Primary Care Management of Children with Attention-Deficit/Hyperactivity Disorder Appears More Assertive Following Brief Psychiatric Intervention Compared with Single Session Consultation

Abstract

Objectives: We examined primary care providers' (PCPs') management of attention-deficit/hyperactivity disorder (ADHD) during and following families' participation in two arms of the Children's ADHD Telemental Health Treatment Study. We hypothesized that more intensive treatment during the trial would show an “after-effect” with more assertive PCPs' management during short term follow-up.

Methods: We conducted a pragmatic follow-up of PCPs' management of children with ADHD who had been randomized to two service delivery models. In the Direct Service Model, psychiatrists provided six sessions over 22 weeks of pharmacotherapy followed by behavior training. In the Consultation Model, psychiatrists provided a single-session consultation and made treatment recommendations to PCPs who implemented these recommendations at their discretion for 22 weeks. At the end of the trial, referring PCPs for both service delivery models resumed ADHD treatment for 10 weeks. We performed intent-to-treat analysis using all 223 original participants. We applied linear regression models on continuous outcomes, Poisson regression models on count outcomes, and logistic regression models to binary outcomes. Missing data were addressed through imputations.

Results: Participants in the Direct Service Model had more ADHD visits than those in the Consultation Model across the full 32 weeks (mean = 7.05 visits vs. 3.36 visits; adjusted rate ratio = 2.1 [1.85–2.38]; p < 0.0001). During follow-up, participants in the DSM were more likely to be taking ADHD-related medications (82% vs. 61%; adjusted odds ratio = 2.44 [1.24–4.81], p = 0.01). At 32 weeks, participants in the Direct Service Model had higher stimulant dosages (adjusted difference = 5.64 [0.12–11.15] mg; p = 0.046).

Conclusion: These results from a pragmatic follow-up of a randomized trial suggest an “after-effect” for brief intensive treatment in the Direct Service Model on the short term follow-up management of ADHD in primary care.

Introduction

Numerous studies have documented that 16%–20% of children and adolescents experience a mental or behavioral health condition (American Academy of Child and Adolescent Psychiatry 2019). The chronic shortage and maldistribution of child and adolescent psychiatrists available to treat these youth (Thomas and Holzer 2006; National Institute of Mental Health 2011; American Academy of Child and Adolescent Psychiatry 2019; American Association of Medical Colleges 2019) have been exacerbated by implementation of mental health parity laws making more children eligible for services without creating new service resources (U.S. Department of Health and Human Services 2010). New models of care are being examined to address the need for more—and more equitable distribution of—psychiatric services for children. Many of these models emphasize support for primary care providers (PCPs) in assuming a central role in children's mental health care (Olfson et al. 2014). How much and the type of support remain unresolved.

Attention-deficit/hyperactivity disorder (ADHD) provides an excellent clinical condition to address this issue. Clear guidelines of care have been developed for the management of ADHD within primary care (American Academy of Pediatrics subcommittee on ADHD 2011), and treatment entails a major role for medication. However, prior work indicates that PCPs vary in their management of ADHD (Epstein et al. 2014, 2017; Silverstein et al. 2015; Brinkman et al. 2016). Cited concerns include reluctance to prescribe stimulant medication, failure to adjust medication according to the child's response, variability in timeliness of follow-up, and lack of access to psychosocial and other supportive interventions (Epstein et al. 2010, 2014; Brinkman et al. 2016). Prior investigations indicate that follow-up within a month after the initial prescription has been associated with PCPs' continuity of treatment, ongoing medication supply, and medication adjustment (Brinkman et al. 2016). Whether psychiatric support can affect these practices, and potential models for such support, is an important next step in developing models of care for PCPs' management of ADHD.

The current investigation compares two psychiatric service delivery models with respect to their effect on PCPs' assertiveness in managing ADHD during and following families' participation in the Children's ADHD Telemental Health Treatment Study (CATTS) (Vander Stoep and Myers 2013; Myers et al. 2015; Rockhill et al. 2016). This analysis compared a brief Direct Service Delivery Model to a Consultation Model. We hypothesized that during the initial 22-week CATTS comparative effectiveness trial and during the 10-week follow-up (22 to 32-weeks), the Direct Service Model would be associated with more assertive ADHD management than the Consultation Model. To our knowledge, this is the first study to compare the management of ADHD in primary care following two different models of psychiatric intervention.

