Differentiating ADHD from oral language difficulties in children: role of movements and effects of stimulant medication

Thirty-three participants were between the ages of 6 and 13 years (15 = male; 18 = female) from an original sample of 67. The participants in the study attended a specially designed early language intervention program at the Shelton School [27], a large private school for children with learning differences, in Dallas, Texas. To be included in the study, the participant had to have a primary diagnosis of Oral Language Disability (OLD) diagnosed by a team of speech-language pathologists, licensed psychologists, and educational diagnosticians at Shelton School and enrolled in the school’s specially designed early intervention program. OLD children had: (a) low average (85 – 89) or below average (<85) verbal IQ, (b) below average (<85) auditory processing, processing speed, visual perceptual ability, reading comprehension, spelling, or handwriting, and (c) average reading rate and accuracy (85 – 115). In addition, their receptive and expressive language performance was in the moderate to severe range of impairment or below 85. Those participants with a comorbid ADHD diagnosis had one of the four subtypes of Predominantly Inattentive Type, Predominantly Hyperactive-Impulsive Type, Combined Type, or ADHD Not Otherwise Specified. Participants in the study were not excluded for taking a stimulant ADHD medication, but were asked to be able to be tested on and off their medication (i.e., parent and child provided consent and assent). The choice to be on medication was based on parent and community physician decisions for treatment. School and study personnel had no control over that, rather we could only indicate if they were on the medication or not. The school nurse knew what medication they were taking, but not the doses. All testing was done within the first two hours of the school day. One on non-stimulant psychotropic medication was excluded. Participants with a history of head injury (a period of unconsciousness followed by lasting impairment) or a neurological disorder, such as a seizure disorder, were excluded (Figure 1).

To determine if the child had ADHD, the parent and the child separately were administered the Kiddie Schedule for Affective Disorders and Schizophrenia in School Age Children, Present and Lifetime (K-SADS-P/L) [28] which uses the same diagnostic criteria as the DSM-IV-TR. All items of ADHD syndrome were administered to both the parent and the child. Specifically, to be diagnosed with ADHD, a participant had to receive ratings of “three - definitely present” for six of the nine inattentive symptoms (ADHD-IA), or six of the nine hyperactive symptoms (ADHD-HI), or a minimum of six inattentive and six hyperactive/impulsive symptoms (ADHD-C) prior to the age of seven. Children were diagnosed with ADHD NOS if ratings did not meet full symptom criteria (e.g., a minimum of 6 criterion symptoms rated as a “three”), but had prominent symptoms of inattention and/or hyperactivity/impulsivity present and causing impairment (e.g., a minimum of four symptoms but < 6 criterion symptoms rated as a “three”). Reliability studies of the K-SADS-P/L report high inter-rater reliability (range: 93% to 100%) and excellent test-retest reliability [28].

This study was reviewed and approved by the Shelton School Internal Review Board (IRB) and the University of Texas Southwestern Medical Center IRB. Both the child and parent gave informed verbal and written assent and consent to participate in the study.

Baseline data Time 1 (T1) data were collected during the Spring [30] and Time 2 (T2) data were collected again during the following Spring after completing a specially designed year-long oral language school intervention [31]. All testing was conducted at the Shelton School in a quiet, secluded testing area during the first periods of school in the morning. Participants were tested one at a time. During the 15 minute test, the administrator stayed in the testing room to notate observations, without talking to the participant and to assure that they remained on task. Participants off of their medication were tested after arriving at school, and then upon completion were taken to the school nurse for medication administration and then onto class. Testing was conducted on different days for those who participated in the either on or off of their medication condition (the counterbalanced order of testing was random). That they had taken, or not taken their medication prior to arriving to school, was verified by the study staff prior to testing. Upon completion of the test, data were transmitted and analyzed by BioBehavioral Diagnostics on a central server. Data were then compared to a normative group by age and gender and a report was produced including statistical and graphical information about attention and movement variables.

Demographic characteristics for the overall sample are described using the sample mean and standard deviation for continuous variables and the frequency and percentage for categorical variables. To test the Quotient® ADHD attention and movement variables, the basic design was a two-group (OLD vs. OLD/ADHD), by Time (T1 and T2) repeated measures Analysis of Variance (ANOVA) design using SPSS Version 19. A secondary analysis of medication (ON or OFF) by Time analysis was conducted for an OLD/ADHD subgroup of 16 participants taking stimulant medication at both time points (the four OLD participants on medication were excluded for the medication analyses).

The means and standard deviations for the various Quotient® movement variables are reported in Table 2. The OLD/ADHD group had a significantly greater number of movements at both Time 1 and Time 2 than the OLD group F(1,29) = 8.54, p < .007, η² = 0.23, a large effect size. Effect size is a partial eta-squared (η²) where values of .01, .06, and .14 represent small, medium, and large effects sizes [32]. Likewise, children with OLD/ADHD had significantly greater distance in their movements (“displacement”) at both Time 1 and Time 2 than the OLD group F(1,29) = 7.76, p < .01, η² = 0.21) and greater area represented in their movements at both Time 1 and Time 2 than the OLD group F(1,29) = 10.64, p < .003, η² = 0.27. Finally, children with OLD/ADHD had a higher frequency of movements (“temporal scaling”) at both T1 and T2 than the OLD group F(1,29) = 8.30, p < .01, η² = 0.22, but the spatial complexity of the movements did not differ between groups. There were no interactions or effect of T1 versus T2. The test-retest reliability for the movement measures was good to excellent with all p values less than 0.001: immobility (r = .75), movement (r = .82), displacement (r = .75); area (r = .66); spatial complexity (r = .65) and temporal scaling (r = .86).

