Predictors of Stimulant Medication Continuity in Children with Attention-Deficit/Hyperactivity Disorder

Abstract

Objective:

To examine the simultaneous impact of patient and parent factors, medication-related and health care system factors on ADHD medication continuity.

Method:

Stimulant-naïve children (N=144, M_age=8 years, 71% male) with ADHD completed a methylphenidate (MPH) trial and were followed for one year after trial completion and return to community care. Multivariable analysis investigated predictors of 1) having at least 1 filled ADHD prescription after return to community care versus none, and 2) having more days covered with medicine after return to community care. Predictors included race, age, sex, income, baseline ADHD symptom severity, MPH trial experience, child and parent mental health conditions, and parent beliefs about ADHD, ADHD medications, and therapeutic alliance.

Results:

One hundred twenty-one (84%) children had at least 1 filled ADHD medication prescription (mean=178 days covered by medication) in the year after return to community care. Multivariable models found that a weaker perceived clinician-family working alliance predicted not filling any ADHD prescriptions. Among those who filled ≥1 prescription, factors linked to fewer days of ADHD medication coverage included child sociodemographic factors (non-White race, older age, being female, lower income), lower parent beliefs that the child’s ADHD affects their lives, and higher parent beliefs that medication is harmful, while child oppositional defiant disorder and parental ADHD predicted having more days of medication coverage.

Conclusion:

Child demographic factors, parent beliefs, and medication-related factors are associated with continuation of ADHD medication. These findings may facilitate development of effective strategies to improve ADHD medication continuity for children from diverse groups.

INTRODUCTION

Attention-deficit/hyperactivity disorder (ADHD), the most common neurodevelopmental disorder in children, is linked to significant long-term impairments that commonly continue into adolescence and adulthood.¹ Evidence-based treatments for ADHD include medications and behavioral therapies.² There is strong evidence that medications decrease core ADHD symptoms³ and negative long-term outcomes, including reducing academic failure,⁴ criminality,⁵ and car accidents.⁶ Therefore, consistent medication treatment is crucial to improving outcomes for individuals with ADHD. Despite these benefits, pediatric ADHD medication continuity remains suboptimal.⁷

Research investigating pediatric ADHD medication continuity supports the Pediatric Self-Management Model, a comprehensive conceptual model that articulates the individual, family, community, and health care system factors influencing adherence and treatment continuity.⁸ Prior studies have elucidated the roles of health care and medication-related factors in ADHD medication continuity, and found that limited insurance coverage,^9,10 lack of mental health providers in the community,¹¹ adverse effects of medications,¹² and medication ineffectiveness¹² were linked to decreased ADHD medication continuity. However, the roles of patient and parent (e.g., condition-related, sociodemographic) factors are less clear. For example, the relationship between comorbid conditions and ADHD medication continuity are mixed. One study reported that the presence of child comorbid Oppositional Defiant Disorder (ODD), depression, and anxiety predicted increased medication continuity,¹³ but two studies found the opposite.^14,15 In addition, although prior studies found that parent mental health diagnoses (e.g., ADHD, anxiety) played an important role in the decision to start ADHD treatment in children,¹⁶ little is known about the link between parent mental health diagnoses and ADHD medication continuity. There is also a paucity of studies investigating whether parent beliefs about ADHD medications and their relationship with their child’s prescribing provider impact continuity of treatment.¹⁷ Furthermore, although a significant relationship between socioeconomic (SES)-related factors (e.g., income) and ADHD medication continuity has been reported,¹⁸ it has recently been recognized that traditional markers of low SES (such as poverty threshold income levels) may not adequately capture the breadth of family strain compared to other measures of hardship.¹⁹ Hence, hardship measures may be more helpful than traditional SES measures in disentangling the relationship between income and race as they relate to the outcome of interest. For example, Beck et al¹⁹ reported that race-related disparities in asthma re-admissions were not explained by traditional SES variables; however, social/financial hardships did explain >40% of racial disparities.