Methods

Description of the CATTS trial and outcomes

The CATTS trial was a 5-year, community-based, randomized controlled trial funded by the National Institute of Mental Health that compared the effectiveness of two telehealth service delivery models for improving the mental health care and outcomes for children with ADHD and their families (Vander Stoep and Myers 2013; Myers et al. 2015; Rockhill et al. 2016; Vander Stoep et al. 2017). The trial was approved by the Institutional Review Board of Seattle Children's Research Institute and monitored by a data safety and monitoring board. Informed consent of parents and assent of youth participants were obtained, including for the research and permission to publish.

The design of the CATTS trial has been reported previously (Vander Stoep and Myers 2013). Participants included 223 boys and girls, 5.5 to 12 years old referred to the trial by 88 PCPs in seven communities. PCPs agreed to resume ADHD treatment following the trial. Diagnoses of ADHD and comorbid oppositional defiant disorder (ODD) and generalized anxiety disorder were determined with the Computerized Diagnostic Interview Schedule for Children, Version IV (Shaffer et al. 2000). All participants met criteria for ADHD, and 75% (168/223) met criteria for one or both comorbidities.

Participants were randomized to either a Direct Service Model of care or a Consultation Model. The Direct Service Model consisted of six sessions spaced 3–4 weeks apart over 22 weeks. Child psychiatrists used video teleconferencing to deliver pharmacotherapy and psychoeducation on the neurobiological model of ADHD. They adjusted medications according to the Texas Medication Algorithm Project (Pliszka et al. 2006; Rockhill et al. 2016). Each telepsychiatry session was followed immediately by a session of a caregiver behavior training protocol delivered in-person by community therapists who were supervised remotely (McCarty et al. 2015). Psychiatrists' decision-making was described in a written report that was shared with the PCPs after each combined intervention session. Participants randomized to the Consultation Model received a single session with a child psychiatrist who summarized treatment recommendations in a written report and telephone call. At conclusion of 22 weeks of Direct Service or following the Consultation session, children's treatment was returned to the care of their referring PCP with written recommendations for continuing ADHD treatment. The PCPs implemented the recommendations at their discretion.

Caregivers completed outcome measures at baseline, 4, 10, 19, and 22 weeks following randomization (Myers et al. 2015). The primary outcome measure was the Vanderbilt ADHD Parent Rating Scale (VADPRS) (Wolraich et al. 2003). Results of the comparative effectiveness trial (Myers et al. 2015) and of the psychiatrists' pharmacological management (Rockhill et al. 2016) have been reported previously. The intent-to-treat analyses indicated that in both conditions participants' ADHD symptoms, ODD behaviors, and functioning improved. The Direct Service Model demonstrated significantly greater reductions in ADHD and ODD and greater improvement in functional outcomes. Children in the Direct Service Model, particularly those with comorbid disorders, were significantly more likely to demonstrate 50% reduction in their ADHD symptoms during the trial (Rockhill et al. 2016).

Methods for current follow-up investigation

The follow-up investigation of PCPs' ADHD treatment included all participants in the original intent-to-treat analysis. The Consolidated Standards of Reporting Trials (CONSORT) diagram is included in Figure 1. This set of analyses compares assertiveness of service delivery of the two models both during the trial (baseline to 22 weeks) and during short-term (22 to 32 weeks) follow-up. Assessments include (1) the number of visits for ADHD management; (2) the likelihood of taking psychiatric medication, particularly stimulant medication; (3) stimulant dosage; and (4) whether two or more psychiatric medications were prescribed (co-prescription), including a medication for sleep.

FIG. 1.

CONSORT diagram. ADHD, attention-deficit/hyperactivity disorder; CATTS, Children's ADHD Telemental Health Treatment Study; CBCL, Child Behavior Checklist; CDISC-IV, Computerized Diagnostic Interview Schedule for Children, based on Diagnostic and Statistical Manual, 4th Edition; CONSORT, Consolidated Standards of Reporting Trials; PCP, primary care providers.

We extracted data from PCP charts relevant to ADHD visits and medication management from baseline to 32 weeks following randomization in the original trial. We included visits that documented ADHD management, but not for other purposes. We categorized medications into four groups: any ADHD-related medication; ADHD-related stimulant only; ADHD-related nonstimulant only; and two or more psychiatric medications. The last two groups included medications to help children settle at night and to sleep such as an alpha-₂ agonist. We also recorded dosages of stimulants. For cases in which there was evidence of a stimulant prescription followed by 1–3 months without prescriptions then resumption, we assumed that the child had received a 90-day supply of medication. If no visits occurred for more than 3 months, we assumed that the child was no longer taking ADHD medication. These decisions followed general guidelines for limiting stimulant prescriptions to 3 months' supply and for the possibility of medication holidays during the summer. If the record was unclear, two study psychiatrists conferred and came to consensus.

