Trajectories of Response to Treatments in Children with ADHD and Word Reading Difficulties

Abstract

This study investigated patterns of response to intervention in children with co-occurring attention-deficit/hyperactivity disorder (ADHD) and reading difficulties (RD), who participated in a randomized clinical trial examining the efficacy of reading intervention, ADHD treatment, or combined treatments. Growth Mixture Modeling (GMM) was used to investigate trajectories of parent and teacher academic impairment ratings and child oral reading fluency, and whether trajectories were predicted by pre-treatment covariates (ADHD severity, reading achievement, phonemic awareness, rapid letter naming, anxiety, oppositional defiant disorder), for 216 children with ADHD/RD in 2^nd–5^th grade (61.1% male; 72.2% African American; 8.8 ± 1.3 years of age). GMM revealed three trajectories for academic impairment (6.9–24.2% stable, 23.7–78.7% moderately improving, and 14.1–52.1% steeply improving) and oral reading fluency (20.8% low improving, 42.1% moderate improving, and 37.1% high improving). Children in the reading intervention were more likely to be in the stable or moderately improving trajectory than those in the ADHD and combined treatments, who were more likely to be in the steeply improving trajectory for academic impairment. Relative to the ADHD intervention, children in the reading intervention were more likely to be in the high improving trajectory than the moderate or low improving trajectory for oral reading fluency. Children without comorbid anxiety and with better reading skills showed a more positive treatment response for teacher-rated academic progress and oral reading fluency. Results highlight the importance of examining individual differences in response to reading and ADHD interventions. Intervention modality predicted differences in parent/teacher ratings of academic progress as well as reading fluency.

Attention-Deficit/Hyperactivity Disorder (ADHD) and word-reading difficulties and disabilities (RD) are highly prevalent childhood disorders which often co-occur, with rates of RD ranging from 25 to 40% among individuals with ADHD (Willcutt et al., 2005). Children with comorbid ADHD/RD exhibit more severe reading problems, greater academic impairment and lower achievement, increased likelihood of placement in special education classes, higher rates of grade-retention, and overall more pervasive negative social and occupational outcomes than children with either disorder alone (Germano et al., 2010; Purvis & Tannock, 2000; Rucklidge & Tannock, 2002; Willcutt et al., 2010). Further, comorbid ADHD/RD is associated with more severe reading problems (Lyon, 1996) and lower grades than RD alone (McNamara et al., 2005), and more severe attention problems than ADHD alone (Mayes & Calhoun, 2007).

There are evidence-based treatments for both disorders [e.g., medication and behavioral therapy are recommended for ADHD (Wolraich et al., 2019), whereas intensive, systematic instruction in phonics and word identification and practice reading connected text is recommended for RD (Fletcher et al., 2019)], however there has been little research examining appropriate intervention strategies for children who have comorbid ADHD/RD. Further, most response to intervention (RTI) models (i.e., a multi-tier approach to the early identification and support of students with learning and behavior needs ranging from universal screening to intensive individualized intervention) that have been investigated are limited to early elementary school-age students and for reading interventions (e.g., Denton et al., 2013; Vaughn et al., 2010). There is little exploration in how evidence-based interventions for ADHD or combined ADHD + RD interventions can be applied within the RTI framework (see Vujnovic et al., 2014). This is troublesome as early identification and effective intervention are particularly important for elementary school-aged children who have comorbid ADHD + RD due to the implications for future social, academic, and occupational attainment (Willcutt et al., 2005; Willcutt et al., 2007a, b).

Only three published studies were found that investigated the relative efficacy of single and combined treatments for ADHD in elementary school-aged children with comorbid ADHD/RD. One assessed the impact of providing methylphenidate (MPH) at various doses or placebo along with reading intervention and reported inconclusive results (Richardson et al., 1988). Another investigated the relative efficacy of medication and strategy-based reading instruction for children with comorbid ADHD/RD and reported stimulant medication improved ADHD symptoms but not reading, and reading instruction improved reading but not behavior (Tannock et al., 2018). The third compared the effects of providing ADHD treatment (parent training and medication), intensive reading intervention, or simultaneous RD intervention and ADHD treatment. Consistent with Tannock et al. (2018), students who received RD intervention performed significantly better than those who received only ADHD treatment on untimed word reading and phonological decoding (Tamm et al., 2017) and timed phonological decoding (Denton et al., 2020). Conversely, youth who received ADHD treatment had significantly better reading comprehension outcomes than those in the RD intervention, likely due to the impact of medication on executive functions supporting comprehension, e.g., cognitive flexibility, planning and inhibition (Denton et al., 2020). ADHD symptoms were significantly more improved in the groups that received ADHD treatment than those that received RD intervention only (Tamm et al., 2017). The combination of treatments did not enhance reading skills or ADHD symptoms beyond the specific effects of either unimodal treatment (Tamm et al., 2017). There may however be distinct subgroups with unique treatment response patterns, which we explore using data from the third study (Denton et al., 2020; Tamm et al., 2017).

Rationale for Exploring Response to Intervention Trajectories

Examining differences at the group level characterizes the average amount of change, but provides less information about patterns of individual response, intra-individual differences in change over the course of treatment, or the extent to which a given child is going to benefit from a particular intervention (Muthen et al., 2002). Alternative statistical approaches such as growth mixture modeling (GMM) may be useful for identifying longitudinal patterns (i.e., trajectories or classes) that help parse inter- and intra-individual heterogeneity, such that latent classes of children may benefit differentially from an intervention relative to others. Further, if an intervention is not universally efficacious, it is important to understand the characteristics of individuals who respond particularly well and of those who may not respond as well to intervention. Approaches like this have been used to characterize individual differences in treatment response trajectories in a range of mental health and academic interventions (Peris et al., 2015; Rubel et al., 2015; Warren et al., 2012). Indeed, such studies in youth with ADHD or RD suggest that not all youth respond equally well or at all to interventions (Breaux et al., 2019b; Denton, 2012; Frijters et al., 2011; Langberg et al., 2016; Swanson et al., 2007; Torgesen et al., 2001; Vaughn & Wanzek, 2014). This is important in an era of precision medicine (Collins & Varmus, 2015) where increased focus is being placed on identifying treatment responders and non-responders and potentially malleable predictors of treatment response. Such knowledge could be utilized to improve targeting, sequencing, and individualization of treatments. As an example, if a child with ADHD/RD and comorbid oppositional defiant disorder (ODD) responds more poorly to an intervention, then a clinician might recommend treatment of the oppositional symptoms prior to the child engaging in the ADHD/RD intervention. We are not aware of studies that have implemented GMM to estimate patterns of inter- and intra-individual change in treatment response to medication, parent training, or intensive reading instruction for children with ADHD/RD.

