A Multimodal Intervention for Children with ADHD Reduces Inequity in Health and Education Outcomes

Abstract

Objective:

To evaluate whether a multimodal intervention for children with attention-deficit hyperactivity disorder (ADHD) resulted in better long-term health and education outcomes and reduced inequity across the socioeconomic gradient.

Method:

We analyzed administrative data held in the Manitoba Population Research Data Repository describing recipients of a combined pharmacological/behavioural intervention for ADHD. The study cohort included children aged 5 to 17 years who visited the Manitoba Adolescent Treatment Centre’s ADHD intervention service at least 3 times (2007-2012). Controls were matched on age, sex, year of ADHD diagnosis, and income quintile. We compared rates of hospital and emergency department visits, medication use and adherence, contact with child welfare services, and whether children were in their age-appropriate grade. We used concentration curves to estimate differences in outcomes between children from high- and low-income families.

Results:

Children who received the intervention (n = 485) had higher rates of medication use (rate ratio [RR], 1.21; 95% CI, 1.08 to 1.36) and adherence (RR, 1.42; 95% CI, 1.03 to 1.96) and were more likely to be in their age-appropriate grade (RR, 1.33; 95% CI, 1.09 to 1.63) compared with controls (n = 1884). The intervention was also associated with reduced inequity in these outcomes across income deciles. There was no difference in the rates of hospital or emergency department visits or contacts with child welfare services.

Conclusions:

A multimodal ADHD intervention was associated with increased medication use and adherence and higher academic achievement. It was also related to lower inequity across the socioeconomic gradient. These results suggest that multimodal approaches may provide more equitable health and education outcomes for children with ADHD.

Attention-deficit hyperactivity disorder (ADHD), characterized by pervasive and age-inappropriate behavioural development, impulsivity, and hyperactivity, is the most common neurobehavioural disorder among children and youth.^1,2 The disorder is a cause of significant morbidity: children with ADHD are often diagnosed with psychiatric comorbidities,^3,4 and ADHD is associated with higher rates of hospital visits for accidental injuries, substance use, and suicidal behaviour.⁵ ADHD diagnoses also correlate with serious behavioural, academic, and social difficulties, which often persist from childhood and adolescence into adulthood.⁶ In school-aged children, ADHD has been linked to parent-child conflicts and decreased academic productivity,⁷ while older individuals with ADHD tend to have poor peer relationships⁸ and lower rates of high school graduation and completion of postsecondary education.⁹ The long-term sequelae of academic underachievement (e.g., fewer employment opportunities, lower earning potential) perpetuate a cycle of disadvantage, particularly among populations of lower socioeconomic status.¹⁰

Clinical and research efforts have focused on improving mental health and educational outcomes for several decades.¹¹ The Canadian ADHD Resource Alliance recommends long-acting psychostimulants as first-line pharmacotherapy for children and adolescents.^12,13 These have been shown to reduce the core symptoms of ADHD.¹⁴ Another approach is behavioural interventions based on strategies to eliminate problem behaviour and promote positive behaviour.¹⁵ The effectiveness of these nonpharmacological approaches for reducing ADHD symptoms has been well established, at least in the short term and for younger children.^16,17 However, studies focusing on short-term changes do not address the underlying goal of early intervention (i.e., preventing the social, academic, and vocational sequelae of ADHD), and the literature suggests that short-term improvement in ADHD symptoms is not maintained over the long term.¹⁸ Moreover, there is little evidence on the efficacy of interventions combining medication and behavioural modification into multimodal approaches,¹⁹ which have been shown to provide superior benefits, as they have a number of complementary effects.^20,21

Over the past several decades, ADHD diagnosis in school-aged children has not varied significantly, with a mean prevalence estimate just over 5% from 1985 to 2012.²² However, children from lower-income families are often disproportionately represented in this group.^23,24 Health policies and programs that have greater impact on (or are more accessible to) wealthy population sectors can widen the gap between the wealthy and the poor, contributing to a cycle of disadvantage in the lives of vulnerable individuals. Therefore, it is critical that interventions be evaluated for their effects not only on health but also on the distribution of outcomes across the socioeconomic gradient.²⁵