Statistical analyses

We performed an intent-to-treat analysis, and all patients who had been randomized in the original randomized controlled trial were included in this analysis. We first summarized baseline sociodemographic and clinical characteristics in both the Direct Service Model and Consultation Model with frequencies, means, and standard deviations. We then examined missing data. Among the 223 patients who were randomized and completed baseline assessments, 15 (6.7%) had missing values in the primary outcomes (number of visits, ADHD medications). The reason for missing values was loss to follow-up. We compared baseline characteristics between patients with complete data and those with missing values. There were no differences in age, gender, comorbidity, and baseline VADPRS summary score. We also compared end of study VADPRS scores, the primary outcome for the CATTS trial, between those with and without missing data and found no significant difference. Missing data were then imputed by missForest (Stekhoven and Bühlmann 2011), an R package for a nonparametric iterative imputation method using random forest, based on the outcomes (number of visits/medications taken/stimulant dosage) from both periods, as well as covariates such as group, gender, age, baseline comorbidity, and VADPRS scores from both periods (R Core Team 2018). All regression analyses were conducted based on imputed data.

Our primary aim was to examine differences between the two service models in number of ADHD-related visits, medication class and dosage, and co-prescription of a second medication with primary ADHD medication. To this end, we fit Poisson regression models and used rate ratios to compare the mean number of ADHD-related visits between the two service delivery models in each period separately (period 1: randomization to 22 weeks and period 2: 22 to 32 weeks, and the entire period from baseline to 32 weeks). Since count data often exhibit overdispersion features in practice, we also fit quasi-Poisson models for this outcome but found that our data were actually underdispersed. To be conservative we kept results from the Poisson models. For each class of ADHD-related medications prescribed (yes/no) at each follow-up point (22- and 32 weeks) we fit logistic regression models to compare odds of taking each class of medication between the two service models. For medication dosage for period 2, measured at 32 weeks, we fit a linear model while adjusting for baseline medication dosage. To increase precision, we adjusted for baseline comorbidity and nearest VADPRS summary score in all the regression models. A significance level of 0.05 and two-sided tests were applied throughout. All statistical analyses were conducted in R (version 3.4.4; R Core Team 2018).

Results

Characteristics of participants

Number of children lost to follow-up after initial randomization did not differ significantly for the Direct Service Model and the Consultation Model (7% vs. 9%, respectively). Table 1 summarizes sociodemographic and clinical features for participants initially randomized to each service model (Myers et al. 2015). Participants in the Direct Service Model had a higher occurrence of any comorbidity at baseline than those in the Consultation Model (82% vs. 69%). Baseline VADPRS scores were slightly higher for youth in the Direct Service Model (2.0 vs. 1.9). Dosages of stimulant medication and the occurrence of a co-prescription of a second medication with a stimulant were similar for the two service models. VADPRS scores completed after the 22-week intervention were significantly lower in the Direct Service Model (1.1 vs. 1.4) demonstrating greater efficacy on children's outcomes of this model. These descriptive findings influenced our decisions regarding the covariates to be included in all statistical models.

Table 1.

Patient Characteristics by Service Delivery Model

Variable	N	Direct service	Consultation
Total no. of participants	223	111	112
Age, mean (SD)	221	9.3 (2.0)	9.2 (2.0)
Sex, N (%)	223
Male		76 (68)	87 (78)
Female		35 (32)	25 (22)
Comorbidity (three levels), N (%)	223
ADHD alone		20 (18)	35 (31)
One comorbidity		51 (46)	55 (49)
Two comorbidities		40 (36)	22 (20)
Comorbidity (two levels), N (%)	223
None		20 (18)	35 (31)
One or more		91 (82)	77 (69)
VADPRS score, mean (SD)
Baseline	223	2.0 (0.5)	1.9 (0.5)
22 weeks	204	1.1 (0.5)	1.4 (0.6)
Stimulant dosage at baseline, mean (SD)	205	23.8 (26.4)	21.2 (27.7)

ADHD, attention-deficit/hyperactivity disorder; SD, standard deviation; VADPRS, Vanderbilt ADHD Parent Rating Scale.

Visits for ADHD management

As noted in Table 2, participants in the Direct Service Model had significantly more ADHD visits than those in the Consultation Model across the full 32 weeks (M = 7.05 vs. M = 3.36; adjusted rate ratio = 2.10 [1.85–2.38]; p < 0.0001). This included more visits during both the initial trial to 22 weeks (M = 6.11 vs. M = 2.94, adjusted rate ratio = 2.08 [1.82–2.38]; p < 0.0001) and during follow-up from 22- to 32 weeks (M = 0.94 vs. M = 0.42, adjusted rate ratio = 2.20 [1.52–3.19]; p < 0.0001).