Differential Responses to Interventions for ADHD

Studies evaluating trajectories of response to interventions for ADHD suggest wide heterogeneity in how youth with ADHD respond. In the Multimodal Treatment of ADHD (MTA) study which tested unimodal and multimodal ADHD treatment strategies, three latent trajectories of ADHD symptoms at baseline, 14-month, 24-month, and 36-month follow-up were identified across treatment groups using parent ratings of ADHD: (a) 52% of youth made large improvements that maintained, (b) 34% made small initial gains and gradually improved, and (c) 14% made large initial improvements followed by deterioration (Swanson et al., 2007). Further, the most favorable trajec tory had a significantly greater percentage of cases assigned to the combined (medication + behavior) treatment (62%) and medication only (55%) conditions relative to the behavior treatment only (46%) and community control (45%) conditions. Langberg and colleagues examined response trajectories in the Challenging Horizons Program (a comprehensive school-based psychosocial program targeting academic and social functioning) from baseline to six-month follow-up for adolescents with ADHD (Langberg et al., 2016). For parent ratings of academic progress: (a) 69% started in the impaired range and made either no (26%) or small improvements (43%) and stayed in the clinical range; (b) 18% made rapid improvements into the normal range that plateaued or slightly increased from post to follow-up; (c) 8% had steady improvement; and (d) 5% had no impairment during the intervention. Similar trajectories were identified among adolescents with ADHD who received either the Homework, Organization, and Planning Skills (HOPS) or a contingency-management homework intervention (Breaux et al., 2019a), including adolescents who did not respond (19–32%) or who responded poorly (12–18%); however, treatment assignment did not differentiate trajectories.

Differential Responses to Interventions for Reading

Differential treatment response has also been demonstrated for reading interventions (Fletcher et al., 2011; Stuebing et al., 2015; Vellutino et al., 2000). For example, Denton and colleagues categorized students as either adequate or inadequate responders following Tier 3 intensive reading instruction, and found that while 72% of students performed within the average range in basic reading skills, only 43% met benchmarks for adequate word reading fluency and only 36% for adequate passage comprehension (Denton et al., 2013). Piecewise growth modeling with time-series data during a Tier 2 targeted intervention for Kindergarten students at-risk for literacy problems revealed variable, nonlinear patterns of change in letter-naming fluency (Zvoch, 2016); specifically, trajectories generally increased during periods of instruction, and declined during school breaks. In a longitudinal follow-up of students who received intensive reading intervention in 1^st grade, latent growth models revealed that most students continued to benefit, although 27–36% were “less responsive” in decoding and fluency (Vadasy et al., 2008).

Limitations of Prior Studies Examining Response to Intervention Trajectories

Although research on treatment response trajectories of reading interventions has used repeated assessments and time series growth modeling (Denton et al., 2010, 2013; Frijters et al., 2011), response to interventions targeting ADHD is often based on only two (i.e., pre, post) or three (e.g., baseline, post, follow-up or baseline, mid, post) time points. Designs that are limited to outcome data collected at three or fewer waves are often restricted to linear models (Breaux et al., 2019a), which imply change is constant with respect to time. Greater frequency of assessment permits estimating other forms of change (e.g., freed, quadratic, piecewise) and allows for interpreting differential change at across time (Biesanz et al., 2004), which may be more appropriate for examining treatment response spanning active intervention and follow-up. Further, no study has examined patterns of treatment response to ADHD treatment (medication and parent training) or intensive reading instruction for children with comorbid ADHD/RD. Given the unique academic impairments faced by children with comorbid ADHD/RD (Willcutt et al., 2005), and the limited evidence-based for combined ADHD/RD treatment, it is especially important to examine the impact of predictors of treatment response in this context. To date, the majority of evidence on predictors of treatment response in reading and ADHD intervention has focused on largely non-malleable factors such as child’s gender, IQ, parent education status (e.g., Denton et al., 2013; Owens et al., 2003) and this is the first to focus specifically on potentially malleable predictors of response to ADHD and/or RD intervention.

Possible Predictors of Intervention Response

Several potentially modifiable pre-treatment covariates have been identified as important treatment moderators or predictors of group-level response to RD and/or ADHD treatments. For example, a large body of literature supports the contributions of phonemic awareness and rapid automatized naming of letters as predictors of children’s responsiveness to reading interventions (Nelson et al., 2003). Further, those with more severe RD tended to respond less well to reading treatment than those with milder RD (Galuschka et al., 2014). Similarly, among youth with ADHD, higher reading achievement scores were a consistent predictor of parent- and teacher-rated treatment response (i.e., greater improvements in homework performance and grades; Breaux et al., 2019a). Children with more severe ADHD have been shown to respond more poorly to ADHD treatments than those with less severe ADHD who demonstrate steeper response slopes and lower post-treatment impairment (MTA Cooperative Group, 1999b; Owens et al., 2003; Buitelaar et al., 1995; Sciberras et al., 2020). There is also mixed evidence for the role of comorbid anxiety disorder, with some studies demonstrating co-occurring anxiety is associated with better ADHD treatment response (MTA Cooperative Group, 1999b; van der Oord et al., 2008a) and others finding anxiety is associated with poorer response to medication and combined treatments as rated by parents (MTA Cooperative Group, 1999b). The evidence is mixed for comorbid ODD, with some studies indicating no association with ADHD treatment response (Langberg et al., 2016; MTA Cooperative Group, 2004; Sidol & Epstein, 2020), and others suggesting comorbid ODD is associated with poorer treatment response to ADHD treatments, predicting less steep slopes of improvement (Chazan et al., 2011; Langberg et al., 2018; Murray et al., 2008; Owens et al., 2016). Although not a diagnosis of ODD, problem behaviors and externalizing symptoms also predict reduced response to reading intervention (Roberts et al., 2019b). No published studies were found investigating predictors of combined ADHD/RD treatment.

