Abstract
OBJECTIVE: Whether and to what extent racial/ethnic disparities in attention-deficit/hyperactivity disorder (ADHD) diagnosis occur across early and middle childhood is currently unknown. We examined the over-time dynamics of race/ethnic disparities in diagnosis from kindergarten to eighth grade and disparities in treatment in fifth and eighth grade.
METHODS: Analyses of the nationally representative Early Childhood Longitudinal Study, Kindergarten Class of 1998–1999 (N = 17 100) using discrete-time hazard modeling.
RESULTS: Minority children were less likely than white children to receive an ADHD diagnosis. With time-invariant and -varying confounding factors statistically controlled the odds of ADHD diagnosis for African Americans, Hispanics, and children of other races/ethnicities were 69% (95% confidence interval [CI]: 60%–76%), 50% (95% CI: 34%–62%), and 46% (95% CI: 26%–61%) lower, respectively, than for whites. Factors increasing children’s risk of an ADHD diagnosis included being a boy, being raised by an older mother, being raised in an English-speaking household, and engaging in externalizing problem behaviors. Factors decreasing children’s risk of an ADHD diagnosis included engaging in learning-related behaviors (eg, being attentive), displaying greater academic achievement, and not having health insurance. Among children diagnosed with ADHD, racial/ethnic minorities were less likely than whites to be taking prescription medication for the disorder.
CONCLUSIONS: Racial/ethnic disparities in ADHD diagnosis occur by kindergarten and continue until at least the end of eighth grade. Measured confounding factors do not explain racial/ethnic disparities in ADHD diagnosis and treatment. Culturally sensitive monitoring should be intensified to ensure that all children are appropriately screened, diagnosed, and treated for ADHD.
- ADD —
- attention-deficit disorder
- ADHD —
- attention-deficit/hyperactivity disorder
- ECLS-K —
- Early Childhood Longitudinal Study, Kindergarten Class of 1998–1999
- SES —
- socioeconomic status
What’s Known on This Subject:
Minority children are less likely than white children to be diagnosed and treated with attention-deficit/hyperactivity disorder. However, diagnosis patterns over time in early and middle childhood and whether confounding factors explain these disparities are not currently well understood.
What This Study Adds:
Racial/ethnic disparities in attention-deficit/hyperactivity disorder diagnosis occur by kindergarten and continue until eighth grade. Racial/ethnic disparities among diagnosed children in medication use occur in both fifth and eighth grades. These disparities are not attributable to confounding factors.
Attention-deficit/hyperactivity disorder (ADHD) is the most commonly diagnosed mental health disorder in US children.1–3 An ADHD diagnosis allows eligible children to receive specialized, adaptive educational programming in school, as well as medication that can help mitigate the disorder’s negative impact on learning and behavior.4–6 Yet, not all groups of children are equally likely to be diagnosed and receive treatment. Racial/ethnic minorities in particular have been reported to be diagnosed with ADHD at lower rates than white children,7 and therefore may have unmet treatment needs.8 Children who are African American are diagnosed with ADHD at only two-thirds the rate of white children despite displaying greater ADHD symptomatology,9 and Hispanic children have also been reported to be underdiagnosed.1 Less is known about diagnosis disparities for children of other races/ethnicities.9 In addition, racial/ethnic minorities who are diagnosed with ADHD have been reported to be less likely to use prescription medication.2,10 Hypothesized mechanisms for ADHD diagnosis and treatment disparities include less frequent access to health professionals and service utilization11–14 and, when seen, less frequent solicitation by professionals of developmental concerns.15 Additional hypothesized mechanisms include a lower likelihood of referral by school professionals, limited ability to pay for health care, and negative attitudes toward disability identification and treatment in some racial/ethnic subcultures.16,17
Research investigating ADHD diagnosis and treatment has typically used cross-sectional designs or convenience samples2,18 or focused on limited time frames.7 The onset and over-time dynamics of racial/ethnic disparities in ADHD diagnosis throughout early and middle childhood are unknown. Yet, early diagnosis and treatment are important because ADHD symptomatology persists by early childhood.19–21 In addition, the extent to which other factors confound reported racial/ethnic disparities in ADHD diagnosis and treatment is unclear. Minority children are more likely to be exposed to risk markers for ADHD, including low birth weight, low maternal education, low household income, greater frequency of classroom problem behaviors, and lower academic achievement.1,22–24 The inclusion of confounding factors is therefore necessary to obtain estimates uniquely attributable to children’s status as racial/ethnic minorities. Identifying risks attributable to these confounding factors should also inform efforts to target ADHD screening, diagnostic, and treatment practices, as well as to clarify potential causes of reported racial/ethnic disparities. Identification and treatment at ages when ADHD symptomatology is still newly emergent, and so more amenable to adaptive services and treatment, should help maximize their effectiveness.