To date, no study has considered the simultaneous impact of health care system, medication-related, and patient- and parent-related factors on ADHD medication continuity. Thus, in a sample of children who were diagnosed with ADHD and received initial ADHD medication in a one month randomized controlled trial, we examined the relationship between this comprehensive set of factors and ADHD medication continuity assessed one year after leaving the trial and returning to their community-based pediatricians for ADHD management. Our outcome measures included 1) having at least one filled ADHD prescription after return to community care versus none, and 2) having more ADHD prescriptions filled after return to community care. We hypothesized that factors linked to medication discontinuation after return to the community for ADHD care would include health care system factors (e.g., lack of private health care insurance, perceived weakness of working alliance between patient and providers),^9,17 medication-related factors (e.g., medication ineffectiveness, adverse events),¹² patient factors (e.g., non-White race, Hispanic ethnicity, female gender, child comorbid conditions, lower baseline severity of ADHD symptoms),^20,21 and parent factors (e.g., family hardships, low income, parent mental health diagnoses [ADHD, depression], negative beliefs about ADHD medication).²²

METHODS

Participants

All participants provided written and informed consent/assent according to the Institutional Review Board-approved protocol. The study included 144 stimulant-naïve children (aged 7–11; mean=8 years) who had participated in a 4-week, double-blind, crossover trial of long-acting osmotic release oral system (OROS) methylphenidate (Concerta®) during which we randomly assigned each subject to receive one of six dosing schedules comprised of three active dosage weeks (18mg, 27mg, 36mg for children ≤25kg; 18mg, 36mg or 54mg for children >25kg; sample M_{max dose} =1.57 mg/kg/day) and one week of placebo (see Froehlich et al²³ for additional trial description). We recruited children for the methylphenidate trial through multiple community and clinical sources, including schools and local pediatric offices from September 2006 to June 2013.²³ We determined ADHD diagnosis using well-established methodology (e.g., similar to that of the Multimodal Treatment Study of Children with ADHD). All participants met ADHD DSM-IV criteria for onset age, pervasiveness, and impairment as reported on the Diagnostic Interview Schedule for Children-Parent Report (DISC-P), and were confirmed to be “moderately ill” with ADHD as defined by a clinician (pediatrician or psychologist) rating of ≥4 on the Clinical Global Impression of functional severity (CGI-S). The DISC-P was administered by a trained research personnel (e.g., clinical research coordinator, psychology graduate student) for approximately half of the participants and by a developmental-behavioral pediatrician for the other half. To determine ADHD presentation/subtype, we used a methodology similar to that employed by the Multimodal Treatment Study of ADHD. Specifically, children met criteria for an ADHD symptom domain (i.e., inattention and/or hyperactivity/impulsivity) if the parent (on the DISC-P) and the teacher (on the Vanderbilt ADHD Diagnostic Teacher Rating Scale) reported at least 6 non-overlapping DSM-IV symptoms in a symptom domain. In addition, as evidence of symptoms across settings, both parent and teacher were required to report a minimum of 4 symptoms in that domain. We oversampled children with ADHD predominantly inattentive presentation (ADHD-I, n=104) to allow adequate power for examination of medication response in children with ADHD-I for the parent trial.²³

To participate in the initial ADHD medication clinical trial, we required children to be medication-naïve at enrollment, with no previous psychoactive medication treatment. We excluded children with a full-scale IQ below 80 (on the Wechsler Abbreviated Scale of Intelligence [WASI]) and children with possible learning disorders (estimated by having a standardized achievement scores below 80 on the Wechsler Individual Achievement Test, 2^nd Edition [WIAT-II] Reading or Numerical Operations subtests) from participation in the study due to concerns that children with learning or intellectual disorders may have a differential response to methylphenidate. In addition, we excluded children if their medical history suggested organic brain injury (e.g., traumatic brain injury). We also evaluated potential participants for psychiatric comorbidities using the DISC-P, and we excluded those with mania/hypomania. We allowed those with comorbid oppositional defiant, conduct, depression, and anxiety disorders to participate unless determined to be the primary cause of ADHD symptomatology or necessitating different treatment.

Measures

Primary Outcomes

Medication Continuity.