Given the wide-reaching negative consequences of ADHD, managing this disorder has been identified as an important public health priority.²⁶ In 2007, the Manitoba government announced a new investment to develop a comprehensive multidisciplinary program for children and youth (aged 5-17 years) living with ADHD.²⁷ This program, the ADHD Service, supports children and youth, their families, and their schools, including ongoing consultations with mental health professionals and family intervention specialists. Based on the literature, we would expect to see reduced ADHD symptoms in program participants, but whether these changes translate into long-term improvements in health and education outcomes remains unknown. Similarly, whether the intervention’s impact is equitably distributed across the socioeconomic gradient has not been explored.

Thus, we conducted analyses using administrative data to evaluate the effects of the multimodal ADHD Service with the goal of addressing a number of gaps in the ADHD literature, including examining the long-term impact of such interventions on health and educational outcomes, the efficacy of mixed pharmacological and behavioural approaches in children and youth, and the impact on the distribution of outcomes across the socioeconomic gradient.

Method

Administrative Data Sources

This study used administrative data routinely collected from publicly funded services in Manitoba, describing health care use, social services use, and education outcomes of virtually all Manitobans (1.3 million individuals). The data are compiled in the Manitoba Population Research Data Repository housed at the Manitoba Centre for Health Policy (MCHP), which has the unique ability to link individual-level data using scrambled identifiers across multiple databases. The PATHS Data Resource used in this study was created from the Repository²⁸ and includes data files for children born from 1984 to 2012. Studies using Repository data have demonstrated high validity and reliability over time for measuring health outcomes.^29–32

Program data from the ADHD Service were acquired from the Manitoba Adolescent Treatment Centre (MATC), through which the intervention is implemented. The ADHD Service provides assessment, treatment, and consultative services for children and youth aged 5 to 17 years and their families. Services may include individual therapy, parent support, group therapy, education, and medication management with a multidisciplinary team. The typical length of participation in the program is 3 to 6 months; however, depending on the recipient’s needs, this period may be longer. Referrals to the program by physicians occur through the Regional Health Authority’s Child and Adolescent Mental Health Service and are drawn from a wide catchment area that includes Manitoba and Nunavut. Our evaluation used data for residents of Manitoba only. Data on specific components of the program, such as how many hours and what types of therapy participants received, were not available for this evaluation.

We linked the program data from MATC with the following Repository databases: 1) physician claims, which are filed by a physician and include a single diagnosis code; 2) hospital discharge abstracts, submitted when an inpatient is discharged from hospital and include up to 25 diagnosis codes; 3) the emergency department information system, which tracks patients during their stay in the emergency department; 4) the Drug Programs Information Network, which contains information on prescription drugs dispensed from community pharmacies; 5) education data, which include school enrollment, graduation, and academic achievement data; and 6) family services data, which include reports of child welfare services involvement and family receipt of income assistance.

We used publically available Statistics Canada census data to construct income quintiles for each census dissemination area in Manitoba (400-700 individuals). The dissemination areas were sorted by average household income and divided into quintiles of equal population size. Previous research from Manitoba has demonstrated that area-level and individual-level income measures are well correlated.³³ About 1% of individuals in Manitoba were excluded from the income quintiles because their postal code did not link with a dissemination area or they lived in a dissemination area where 90% or more of the population was institutionalized (i.e., prison, nursing home). The study was approved by the University of Manitoba Health Research Ethics Board and the Manitoba Health Information Privacy Committee.

Study Design

We identified individuals in the PATHS Resource who had been diagnosed with ADHD and had been treated by the ADHD Service between 2007 and 2012. Individuals in the intervention group needed to have had at least 3 ADHD treatment sessions during the study period to ensure adequate exposure to the intervention. For each person in the intervention group, up to 5 controls with a diagnosis of ADHD but who did not have contact with the ADHD Service were matched on age, sex, year of ADHD diagnosis, and income quintile. Matches were identified separately in urban and rural income quintiles.

