Racial Disparities in Diagnosis of Attention-Deficit/Hyperactivity Disorder in a US National Birth Cohort

Key Points Question Are there racial and ethnic disparities in the diagnosis and treatment of attention-deficit/hyperactivity disorder (ADHD) in children? Findings This cohort study of 238 011 children examined the association between race/ethnicity and the diagnosis of ADHD. Asian, Black, and Hispanic children were significantly less likely to be diagnosed with ADHD compared with White children. White children were also more likely to receive treatment for ADHD. Meaning These findings suggest that racial and ethnic disparities in the diagnosis and treatment of ADHD are evident.


Introduction
Attention-deficit/hyperactivity disorder (ADHD) affects a large number of children and has long-term effects on their health and learning. [1][2][3] For example, ADHD is associated with poorer quality of life and higher medical costs. 2,4 Based on data from national surveys, the prevalence of ADHD in the US appears to have increased in the last 2 decades. 5,6 However, little is known about the incidence of ADHD diagnosis at the national level. Knowledge about the age at initial diagnosis and the number of children being diagnosed at a certain age could support the formulation of public health policy to improve early interventions such as medical treatments and school services. [7][8][9] Previous studies suggest that racial and ethnic disparities exist in the diagnosis of ADHD.
Although most studies have reported differences among prevalence of ADHD between racial and ethnic groups, there are discrepancies in the direction of the inequality. Black children may have a higher or lower rate compared with non-Hispanic White children depending on the study. [10][11][12][13][14][15] The underlying mechanism of these disparities is unclear, but it may include socioeconomic and cultural factors, variations in the interpretation of children's behavior, and the application of diagnostic criteria. 11,[16][17][18] Children from different racial groups may also have different rates of psychiatric comorbidities. 19 Racial and ethnic disparities may also exist in the treatment of ADHD. According to guidelines for treatment of ADHD, any child aged 4 through 18 years who presents with academic or behavioral problems suggesting ADHD should be evaluated by a primary care clinician. 20 Once a diagnosis of ADHD is made, behavioral therapy should be the first-line treatment for preschool-aged children and continue to be part of the treatment plan, along with medication starting in elementary school-aged children. 20 However, it is unclear how many children are treated according to this guideline. Previous reports have indicated that non-Hispanic White children were more likely to take medication for ADHD compared with children in other racial and ethnic groups. 21 It is unknown if such disparities in treatment occur immediately after the initial diagnosis. Further understanding of how treatment patterns for ADHD may differ based on race, at the time of initial diagnosis and in the early stages of treatment, may help all children receive appropriate evidence-based care.
In this study, we constructed a birth cohort of children from a national commercial insurancebased data set to test the hypotheses that Non-Hispanic White children have higher cumulative incidence of ADHD diagnosis and that they are more likely to be treated with medications within the first year after diagnosis compared with children from other racial or ethnic groups. We also explored whether children had different psychiatric comorbidities associated with ADHD based on race and ethnicity. Unlike previous studies that relied on parental reports, we used clinical diagnostic and billing codes as outcome measures.

Methods
The Mayo Clinic institutional review board exempted this study from review and informed consent requirements because the study used preexisting and deidentified data. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies. date of follow-up. In the analysis of treatment for ADHD in the first year after the initial diagnosis, only children who had a least 1 year of insurance coverage beyond the date of ADHD diagnosis were included.

Outcomes
The primary outcome was the diagnosis of ADHD, which is defined by International Classification of Diseases, Ninth Revision (ICD-9) 22 codes of 314.X or ICD-10 23 codes of F90.X. ADHD with predominantly inattentive type was defined as having ICD-9 code of 314.00 or ICD-10 code of F90.0.
If a child was diagnosed with ADHD before 3 years of age but did not have any subsequent visits related to ADHD, they were classified as not having ADHD. This decision was made based on the American Academy of Pediatrics clinical practice guideline 20 that there is insufficient evidence to recommend diagnosis and treatment for children younger than 4 years of age.
Secondary outcomes included both medications and behavior therapies for treatment of ADHD in the clinical setting after the initial date of diagnosis. Medications were identified by having a filled prescription approved for the treatment of ADHD. Psychological behavioral therapy was identified by Current Procedural Terminology codes, Healthcare Common Procedure Coding System codes, and revenue codes.
Psychiatric comorbidities among the children who were diagnosed with ADHD, including common internalizing and externalizing disorders, were evaluated separately from date of birth to the date of ADHD diagnosis and from the date of diagnosis to the end of follow-up. Comorbidities were identified using ICD-9 and/or ICD-10 codes anywhere within the claim. Only the comorbidities with statistically significant difference based on race were reported.