Present Study

The current study focused on examining trajectories of response to interventions for elementary school-aged youth with co-occurring ADHD/RD. GMM was used to investigate longitudinal trajectories of individual growth in parent and teacher ratings of academic progress as well as child reading fluency assessed at six time points. The present study utilized a multi-method, multi-informant assessment of treatment outcomes, including both parents’ and teachers’ perspectives of academic progress and students’ oral reading fluency performance. Importantly, parents and teachers likely observe different academic behaviors and functioning, across home and school contexts, and each perspective provides unique contributions for understanding students’ academic functioning. A secondary aim was to examine which factors differentially predicted treatment response. We focused on potentially modifiable pre-treatment variables, which have shown to be an important treatment predictor in ADHD and/or RD outcomes including ADHD severity, comorbid anxiety, comorbid ODD, reading skills, phonemic awareness, and rapid automatized naming. We predicted children with poorer reading skills and/or more severe inattention and hyperactivity/impassivity to be in trajectories associated with poorer treatment response. Given prior studies report mixed findings depending on the outcome, informant, or measure of comorbid anxiety and comorbid ODD, we did not make hypotheses regarding the effects of these constructs on specific academic outcomes. We intentionally examined potentially malleable factors to inform potential targets, approaches, or prevention strategies for future intervention refinement intended to improve treatment response.

Method

Participants

Children (n = 216) in grades 2 to 5 (M = 8.8 years of age, SD = 1.3) were primarily recruited in schools, but also from outpatient clinics and the community. This study took place at two sites, one in the Houston metropolitan area and the other in the greater Cincinnati area. Children who were initially identified by their teachers or parents as having both reading and attention problems were assessed for eligibility. Eligible children met DSM-IV diagnostic criteria for ADHD (Combined or Inattentive) on the Diagnostic Interview Schedule for Children 4.0 (DISC) (Shaffer et al., 2000) which was supplemented by parent and teacher ratings on the SNAP-IV ADHD Rating Scale (Swanson et al., 2012). Specifically, if a child was rated by the teacher as “quite a bit” or “very much” on ADHD symptoms that did not overlap with those identified by the DISC, those symptoms were counted in the eligibility determination (MTA Cooperative Group, 1999a). Children were considered to have RD if they had a standard score ≤ 25^th percentile on Woodcock-Johnson III Letter-Word Identification and/or Word Attack subtests or the Basic Reading Skills composite (Woodcock et al., 2001). Participants were severely impaired word readers; at baseline, mean word reading and decoding scores ranged from the 3^rd to 5^th percentiles across treatment groups. Children were predominantly male (61.1%), and economically disadvantaged (75.9%) as indexed by receiving free or reduced lunch at school. The majority were African American (72.2%), Caucasian (19.1%), biracial (6.5%) or other (1.9%), and 11.1% identified as Hispanic. In terms of grade, 30.6% were in the 2^nd grade, 25.5% were in the 3^rd grade, 24.1% were in the 4^th grade, and 19.9% were in the 5^th grade. Approximately 31.5% of the sample met diagnostic criteria for an anxiety disorder (excluding specific phobia), and 32.9% met criteria for ODD. Exclusionary criteria were Full Scale or Nonverbal IQ estimates <70; cardiovascular problems; chronic tics; taking non-stimulant medication with potential to affect ADHD (e.g., Wellbutrin); severe psychopathology, autism, or sensory disabilities.

Procedures

Institutional Review Board approval was obtained from Cincinnati Children’s Hospital Medical Center and the University of Texas Health Science Center at Houston. After providing informed consent/assent at the initial visit, families were screened for eligibility including parents completing the DISC, teachers completing ADHD rating scales, and children receiving IQ and academic assessments. Children taking ADHD medication at the time of enrollment (n = 53) did not take those medications during the baseline assessment (24-h washout for stimulants, 2-week washout for nonstimulants). Eligible children were randomized, stratified by grade, to: (a) reading treatment only, (b) ADHD treatment only, or (c) combined reading and ADHD treatment.

The ADHD treatment consisted of a medication titration trial to identify optimal dose followed by maintenance on that dose, and a concurrent 9-week group behavioral parent training. The parent training intervention, which was delivered by licensed psychologists, was adapted from that utilized in the MTA (MTA Cooperative Group, 1999b). In the reading intervention children were provided with explicit, systematic instruction in word reading, decoding, and spelling using published programs (Vadasy & Sanders, 2007; Vadasy et al., 2005), supplemented with a set of hands-on practice activities that included the use of manipulatives. To build oral reading fluency, students engaged in repeated reading practice of nonfiction texts with monitoring and feedback (Chard et al., 2002) using QuickReads (Hiebert, 2003). Intervention was provided primarily by certified teachers to one or two students at a time for 45 min, four days per week.

After randomization, all participants received 16 weeks of treatment and participated in a post-test evaluation within 1 month of completing the intervention. Follow-up evaluations were conducted on average 3 months after post-test (M = 3.0, SD = 1.52). Impairment and reading fluency were assessed at each of the six time points (pre, after 1, 2, and 3 months of treatment, post, and follow-up). Children in the ADHD and combined treatment groups completed the post-test while on study medications. Interventionist treatment fidelity was excellent (M = 95% ± 3%) and > 85% of participants received at least some of the treatment they were assigned (Tamm et al., 2017). Neither reading intervention attendance, parent training attendance, nor medication adherence differed significantly between treatment groups (Tamm et al., 2017).

Measures

Treatment Outcomes

Impairment Rating Scale (IRS)

The IRS assesses areas of functioning that typically characterize children with ADHD (Fabiano et al., 2006). The IRS items are rated by parents or teachers on a 0 (signifying no impairment) to 6 (signifying extreme impairment) scale. Scores of ≥ 3 on any IRS item reflect significant impairment or clinical range (Fabiano et al., 2006). The IRS demonstrates excellent temporal stability (i.e., individuals maintain similar rank ordering of levels of impairment), convergent, and discriminant validity in clinic and community samples (Fabiano et al., 2006). In the current study, parent and teacher ratings of impairment on the academic progress item was examined as an outcome at each of the 6 time points (rs across time between raters ranged from 0.23—0.30, ps < 0.05). This item has been used in previous GMM analyses examining treatment response (Langberg et al., 2016).

Dynamic Indicators of Basic Early Literacy Skills Oral Reading Fluency (DORF)

Oral reading fluency in connected text was measured with the DORF (Good & Kaminski, 2002). Students read passages orally, and the score is the number of words read correctly in one minute. Both alternate form and interrater reliability for the DORF exceeds 0.80 (Goffreda & DiPerna, 2010). In this study, students read two unique grade-level passages at each administration, and the dependent variable was the mean score for the two passages at each of the six time points (Cronbach’s αs between the 2 passages across time points were 0.91—0.96).