A MANOVA at baseline to minimize possibility of Type I error, found no significant group differences on the attention variables so they are not reported, Wilk’s Λ = 0.915, F(4, 46) = 1.07, p = .38, [30], or again at T2 [32]. The attention scores were impaired compared to norms (e.g., mean accuracy scores for OLD of 81% and 78% for OLD/ADHD whereas normal performance would be in the low to mid 90s [23]. Separate power analyses indicated that much larger group sizes would be required (>200) for differences at p < .05 and power of 80.

A discriminant function analysis was conducted using all of the baseline movement scores to predict OLD versus OLD/ADHD group membership. The Wilk’s Lambda chi square was significant at 16.3 (df = 6), and p = .01, with 82.8% of the groups correctly classified. The Kappa was 0.66, p < .001. A separate sensitivity/specificity analysis indicated 73.3% sensitivity and 92.9 specificity, with a positive predictive value of 91.7%. The findings provide strong support that body movements as measured by the Quotient® during a CPT task can significantly differentiate those children with OLD only diagnoses from those with OLD and ADHD and are stable indices whereas by contrast, there were no differences based on classic CPT attention measures.

To assess the effects of medication, a secondary set of analyses utilized a within subject two group (On Medication versus Off Medication) repeated measures ANOVA based on Time 1 and Time 2 (a year later) for a subgroup who were on medication at both time periods. The means and standard deviations for the various Quotient® ADHD System movement variables are reported in Table 3. All movement measurements were significantly different when on medications. There were no interactions or effect of T1 versus T2. Children spent more time sitting still (“immobility duration”) compared to their performance off medication at both T1 and T2 F(1,15) = 20.54, p < .00, η² = .58, representing a large effect size. They had fewer position changes (less “movement”) when tested on medication compared to their performance off medication at both T1 and T2 F(1,15) = 26.10 p < .00, η² = .64, and shorter total distance of movements (“area”) when tested off medication F(1,15) = 22.45 p < .00, η² = .60. Children had a lower “area” score (i.e., the space they moved in was smaller) when tested on medication compared to their performance off medication at both T1 and T2 F(1,15) = 31.08 p < .00, η² = .67. They also had a higher “spatial complexity” score (i.e., movements were qualitatively more complex) when tested on medication compared to their performance off medication at both T1 and T2 F(1,15) = 17.54, p < .00, η² = .54. Children tested on medication had significantly less frequent movements (“temporal scaling”) at both T1 and T2 compared to more frequent movements when tested off medication F(1,15) = 24.49 p < .00, η² = .62. Using the Quotient® ADHD System to measure movement, children tested on medication spent significantly more time sitting still, had fewer position changes, traveled less distance in their movements, had smaller area of movement, had less frequent movements, and had more complex movements at both T1 and T2.

The means and standard deviations for the various Quotient® ADHD System attention variables are reported in Table 4. Children had better “accuracy” scores at both time points while medicated F(1,15) = 8.14, p < .01, η² = .35. Likewise they had fewer “omissions” (i.e., fewer missed targets) when tested on medication compared to their performance off medication F(1,15) = 9.90, p < .01, η² = .40, but no differences for incorrect responses to non-targets (“commissions”). “Variability” (defined as the standard deviation of response time to target) was significantly lower on medication compared to their performance off medication F(1,15) = 33.62, p < .00, η² = .69. For “latency”, children responded significantly faster when they were on medication compared to their performance off medication F(1,15) = 28.03, p < .00, η² = .65. For Coefficient of Variance (COV) [a more stringent measure of response consistency: (100 x variability)/latency], response consistency was significantly greater when children were tested on medication compared to their performance off medication F(1,15) = 32.28, p < .00, η² = .68. These attention scores clearly indicate that stimulant medication improves attention performance during a fifteen minute continuance performance task with significantly better accuracy, fewer omission errors, faster response time, and less variability.

Additional measures calculated by the Quotient® ADHD System are known as attention state variables. For example, attention shifts and the number of shifts in attention state are defined as On Task (percent of time hit many targets and few non-targets), Distracted (percent of time hits some targets and some non-targets; accuracy is better than chance), Impulsive (percent of time hits many targets and some non-targets), Random Responding (hits most targets and non-targets; accuracy of responding is as good as chance), Minimal Responding (misses most targets and non-targets; accuracy is about as good as chance), and Contrary (response accuracy is significantly worse than chance). The analyses of these variables indicate that the children spent more time responding “on task” when they were tested on medication F(1,15) = 4.66, p < .05, η² = .24, spent a lower percentage of time in a “distracted” attention state, and less time in the “minimal responding” attention state profile (i.e., children were less likely to make both omission and commission errors) F(1,15) = 4.92, p < .04, η² = .25.

Despite significant findings for differentiating diagnostic categories and effects of stimulant medication as identified by the Quotient® ADHD System, some limitations for the current study are considered. As noted earlier, the small sample size may not have detected important group, time, and interaction differences. Power analyses for the attention measures suggested that a fairly large number of subjects would be required (typically > 200 per group). Nonetheless, for the movement measures it is important to note that a number of significant differences had large effects sizes supporting that the differences were large enough to be detected with small sample sizes which supports that the findings are nonetheless robust. Likewise, the findings were corroborated when replicated a year later.

A second limitation suggests caution in generalizing findings. Children in the study represented a sample of opportunity selected from a pool of participants enrolled in a specially developed language intervention program [27] at a large private school for children for learning differences. Nonetheless, the importance is the unique opportunity to study what would be considered a fairly large homogenous sample of children with well identified oral/language differences involved in a structured learning intervention program, a subgroup which also had comorbid ADHD. And finally, secondary analyses related to the effects of medication merit caution given the limitations inherent in the design.