Abstract
Objectives:
The National Survey of Children's Health reported a concerning increase in children 2–5 years being diagnosed with attention-deficit/hyperactivity disorder (ADHD) in 2016. Concerns include both the increase in diagnosing and potential deviations from published guidelines for the treatment of ADHD in preschoolers. The present study aims to describe the epidemiology and factors associated with receiving the diagnosis and treatment types for low-income preschoolers.
Methods:
Using Kentucky Medicaid claims from 2012 to 2017, a retrospective cohort study of children 2–5 years of age (n = 337,631) with a diagnosis of ADHD (n = 11,712) was completed. Trends in demographics, comorbidities, and treatment and provider types are presented. Multinomial logistic regression was used to determine predictors of receipt of the diagnosis and treatment type (a stimulant only, an alpha-2 agonist [A2A] only, both, or neither) based on nonmissing 2017 data (n = 2394).
Results:
The number of children in the cohort diagnosed with ADHD and receiving a stimulant decreased from 2012 to 2017, but the use of A2As increased. Primary care physicians were the most frequent prescribers of both medications. The adjusted odds ratios (AORs) of receipt of an A2A alone, stimulant alone, or both medications over receiving no ADHD medication were associated with specific demographics and comorbid conditions for each medication regimen. Race/ethnicity is associated with receiving the diagnosis of ADHD and treatment with A2A. Comorbid mental health conditions and provider type are associated with treatment type.
Conclusion:
Use of stimulants for preschoolers in Kentucky has decreased and A2A use has increased since 2012. Continued vigilance and long-term follow-up of preschoolers with ADHD are warranted. The appropriateness of the diagnosis and treatment type could not be determined.
Introduction
Attention-deficit/hyperactivity disorder (
The American Academy of Pediatrics (AAP) recommends behavioral health interventions as first-line treatment followed by administration of methylphenidate if an adequate response is not achieved for children 4 and 5 years of age (Wolraich et al. 2011). There are no recommendations for children younger than 4 years. Dextroamphetamine is the only stimulant medication with an indication by the Food and Drug Administration (FDA) for the treatment of ADHD in children younger than 6 years. There is some evidence that stimulant use in preschoolers is declining (Blum et al. 2018; Davis et al. 2019); however, in general, practice guidelines are not being followed (Moran et al. 2019).
In addition to the use of stimulants, alpha-2 agonists (A2As) have been prescribed for ADHD and other psychiatric disorders in children in recent years (Sallee et al. 2013; Hirota et al. 2014; Pringsheim et al. 2015; McCracken et al. 2016). Like stimulants, they have not been approved for use in young children. In a sample from the Developmental Behavior Pediatrics Network (DBPNet), Blum et al. (2018) reported that stimulants were more likely to be used for the treatment of ADHD in older preschoolers, and A2As were used more often for younger preschoolers. Both medications were found to be used in the presence of comorbid conditions (Blum et al. 2018). This study was a cross-sectional study of children (n = 984) seen in 2 developmental/behavioral pediatric practices.
The purpose of the present study is to describe the diagnosis and prescribing patterns for stimulants and A2As in children 2–5 years of age in a state-wide sample of children receiving Kentucky Medicaid from 2012 to 2017, examine stimulant and A2A use patterns over time, and identify factors associated with the receipt of the ADHD diagnosis and treatment type for this age group.
Methods
Data source and sample
The data consisted of Kentucky Medicaid medical and pharmacy claims with a date of service between January 1, 2012, and December 31, 2017. The data were provided by the Kentucky Cabinet for Health and Family Services, Department of Medicaid Services. Kentucky Medicaid claims include billing information on all health services billed through Kentucky Medicaid insurance. The service date, prescription, type of claim, location, provider, and associated International Classification of Diseases, 9th ed. [ICD-9] code and ICD-10 code, as well as patient demographic information such as age, gender, self-identified race, and postal code for all enrollees, are contained in these data. For the purpose of this study, the sample only included data for children 2–5 years of age (N = 340,833) with a diagnosis of ADHD (n = 12,528). Children in the foster care system with a diagnosis of ADHD (n = 816) were omitted, resulting in the final analytic sample across 2012–2017 (n = 11,712). Children in foster care were omitted because of potential underestimation of the receipt of behavioral interventions due to the practice of “bundling” the payments for some services for certain children. Because of the “bundled” payment, no claim would be filed for the services provided. Foster care status was provided by yearly enrollment data files from Kentucky Medicaid.