We selected confounders that reflected use of health services (medical claim and hospital discharge diagnoses for ADHD, drug dispensations in the 2 years before diagnosis), social services use (receiving income assistance), and education outcomes (school readiness³⁴ and special needs funding) to ensure that the severity of ADHD among children in our study was comparable. Health services use variables demonstrate the level of contact with the health care system for ADHD-related treatment, social services use variables capture the association between low income and the prevalence of ADHD, and education variables such as special education funding and school readiness are among the few early childhood indicators for this disorder.

Confounders were used to calculate propensity scores that captured children’s probability of having contact with the ADHD Service. Propensity scores are the probability of exposure to the ADHD service obtained from a logistic regression analysis where ADHD service receipt was the outcome variable. To limit the analytic sample to areas of common support (i.e., where the probability of exposure or nonexposure was sufficiently different from 0), propensity scores were trimmed asymmetrically at the 2.5th percentile among children who received treatment through the ADHD Service and at the 97.5th percentile in the matched control group.³⁵ Trimming following logistic regression is akin to a matching procedure in which cases with no similar controls (and vice versa) are excluded, improving the accuracy and precision of final parameter estimates.³⁶ We constructed the inverse probability of treatment weights using these propensity scores and applied them to the data to adjust for the confounding variables.³⁷ We then assessed the standardized differences in confounders between the intervention and control groups using an a priori cutoff of 0.10 to establish whether the groups were balanced.³⁸

Outcome Measures

Health and Social Services Use

Primary outcomes included the rates of hospital admissions, emergency department visits (all visits and injury-related visits), medication use (the number of patients who had 1 or more ADHD medications dispensed), and medication adherence (number of patients who had a medication possession ratio of at least 0.8). The medications for ADHD included in our definitions of medication use and adherence were as follows: mixed salt amphetamine, methylphenidate, methamphetamine, dextroamphetamine sulfate, pemoline, modafinil, lisdexamfetamine, and atomoxetine. In Manitoba, the universal Pharmacare plan provides medication coverage for low-income families.

We measured the health outcomes during the 2 years following the patients’ last recorded visit to the ADHD Service. Primary outcomes were tested at α = 0.05 due to the correlated nature of the outcome variables. Even though Bonferroni adjustment in this case would be overly conservative, we reported our results at P < 0.01 to ensure that we met the criteria. We also determined whether the study cohort had any new contacts with the child welfare system in the 2 years following their treatment and whether they were in the school grade appropriate to their age (as a proxy for grade repetition) in the academic year following their last visit to the ADHD Service. We measured outcomes in controls using this same time interval but starting at the control’s own ADHD diagnosis date.

Inequities in Health and Social Outcomes

Inequities related to socioeconomic status were measured by calculating concentration curves and concentration indices. Concentration curves are a measure of how evenly an outcome is distributed across the socioeconomic gradient.^39,40 The “line of equity” represents a situation in which an outcome is evenly distributed across the socioeconomic gradient. Concentration curves that lie above the line of equity indicate that an outcome tends to occur more in lower income deciles (i.e., is more “concentrated” in lower income deciles), while curves that lie below the line of equity demonstrate that an outcome is more concentrated among higher income deciles.

Concentration indices are determined by calculating the area between the concentration curve and the line of equity. A concentration index approaching +1 indicates that the outcome is concentrated among the wealthy deciles, whereas a value approaching –1 means it is more common among the poor deciles. A concentration index not significantly different from 0 indicates that the outcome is evenly distributed across income deciles. Calculating the difference in concentration indices between two groups allows an estimate of change in equity. Therefore, when comparing an intervention group to a control group, this method can be used to demonstrate whether a program or intervention increases or decreases inequity related to socioeconomic status in outcome distribution.