The objective of this study was to investigate the onset and over-time dynamics of race/ethnic disparities in ADHD diagnosis across early and middle childhood in the United States. We extend previous work by modeling these disparities over children’s early life course rather than only at a particular time point, by extensively adjusting for plausible confounding factors (eg, lower socioeconomic status [SES] and lower behavioral and academic functioning), and by analyzing a nationally representative cohort of US schoolchildren (ie, the Early Childhood Longitudinal Study, Kindergarten Class of 1998–1999 [ECLS-K]). Our longitudinal modeling of ADHD diagnosis in the United States capitalizes on the richness of the ECLS-K data set, which includes the timing of ADHD diagnosis as well as detailed information about family sociodemographic characteristics and other potential confounders. We also use this detailed information to examine racial/ethnic disparities in prescription medication use.
Methods
Sample Description
The ECLS-K is a nationally representative cohort of US children who entered kindergarten in the fall of 1998. This database is maintained by the National Center for Education Statistics (http://nces.ed.gov/ecls/kindergarten.asp). Data were subsequently collected from children, parents, and teachers in spring of 1999, fall of 1999, spring of 2000, spring of 2002, spring of 2004, and spring of 2007. Our initial analytical sample consisted of 17 100 kindergarten children with information on race/ethnicity and who were (1) followed until eighth grade and had not been diagnosed with ADHD before that time point or (2) censored from the study either because they were diagnosed with ADHD before eighth grade or their data were missing before this time point. We used this sample in our initial (ie, model 1) analysis, which included only race/ethnicity and time as predictors of an ADHD diagnosis. Our second analysis (model 2) included additional child- and family-level predictors measured in kindergarten, as well as time-varying measures of children’s behavioral and academic functioning. Missing data on these variables reduced the analytical sample size to 15 100. Table 1 shows the means of the criterion and predictor variables for each of these analyses. The similarities of these distributions across the analytical samples suggest that sample attrition had relatively little effect on our estimates. Additional detail regarding the ECLS-K’s study design is available from the National Center for Education Statistics.25 We obtained Penn State institutional review board approval for the study.
Descriptive Statistics
Variables of Interest
ADHD Diagnosis
We identified children as diagnosed with ADHD if they were reported by their parent in kindergarten, first, third, fifth, or eighth grade as having received a formal diagnosis of attention-deficit disorder (ADD), ADHD, or hyperactivity. Specifically, we identified children as having an ADHD diagnosis on the basis of a parent’s response of “yes” to each of 3 separate survey questions (ie, “yes” to the child being evaluated by a professional in response to a problem in paying attention, learning, behaving, or in activity level; “yes” to receiving a diagnosis by this professional; and “yes” to the diagnosis being for ADHD, ADD, or hyperactivity). Although medical record documentation of the presence or absence of ADHD is preferable, parents have been repeatedly found to be valid and reliable reporters of ADHD diagnosis, symptoms, and receipt of treatment.26–31 Supplementary analyses (not shown but available from the study’s first author) of special education teacher– rather than parent-reported ADHD diagnosis were consistent with this study’s findings. Table 1 shows weighted descriptive statistics for the analytical samples we used. In the first and largest of these analytical samples, 6.6% of the students reportedly received a diagnosis of ADHD by the eighth grade.