Predictors

Swanson, Nolan, and Pelham (SNAP-IV) DSM-IV ADHD Rating Scale

Raters evaluate how often a child evidences each of the 18 DSM-IV ADHD symptoms on a four-point Likert scale (0 = Not at all, 1 = Just a little, 2 = Quite a bit, 3 = Very much). The measure shows adequate internal consistency (0.94) and test–retest reliability (Bussing et al., 2008; Gau et al., 2008). Averages of the baseline ADHD symptom inattention (α = 0.88) and hyperactivity-impulsivity (α = 0.92) scores (derived by averaging the 9 individual items in each subscale) for parent and for teacher were computed to index ADHD severity.

Woodcock Johnson Tests of Achievement, Third Edition (WJ-III)

The WJ-III Letter-Word Identification (naming letters and reading words aloud from a list) and Word Attack (reading nonsense words aloud to test phonetic word attack skills) subtests, comprising the Basic Reading Skills score, were administered. These subtests have good reliability (> 0.80) and validity (Woodcock et al., 2001). The baseline Basic Reading Skills score indexed reading achievement.

The Diagnostic Interview Schedule for Children, Parent Version 4.0

The DISC (Shaffer et al., 2000) is a structured diagnostic interview designed for use in epidemiological and clinical studies by lay interviewers. It contains algorithms to generate diagnoses, based on rules similar to those in the DSM-IV. The DISC has good test–retest reliability (kappa = 0.82), interrater reliability (kappa = 0.7), and good convergent validity with behavior rating scales (Hersen & Turner, 2003). Modules administered included: ADHD, ODD, Conduct Disorder, Social Phobia, Separation Anxiety Disorder, Specific Phobia, Generalized Anxiety Disorder, Obsessive Compulsive Disorder, Major Depression/Dysthymia, and Mania/Hypomania. For this study, anxiety (i.e., met criteria for any anxiety disorder other than specific phobia) and ODD were examined as predictors.

Comprehensive Test of Phonological Processing (CTOPP)

The Rapid Letter Naming (i.e., ability to quickly name letters) and Elision (i.e., ability to hear, identify and manipulate phonemes) subtests of the CTOPP (Wagner et al., 1999) were included to assess rapid letter naming and phonemic awareness, respectively. These two subtests have good reliability (internal consistency > 0.88) and construct validity (Wagner et al., 1999).

Analytic Approach

GMMs were computed in Mplus, Version 8.3 (Muthen & Muthen, 2016) to explore the differential trajectories of parent and teacher ratings of academic impairment and child reading fluency at six time points. GMM attempts to capture sample heterogeneity by examining multiple latent subpopulations that differ in model parameters (intercepts and slopes) and allowing variability around these parameters within each class (Jung & Wickrama, 2008; Muthen & Muthen, 2000). GMMs examine latent classes that can differ in intercepts and slopes, as well as allow for class-specific variations in these parameters (Jung & Wickrama, 2008; Lubke & Muthén, 2007; Ram & Grimm, 2009). Of note, GMMs do not necessarily assume that growth exists; rather latent classes may exhibit positive slopes, negative slopes, or no change over time. A visual inspection of individual trajectories indicated a high level of variability between individuals and non-linear trajectories of ratings of academic impairment and an objective measure of reading fluency. To account for the rapid increases that occurred for many across waves, we used a freed loading growth model (Cudeck & Harring, 2007), with the first and last time points fixed and all intermediate time points freely estimated. As such, the estimated factor loadings for the intermediate time points represent the total change observed between the first and last time point. This approach accounts for the rapid improvement (i.e., acceleration) during the first 1–2 months of treatment.

Consistent with these recommendations for GMMs and similar to recent studies using GMM to examine heterogeneity in treatment response (Breaux et al., 2019a; Lutz et al., 2014), we performed these analyses in two steps. First, for each outcome variable model fit indices compared solutions with an increasing number of latent classes K (in which K = 1, 2, 3, 4) to determine best fit for the data, until adding additional classes no longer improves model fit. Model fit was determined using a combination of empirical criteria (i.e., sample size adjusted Bayesian Information Criterion [BIC], Akaike Information Criteria [AIC], Lo-Mendell-Rubin [LMRT] adjusted likelihood ratio test, bootstrapped parametric likelihood ratio test [BLRT]), classification probabilities (how distinct each class is from the other classes) (Muthen & Muthen, 2000). Better model fit was indicated by having the majority of model-fit indicators in a model’s favor: AIC and BIC decreasing, the LMRT and BLRT remaining significant, and entropy and classification probabilities remaining > 0.80. We also examined class sizes, considering classes with no less than 5% of the total count (Jung & Wickrama, 2008). Prior simulation studies support this sample size is sufficient for identifying K = 1 to 4 classes (Lubke & Muthén, 2007; Ram & Grimm, 2009). Once the best fitting model was determined, models were examined to determine if trajectories had significant slopes (i.e., indicating either improvement or worsening; nonsignificant slopes indicated stable functioning). To investigate the stability of each solution, we re-estimated the model with different starting values for the growth parameters. In each case, the solution proved robust for different starting values, suggesting optimization was not achieved through identification of a local maximum (Hipp & Bauer, 2006).

Second, models were run with the AUXILIARY function in Mplus using Vermunt’s three-step approach (Vermunt, 2010), with treatment condition included as a predictor. If treatment condition was a significant predictor, we further explored key baseline characteristics (i.e., inattention and hyperactivity/impulsivity severity, Basic Reading Skills, co-occurring anxiety or ODD, rapid letter naming, and phonemic awareness) as additional predictors of class membership. Missing data were minimal (see Supplemental Materials) and addressed using maximum-likelihood estimation with robust standard errors (MLR). To get an estimate of effect size for any significant predictors, estimates of best classification were used to determine class membership; Cohen’s d was then calculated based on means and standard deviations for each class.

Results

Parent-Rated Impairment in Academic Progress on the IRS

A three-class model was determined to be the best fit for parent-rated impairment in academic progress (see Table S1 in supplementary materials). The AIC and BIC evidenced diminishing gains in estimating beyond a 3-class model. A non-significant adjusted likelihood ratio statistic for the 4-class model and significant adjusted LRT and bootstrapped LRT for the 3-class model provide further support for the 3-class model. Additionally, the smallest class size for the 3-class model represented 6.9% of the sample, however in the 4-class model there were one class representing less than 5% of the sample. We retained the 3-class model solution as the preferred unconditional model. This 3-class model adequately discriminated between classes with class probabilities ranging from 0.85—0.95. The trajectories (Fig. 1a) consisted of a class that started with high impairment (M = 5.75, SE = 0.12) and remained high (stable/low improving class; 6.9%, n = 15) and well above the clinical range [i.e., scores > = 3 on the IRS (Fabiano et al., 2006)] throughout the intervention period (μ_slope = −0.26, p = 0.10). Two additional classes demonstrating improving patterns in academic impairment were also observed: a moderately improving class (78.7%, n = 170) that started with moderate impairment (M = 4.31, SE = 0.15) that decreased slightly (μ _slope = −1.55, p < 0.001) ending just below the clinical range, and a high/rapid improving class (14.4%, n = 31) that started with high impairment (M = 5.42, SE = 0.15) and demonstrated rapid decreases to below the clinical range, particularly after the first month with continued improvements through follow-up (μ _slope = −5.07, p < 0.001).