After completing the 4-week methylphenidate clinical trial, children returned to their primary care physicians for continued ADHD management and ADHD medication prescription if desired. 141 (98%) participants signed an authorization form to obtain pharmacy dispensing records to determine the number of ADHD medication prescriptions filled in the year after returning to ADHD community care. Medication dispensing records provides information on prescription written (i.e., medication, dosage, amount dispensed) and they are an objective, unobtrusive, reliable measure that is a well-accepted proxy for medication consumption. The number of days covered with medication is the sum of the number of pills dispensed in the 12 months after leaving the trial. We investigated two outcomes: 1) having at least 1 filled ADHD prescription versus none after return to community care, and 2) having more ADHD prescriptions filled after return to community care.

Predictors

Socio-Demographic Characteristics.

Before beginning the methylphenidate clinical trial, parents completed a baseline demographic questionnaire to report child race/ethnicity, health insurance status, household income, receipt of public assistance (e.g., welfare, food stamps, Women, Infants, Children [WIC] program), and whether they had been diagnosed with any mental health or learning disorders (e.g., ADHD, anxiety, depression, learning disorders). In addition, to further examine social/financial hardships, we queried parents about whether “[they] did not purchase medications due to lack of money”, “a creditor has called”, and/or “a gas/electric company has turned off services.” We considered an answer of “yes” to any of these three questions an indicator of hardship.

Baseline ADHD Symptoms and Impairments.

Parents and teachers completed the Vanderbilt ADHD Diagnostic Parent Rating Scale (VAPRS, r=0.79), the Vanderbilt ADHD Diagnostic Teacher Rating Scale (VATRS) at baseline. Both the VAPRS and the VATRS ask parents/teachers to rate the child on each of the 18 DSM-IV ADHD symptoms (0=none, 1=occasionally, 2=often, 3=very often). An ADHD total symptom score (TSS; range=0–54) is derived by adding the 18 symptom scores. The scales also include 8 scale-specific items measuring the degree of impairment across several domains (e.g., peer relationships, relationship with parents, participation in activities, overall school performance, following directions, disrupting class, and assignment completion). These items ask whether the child performs “Excellent-1”, “Above Average-2”, “Average-3”, “Somewhat of a Problem-4”, or “Problematic-5” in each domain. The VAPRS/VATRS impairment score is derived by counting the total number of questions scored 4 or 5. We calculated the baseline ADHD TSS and baseline impairment score.

Child Mental Health Comorbid Conditions.

The VAPRS/VATRS also includes items about common ADHD comorbidities (i.e., ODD, conduct disorder, Anxiety, and Depression) which are rated on the same scale as the ADHD symptom items. The VAPRS includes ODD (must score a 2 or 3 on 4 out of 8 items), conduct disorder (must score a 2 or 3 on 3 out of 14 items), and Anxiety/Depression (must score a 2 or 3 on 3 out of 7 items) comorbidity screening scales; whereas the VATRS includes ODD/conduct disorder (must score a 2 or 3 on 3 out of 10 items) and Anxiety/Depression (must score a 2 or 3 on 3 out of 7 items).

Response to Treatment.

Parents and teachers completed the VAPRS/VATRS and the 13-item Pittsburgh Side Effects Rating Scale at baseline and at the end of each trial week. We also collected side effects measures at baseline because some children experience these putative adverse event symptoms (i.e., headache, stomachache, etc.) before starting medicine. To determine methylphenidate responder status and optimal week, we graphed ratings of behavior and side effects by parents and teachers from the 4 titration trial weeks and two study physicians (TF, WB) blindly and independently reviewed the graphs to judge if there was an optimal trial week (e.g., optimal balance between improvement in ADHD symptoms and the experience of adverse effects), and if so, which week(s) were optimal. We calculated the reduction in ADHD TSS from baseline to optimal week of the methylphenidate trial. We also calculated the number of moderate or severe side effects experienced before starting medicine and the average number experienced during the methylphenidate weeks of the trial.

Parent Beliefs about ADHD.