The descriptive sample was stratified by year and treatment type to assess trends in children treated with only a stimulant prescription during the calendar year, only an A2A prescription, both a stimulant and A2A, and those with no medication within the calendar year. We did not remove children from the study sample due to noncontinuous enrollment. Understanding that this could potentially bias the estimates provided since shorter enrollments provide less time to accumulate a claim for ADHD or a medication, we compared our estimates from regression analysis with and without the exclusion of partial enrollees, finding them nearly identical.
A secondary data set was created for the regression models using only 2017 data, which are the most recent data available, to assess possible predictors of receiving an ADHD diagnosis and, having received it, the possible predictors of receiving an A2A, stimulant, both, or no medication. By using only 2017 for the regression models, we ensured that all children were in the sample for the same number of years and the more recent data are likely to be most reflective of the present practice and policy environments. Children with incomplete records of race/ethnicity, gender, or geography were omitted, decreasing our 2017 sample from 137,997 to 123,957 children.
Definition of behavioral health care, utilization, and diagnostic categories
We defined a child as having received behavioral health care if at least one medical claim within the year indicated one of the following current procedural terminology codes: 90832-34, 90836-40, 90840, 90845-49, 90853, 90875-76, and 90880. These codes include both outpatient and inpatient visits for individual and group therapy.
To identify comorbidities, we used the presence of relevant ICD-9 (before October 1, 2015) or ICD-10 codes for psychiatric disorders in any of the four available diagnostic code columns on either inpatient or outpatient claims (for a full description of diagnoses that were used and ICD codes, refer Supplementary Table S1). ADHD was defined by ICD-10 codes F90.0, F90.1, F90.2, F90.8, and F90.9, and ICD-9 codes 314.00, 314.01, 314.1, 314.2, 314.8, and 314.9. A specific subset of the psychiatric comorbidities presented in the descriptive data table was included as predictors in our treatment outcome model, categorized as follows: Conduct disorder and oppositional defiant disorder Impulse control disorder Language disorder Sleep-wake disorder Anxiety and trauma disorders, which included anxiety disorders, posttraumatic stress disorder (PTSD), acute stress disorder, obsessive-compulsive disorder and related disorders, adjustment disorders, and reactive attachment disorder Other neurodevelopmental disorders, which included intellectual disability, tic disorders, stereotypical movement disorder, and developmental delays.
Diagnoses from Table 1 that were excluded from the regression analysis are autism and pervasive developmental disorders (PDD) not otherwise specified (NOS); mood disorders; other mental disorders; schizophrenia and other psychotic processes; and seizures/convulsions. These diagnoses were excluded because they are not disorders that are typically treated with stimulants and/or A2As without the additional diagnosis of ADHD.
Demographic Characteristics of Children Age 2–5 Years with Attention-Deficit/Hyperactivity Disorder
ADHD, attention-deficit/hyperactivity disorder; IQR, interquartile range; NOS, not otherwise specified; PDD, pervasive developmental disorders; SD, standard deviation.
National Drug Codes and prescribing provider categories
Treatment with A2A and stimulants was identified using National Drug Codes (NDC) on a pharmacy claim within each given calendar year (see Supplementary Table S2 for medications). For the purposes of the present study, other psychotropic medications were not considered with the exception of Strattera.
To create the provider categories, we merged the National Provider Identifier (NPI) from the billing claim and the classification from the data provided by the National Uniform Claim Committee (NUCC), using taxonomy codes found in the National Plan and Provider Enumeration System (NPPES). The final classification included (1) primary care providers (pediatrics and family medicine), (2) psychiatrists/neurologists, (3) nurse practitioners and physician assistants, and (4) other. Across all study years, on average, NPI was missing from 1.5% of prescription claims (range 0%–4%).
