s alpha coefficient for the total 18 items of SNAP–IV, as well as predominantly inattentive (first 9 items) and predominantly hyperactive/impulsive (second 9 items) subscales of it, were measured as 0.99, 0.97, and 0.98, respectively, indicating excellent internal consistencies. Test-retest reliabilities for the total 18 items and predominantly inattentive (first 9 items) and predominantly hyperactive/impulsive (second 9 items) subscales of SNAP–IV were 0.72, 0.77, and 0.68, respectively, indicating acceptable test-retest reliabilities for it. For CPRS–R, Cronbach’s alpha coefficients were obtained as 0.95, 0.89, and 0.88, 0.90 for cognitive (6 items), hyperactive (6 items), oppositional (6 items), and total (9 items) subscales, respectively. Test-retest reliabilities for cognitive, hyperactive, oppositional, and total subscales were computed as 0.94, 0.74, and 0.77, 0.86 respectively. ROC analyses also presented an excellent to acceptable sensitivity and specificity for all the subscales of the CPRS–R, and SNAP–IV rating scales. All sensitivities ranged between 0.70 and 0.85, all specificities raged between 0.70 and 0.90; the AUC values of SNAP–IV and CPRS–R lied between 0.80 and 0.89, which indicated their excellent diagnostic power with low FPR and FNR rates. 

Our results suggested that the diagnostic accuracy of the Persian versions of the CPRS–R, and SNAP–IV subscales were excellent to acceptable. The CPRS–R and SNAP–IV indicated excellent diagnostic accuracy in this study; however, using them in isolation is not recommended.