At the end of the medication trial, but prior to un-blinding the dosing schedule, parents completed the Brief Illness Perceptions Questionnaire (BIPQ, Cronbach’s α=0.80–0.85, test-retest reliability r=0.42–0.75). The BIPQ is a 10-item scale that uses a 0 to 10 response scale for each item and is designed to assess the parent’s representation of ADHD (e.g., expected time course, consequences, etc.). We used all the items for analysis. Higher scores indicate stronger beliefs (e.g. that ADHD affects the child’s life; that the parent is concerned about the child’s ADHD).

Parent Attitudes and Beliefs about Medicine.

At the end of the medication trial, but prior to un-blinding the dosing schedule, parents completed the Beliefs about Medicines Questionnaire (BMQ, Cronbach’s α>0.7, test-retest reliability r>0.6). The BMQ is a 25-item scale that has four subscales to assess beliefs about medication. The four subscales include 1) Necessity (parent beliefs about the necessity of prescribed medication), 2) Concerns (parent concerns about potential adverse consequences of using medication), 3) Overuse (parent beliefs about overuse of medicine by doctors), and 4) Harm (parent belief about the intrinsic harmfulness of medicine). For all items, respondents indicate their degree of agreement with each individual statement on a 5-point Likert-type scale (e.g., 1=strongly disagree to 5=strongly agree). Scores obtained for the individual items within each subscale are summed to give a subscale score. Mean scores are also calculated for each subscale (range of 1 to 5).

After completion of the medication trial, the study physician and parents met for a debriefing session during which the dosing schedule during the trial was un-blinded and reviewed. After this debriefing, parents completed the Satisfaction with Information about Medicine Scale (SIMS, Cronbach’s α=0.81–0.91, test-retest reliability r>0.6, predictive validity r=0.31). SIMS is a 14-item scale that provides a profile of a respondent’s satisfaction with information they have received about medicine. Respondents indicate their satisfaction with a particular aspect of medication information received by answering “satisfied-1” or “dissatisfied-0.” Total score is calculated by summing the scores for each item, with higher scores indicating greater satisfaction (range of 0 to 14).

Parent Perception of Working Alliance and Decisional Conflict.

After the debriefing, parents also completed the Working Alliance Inventory (WAI, Cronbach’s α=0.98) and the Decisional Conflict Scale (DCS, Cronbach’s α=0.78 to 0.92, test-retest reliability r=0.81). The WAI is a 12-item measure of a parent’s perceived agreement with their healthcare provider on the goals/tasks of treatment and the extent to which they have a strong personal bond with their provider. Respondents rate their responses on a scale of 1 (strongly disagree) to 5 (strongly agree). Item scores are summed (range 12–60), with higher scores indicating a better parent-provider alliance. The DCS is a 16-item scale eliciting the parent’s preference for treatment options and parent-perceived sources of uncertainty about their preferred option. Respondents rate their responses on a scale of 0 (strongly agree) to 4 (strongly disagree). Items scores are summed, divided by 16, and then multiplied by 25; with higher scores indicating more conflict (range from 0 to100).

Statistical Analysis

We summarized patient and parent sociodemographic (i.e., age, sex, race, health insurance status, household income, public assistance, social/financial hardships), parent mental health diagnoses, patients’ clinical characteristics, and parents’ post-trial beliefs using means and standard deviations for continuous variables that were normally distributed, medians and inter-quantile range for non-normally distributed variables, and proportions for categorical variables. Covariates missing pattern was described and correlation structure was analyzed to validate missing at random assumption. We assumed missing values using multiple imputations. We used Fully Conditional Specification regression and discriminant function methods for continuous and discrete missing variables, respectively. After twenty burn-in iterations, we generated a random sample (n=15) of the missing values. Sensitivity analyses of using complete data were conducted to confirm imputation effects.

