Racial and Ethnic Disparities in Childhood ADHD Treatment Access and Utilization: Results From a National Study

We used multiple cross-sections from the 2011–2019 Household Component of the MEPS (9). Multiple years were pooled to increase the precision of our estimates. The MEPS is administered by the Agency for Healthcare Research and Quality and captures detailed sociodemographic, clinical, and health care information from a nationally representative, community-based sample of noninstitutionalized U.S. civilians. The MEPS is administered in both English and Spanish, which is especially important given that non-English speakers are more likely to face barriers to mental health treatment because of the lack of interpreter services and translated materials (10). Previous studies have largely relied on self-reported measures of treatment use. Because of its follow-up verification of utilization with medical providers and insurance companies, the MEPS offers more accurate estimates of treatment utilization. ADHD diagnosis status was obtained by asking caregivers whether they had ever been told by a doctor or other health professional that their child or children ages 5–17 had ADHD or attention deficit disorder (ADD). This resulted in a final sample of 5,838 participants with ADHD. This study was deemed exempt by the Cambridge Health Alliance Institutional Review Board.

Primary outcomes were access to and utilization of ADHD treatment, separated into treatment visits, ADHD medication prescriptions, and total ADHD treatment–related expenditures. We defined access as any treatment use, which captured enabling factors, such as coverage and availability of services, that allowed an individual to use care (11), and we defined utilization as the amount of care used once treatment was accessed. We constructed dichotomous and count variables that represented health care provider (e.g., psychiatrist, psychologist, social worker, or primary care provider) visits for an ADHD diagnosis and ADHD medication prescription fills in the past year. Any visit associated with ICD-9 code 314.X or ICD-10 code F90.X was considered to involve ADHD treatment. ADHD-specific medications were identified by using the National Institute of Mental Health, Healthline, and GoodRx websites (a table listing the names of these medications is included in an online supplement to this article). We combined total direct payments associated with provider visits to assess ADHD treatment–related expenditures in the past year, which included all direct payments by private insurance, Medicaid, and other insurance sources, as well as out-of-pocket payments (12).

Given the high level of comorbid psychiatric diagnoses among children with ADHD (13, 14), we conducted secondary analyses examining access to and utilization of treatment for any psychiatric diagnosis, including but not limited to ADHD—hereinafter referred to as mental health treatment. Examining receipt of treatment for any psychiatric diagnosis may elucidate whether disparities for children with ADHD from racial-ethnic minority groups persist beyond ADHD treatment. We assessed access to and expenditures for mental health treatment, defined as outpatient or inpatient visits or medication prescription fills associated with any psychiatric diagnosis in the past year. A visit was considered to involve mental health treatment if it was associated with a mental disorder diagnosis using ICD-9 codes 291, 292, or 295–314 (15) or ICD-10 codes F01–F99 or if the treatment was coded as psychotherapy or mental health counseling. Psychotropic medications were identified by using the Multum classification system (16).

We conducted our main analyses in accordance with the Institute of Medicine (IOM) definition of health care disparity as described in Unequal Treatment (17) (the IOM is now named National Academy of Medicine [NAM]; per NAM recommendation, we use the original acronym for this definition). The IOM defined disparities as observed differences that are not explained by differences in clinical appropriateness, treatment need, or patient preferences. Differences due to socioeconomic status (SES) factors (e.g., income and health insurance status) are considered contributors to disparity (17). As such, our main analysis did not adjust for SES covariates. We included measures of self-rated physical and mental health status as well as impairment as proxies for treatment need. The MEPS does not capture patient preferences or clinical appropriateness, and our main analyses did not adjust for these factors.

Although we regard implementing the IOM definition of disparities as best practice, using a narrower view of disparities may elucidate the specific contribution of race-ethnicity to disparities. The residual direct effect (RDE) method implicitly defines a disparity to be any difference between racial-ethnic groups that is evident after the analysis controls for all possible confounders (18), which contrasts with the IOM definition by the inclusion of adjustment for SES. Given the complex relationship between race-ethnicity, SES, and health, we conducted exploratory analyses by using the RDE definition to characterize disparities after additionally adjusting for racial-ethnic differences in SES (see tables in the online supplement).

The primary variable of interest was self-reported race-ethnicity, which was categorized according to the U.S. Census as non-Hispanic White, non-Hispanic Black, Hispanic, and non-Hispanic Asian, including Native Hawaiian and Pacific Islanders (hereinafter referred to as White, Black, Hispanic, and Asian, respectively). Individuals of another race or of multiple races were excluded because of the small sample sizes. Covariates representing treatment need were scores on the Columbia Impairment Scale (CIS) (19), caregiver-rated perceived physical health, and caregiver-rated perceived mental health. Age and sex were also included as proxies for treatment need, given the reported differences in prevalence of psychiatric disorders across these factors (20, 21). Indicators for survey year were included to account for differences by sampling year. SES covariates, included in exploratory RDE analyses only, were U.S. region of residence (Northeast, Midwest, South, or West), insurance status (private, public [Medicaid or Children’s Health Insurance Program], other, or uninsured), and federal poverty level (FPL) (poor, <100% FPL; near poor, 100%–124% FPL]; low income, 125%–199% FPL; middle income, 200%–399% FPL; or high-income, ≥400% FPL).