Fig. 1.

Trajectories for the Three-Class Growth Mixture Models for Parent and Teacher Ratings of Academic Progress Impairment and Child’s Oral Reading Fluency. Note: Solid lines are model estimated trajectories and dashed lines are observed mean scores in the respective classes. Lower scores for parent and teacher rated impairment in academic progress represent better functioning, whereas higher scores for oral reading fluency represent better functioning. Both parent and teacher-rated trajectories control for baseline covariates: ADHD symptoms, reading achievement, phonemic awareness, rapid letter naming, presence of co-occurring anxiety disorder and oppositional defiant disorder

Note. Solid lines are model estimated trajectories and dashed lines are observed mean scores in the respective classes. Lower scores for parent and teacher rated impairment in academic progress represent better functioning, whereas higher scores for oral reading fluency represent better functioning. Both parent and teacher-rated trajectories control for baseline covariates: ADHD symptoms, reading achievement, phonemic awareness, rapid letter naming, presence of co-occurring anxiety disorder and oppositional defiant disorder.

Treatment condition significantly predicted parent-rated academic impairment trajectories, such that relative to the reading intervention only group, children who received either the ADHD treatment only or combined treatment were more likely to be in either of the two improving trajectories (high/rapid improving or moderately improving) and significantly less likely to be in the stable/low improving impairment trajectory (ds from 0.68 – 1.11; Table 1). Of note, relative to the ADHD intervention only, children in the combined intervention were more likely to be in the high/rapid improving than the moderate improving trajectory, however this was not significant (χ² (1) = 3.46, p = 0.063, d = 0.37). Baseline Basic Reading Skills predicted response such that those in the high/rapid improving trajectory had significantly higher reading achievement than those in the moderately improving trajectory (d = 0.36). Inattentive symptoms predicted response such that those in the high/rapid improving trajectory had significantly lower inattention than those in the stable/low improving trajectory (d = 0.53). Hyperactivity/impulsivity, anxiety, ODD, rapid letter naming and phonemic awareness were not significant predictors.

Table 1.

Wald Chi-Square Tests of Mean Equality of the Auxiliary Analyses of Predictors

	Class Specification Means			Wald Chi-Square Tests of Mean Equality
	M (SE)	M (SE)	M (SE)	χ 2	d	χ 2	d	χ 2	d
PR Academic Impairment
Predictors	High/Rapid Improving	Moderate Improving	Stable/Low Improving	High/Rapid Improving vs Moderate Improving		Stable/Low Improving		Moderate vs Stable/Low Improving
Treatment Condition	0.04(0.15)	0.54 (0.05)	0.56 (0.17)	0.01	0.03	9 17**	0.68	5.03*	0.74
Reading vs ADHD Reading vs Comb	0.03 (0.09)	0.61 (0.05)	0.63 (0.17)	0.01	0.03	29.04***	1.11	9.05**	0.92
ADHD vs Comb	0.36 (0.11)	0.60 (0.05)	0.57 (0.20)	3.46	0.37	0.88	0.32	0.02	0.05
PR Inattention	2.04(0.13)	2.73 (0.22)	2.40 (0.16)	1.96	0.26	4.14*	0.53	0.33	0.12
PR Hyp/Imp	1.63 (0.16)	1.56 (0.07)	1.74 (.24)	0.13	0.08	0.15	0.10	0.48	0.18
ODD	0.42(0.11)	0.31 (0.04)	0.37 (0.15)	0.87	0.21	0.10	0.09	0.13	0.12
Anxiety	0.29 (0.10)	0.34 (0.05)	0.55 (0.16)	0.16	0.08	2.10	0.46	1.68	0.33
Reading Achievement	84.57(1.61)	80.01 (1.03)	81.12(3.07)	3.93*	0.36	1.50	0.35	0.11	0.09
Rapid Letter Naming	7.53 (0.34)	8.08 (0.18)	7.52 (0.41)	1.87	0.24	0.01	0.01	1.51	0.25
Phonemic Awareness	7.03 (0.53)	6.32(0.16)	6.85 (0.71)	1.52	0.32	0.04	0.06	0.52	0.24
TR Academic Impairment
Predictors	High/Rapid Improving	Moderate Improving	Stable/Low Improving	High/Rapid Improving vs Moderate Improving		Stable/Low Improving		Moderate vs Stable/Low Improving
Treatment Condition	0.47 (0.11)	0.51 (0.06)	0.58 (0.09)	0.09	0.04	0.56	0.11	0.39	0.13
Reading vs ADHD
Reading vs Comb	0.38 (0.09)	0.54 (0.09)	0.63 (0.07)	1.38	0.19	3.78*	0.30	0.68	0.16
ADHD vs Comb	0.51 (0.08)	0.62 (0.07)	0.46 (0.10)	0.95	0.15	0.15	0.06	1.74	0.26
TR Inattention	2.15 (0.12)	2.38 (0.08)	2.71 (0.40)	0.88	0.21	2.52	0.28	0.73	0.15
TR Hyp/Imp	1.06 (0.68)	1.35 (0.13)	1.37 (0.15)	0.87	0.05	0.99	0.05	0.01	0.02
ODD	0.39 (0.07)	0.26 (0.05)	0.41 (0.07)	1.80	0.20	0.06	0.03	2.77	0.35
Anxiety	0.18 (0.06)	0.35 (0.05)	0.35 (0.07)	5.39*	0.30	3.30	0.29	0.01	0.01
Reading Achievement	84.39 (1.00)	81.00(1.10)	78.69 (1.47)	6.40**	0.35	11.17***	0.54	1.47	0.25
Rapid Letter Naming	8.44 (0.32)	7.87 (0.21)	7.55 (0.32)	1.97	0.19	3.91*	0.29	0.65	0.17
Phonemic Awareness	7.45 (0.35)	6.11 (0.30)	6.09 (0.20)	8.66**	0.41	10.28***	0.43	0.01	0.01
TR Academic Impairment
Predictors	High/Rapid Improving	Moderate Improving	Stable/Low Improving	High/Rapid Improving vs Moderate Improving		Stable/Low Improving		Moderate vs Stable/Low Improving
Treatment Condition	0.92 (0.13)	0.60 (0.08)	0.41 (0.06)	4.23*	0.33	3.42*	0.53	1.71	0.29
Reading vs ADHD Reading vs Comb	0.79 (0.22)	0.53 (0.06)	0.50 (0.07)	1.19	0.18	1.61	0.18	0.07	0.05
ADHD vs Comb	0.49 (0.22)	0.62 (0.06)	0.45 (0.07)	0.28	0.09	0.03	0.03	2.79	0.32
PR Inattention	1.81 (0.21)	2.04 (0.29)	2.33 (0.11)	1.13	0.10	2.91	0.33	1.08	0.13
PR Hyp/Imp	1.50 (0.09)	1.43 (0.30)	1.73 (0.09)	0.04	0.04	3.81*	0.31	0.89	0.12
TR Inattention	2.11 (0.07)	2.82 (0.30)	3.24 (0.48)	3.70*	0.34	4.64*	0.57	1.43	0.14
TR Hyp/Imp	1.26 (0.11)	2.38 (0.58)	2.48 (0.39)	2.66	0.28	1.60	0.70	0.01	0.02
ODD	0.17 (0.13)	0.32 (0.05)	0.37 (0.06)	1.06	0.17	2.14	0.21	0.53	0.11
Anxiety	0.19 (0.05)	0.45 (0.08)	0.97 (0.40)	2.75	0.32	9.61**	0.48	1.29	0.46
Reading Achievement	85.36 (0.65)	78.74 (4.84)	74.26 (1.62)	0.66	0.20	35.57***	1.40	0.32	0.12
Rapid Letter Naming	8.75 (0.54)	8.49 (0.19)	7.07 (0.24)	0.18	0.07	8.04**	0.42	18.69***	0.82
Phonemic Awareness	6.93 (0.20)	6.41 (0.47)	5.81 (0.25)	0.93	0.15	11.22***	0.65	1.30	0.16