Explanatory Variables
Race and ethnicity were based on self-report and were combined as 1 variable and classified as non-Hispanic White (ie, White), non-Hispanic Black (Black), Hispanic, Asian, or other/unknown. 24 Geographic regions were classified as Midwest, Northeast, South, and West according to the US census. Annual household income in US dollars at the time of last follow-up was categorized as less than $40 000, $40 000 to $74 999, $75 000 to $124 999, $125 000 to $199 999, $200 000 or more, and other/unknown. 24 Patients with missing race/ethnicity or annual household income were classified into the respective other/unknown categories. A sensitivity analysis was performed by identifying patients who had missing variables and then removing them from the cohort entirely.
Removal of these patients did not significantly change the results.

Statistical Analysis
Differences in baseline characteristics according to ADHD status were examined using a χ 2 test for categorical variables and a t test for continuous variables. In the analysis of ADHD incidence, a child contributed to person-time until the first date of ADHD diagnosis, the end of insurance plan coverage, or the end of the follow-up period (ie, June 30, 2019). Cumulative incidence was calculated based on the number of children who were diagnosed with ADHD over the number of children who contributed to person-time. Incidence over time was depicted using Kaplan-Meier curves. Cox proportional hazards regression was performed to calculate the hazard ratio (HR) of ADHD comparing other racial/ethnic groups with the results from White children. Multivariate Cox regression was then performed to adjust for sex, region, and household income.
In the analyses of psychiatric comorbidities and treatment within the first year of ADHD diagnosis, we restricted analysis to children who had at least 1 year of follow-up after their initial diagnosis. A χ 2 test was performed when making comparisons among racial groups. P values were based on F tests. A P < .01 was considered statistically significant in 2-sided tests. Analyses were performed using SAS Enterprise Guide version 7.13 (SAS Institute Inc) and Stata version 16.1 (StataCorp). Data analysis was conducted from October 2019 to December 2020.

Results
Among the 238 011 children in the cohort, 116 093 (48.8%) were girls; 15 183 (6.7%) were Asian, 14 792 (6.2%) were Black, 23 358 (9.8%) were Hispanic, and 173 082 (72.7%) were White children ( Table 1). In this cohort, 11 401 children (4.8%) were diagnosed with ADHD during the follow-up period with a mean (SD) age of diagnosis of 6.5 (1.9) years. The overall incidence of ADHD was 69 (95% CI, 68-70) per 10 000 person years (eTable 1 in the Supplement). Compared with children who were not diagnosed with ADHD, children with ADHD had more years of coverage in the data set Children were often diagnosed with other disorders prior to their ADHD diagnosis (

Discussion
The major contributions of this study to the current literature include providing findings that suggest (1) significant racial disparities in ADHD diagnosis in privately insured children, including Asian children; (2) racial disparities in treatment, including evidence for gaps between guidelines and clinical practice; and (3) differing patterns of comorbidities accompanying ADHD diagnosis according to race.
A limited number of previous studies have reported the incidence of diagnosis of ADHD. [25][26][27] In our study, the cumulative incidence increased steadily as children entered school, with 10% of children diagnosed before age 10 years. Our study is consistent with a 2004 US population-based study by Barbaresi et al. 25 Using a cohort in Olmsted County, Minnesota, their cumulative incidence at age 13 years was approximately 10%, which is comparable with our estimate of 13% at age 12 years.
Our findings are also consistent with previous studies that reported boys are more than twice as likely to be diagnosed with ADHD than girls and that there are higher rates of ADHD in children from families with lower income. 13,14 Regional differences in diagnosis rates were also similar to previous studies, which also found higher rates of ADHD diagnosis in the South and Midwest regions of the US. 6,15 Racial disparities in prevalence of ADHD diagnosed based on parental reports have been reported in several national surveys. In both the National Health Interview Survey 14 and National Survey of Children's Heath, 15 Black children had higher prevalence compared with White children, whereas Hispanic children were 35% less likely to be diagnosed compared with White children.
According to the Early Childhood Longitudinal Study, the incidence of diagnosis of ADHD for Hispanic children was also significantly lower than that for non-Hispanic children. 27 However, in contrast to the National Health Interview Survey and National Survey of Children's Health, the HR comparing ADHD diagnosis for Black with White children was 0.3. The racial disparities in these national surveys persisted after controlling for socioeconomic factors. When ADHD diagnosis was made according to

Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) 28 using information obtained
from parental interview, a study with the National Health and Nutrition Examination Survey also showed higher prevalence of ADHD in White children (12.65%) aged 12 to 15 years compared with Hispanic (7.11%) or Black (7.69%) children. 29 In comparison, we found a smaller but consistent difference between White and both Black and Hispanic children, with the latter being less likely to be diagnosed. In our study, all children had coverage by commercial insurance, which decreased the potential effect of access to health care on ADHD diagnosis. Fewer studies of ADHD have examined Asian children as a separate racial group. Our finding that this group had the lowest incidence of ADHD is consistent with 2 studies of children covered by the Kaiser Permanente health plan in California. 12,13 The cause of the disparities in ADHD diagnosis according to race and ethnicity is not fully understood. Eiraldi et al 30 proposed a model on the care-seeking pathway of children with ADHD and provided a conceptual framework for understanding racial differences in ADHD diagnosis and treatment. This model included multiple factors that may influence problem recognition, making the decision to seek help, service selection, and service utilization. For example, Coker et al 31 suggested that Black children had more symptoms consistent with ADHD based on questionnaires but were less likely to have been clinically diagnosed. Because of the variation in symptoms of ADHD, cultural values may impact the perceptions of such behaviors. 32,33 Our findings provide some evidence that Asian parents brought their children for clinical evaluation for reasons that differed compared with other racial groups, as reflected by the differences in comorbidities preceding or subsequent to ADHD diagnosis among racial and ethnic groups: Asian children were found to have higher rates of speech language disorder and autism spectrum disorder while White children had more anxiety and adjustment disorders. It is also likely that patients' concerns about racism play some role in influencing their willingness to approach the health care system. 34 The disparities in treatment among children who were diagnosed with ADHD may also reflect the results of parental care-seeking preferences. Asian and Hispanic children were less likely to receive medication treatment than White and Black children. However, the percentage of Asian children receiving psychotherapy was not significantly lower than other groups, which is different than a 2013 study 13 finding that Asian children with ADHD were less likely to use mental health services. Disparities in mental health service in children continue to be a public health issue. 35 Health care professionals may also contribute to the racial disparities in diagnosis and treatment. Stereotype and bias, both explicit and implicit, have been increasingly recognized as factors potentially contributing to physicians' clinical decision-making. [36][37][38] It is possible, for example, that identical behavior displayed by Black and non-Hispanic White children may be interpreted differently based on race-based expectations for the behavior of children, and thus, behavior that is identified as disordered in White children might be inappropriately interpreted as normal in Black children.
In our cohort, more than half of children who were diagnosed during preschool years were taking ADHD medications. Although it is unknown whether those children had received failed behavioral therapy treatment before the initiation of medications, this finding may suggest deviation from the American Academy of Pediatrics clinical practice guidelines. Most school-aged children were treated with medications, which is consistent with previous reports. 15,21 Limitations This study has several limitations. First, we used ICD codes to identify ADHD cases. Because administrative data were collected for billing purposes rather than research, the presence of ICD codes may not always indicate true clinical diagnosis. The utility of ICD codes in ascertaining ADHD has been evaluated in previous studies, [39][40][41][42] where it was shown to have a high sensitivity, specificity, and positive predictive value. 40,42 However, we still might have mistakenly assigned ADHD status to a small number of children. Second, our birth cohort of children was from a national commercial insurance database and may not be representative of all children in the US. However, a 2018 report 21 on children aged 2 to 17 years who were covered by Medicaid in New York state suggested 5.4% were diagnosed with ADHD, which is similar to our finding of 4.8%. Third, we had no data on child-level information, such as behavioral symptoms, and were unable to comment on the possibilities of under-or overdiagnosis of ADHD or any of the psychiatric comorbidities. Fourth, in the evaluation of psychotherapy for ADHD, we had no data on any therapies that were not covered by the insurance plan and thus did not result in a claim in the data set. As a result, the number of children in therapy is likely underestimated. We are also unable to comment on any services that children received from their schools. Fifth, we were only able to control for a limited number of variables in our multivariate analysis and could not rule out the effect of unmeasured confounders.