Covariates
The child's age, gender, race, geographic area of residence (metro vs. nonmetro), number of and specific comorbidities, whether the child received behavioral health care, and whether he or she had at least one visit with a psychiatrist/neurologist were included in our predictive model. The children diagnosed with ADHD were divided into four subcategories relating to pharmaceutical treatment. The groups were as follows: (1) children with A2A medication claims alone [no stimulants], (2) children with stimulant medication claims alone [no A2A], (3) children with both A2A and stimulant medication claims, and (4) children without any claims for A2A or stimulant medications.
Age, gender, and race were provided by yearly enrollment data files from Kentucky Medicaid. Geography was determined by mapping individual postal codes to the 2013 Rural Urban Continuum Codes (RUCC) as defined by the U.S. Department of Agriculture (USDA). Individuals' geographic variables were categorized as metropolitan ([metro], RUCC code of 1–3), and nonmetropolitan ([nonmetro], RUCC code of 4–9). Comorbidities included as covariates were created, as defined above, and an additional variable was created to account for the number of comorbidities present.
Statistical analysis
The demographic characteristics were summarized for 2012–2017, and differences in our sample groups were assessed, along with annual trends in A2A and stimulant use. Descriptive tables are included to illustrate the yearly trends in prescriber type and behavioral health utilization, and to illustrate the differences related to age, race, and common comorbidities. We examined group differences in proportions and yearly trends by using a chi-square test for proportion and Cochran/Armitage test for trend.
A logistic regression model was used to identify the risk factors for ADHD. The potential risk factors included gender, race/ethnicity, geography, and age. We report odds ratios (ORs) with 95% confidence interval (CI) and 2-sided p-values, assessing important individual and environmental characteristics.
Next, we created a multinomial logistic regression model to assess the predictors associated with receiving an A2A alone, a stimulant alone, or both medications compared with receiving no ADHD medications among the children who received an ADHD diagnosis in 2017 (n = 2394). Using no ADHD medication (see Supplementary Table S2 for study medications) as the reference category, we report the adjusted odds ratios (AORs) for receiving an A2A alone [no stimulants], a stimulant alone [no A2A], or both medication types, the 95% CI, and 2-sided p-values. A 2-sided p-value of ≤0.05 was considered statistically significant. Data preparation and analyses were performed with R statistical software, version 3.6.0 (July 2, 2018). The study was approved by the Institutional Review Board at the [BLINDED].
Results
Descriptive analyses
The gender and age proportion of children with ADHD did not change over time. However, the percent of children with ADHD living in a metro area has increased to almost half in 2017, up from 39% in 2012 (Table 1). Language, impulse control, seizure/convulsive, and other mental disorders decreased; schizophrenia and mood disorders remained constant; and anxiety and trauma-related disorders, autism and PDD, NOS, other neurodevelopmental, conduct, oppositional defiant, and sleep-wake disorders increased.
Among preschoolers with ADHD, the use of stimulants alone decreased from 27% in 2012 to 15% in 2017, while treatment with A2A alone increased from 10% to 23%, with the steepest increase occurring from 2015 to 2017 (Fig. 1). The receipt of both a stimulant and an A2A followed the decreasing trend in stimulant use (27% in 2012 and 18% in 2017). While the decrease in stimulant use for children with a diagnosis of ADHD seems to be partially mitigated with an increase in A2A use, there is a modest increasing trend of no treatment with either of these medication types, increasing from 37% in 2012 to 44% in 2017 (Fig. 1).
Proportion of children <6 with ADHD receiving treatment by year. ADHD, attention-deficit/hyperactivity disorder.