Time-Invariant Measures
Parents reported child gender and race/ethnicity. The sample contained approximately equal proportions of males and females (see Table 1). Racial/ethnic categories included Hispanic (∼19%), non-Hispanic African American (16%), non-Hispanic white (57%), and other races/ethnicities (including Asian, Native Hawaiian, Pacific Islander, American Indian, and Alaskan Native and those described as being more than 1 race).
The child’s and mother’s ages at the date of interview in the spring of kindergarten were recorded and used to calculate the mother’s age at time of birth. If the child’s mother was ≤17 years of age at her child’s birth, she was classified as being a younger mother (∼6%). If the mother was ≥39 years old, she was classified as an older mother (4%). Parents reported whether the child was born with low birth weight (4% of children weighed <5 lb)32 and whether health insurance was available for the child (9% of children did not have health insurance). Whether the interview was conducted in English (∼92% of the sample) was used as a measure of whether the parents spoke English. ECLS-K staff combined reports of the mother’s and father’s education and occupation, along with family income, to create a scale measuring the family’s SES. Dummy variables for SES quintiles were used in our analyses. Time-invariant data were collected in the spring of kindergarten.
Time-Varying Measures
General education teachers in kindergarten, first, third, and fifth grades independently rated the frequency of children’s externalizing and learning-related behaviors by using the Social Rating Scale, a psychometrically validated behavioral measure. The Externalizing Problem Behaviors subscale items measure acting-out behaviors (eg, argues with a teacher, fights, shows anger, acts impulsively, disturbs the classroom). The scale’s reliabilities at each wave ranged from 0.86 to 0.89. The Approaches to Learning subscale items measure learning-related behaviors (ie, pays attention, keeps belongings organized, works independently, shows eagerness to learn new things, easily adapts to changes in routine, and persists in completing tasks). This scale’s reliabilities ranged from 0.89 to 0.91. We used the average if 2 teachers provided ratings of the student in a given year. These survey items were not collected in the eighth grade, so we estimated children’s eighth grade behavioral functioning by using ratings obtained from the previous (ie, fifth grade) survey wave.
Children’s academic achievement was estimated by using the average of their scores from the ECLS-K’s Reading and Mathematics Tests during each survey wave. These individually administered and untimed achievement tests were psychometrically validated. Both measures use item-response theory and routing procedures to maximize academic content coverage and to derive scale scores that are comparable across different grade levels (reliabilities ranged from 0.87 to 0.96). All time-varying measures were standardized.
Prescription Medication Use
In the fifth- and eighth-grade surveys, parents were asked whether their child was currently taking any prescription medication (eg, methylphenidate, amphetamine, atomoxetine) related to ADD, ADHD, or hyperactivity. We conducted supplementary analyses investigating whether and to what extent racial/ethnic minority children diagnosed with ADHD in the fifth grade were also less likely to be taking medication for the disorder in the fifth and eighth grade. “Yes” or “no” responses were coded as 1 or 0, respectively. Approximately 650 of 780 children with an ADHD diagnosis were identified by their parents as using prescription medication for ADHD in the fifth grade.
Analytic Method
We used discrete-time logit (hazard) models for event history analysis33 to identify factors predictive of the timing of ADHD diagnosis from kindergarten to eighth grade by using the program “dthaz” in Stata version SE 12.1 (StataCorp, College Station, TX). These models are advantageous over analyses predicting ADHD diagnosis at a particular time point because they situate the diagnoses in time, appropriately handle the censoring of children who are lost to attrition or who were not diagnosed with ADHD by the end of the observation period, and allow for time-varying covariates (eg, externalizing problem behaviors, learning-related behaviors, academic achievement). Hazard models have been used to investigate racial/ethnic disparities in conditions and disorders other than ADHD.34–36 We also used logistic regression modeling to estimate racial/ethnic disparities in prescription medication use. Table 3 shows these estimates, without and with adjustment for confounding factors.