Statistical Analyses

We applied a generalized linear model using a negative binomial distribution and a log-population offset to generate rate ratios (RRs) comparing health and social services use between the intervention and control groups. We used logistic regression models to calculate odds ratios (ORs) for the relationship between treatment through the ADHD Service and whether the child was in his or her age-appropriate grade. Statistically significant relationships were identified using 95% confidence intervals. All data management, study design, and data analyses were performed using SAS version 9.3.⁴¹

Results

Characteristics of the study cohort are described in Table 1. There were no significant differences in baseline characteristics between the intervention and control groups. The majority (61%) of individuals in the study cohort were younger than 10 years. Median ADHD Service treatment duration was about 8 months, with an average of 24 visits per individual. Matching ensured that boys/girls and different age groups were equally represented in the intervention and control groups (Supplemental Table S1).

Table 1.

Descriptive Characteristics of the Study Cohort (2007-2012).

	Intervention		Control
	n	%	n	%
Total n	485	100	1884	100
Sex
Male	401	83	1604	85
Female	84	17	280	15
Age at first visit (y)
≤6	76	16	313	17
7	79	16	346	18
8	71	15	297	16
9	70	14	266	14
10	49	10	190	10
11	38	8	136	7
12	38	8	148	8
≥13	64	13	188	10
Income quintile
Q1	93	19	399	21
Q2	94	19	353	19
Q3	92	19	333	18
Q4	88	18	338	18
Q5	109	22	429	23
Not found	9	2	32	2
Number of treatment visits
Mean	24.4		n/a
Median	17		n/a
Duration of treatment (d)
Mean	273.8		n/a
Median	245		n/a

n/a = not applicable.

The study cohort’s medication use is described in Table 2, and health service use, social service use, and education outcomes are shown in Table 3. Hospitalization and emergency department visit rates were not different between groups. However, treatment in the ADHD Service was associated with increased rates of prescribing and filling of prescriptions. Children who received treatment through the ADHD Service were also more likely to be enrolled in their age-appropriate grade following the intervention.

Table 2.

Attention-Deficit Hyperactivity Disorder Medication Use by the Study Cohort.^a

Medication	Intervention	Control
	n (%)	n (%)
Atomoxetine	47 (9.1)	100 (6.9)
Dextroamphetamine sulfate	43 (8.3)	116 (8.1)
Lisdexamfetamine	46 (8.9)	19 (1.3)
Methylphenidate	363 (70.0)	1183 (82.2)
Mixed salt amphetamine	20 (3.9)	21 (1.5)
Modafinil	s	s

^a“s” indicates that the data were suppressed due to small numbers.

Table 3.

Health and Social Service Use Outcomes for Children with Attention-Deficit Hyperactivity Disorder.^a

	Crude Rates		Adjusted Rate Ratio (95% CI)	P Value
	Intervention (n = 485)	Control (n = 1884)
Health services (per person)
Hospital admissions	0.02	0.02	1.29 (0.68 to 2.46)	0.43
Visits to emergency department
All	0.25	0.24	1.03 (0.75 to 1.41)	0.87
Injury related	0.09	0.10	1.00 (0.68 to 1.46)	1.00
Medication use (1+ medication)	0.64	0.50	1.21 (1.08 to 1.36)	<0.01
Medication adherence (MPR ≥ 0.8)	0.36	0.23	1.42 (1.03 to 1.96)	<0.05
Social services
Children with child welfare contact	0.03	0.02	1.34 (0.54 to 3.35)	0.53
Children in age-appropriate grade	0.89	0.88	1.33 (1.09 to 1.63)	<0.01

^aRate ratio is adjusted for variables in Supplemental Table S1. Medication use is reported as the proportion of participants who were dispensed at least one medication. Medication adherence is reported as the proportion of participants who had an MPR of at least 0.8. Significant values (P < 0.05) are in bold text. CI = confidence interval; MPR = medication possession ratio