The prescribing of an A2A instead of a stimulant is greater for children 2–4 years of age, while the use of stimulants over A2As is greater in 5-year-old children (Fig. 2). The use of A2A alone peaks around 3 years of age and then decreases (Fig. 2). The use of stimulants steadily increases with age (Fig. 2). The percent of young children without a prescription for either medication decreases with age with 83% of 2-year olds and 34% of 5-year olds receiving neither medication type (Fig. 2).
Proportion of children <6 with ADHD receiving treatment by age. ADHD, attention-deficit/hyperactivity disorder.
Primary care providers are the highest prescribers of both medications across all years, accounting for approximately half of the prescriptions. Prescriptions from psychiatrists decreased over the years from 36% in 2012 to 27% in 2017; increased from nurse practitioners from 10% in 2012 to 18% in 2017; and other providers from 6% in 2012 to 13% in 2017.
Mirroring the increase in receiving no medication, children who received only behavioral health services grew from only 7% in 2012 to 22% in 2017 (Fig. 3). The percent of children with ADHD who were treated with both behavioral health and medication also increased during this time (17% in 2012 vs. 28% in 2017; Fig. 3).
Proportion of children <6 with ADHD receiving behavioral health treatment. ADHD, attention-deficit/hyperactivity disorder.
Multivariate analyses
Our multivariate regression analyses for the most recent calendar year (2017) largely supported the results in our cross-sectional analyses. In 2017, the odds of receiving a diagnosis of ADHD (Table 2) were higher for males (OR = 2.84, 95% CI = 2.58–3.11), children living in a nonmetro area (OR = 1.12, 95% CI = 1.02–1.22), increasing child age (OR = 3.09, 95% CI = 2.92–3.26), and children who are non-Hispanic white (Table 2).
Odds Ratios for Receiving an Attention-Deficit/Hyperactivity Disorder Diagnosis in 2017
CI, confidence interval; OR, odds ratio.
From the multinomial logistic regression model with our four treatment levels (Table 3), we found that receiving an A2A medication alone was significantly predicted by an additional diagnosis of impulse control (AOR = 4.78, 95% CI = 1.98–11.53), oppositional defiant disorder (AOR = 1.6, 95% CI = 1.09–2.34), and for each additional comorbidity (COR = 1.61, 95% CI = 1.28–2.04). In line with our univariate trend analysis, for every year that a child grew older, the odds of receiving a stimulant instead of no medication increased (OR = 3.96, 95% C I = 2.98–5.27) as did the odds of receiving an A2A instead of no medication (OR = 1.21, 95% CI = 1.04–1.4). Age remained an important predictor in the same direction for receiving both medications. Children of black race (OR = 0.36, 95% CI = 0.23–0.57) or Hispanic ethnicity (AOR = 0.47, 95% CI = 0.24–0.93) were less likely to receive an A2A instead of no medication. No other racial/ethnic differences were noted in receiving A2A, both medications, or stimulant medication.
Multinomial Logistic Regression Comparing the Odds of Being Treated with (1) Alpha-2 Agonist Alone, (2) Stimulant Alone, (3) Both Medications, and (4) No Medication, Using No Medication as the Baseline Treatment
AOR, adjusted odds ratio; CI, confidence interval.
The total number of comorbidities was also a significant predictor of receiving both medications over neither medication (AOR = 1.69, 95% CI = 1.31–2.19). The odds for receiving both medications increased by 6.23 (95% CI = 2.52–15.39) with the presence of an impulse control disorder and 2.33 (95% CI = 1.56–3.48) for oppositional defiant disorder (ODD). The odds for receiving stimulants alone compared with neither medication decreased with the presence of language disorder, but only approached significance (AOR = 0.54, 95% CI = 0.29–1.01; p = 0.054). However, the odds for receiving stimulants alone compared with neither medication increased with the presence of ODD (AOR = 2.36, 95% CI = 1.45–3.85), conduct disorder (AOR = 2.07, 95% CI = 1.32–3.27), impulse control disorder (AOR = 4.1, 95% CI = 1.36–12.35), and other neurodevelopmental disorders (AOR = 1.64, 95% CI = 1.02–2.65). The presence of behavioral health treatment was associated with a decrease in the odds of receiving a stimulant medication alone (AOR = 0.59, 95% CI = 0.45–0.78), an A2A alone (AOR = 0.68, 95% CI = 0.53–0.86), or receiving both (AOR = 0.72, 95% CI = 0.55–0.94) compared with no medication. However, the presence of a medical claim from a psychiatrist/neurologist increased the odds of receiving a stimulant medication alone (AOR = 1.78, 95% CI = 1.18–2.67), an A2A alone (AOR = 1.59, 95% CI = 1.11–2.27), or receiving both (AOR = 3.3, 95% CI = 2.33–4.67) compared with no medication in 2017.