At the bivariable level, we conducted logistic regression analyses to evaluate the association between all independent variables and dichotomized medication continuity (no medication vs. had at least one filled ADHD prescription). Within participants who had at least one prescription filled, we evaluated the association between number of days covered with ADHD medications (according to fill records) and the independent variables by linear regression analyses. In order to achieve a solid predicted model, we conducted least absolute shrinkage and selection operator (lasso) regression analysis with 10-fold cross-validation to confirm the most important subset of predictors for both dichotomized and continuous outcomes. The lasso regression model is used for data in which the number of predictors may be large relative to the sample size and when predictors may be correlated. Optimal models were selected based on Bayesian Information Criteria. We repeated lasso regression 100 times using bootstrapping technique to account for variation of estimations. For each predictor, we counted the number of models that included the predictor. We selected predictors that were included in more than 70% of models for inclusion in the final model.

RESULTS

Sample Characteristics

Participant sociodemographic and clinical variables can be found in Table 1. Children were primarily boys (n=102). Approximately 85% of the sample was of Caucasian race, 11% was African American, and 4% was of other racial origin. At baseline, children had mean total ADHD symptom scores of 32.8 by parent report and 30.7 by teacher report. On average, medication was associated with a mean reduction of total ADHD symptoms from baseline to optimal week of the titration trial of 16.9 (SD 10.6) as reported by parents and 15.4 (SD 11.0) by teachers. Eighty-four percent (n=121) of children had at least one filled ADHD prescription. Among these children, there was a mean of 178 days (SD=113) covered by medication in the year after leaving the MPH trial and returning to community care (See Figure 1). Descriptive statistics for additional predictor variables are shown in Table 2.

Table 1.

Sociodemographic and Clinical Predictors

	N (%) or Mean (SD) or Median (range) Total N = 144a
Child and parent sociodemographic characteristics
Child age, years	8 (7 – 11)
Child sex, male	102 (70.8%)
Child race/ethnicity, white or non-Hispanic	122 (84.7%)
Child insurance, Medicaid	19 (13.2%)
Household Income	[n = 136]
≤$20,000	8 (5.9%)
$20,001 - $40,000	22 (16.2%)
$40,001 - $60,000	23 (16.9%)
$60,001 - $80,000	25 (18.4%)
>$80,000	58 (42.6%)
Family receipt of welfare	4 (2.8%)
Family receipt of food stamps	11 (7.6%)
Family receipt of WIC	5 (3.5%)
Social/Financial Hardships [Yes]
Did not purchase medications	2 (1.4%)
Creditor has called	14 (9.7%)
Gas/electric company turned off service	7 (4.9%)
Child Clinical Characteristics
ADHD Predominantly inattentive type	103 (72.0%)
Co-occurring disorders
Oppositional defiant disorder	34 (23.8%)
Conduct disorder	4 (2.8%)
Anxiety disordersb	45 (31.3%)
Mood disordersc	1 (0.7%)
VAPRS total symptom score (baseline) [range 0–54, higher = more symptoms]	32.8 (9.8) [n=142]
VATRS total symptom score (baseline) [range 0–54, higher = more symptoms]	30.7 (10.8) [n=143]
VAPRS impairment score (baseline) [range 0–8, higher = more impairment]	2.98 (0.80) [n=143]
VATRS impairment score (baseline) [range 0–8, higher = more impairment]	3.75 (0.00 – 5.00) [n=143]
Reduction in VAPRS total symptom score from baseline to optimal week of the medication trial [lower = greater reduction]	−16.9 (10.6) [n=142]
Reduction in VATRS symptom score from baseline to optimal week of the medication trial [higher = greater reduction]	−15.4 (11.0) [n=133]
No. moderate/severe side effects experienced at baseline [range 0–13, higher = more side effects]	0.00 (0.00 – 9.00) [n=143]
No. moderate or severe side effects experienced during the MPH weeks of the trial [range 0–13, higher = more side effects]	0.75 (0.00 – 4.00) [n=143]
MPH Responder	122 (85.9%)

^aThose variables with missing data have the correct N in brackets.

^bIncludes social phobia, separation anxiety, specific phobia, panic agoraphobia, gad, ocd, and ptsd.

^cIncludes depression and mania (no mania disorders reported).

Figure 1.

Number of Days Covered with ADHD Medications after Leaving the Trial

Table 2.