PR parent rated, TR teacher rated. Comb Combined Treatment, Hyp/Imp hyperactivity/impulsivity, Wald multivariate Wald χ² (1) and represents differences in the likelihood of having being in either treatment condition or having higher (or lower) levels of baseline functioning

^*p < 0.05.

^**p < 0.01.

^***p < 0.001

Teacher-Rated Impairment in Academic Progress on the IRS

A three-class solution was the best fitting model for teacher-rated impairment in academic progress (see Table S1 in supplementary materials). Although the adjusted LRT was nonsignificant and the bootstrapped LRT was significant for both the 3-class and 4-class solutions, the BIC decreased in the 4-class solution, supporting the 3-class model. Further, a visual inspection of the 4-class solution demonstrated two classes with parallel patterns (similar intercepts and slopes) that were not qualitatively different from the stable/low improving class in the 3-class solution. We retained the 3-class model solution as the preferred unconditional model, which adequately discriminated between classes with class probabilities ranging from 0.92—0.93. All three trajectories started with high levels of teacher-rated academic impairment at baseline. The first trajectory (Fig. 1b) had high initial levels of academic impairment (M = 5.86, SE = 0.05) which remained high and above the clinical range (stable/low improving class; 24.2%, n = 53; μ _slope = −0.21, p = 0.08). A second class (52.1%, n = 112; moderately improving class) started with high impairment (M = 4.71, SE = 0.20) that decreased moderately (μ _slope = −2.23, p < 0.001) although remained above the clinical range. A third class (23.7%, n = 50; high/rapid improving class) had high initial levels of impairment (M = 5.13, SE = 0.15) but demonstrated rapid decreases, particularly in the first three months, which leveled off by follow-up (μ _slope = −0.81, p < 0.001), with impairment below the clinical range.

Treatment condition significantly predicted teacher-rated academic impairment trajectories, such that children who received combined intervention were significantly more likely to be in the high/rapid improving trajectory than the stable/low improving impairment trajectory (d = 0.30; see Table 1) relative to children in the reading intervention only who were more like to be in the stable impairment trajectory. There were no significant differences between reading intervention only relative to the ADHD treatment only condition, or between the ADHD treatment only and combined treatments, across the trajectories. Co-occurring anxiety was significantly associated with being moderate and stable/low improving trajectories, relative to the high/rapid improving trajectory (ds = 0.30 and 0.29, respectively). Baseline Basic Reading Skills, rapid letter naming, and phonemic awareness were also significant predictors, such that reading achievement and phonemic awareness was significantly higher in the high/rapid improving trajectory relative to the moderate improving (ds = 0.35–0.41) and stable/low improving (ds = 0.43–0.54) trajectories. Those in the high/rapid improving trajectory also had significantly higher scores on rapid letter naming than those in the stable/low improving trajectory (d = 0.29). Baseline inattention and hyperactivity/impulsivity severity and ODD were not significant predictors.

Oral Reading Fluency

A three-class model proved to be the best fit for children’s performance on the DORF (see Table S1 in supplementary materials). The increased BIC, nonsignificant the adjusted LRT statistic and bootstrapped LRT for the 4-class model, and significant adjusted LRT and bootstrapped LRT for the 3-class model provides support for the 3-class solution. Further, the smallest class size for the 3-class model represented 20.8% of the sample, however in the 4-class model there was one class representing less than 5% of the sample (2.4%, n = 3). We retained the 3-class model solution and this model adequately discriminated between classes with class probabilities ranging from 0.88—0.95. The first trajectory (Fig. 1c) started with relatively high reading fluency (M = 69.62, SE = 3.70) which steadily increased throughout the intervention (high improving class; 37.1%, n = 80; μ _slope = 21.35, p < 0.001). A second trajectory started with moderate reading fluency (M = 33.50, SE = 3.46) that increased particularly in the first month of intervention and during follow-up (moderate improving class; 42.1%, n = 91; μ _slope = 14.56, p < 0.001). A third trajectory started with low reading fluency (M = 10.97, SE = 1.39) that remained fairly stable, only increasing slightly during the intervention (stable/low improving class; 20.8%, n = 45; μ _slope = 9.66, p < 0.001).