Equity analyses were conducted for the outcomes found to be significantly altered by the intervention. Concentration curves were constructed for medication adherence (Figure 1), medication use (data not shown), and age-appropriate grade (Figure 2). In Figure 1, the control group curve falls below the line of equity, and the concentration index is significantly greater than 0 (Table 4), demonstrating that in this group, rates of having a medication possession ratio of at least 0.8 were higher in children from high-income families. The concentration index for the intervention curve is not different from zero, indicating that the intervention was associated with reduced inequity in the distribution of medication adherence across the income gradient. The same pattern is evident for children with ADHD being in their age-appropriate grade (Figure 2). In the control group, the outcome was more concentrated in the wealthier income deciles, but in the intervention group, the curve was not different from the line of equity (Table 4), indicating that the intervention was associated with an equal distribution across the income gradient of children being in their age-appropriate grade.

Figure 1.

Concentration curves for medication adherence in children with attention-deficit hyperactivity disorder.

Figure 2.

Concentration curves for age-appropriate grade in children with attention-deficit hyperactivity disorder.

Table 4.

Postintervention Distribution of Health and Social Service Outcomes across the Socioeconomic Gradient.^a

Outcome	Concentration Index (95% CI)
	Intervention (n = 485)	Control (n = 1884)	Absolute Difference
Medication use	–0.045 (–0.096 to 0.007)	0.025 (–0.003 to 0.054)	0.070 (0.013 to 0.127)
Medication adherence	–0.052 (–0.107 to 0.003)	0.060 (0.031 to 0.090)	0.112 (0.052 to 0.173)
Age-appropriate grade	–0.023 (–0.082 to 0.036)	0.062 (0.032 to 0.091)	0.084 (0.017 to 0.152)

^aSignificant values (P < 0.05) are in bold text. CI = confidence interval.

The absolute difference in concentration indices between the intervention and control groups was significant for medication use, medication adherence, and being in the age- appropriate grade (Table 4). This measure demonstrates that the intervention was associated with a more equitable distribution of these outcomes across the socioeconomic gradient.

Discussion

This study demonstrates that the ADHD Service, a multi-modal intervention available to Manitoban children with ADHD, was associated with positive health and education outcomes. Specifically, participation in the ADHD Service was associated with increased prescribing and filling of prescriptions, and children in the intervention group were more likely to be in the school grade appropriate to their age. Moreover, participation in the intervention was associated with a reduced tendency for only children from higher income families to achieve these outcomes, thereby improving equity in these outcomes across the socioeconomic gradient.

Pharmacologic treatment is a key component of evidence-based care for ADHD patients. Many individuals with ADHD require long-term treatment because of the chronic nature of ADHD symptoms. Despite the demonstrated effectiveness of current ADHD medications, treatment discontinuation is fairly common: 50% of patients fail to adhere to treatment guidelines⁴² or discontinue treatment within 2 to 3 years of starting pharmacologic therapy.^43,44 Such behaviours may lead to suboptimal control of ADHD symptoms, increasing the likelihood of serious deleterious effects that can accrue over time.⁴⁵ Our study demonstrated a significant increase in the rates of medication use and adherence in children and youth who received the intervention. This finding suggests that the program played a role in encouraging positive behaviours in young people (and/or their parents) that may to lead to better management of ADHD symptoms and long-term outcomes. However, it must also be acknowledged that whether long-term medication adherence translates to better long-term health outcomes is really not known. The Multimodal Treatment Study of Children with ADHD has been one of the only randomized controlled trials to evaluate pharmacotherapy and behavioural therapy separately and in combination.⁴⁶ Although initial findings suggested that medication alone was the superior treatment option with respect to ADHD symptom improvement, these improvements gradually abated over the next 2 years.^47–49 Fifteen years of follow-up data revealed that adult-rated symptoms and functional outcomes (academic achievement, social skills, and parenting practices) in the combination treatment arm were optimal.^21,50

Impairments in academic and social functioning are defining features of ADHD.⁵¹ Interventions to remediate these issues may help to address the well-documented sequelae of grade retention, academic underachievement, and school dropout,^52–54 and there is research showing that multimodal treatment programs can have a positive effect on educational outcomes.⁵⁵ Our finding that our intervention group was more likely than the control group to be in their age-appropriate school grade speaks to the positive impact of multimodal treatment programs such as the ADHD Service. In fact, in-school functioning has been shown to be more predictive of long-term executive functioning than measures of ADHD symptoms.⁵⁵ Keeping children and youth in the same grade as their peers is a strong indicator of treatment effectiveness in this population, as it demonstrates that they are able to fulfill academic requirements, and may support their success in achieving other important milestones.