Parent beliefs about ADHD, medicine, and their working alliance with the child’s doctor

	Mean (SD) or Median (range) Total N = 144a
BIPQ: (range 0–10 on each item, higher = stronger belief)	[n=141]
How much do you think your child’s ADHD/ADD affects his/her life?	6.74 (1.80)
How much does your child’s ADHD/ADD affect your life?	6.00 (1.00 – 10.00)
How long do you think your child’s ADHD/ADD will continue?	7.00 (2.00 – 10.00)
How much control do you feel your child has over his/her ADHD/ADD?	3.00 (0.00 – 9.00)
How much control do you feel you have over your child’s ADHD/ADD?	3.00 (0.00 – 9.00)
How much do you think your child’s treatment can help his/her ADHD/ADD?	8.00 (0.00 – 10.00)
How much does your child experience symptoms from his/her ADHD/ADD?	6.08 (2.05)
How concerned are you about your child’s ADHD/ADD?	8.00 (3.00 – 10.00)
How well do you feel you understand your child’s ADHD/ADD?	8.00 (0.00 – 10.00)
How much does your child’s ADHD/ADD affect you emotionally?	7.00 (0.00 – 10.00)
BMQ	[n=141]
Necessity scale, [range 1–5, higher = stronger belief]	2.39 (0.76)
Concerns scale, [range 1–5, higher = stronger belief]	2.69 (0.67)
Overuse, [range 1–5, higher = stronger belief]	2.71 (0.70)
Harm, [range 1–5, higher = stronger belief]	1.95 (0.60)
Satisfaction with Info about Medicine Scale	[n=139]
Total score [range 0 to 14, higher = more satisfied]	14.0 (0.0 – 14.0)
Working Alliance Inventory	[n=143]
Total Score [range 12 to 60, higher = greater alliance]	52.0 (0.0 – 60.0)
Decisional Conflict Score	[n=142]
Total Score [range 0 to 100, higher = more conflict]	15.6 (0.0 – 50.0)

^aThose variables with missing data have the correct N in brackets.

BIPQ = Brief Illness Perceptions Questionnaire, BMQ = Beliefs about Medicines Questionnaire

Initiation of ADHD medication in the clinical setting

Nine variables (i.e., non-White race, receiving WIC support, endorsing at least one financial hardship, not having comorbid diagnoses, lower baseline VATRS symptoms, lower baseline VATRS impairment score, lower belief that treatment can help ADHD, higher belief that medication is harmful, and weaker perceived clinician working alliance) demonstrated significant (p < .05) bivariate relationships with not having any filled ADHD prescription in the community setting. In the multivariable models, only weaker working alliance with the clinician was identified as a predictor of not filling any ADHD prescriptions in the year after leaving the clinical trial and embarking upon ADHD management in the community (β = 0.0319).

ADHD medication continuity after returning to community care

The multivariable analysis showed that among those who filled at least one ADHD prescription, factors linked to suboptimal medication continuity (i.e. fewer days covered with medication) included patient- and parent-related factors. On average, non-White patients, females, and low-income families had 67, 23, and 18 fewer days covered with medications compared to White patients, males, and high-income families, respectively. Moreover, as age increased by 1 year, patients showed less improvement in ADHD symptoms during the RCT, parents reported lower beliefs that ADHD affects the child’s life, and parents reported higher belief that medication is harmful; medication coverage decreased on average by 13, 1, 10, and 36 days, respectively. On the other hand, we found that the child having comorbid ODD and the parent having a prior ADHD diagnosis each were associated with more days covered with medication. (See Table 3).

Table 3.

Predictors of ADHD Medication Continuity

	Multivariable Lasso (estimate)1
Child/adolescent Factors
Non-white race	−66.79
Older age	−13.44
Female	−23.15
No comorbid ODD	−26.76
Less improvement in ADHD symptoms during RCT	−0.99
Family/Parent Factors
Lower income (20–40 k)	−18.29
Parent did not have a diagnosis of ADHD	−26.53
Lower parent belief that the child’s ADHD affects their lives	−9.70
Higher parent belief that medication is harmful	−36.34

¹A negative Lasso estimate denotes fewer days of covered with ADHD medication while a positive estimate denotes more days covered.