Discussion
The present study represents the first study to examine stimulant and A2A agonist use patterns in preschool-age children over time. The children were receiving Kentucky Medicaid and likely represent the most vulnerable children in terms of behavioral and mental health concerns. Similar to Blum et al. (2018), we demonstrated that younger preschool-age children with ADHD were more likely to receive an A2A instead of stimulant medications. This increase in use of A2A is most dramatic in children 2–3 years of age. The reason for the increase in use of A2A is unknown, although it may relate to the FDA approval of two long-acting A2As for the treatment of ADHD in 2010. In addition, we found that the use of A2A increased over time, while the use of stimulants in these preschoolers decreased during the same time period.
In our predictive models, we found that male gender and increasing age were positive predictors of receipt of the ADHD diagnosis. Children of races/ethnicities other than non-Hispanic white and those living in metro areas were less likely to receive the diagnosis.
Similar to other studies, we found that children with comorbidities were more likely to be prescribed medication (Blum et al. 2018). However, the specific comorbidities that led to increased A2A prescribing varied. For example, in this study, impulse control disorders and opposition defiant disorder increased the likelihood of being prescribed an A2A, whereas in the study by Blum et al., disruptive behavior disorders increased the likelihood of being prescribed a stimulant. These variations may occur due to lack of data or guidelines about when to use stimulants versus A2As in preschool-age children with ADHD.
Recent literature has shown A2A to be effective in treating core symptoms of ADHD in children 6–17 years of age (Sallee et al. 2013; Hirota et al. 2014; Pringsheim et al. 2015; McCracken et al. 2016). Mild-to-moderate side effects included somnolence, bradycardia, fatigue, hypotension, and headache (Hirota et al. 2014; McCracken et al. 2016). In overdoses, significant bradycardia can occur. Safety and efficacy concerns were also noted for combination therapy that included a stimulant and an A2A (Sallee et al. 2013; Hirota et al. 2014; McCracken et al. 2016). However, there are no data related to the use of A2A in children younger than 6 years. In addition to the paucity of data related to safety and initial efficacy of these psychotropic medications in preschool-age children, very little is known about the long-term effects of drug exposures on brains undergoing rapid development and differentiation in early childhood. This lack of evidence suggests a need to proceed with caution in the treatment of ADHD in preschool-age children, especially those younger than 4 years. Adherence to practice guidelines to provide first-line psychosocial care for children younger than 6 years is critically important. Exceptions may be indicated for severe and complex cases, but caution is warranted. For those hwo are prescribed A2A in the preschool period, long-term monitoring would be essential for determining the effects of using this intervention in young children. In addition, targeted outreach and education to those providers prescribing the most stimulants and A2As may partially mitigate any concerns about overprescribing.