Table 2.

Number of Visits by Service Models Across Initial Randomized Trial and 10-Week Follow-Up

	Direct service	Consultation	Rate ratio (95% CI)	p-Value
	Mean (SD)	Mean (SD)
Visits: baseline to 22 weeks	6.11 (1.43)	2.94 (1.42)	2.08 (1.82–2.38)a	<0.0001
Visits: 22–32 weeks	0.94 (0.75)	0.42 (0.65)	2.2 (1.52–3.19)a	<0.0001
Visits: baseline to 32 weeks	7.05 (1.72)	3.36 (1.62)	2.10 (1.85–2.38)a	<0.0001

^aResults are from adjusted Poisson regression models. For baseline to 22 weeks and baseline to 32 weeks, models were adjusted for baseline comorbidity and baseline VADPRS item summary scores. For 22–32 weeks, we adjusted for baseline comorbidity and VADPRS item summary scores at 22 weeks.

ADHD, attention-deficit/hyperactivity disorder; CI, confidence interval; SD, standard deviation; VADPRS, Vanderbilt ADHD Parent Rating Scale.

Pharmacotherapy

As shown in Table 3, participants in the Direct Service Model were not significantly more likely than those in the Consultation Model to be taking any ADHD-related medication from randomization to 22 weeks (0.77 vs. 0.72; adjusted odds ratio [OR] = 1.43 [0.76–2.69], p = 0.27). This included no difference in a stimulant alone, a nonstimulant, and two or more psychiatric medications.

Table 3.

Medication Regimens Comparing Service Delivery Models Across Initial Randomized Trial and Follow-Up

	Direct service	Consultation	OR (95% CI)	p-Value
	Proportion (n/N)	Proportion (n/N)
Any ADHD-related medication
Period 1: baseline to 22 weeks	0.77 (79/103)	0.72 (76/105)	1.43 (0.76–2.69)a	0.27
Period 2: 22–32 weeks	0.82 (84/103)	0.61 (65/106)	2.44 (1.24–4.81)b	0.01
ADHD-related stimulant only
Period 1: baseline to 22 weeks	0.70 (72/103)	0.64 (67/105)	1.53 (0.85–2.77)a	0.16
Period 2: 22–32 weeks	0.76 (78/103)	0.54 (57/106)	2.52 (1.33–4.76)b	0.005
ADHD-related nonstimulant
Period 1: baseline to 22 weeks	0.29 (30/103)	0.21 (22/105)	1.40 (0.74–2.68)a	0.30
Period 2: 22–32 weeks	0.30 (31/103)	0.17 (18/106)	2.21 (1.08–4.55)b	0.03
Two or more psychiatric medications
Period 1: baseline to 22 weeks	0.29 (30/103)	0.21 (22/105)	1.23 (0.64–2.38)a	0.53
Period 2: 22–32 weeks	0.33 (34/103)	0.17 (18/106)	2.56 (1.25–5.27)b	0.01

^aOR adjusted for baseline comorbidity and baseline VADPRS summary score.

^bOR adjusted for baseline comorbidity and 22-week VADPRS summary score.

ADHD, attention-deficit/hyperactivity disorder; CI, confidence interval; OR, odds ratio; VADPRS, Vanderbilt ADHD Parent Rating Scale.

However, during follow-up from 22- to 32 weeks, participants in the Direct Service Model were significantly more likely to be taking any ADHD-related medications (0.82 vs. 0.61; adjusted OR = 2.44 [1.24–4.81], p = 0.01). This included a greater proportion of children in the Direct Service Model taking a stimulant only, taking a nonstimulant, and taking two or more medications. This difference appeared at least, in part, related to a larger proportion of children across the four medication groups in the Direct Service Model who continued to take medication than children in the Consultation Model who were more likely to discontinue medication.

During the full study course from randomization to 32 weeks, participants in both models who were prescribed stimulants had an increase in their dose. Those in the Direct Service Model increased their dose from an average of 23.76 (±26.41) mg at baseline to 38.06 (±22.27) mg at 32 weeks (change of 14.30 mg) and those in the Consultation Model increased their dose from an average of 21.17 (±27.72) mg at baseline to 28.95 (±23.45) mg at 32 weeks (change of 7.78 mg). Adjusted analyses indicated that these changes represented a significant difference in stimulant dose between the two models at 32 weeks after adjusting for baseline values (difference in dose at 32 weeks in adjusted analysis = 5.64 [0.12–11.15] mg; p = 0.046).