DISCUSSION

This secondary data analysis of 144 stimulant-naïve, school-age children contributes to and expands our understanding of predictors of ADHD medication initiation and continuity¹⁷ by evaluating the simultaneous impact of health care system-, medication-, patient-, and parent-related predictors of medication continuity in the first year after children were diagnosed with ADHD. We found that a positive partnership and working alliance between the family and the medical provider who first worked with them on initial ADHD medication treatment (in this case in the context of a clinical trial) predicted having at least one medication prescription filled in the first year of ADHD care in the community setting. Among those who filled one or more prescriptions, factors linked to fewer days of medication coverage included child sociodemographic factors (non-White race, older age, being female, lower income), less improvement in child ADHD symptoms during the RCT, lower caregiver beliefs that the child’s ADHD affects their lives, and higher caregiver beliefs that medication is harmful, while child oppositional defiant disorder and parental ADHD were linked to more days of medication coverage.

Our results corroborate prior studies indicating that being from a racial/ethnic minoritized group,¹⁰ lower income, ^18,28 older age (e.g., adolescent versus school-age),²² and being female²⁰ are associated with poor ADHD treatment continuity. Some hypothesized reasons for racial/ethnic disparities in ADHD treatment continuity include uncertainties about medication efficacy and side effects, reduced access to mental health services, and distrust of the health care system.²⁴ In particular, it is reported that mistrust of the health care system by minoritized groups may be a result of their longstanding experience of mistreatment, discrimination and/or systemic racism within the system.²⁵ Further, similar to our study, low income has been linked to a decreased likelihood of receiving ADHD medication treatment in several studies, including the landmark Multimodal Treatment of Study of Children with ADHD.^18,28 The decreased medication use among low-income families may be explained by health services barriers such as limited transportation to obtain services and lack of available providers and pharmacies in the community.²⁸ Surprisingly, however, health insurance type was not related to ADHD medication continuity in our study despite the fact that limited insurance coverage has previously been reported as a predictor of ADHD medication discontinuity due to substantial out-of-pocket payments for families.¹⁰

The gender disparity in ADHD treatment observed in our study may be explained by 1) externalizing behaviors being a strong predictor of ADHD medication use, and 2) girls typically presenting with fewer externalizing behaviors (i.e., hyperactivity/impulsivity and symptoms of conduct disorder) compared to boys.²⁶ In addition, reduced ADHD medication continuity among older children compared to school-age children is likely the consequence of the adolescent’s developmental drive for autonomy (since adolescents are more likely to refuse medication)^14,15,31 and symptom remission¹³ (since adolescents often experience a lessening of hyperactive-impulsive symptoms as part of the natural history of ADHD). Adolescents with ADHD have also expressed concern that using medication made them slow and “boring,” leading to medication cessation due to a desire for self-expression.²⁷

In our study, lower parent belief that the child’s ADHD affects their lives and higher parent belief that medication is harmful were associated with fewer days covered with ADHD medication after treatment was begun in the community setting. Parent beliefs about ADHD and their attitudes toward treatment have been associated with initiation and continuity of ADHD medication in previous studies. For example, parents who view a child’s difficulties as a medical disorder that requires a biological intervention are more likely to accept and continue medication.⁹ Similarly, parents who perceive ADHD medication as safe, effective and socially acceptable are more likely to continue their children’s treatment.²⁰

In our study, having comorbid ODD predicted ADHD medication continuity. The opposite relationship has been reported in a few prior studies, presumably because of child refusal to take medication.^14,15 In contrast, and similar to our results, Atzori et al found that the presence of child comorbidities predicted ADHD medication continuity.¹³ Comorbid conditions (e.g., ODD and conduct disorder) have been associated with worsening ADHD-related impairments and overall prognosis,¹ possibly leading to parental persistence in seeking treatment and therefore obtaining more prolonged courses of treatment. In addition, we found a link between parent ADHD and child ADHD medication continuity, which is supported in the literature.¹⁶