First, unadjusted differences in treatment access and utilization, demographic factors, treatment need, and SES were compared by race-ethnicity. Student’s t tests or Pearson’s chi-square tests were used as appropriate, with White as the reference group. Next, using multivariable regression, we examined whether treatment access and amount of treatment received by children with ADHD differed by race-ethnicity after adjustment for demographic factors and treatment need.

In our primary analysis, we used two-part multivariable regression models to compare treatment disparities between racial-ethnic groups. This analytic approach allowed us to account for the skewness and large number of zeros in the expenditure data (22, 23). First, access disparities were examined by using logistic regression models to estimate the likelihood that a racial-ethnic group would receive any treatment. Then, treatment use disparities were examined by using generalized linear models (GLMs) to estimate differences in treatment use counts and expenditures conditional on treatment access. For count outcomes, we estimated GLMs with a negative binomial distribution and logit link function. This approach accounts for the overdispersion of the count data (24). When assessing expenditures, we estimated GLMs with a gamma distribution and a log link function. The modified Park test was used to determine the optimal mean-variance relationship (22, 23). In secondary analyses, we employed similar statistical methods and examined racial-ethnic disparities in access and expenditures associated with any mental health treatment. In exploratory analyses assessing the impact of SES on disparity estimates, we reestimated regression models using the RDE definition of health care disparities.

We used the predictive margin method to obtain disparity results in the scales of interest (percentages and means) (25) (see tables in the online supplement for full regression model results). As such, disparities in treatment access rates were reported as the absolute percentage-point difference in the predicted probability of a minority group’s receipt of treatment, compared with the baseline predicted probability for White children. Treatment use disparities were reported as the absolute difference in mean number of visits or medication prescription fills or mean expenditures between each minority group and White children.

Variables with missing values, apart from ADHD diagnosis and race-ethnicity, were imputed by using multiple imputation methods. This approach creates five data sets, imputes missing values by using a chained-equations method, analyzes each data set, and uses standard rules to combine estimates and adjust standard errors for the uncertainty attributable to imputation (26, 27). All statistical tests were conducted with a significance level of p<0.05. Model estimates were weighted according to sample design and survey nonresponse. All statistical analyses were conducted with Stata, version 16.1 (28).

Sample characteristics are presented in Table 1. Compared with White children (N=2,549), on average, Black children (N=1,537) had worse caregiver-rated physical and mental health status and were more likely to live in the South, have public health insurance, and live in households with income below the poverty line. Compared with White children, Hispanic children (N=1,681) had worse caregiver-rated physical and mental health status and lower impairment scores (less severe impairment) and were more likely to live in the West, have public health insurance, and live in households with income below the poverty line. Compared with White children, Asian children (N=71) were more likely to live in the West and more likely to live in high-income households.

Black, Hispanic, and Asian children were less likely than White children to have accessed any ADHD-specific treatment visit, ADHD medication, or any mental health treatment. Compared with White children, children from all minority groups filled fewer ADHD medication prescriptions, Black children had fewer ADHD-specific visits and lower ADHD-specific visit expenditures, and Black and Hispanic children had lower total mental health treatment expenditures.

As shown in Table 2, the baseline predicted probability of White children with ADHD to have accessed any ADHD treatment visit in the past year was 51%. Children with ADHD from the three minority groups were significantly less likely than White children to have done so: Black children, −15 percentage points (p<0.001); Hispanic children, −12 percentage points (p<0.001); and Asian children, −21 percentage points (p<0.01). Similarly, compared with White children, children from the minority groups were significantly less likely to have filled any ADHD medication prescription: White, predicted probability, 41%; Black, −8 percentage points (p<0.01); Hispanic, −11 percentage points (p<0.001); and Asian, −30 percentage points (p<0.001).

In terms of utilization, for White children, the predicted number of prescription fills for ADHD-related medications was 4.9. Black and Asian children filled fewer such prescriptions, compared with White children: Black, 0.9 fewer (p<0.05); Asian, 2.7 fewer (p<0.01).

IOM-concordant racial-ethnic disparities were noted in access to and use of treatment for any psychiatric diagnosis. As shown in Table 3, the baseline predicted probability of White children with ADHD to have accessed any mental health treatment in the past year was 66%. Children with ADHD from the three minority groups were significantly less likely than White children to have done so: Black children, −16 percentage points (p<0.001); Hispanic children, −14 percentage points (p<0.001); and Asian children, −27 percentage points (p<0.01). Compared with White children’s total past-year expenditures for any mental health treatment (predicted expenditures, $2,459), expenditures for Black children were $484 lower (p<0.05), and for Hispanic children, they were $415 lower (p<0.05).

Consistent with IOM-concordant results, children with ADHD from the three minority groups were significantly less likely than White children to have had any ADHD treatment visit or to fill an ADHD medication prescription in the past year (see table in online supplement). In contrast to our IOM-concordant results, the only significant disparity in utilization was between Black and White children: Black children filled significantly fewer ADHD prescriptions.

In terms of utilization of any mental health treatment, consistent with IOM-concordant results, children with ADHD from the racial-ethnic minority groups were significantly less likely than White children to have received any mental health treatment (see table in online supplement). In contrast to the IOM-concordant results, no significant disparities were noted in total mental health expenditures.