Equity analyses in our study demonstrated that positive health and education outcomes were distributed evenly across the socioeconomic gradient among children who received treatment through the ADHD Service. Children living in lower income households face numerous disadvantages, including being at higher risk for ADHD.²⁴ A recent study from the United Kingdom demonstrated that ADHD was associated with poverty, housing tenure, income, and maternal mental health, while there was no evidence to suggest that childhood ADHD was a causal factor of socioeconomic disadvantage.⁵⁶ Parental income can be a factor in obtaining an ADHD diagnosis and treatment in some health care systems where specialist services may be a financial burden.⁵⁷ The results from the present study found that access to the ADHD Service was not significantly lower for children from backgrounds of low socioeconomic status, and thus, this intervention may contribute to closing an important income gap. However, with mounting evidence for an association between low income and ADHD prevalence,^24,58 more should be done to remove barriers for disadvantaged families in obtaining a diagnosis and accessing ADHD management strategies. An emerging body of literature investigates how these barriers for marginalized and hard-to-reach families can be overcome.^59,60

The association of ADHD with impulsivity, low executive and social functioning, and subtle developmental delays results in elevated risk of physical injury.^61–64 Maxson et al.⁶⁵ showed that children admitted to the hospital for injuries were more than 3 times more likely to screen positive for ADHD indicators than were those admitted for appendicitis. However, studies examining the association between ADHD and hospitalization typically rely on medical records and are therefore limited to serious injuries that require medical attention. Moreover, the rate of such injuries tends to be very rare, and detection requires a large sample size and a substantial follow-up period. Two studies of pharmacological intervention for ADHD^61,62 failed to demonstrate a significant decrease in injury rates, but because of the relatively low rate of injury, the authors concluded they were underpowered. Although not underpowered, our study likewise did not demonstrate a reduction in the rate of hospital episodes and emergency department visits for injury.

The social outcomes in our study included contacts with child welfare services. Many studies have demonstrated an association between ADHD in children and elevated risk of maltreatment of those individuals,^66–68 but our results show the intervention had no impact on the rate of contacts with child welfare, likely because this is a similarly rare outcome.

Strengths and Limitations

The strengths of this study include its population-based design and our ability to control for a wide variety of potential confounders. Propensity score matching helped to ensure that individuals with ADHD and controls had many similar characteristics, strengthening the import and generalizability of the results. Outcomes were designed to address long-term sequelae of ADHD in children and youth, providing new information to the field of ADHD intervention research. Equity analyses have also not been previously conducted in a young ADHD population. One of the study’s important limitations is the possibility of residual confounding. We could not adjust for factors such as symptom severity or family functioning, as this information was not available. There is a possibility of selection bias, as there may have been additional factors unaccounted for in the propensity score model that could have led individuals to participate in the ADHD Service, and the requirement that the intervention group have 3+ visits to the ADHD Service may also have unintentionally selected for a group of recipients with higher participation. Finally, because the treatment was tailored to each individual child’s needs, it was difficult to define the exact components of the intervention that contributed to the improvements we observed. Thus, there is the possibility that the program might have achieved the same results even if it had supported medication adherence only.

Conclusions

In conclusion, participation in the ADHD Service, a multimodal intervention for children with ADHD, was associated with increased medication use and adherence and improved academic achievement. It was also related to lower inequity in these outcomes across the socioeconomic gradient. With mounting evidence for an association between socioeconomic disadvantage and ADHD, these findings are encouraging, suggesting that interventions that combine multiple approaches can provide equitable health and education outcomes for children with ADHD.