Like all studies, there are limitations to the analyses presented here. Administrative claims data are known to have limitations (Research Data Assistance Center 2018). Some limitations include that data are only captured if a claim is submitted, which could lead to underreporting of actual prescribing practices. Providers may prescribe a medication that the parent chooses not to fill or a provider may not bill for services that are known to be frequently denied or for which reimbursement is low (Research Data Assistance Center 2018). In addition, if payments are “bundled” for groups of services, a claim will not be filed. The bundling of services is known to occur for some children in foster care, such as those in residential facilities. Therefore, children in foster care were omitted from the analytic sample. We did not examine the timing, frequency, or duration of behavioral health interventions, including whether the treatment was first-line. A child was considered to have received behavioral health if he or she had at least one claim per year. Certain details of the disease characteristics and/or the treatment type or quality are not provided in claims data (Research Data Assistance Center 2018). We identified cases if a child had one medical claim with a diagnosis of ADHD. Using only a single claim could result in overreporting due to coding errors or use of codes when ruling out the diagnosis of ADHD. An additional limitation is that changes in classification can occur over time. In the present study, it appears that impulse control disorder decreased as a comorbidity among young children with ADHD. This result may not be a meaningful reduction as changes in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) (American Psychiatric Association 1994) and Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) and in the ICD-9 and ICD-10 classifications occurred. For example, trichotillomania (hair pulling) had previously been considered an impulse control disorder, but is now classified as an anxiety disorder, which may partially explain the decrease in impulse control disorder and the increase in anxiety disorders. Although we note this one classification issue, every effort was made to capture the changes accurately. Another limitation is that race/ethnicity is an optional field in the claims data resulting in missing data. Lastly, the data represent only one state so it is not known whether the findings are generalizable to other states. While Kentucky, in general, is a poor state, the children on Kentucky Medicaid would likely be of a similar socioeconomic status as children on Medicaid in other states.
Despite the limitations, the present study has important implications and contributes new knowledge to the literature. The use of stimulants and A2As for the treatment of ADHD in preschoolers continues despite the lack of empirical evidence for effectiveness, safety, and long-term effects in this population. These treatments are being done “off-label” as most medications do not have an FDA approval for use in children younger than 6 years. While the use of stimulants decreased in our study from 2012 to 2017, the use of A2A increased, with the sharpest increase occurring since 2015 and in children 2–3 years of age. Of the preschoolers with an ADHD diagnosis, 22% received behavioral health intervention alone and 28% received behavioral health services in combination with medication. This low rate of behavioral health claims suggests that the AAP guidelines are not being followed. The AAP recommends behavioral health as the first-line treatment for children 4 and 5 years of age and there is no guideline for children younger than 4 years. It is not known whether parents and/or providers are knowledgeable about the AAP recommendations. In addition, more data are needed to determine the barriers to receiving behavioral interventions. It is not known whether health care providers are not recommending behavioral interventions or if parents are unable to access or not interested in receiving behavioral health services. Eighty-two percent of the counties in Kentucky do not have a child and adolescent psychiatrist and the remaining 18% are underserved (American Academy of Child and Adolescent Psychiatry 2017). The number of other types of child mental health providers is not known. Health Resources and Services Administration designates locations as Health Professional Shortage Areas (HPSA) for Mental Health by geography and underserved populations, but there is no differentiation for professionals specifically trained in child mental health professions. However, since Kentucky has many areas designated as HPSAs, it is likely that the shortage of child mental health professionals is great.
Conclusions
More research is needed on the safety, efficacy, and long-term effects of using stimulants, A2As, and combination therapy in young children. Lastly, policy and practice changes, specifically targeting primary care providers, are needed to encourage the use of first-line psychosocial care.
Clinical Significance
This epidemiological study of ADHD treatment in preschoolers receiving Kentucky Medicaid from 2012 to 2017 demonstrates that, despite limited evidence, the use of stimulants and A2As continues. Specific child factors predict each treatment type. Primary care physicians are the highest prescribers. Receipt of behavioral health was associated with reduced odds of receiving stimulants alone, A2As alone, or both medications over the odds of receiving no medication. Seeing a psychiatrist/neurologist was associated with an increased odd of receiving each medication type, possibly suggesting that these young children who were referred to a specialist may have more complex problems.
Footnotes
Disclaimer
Data were obtained from and approved for use by the Kentucky Department of Medicaid Services. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Cabinet for Health & Family Services, Department for Medicaid Services.
Disclosures
No competing financial interests exist.
Supplementary Material
Supplementary Table S1
Supplementary Table S2
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