Not surprisingly, children who had less improvement in ADHD symptoms during our initial medication trial were more likely to fewer days covered with medicine in the community setting.¹² However, even though adverse effects have been the most commonly cited reason for ADHD medication non-adherence in prior studies,^12,20 no association was found in the current study between adverse events and medication continuity. The lack of association may be related to our measurement of medication adherence after children completed a rigorous titration trial that ended with a recommendation for a medication regimen that optimized ADHD symptom reduction and minimized side effects. This finding highlights the importance of a careful titration and close follow up after starting treatment as recommended in ADHD guidelines to improve medication adherence by minimizing adverse events.⁷

This study’s findings must be interpreted in light of its limitations. First, medicine supply served as a proxy for medicine use. Medication dispensing records do not ensure that the medication was taken according to the prescriber’s instructions, although they do provide an objective, unobtrusive, reliable measure that is a well-accepted proxy for medication consumption. Second, there is a potential for selection bias because all participants attended a rigorous medication trial that may have been difficult for low-resource families to complete. Thus, the study findings are probably most applicable to middle- to high-income families, although it should be noted that recruitment for this trial was through community-based practices and schools.²³ Third, our study did not collect data on reasons why children did not take or stopped taking ADHD medications in the community setting. For example, lack of medication treatment in the community setting may be due to receiving behavioral therapy, and therefore may signify warranted variation since behavior therapy is a recommended, evidence-based treatment for ADHD.² Fourth, although we did not find a significant association between family hardships and ADHD medication continuity, because this is the first report examining this association, additional research is warranted. Future studies may also use additional validated questions to better characterize social/financial hardships¹⁹ and provide a more fine-grained assessment of the factors that may influence ADHD medication use, such as questions eliciting family report of having difficulty making ends meet, “being unable to pay the full amount of rent, mortgage, or utility bills,” “household’s inability to borrow money during times of need”, and things that people might do when money is short (e.g. pawn/sell possessions, move in with other people, or stay overnight in a charity shelter). Fifth, contrary to past research on predictors of medication continuity in ADHD¹⁷ and other chronic conditions (e.g., asthma, diabetes, depression),²⁹ we did not find an association between clinical working alliance and medication continuity after medication initiation in the community setting. This finding may be explained by our study measuring the perceived working alliance with the trial clinician and not with the subsequent primary care physician working with the family in the community setting. In addition, we do not have information on the type of PCP provider (e.g., pediatrician, nurse practitioner, family physician) and type of primary care setting (e.g., private practice, community health center, University-based practice, school-based clinic). Future studies may evaluate whether ADHD medication continuity and/or working alliance with clinician may vary by the type of PCP provider or the type of primary care setting. Lastly, the use of Concerta® during the clinical trial may have impacted the patient/family experience after return to community care if the child was then prescribed generic OROS methylphenidate by the PCP, as some of these preparations have been flagged by the FDA as not being pharmacologically equivalent to the brand-name preparation.

In conclusion, we identified specific individual, parent, and medication-related factors predicting less stimulant medication continuity in community setting for newly diagnosed children with ADHD. Importantly, our study highlighted specific sociodemographic groups – including children who are non-white, low income, and female – whom clinicians should be aware may be at increased risk for reduced adherence to evidence-based medication treatments. Our findings also underscore that ADHD medication receipt and adherence may be increased if families perceive a strong working alliance with their clinicians and if clinicians take the time to address family concerns about medication adverse effects and outcomes. An important next step for the field may be to develop strategies to improve ADHD medication continuity, when appropriate. Some plausible interventions to increase medication continuity, especially for children from low income and minoritized racial/ethnic groups, include patient/family education, behavioral strategies (e.g., parent training), clinician interventions (e.g., use of decision aids, training and supports to implement ADHD clinical practice guidelines, including early medication titration and early contact), peer support models (e.g., family advocates), and health disparity-reducing interventions (e.g., medical-legal partnership, school-based therapy programs).³⁰