Discussion

The current pragmatic follow-up of ADHD management in primary care following a randomized clinical trial was conducted to contribute to the emerging evidence base regarding models of care for supporting PCPs in the increasing expectation that they manage the treatment of children diagnosed with ADHD. We hypothesized that compared to a single-session Consultation Model, a brief Direct Service Model of care would be associated with more assertive ADHD management during the trial and would demonstrate an “after-effect” evidenced in more assertive ongoing management by the referring PCPs who resumed children's treatment during short-term (10-week) follow-up. We found some support for this hypothesis during the trial in terms of a greater number of visits for ADHD treatment but not for medication use. We found support for the hypothesis during short-term follow-up in terms of the number of visits, prescribing of medication, higher doses of stimulant medication, and use of two or more ADHD-related medications.

These findings are consistent with the recommendation for more frequent and timely follow-up (American Academy of Pediatrics Subcommittee on ADHD 2011; Epstein et al. 2014, 2017; Silverstein et al. 2015; Brinkman et al. 2016) and a team-based approach (Wolraich et al. 2019) to the treatment of ADHD. The directionality of the follow-up results is not clear from our findings. On the one hand, leveraging psychiatrists through telepsychiatry to work with community-based therapists provided families with knowledge of the neurobiological model of ADHD and the core role for medication, as well as training in the evidence-based interventions to address the behavioral deficits in ADHD. This knowledge along with the greater reductions in ADHD and ODD symptoms and greater improvement in functional outcomes (Myers et al. 2015; Rockhill et al. 2016) during the trial may have motivated parents in the Direct Service model to continue treatment. On the other hand, the written updates from the psychiatrist after each session may have provided education for PCPs, as well as provided a basis for prescribing more assertive medication regimens during follow-up.

In the 22-week intervention trial, the twofold increase of sessions for children randomized to the Direct Service Model was expected, as the series of six sessions was scheduled at study enrollment. In the Consultation model, PCPs had clinic visits with patients approximately once every two months (2.97 sessions during 22 weeks) following the single psychiatric consultation. While less frequent than the 3–4-week interval between sessions in the Direct Service Model, it is more assertive than simply providing a 3-month supply of medication or not scheduling timely follow-up to adjust care according to the child's response. The lower frequency of visits in the Consultation Model was also seen in the 22- to 32-week follow-up, in part, related to the decreased proportion of children continuing to take medication during the follow-up period.

Finally, it is worth noting that the PCPs and psychiatrists were equally likely to prescribe both stimulant and nonstimulant ADHD medications during the trial, suggesting increasing expertise in primary care to treat ADHD, including with stimulants. It was during follow-up that the models diverged. Overall, the current investigation supports an “after-effect” of a brief expert Direct Service Model compared with a single session Consultation Model for facilitating PCPs' assertiveness in the short-term follow-up management of ADHD.

Limitations

This analysis of primary care provided after a randomized trial aimed to compare differences in PCPs' management of ADHD across two service models. Given that these outcomes were not the outcomes for which we powered the initial trial, we might have limited sample size and power for some of the comparisons. As the Direct Service Model consisted of several active components during the intervention, we cannot attribute findings to a specific component of the model's intervention due to lack of component-specific data. The follow-up data were obtained from medical records that are subject to recording errors and inconsistencies. We had to make some a-priori decisions about coding prescribing practices on the basis of the information contained in medical records. Finally, the samples were derived from a telepsychiatry study of children in underserved communities. Thus, the generalization of findings may be limited to similar geographic settings and families willing to obtain expert care remotely.

Conclusion

The results from this pragmatic follow-up of a randomized trial suggest an “after-effect” for brief intensive treatment in the Direct Service Model on the short-term (10-week) assertive follow-up management of ADHD in primary care. The PCPs who resumed care after the Direct Service Model had more ADHD visits, prescribed more ADHD-related medications, and prescribed higher doses of stimulant medication than those who resumed care after the Consultation Model.

Clinical Significance

The increasing need for psychiatric services for children requires new service delivery models for collaboration between psychiatrists and PCPs. Results of the current study suggest that both a Direct Service Model and a Consultation Model support PCPs in managing ADHD and, thereby, improve the care and outcomes of children with ADHD. A single session Consultation Model may optimally conserve the expert psychiatric workforce and distribute that expertise over the broadest population, but may not provide optimal initial outcomes or prompt optimal PCP management of ADHD over time. The brief Direct Service Model may help to optimize initial outcomes, as well as produce an “after-effect” for PCPs' follow-up care, but may be more difficult for PCPs to access. Hopefully, the current findings will encourage further discussion of the options for collaboration between primary care and psychiatry in optimizing the treatment of children with ADHD.

Disclosures

The authors have no financial relationships relevant to this article to disclose.