Treatment condition significantly predicted oral reading fluency, such that relative to the ADHD intervention only, children who received reading intervention were significantly more likely to be in the high improving trajectory than either the moderate improving or stable/low improving trajectories (ds = 0.33–0.53; see Table 1). There were no significant differences between the combined intervention relative to either reading only or ADHD only, across the trajectories. Baseline teacher-rated inattentive severity was significantly associated with high-improving trajectory, such that children with higher-rated ADHD were more likely to be in the moderate improving (d = 0.34) or stable/low improving (d = 0.57) trajectories relative to the high/improving trajectory. Those with higher parent-rated hyperactivity/impulsivity severity were more likely to be in the stable/low improving trajectory relative to the high improving trajectory (d = 0.31). Of note, children with higher parent-rated inattention severity were more likely to be in the stable/low improving relative to the high improving trajectory, however this was not significant (χ² (1) = 2.91, p = 0.081, d = 0.33). Co-occurring anxiety was significantly associated with being in the stable/low improving trajectory relative to the high improving trajectory (d = 0.45). Basic Reading Skills, rapid letter naming, and phonemic awareness were also significant predictors, such that higher scores at baseline were more likely to be in the high improving trajectory relative to the stable/low improving trajectory (ds = 0.42 – 1.40). Those in the moderate improving trajectory also had significantly higher scores on rapid letter naming than those in the stable/low improving trajectory (d = 0.82). Baseline ODD was not a significant predictor.

Discussion

This study evaluated trajectories of response to intervention in children with co-occurring ADHD/RD who received either a) reading intervention only, (b) ADHD treatment only, or (c) combined reading and ADHD treatment. In addition, predictors of treatment response were evaluated with a focus on determining what potentially modifiable factors distinguished participants who improved greatly from those participants who made only small or negligible improvements. Using GMM, three patterns of treatment response were identified for academic progress ratings (stable, moderately improving, and rapidly improving) and oral reading fluency (low/stable, moderate/improving, and high/improving). Across outcomes the smallest proportion of youth were in the class demonstrating minimal treatment gains (7–24%). Most children were in the moderately improving class (43–79%) followed by the rapidly improving class (14–37%), which made substantial improvements, particularly for academic progress which moved from having impairment in the clinical range prior to intervention to being within the normal range post intervention according to both parent and teacher reports (see Fig. 1a, b). Consistent with other studies of treatment response patterns (Haas et al., 2002; Lutz et al., 2014), those in the rapidly improving class achieved success early (e.g., after the first month of treatment) and maintained their gains over time.

Several predictors were significantly associated with the treatment response trajectories. Treatment condition significantly differed across trajectories for parent- and teacher-rated academic impairment, such that youth in combined treatment were significantly more likely to be in the high/rapid improving trajectory than the stable/low improving trajectory, relative to children in the reading intervention only who were more likely to be in the stable/low improving class. Relative to children in the reading intervention only group, those who received the ADHD intervention only were also more likely to be either the high/rapid improving or moderately improving trajectories relative to the stable/low improving trajectory for parent-rated academic progress. Alternatively, for children’s oral reading fluency, relative the ADHD only intervention, children in the reading intervention only were more likely to be in either improving trajectory (high or moderate) relative to the low improving trajectory. Interestingly, across all outcomes, there were no differences between the combined treatments and ADHD treatment only groups.

These findings are consistent with the group-level analyses comparing the three treatment groups in the RCT at post treatment which found significant differences for parent and teacher ratings of ADHD symptoms (such that ADHD treatment only and combined interventions were similarly greater than the reading intervention only) (Tamm et al., 2017). However, contrary to findings from group-level treatment outcomes that demonstrated no differences post-treatment for oral reading fluency on the DORF (Denton et al., 2020), in the present study the reading intervention was significantly associated with the high/improving reading fluency trajectory relative to the ADHD treatment only which was associated with the moderate and low improving trajectories of oral reading fluency. These results are supported by reviews demonstrating limited group-level effects of ADHD interventions on reading outcomes and limited group-level effects of reading interventions on ADHD symptoms (Froehlich et al., 2018; Roberts et al., 2019a). Furthermore, these findings are consistent with work showing that children receiving medication and/or behavioral parent training treatments for ADHD have demonstrated significant immediate response (Swanson et al., 2007) which may be due to the relatively immediate effect of medication for some children with ADHD (Swanson et al., 1995). Interestingly, the present study did not identify any “honeymoon” trajectories (i.e., initial improvement followed by deterioration over time) identified in prior studies of ADHD treatments, particularly with medication response (Swanson et al., 2007). Notably, the study intervention period was shorter (16 weeks) than the MTA study (14 months), and only a limited set of outcomes were assessed.

Although the majority of the sample fell within either the rapidly improving or moderately improving trajectories across outcomes, it is somewhat surprising that a greater proportion of youth were not in the rapidly improving class given the known efficacy of the ADHD (Wolraich et al., 2019) and RD treatments (Fletcher et al., 2019). Indeed, one group of students were identified for each outcome who started with at least moderate impairment and only made small or negligible gains. Fortunately, only a small proportion of students did not respond optimally (7–24%), however these findings suggest some students with ADHD/RD do not respond well to ADHD and/or reading interventions. These findings are consistent with reading intervention studies indicating variability in the adequacy of response to reading interventions (Torgesen, 2000), which may partially be due to the criteria used to measure adequate response to intervention. Alternatively, some students may take longer to show gains in academic progress and perhaps would benefit from more extended/sustained (e.g., yearlong) reading intervention (Vaughn et al., 2010). Of note, the 16-week reading intervention in the current study was shorter than typically provided to students with RD (Wanzek & Vaughn, 2007; Wanzek et al., 2010). These findings highlight the importance of evaluating individual differences and the need to identify potentially malleable predictors of this differential response.

The presence of a comorbid anxiety disorder was associated with less positive trajectories of treatment response for oral reading fluency and teacher-rated academic progress. Overall these findings are in line with the mixed evidence on impact of co-occurring anxiety, with some studies demonstrating anxiety predicts lower teacher-rated disruptive behavior outcomes in children treated with medication only or medication and behavior therapy (van der Oord et al., 2008b) others finding anxiety is associated with poorer response to medication and combined treatments as rated by parents (MTA Cooperative Group, 1999b). Nonetheless, co-occurring anxiety appears to demonstrate a negative/risk effect for oral reading fluency and teacher ratings of academic progress in the present study. Inattentive symptom severity was a significant predictor for oral reading fluency and parent ratings (but not teacher ratings) of academic progress. Neither parent nor teacher ratings of hyperactivity/impulsivity were significant predictors of parent or teacher ratings of academic progress. It is perhaps not surprising that inattention and not hyperactivity-impulsivity was a significant predictor given that genetic modeling studies show that the association of ADHD with RD is mostly explained by shared genetic influences (Wadsworth et al., 2015), and in particular driven by the association of inattention with reading decoding which largely shares the same genetic etiology in elementary school children (Plourde et al., 2015).Parent-rated hyperactivity/impulsivity severity was significantly lower in the high improving trajectory of oral reading fluency relative to the stable/low improving trajectory. Overall, these findings are also in line with previous studies showing less severe ADHD is associated with better response to medication and combined treatments (MTA Cooperative Group, 1999b; Owens et al., 2003). The null results for ODD as a predictor of treatment response trajectories across all three outcomes suggest that identification of children with ODD may not be necessary, at least for treatment response on academic progress ratings and oral reading outcomes among elementary school-aged children. However, these findings do contradict prior research with ADHD treatments that have shown comorbid ODD is linked to treatment response (e.g., Langberg et al., 2018; Murray et al., 2008). Replication is warranted.

Reading achievement, phonemic awareness, and rapid letter naming were the strongest predictors of positive treatment response, with reading achievement predicting high improving patterns in all three outcomes. Higher scores on phonemic awareness and rapid letter naming predicted optimal response to treatment, in terms of teacher-rated academic progress and oral reading fluency. Notably, effect sizes for trajectory differences in these predictors were largest for oral reading fluency. This is consistent with prior studies reporting youth with more severe reading problems tend to respond less well to reading treatment than those with milder forms of RD (Fletcher et al., 2011; Galuschka et al., 2014; Nelson et al., 2003; Scarborough, 1998).

Limitations and Future Directions

The findings of this study must be interpreted in light of its limitations. First, the measure of academic impairment was a single IRS item rated by parents and teachers which is a broad measure of overall academic functioning. As such parents and teachers could potentially rate students as improving academically based on a number of different factors including their perceived academic productivity, academic performance, classroom behavior, and/or students’ academic ability. Similarly, the DORF, although frequently used as a curriculum-based measure in classrooms, may not be the most ecologically valid measure of reading skills. It would have been ideal to gather additional objective measures of academic performance and productivity such as grades across multiple time points. Further it may be that other important individual and contextual factors not measured or associated with the interventions led to poor treatment response for some youth. Relatedly, the focus on academic outcomes is quite limited; many other constructs (e.g., other functional domains) could have been included as well. Future studies may consider examining the role of neurocognitive, sociodemographic, family, and/or school-level factors, for predicting intervention response.

Our sample size may have limited the number of trajectory classes we were able to examine given small class membership. It is possible that other trajectories of response also exist that were not captured in the present data that could result in different associations. An additional limitation is that parents and teachers were aware of children’s treatment assignment; thus, their ratings may have been influenced by expectations of improvement. Also, there was not a no-treatment control group; the mere act of participating in an intervention may account for some of the improvement reported for all groups. Further, children who were assigned to either the ADHD treatment only or (c) combined reading and ADHD treatment conditions were not asked to stop taking the ADHD medication on the day of a visit or before parents and teachers completed ratings. As such, it is possible that performance on the reading fluency task improved due in part to the effects of ADHD medications for improving children’s focus. Nonetheless, reading fluency is a multifaceted construct dependent on several skills (e.g., seeing the words, decoding words, automatically recognizing some words, vocabulary, focus), and it is important for future efforts to examine these mechanisms more closely as well as the relative role of ADHD medication for improving reading fluency.

While oversampling of low income, African American children provides valuable information about treatment of this understudied population, it may also limit generalizability. Additionally, this study focused on intervention response among youth in grades 2 to 5 as an important developmental period for examining ADHD/RD interventions and the findings may not generalize to later childhood or adolescence. Future studies should consider examining patterns of ADHD/RD intervention response over other developmental periods. Finally, the reading inclusionary criteria capture children with reading difficulties and may not fully generalize to children with reading disorder; it should be noted, however, that the sample had mean baseline word reading and decoding scores at the 4^th and 5^th percentiles.

Conclusions

Overall, the results of this study are consistent with group-level findings suggesting that disorder-specific treatments are required to treat children with comorbid ADHD/RD. The present study results suggest that treatments involving medication and behavioral parent training with or without reading intervention are similarly associated with parents’ and teachers’ perceptions of improving academic progress. Similarly, treatments involving intensive reading intervention are associated with improving oral reading fluency. Findings from this study also suggest that baseline characteristics including anxiety, reading achievement, rapid letter naming, and phonemic awareness can predict which students are likely to respond best to intervention. Specifically, youth with co-occurring anxiety and more significant reading problems were more likely to be in the low responding/stable impairment class for teacher-rated academic impairment and oral reading fluency outcomes. These findings emphasize the importance of considering individual differences in patterns of treatment response.

Our findings also have implications for clinicians making treatment recommendations and for managing expectations related to treatment efficacy. Children with ADHD/RD and co-occurring anxiety and with more significant reading problems may require even greater intensity or longer term intervention, such as more intensive Tier 3 or 4 interventions covered under the Individuals with Disabilities Education Act (IDEA, 2004) and consistent with the multi-tiered systems of support framework (Weist et al., 2014). Generally, Tier 3 or 4 interventions are individualized to address the needs of the student, incorporate additional personnel, are more time intensive, and are implemented when students first demonstrate nonresponse to targeted preventative interventions (e.g., Tier 2). Although more research is clearly needed, the results support the value-added benefits of initial assessments of co-occurring mental health, basic reading skills, rapid letter naming abilities, and phonemic awareness for predicting distinct patterns of response to intensive Tier 3 ADHD and/or reading interventions used in the present study. Consistent with the multi-tiered systems of support framework, RTI approach, and IDEA federal regulations (2004), these findings support the importance of closely monitoring response. For students who are nonresponsive to the interventions examined in the present RCT, schools and interventionists may consider more remedial or intensive skills training and/or may consider pursuing alternative interventions such as a tailored daily report card intervention used in conjunction with a home and school reinforcement system and/or extended individualized reading intervention support. Clinicians may also wish to refer a child with ADHD/RD and co-occurring anxiety for anxiety-focused therapy (e.g., cognitive behavioral therapy) prior to recommending ADHD and reading interventions. Overall, the present study highlights the importance of considering multiple outcomes, measures, and stakeholder perspectives when evaluating interventions.

Data Availability

Data used in the current manuscript is available from the corresponding author, Leanne Tamm, Ph.D., upon request and execution of a data sharing agreement.