Results
Table 2 shows weighted estimates from the discrete-time logit modeling of children’s ADHD diagnosis from kindergarten to eighth grade. Model 1 included race/ethnicity and time as predictors. From kindergarten to eighth grade, children from each racial/ethnic minority group were significantly less likely to receive an ADHD diagnosis than whites. Compared with white children, Hispanic children had odds that were 56% (1–0.44) lower of being diagnosed with ADHD. The odds for African Americans and children of other race/ethnicities were 36% (1–0.64) and 48% (1–0.52) lower than whites, respectively. The time point–specific odds of being diagnosed with ADHD peaked at third grade and declined subsequently.
Discrete-Time Logistic Regression Models (Adjusted Odds Ratios) of ADHD Diagnosis
Model 2 added a number of time-invariant and time-varying controls to the regression equation. Lower odds of ADHD diagnosis among racial/ethnic minority children remained evident in model 2 despite inclusion of potential confounding variables. African American and Hispanic children and those of other races/ethnicities were 69% (1–0.31), 50% (1–0.50), and 46% (1–0.54) less likely than white children, respectively, to be diagnosed with ADHD. Model 2 also indicated that boys had twice the odds of girls of being diagnosed with ADHD (differences by race/ethnicity were similar for boys and girls; results not shown). More frequent externalizing behaviors increased this risk. In contrast, more frequent learning-related behaviors and greater academic achievement decreased this risk. Children whose mothers were older than age 38 at the child’s birth were more likely to receive an ADHD diagnosis, as were children who had access to health care or those raised by English-speaking parents. Those from the western United States were less likely to be diagnosed than those from the Midwest.
These results show that with the most extensive control variables, including time-varying factors relating to behavioral risk indicators of ADHD and to academic achievement, children who are African American, Hispanic, and of other race/ethnic groups are much less likely than children who are white to receive an ADHD diagnosis in the United States across early and middle childhood. To further assess the robustness of these results, in additional models (not shown) we included additional time-varying variables of teacher’s race and age, as well as school-level characteristics (eg, the percentages of racial/ethnic minority students and those receiving free or reduced lunch attending the school). These variables were excluded in the final reported models because they were not, in general, statistically significant predictors and their inclusion in the prediction equation did not change our substantive findings. (Inclusion of these variables also reduced the sample size because of missing values.)
Figures 1 and 2 use model 2’s results to derive time-specific probabilities of ADHD diagnosis by race/ethnicity. The over-time hazard of ADHD diagnosis (Fig 1) increased through third grade and declined thereafter. This hazard (ie, the instantaneous probability of diagnosis) in kindergarten, first grade, third grade, fifth grade, and eighth grade was highest for white children, holding the other variables in Table 2’s model 2 constant at their means. These same probabilities of diagnosis were lower for Hispanics, and children of other races/ethnicities, and lowest for African Americans. Figure 2 shows the survival function, or the cumulative proportion of children not diagnosed with ADHD, at each survey wave by race/ethnicity, again holding model 2’s other variables constant at their sample means. By the spring of eighth grade, ∼7% of white children received an ADHD diagnosis sometime between kindergarten and eighth grade. The comparable rates for African American, Hispanics, and children of other races/ethnicities were ∼3%, 4.4%, and 3.5%, respectively.
Plot of the hazard function of ADHD diagnosis by race/ethnicity. K, kindergarten; 1st, first grade; 5th, fifth grade; 8th, eighth grade.
Plot of the survival function of ADHD diagnosis by race/ethnicity. K, kindergarten; 1st, first grade; 5th, fifth grade; 8th, eighth grade.
Table 3 shows logistic regression (odds ratios) estimates of racial/ethnic disparities in prescription medication use among children diagnosed with ADHD in the fifth grade. The odds ratios shown in model 1 were unadjusted, whereas model 2 included a full set of controls. Before and after covariate adjustment, Hispanic, African American, and children of other races/ethnicities diagnosed with ADHD were much less likely to be using medication than white children diagnosed with ADHD. This disparity occurred in both the fifth and eighth grades.
Race/Ethnicity Odds Ratios from Logistic Regression Models of Prescription Medication Use for ADHD in the Fifth or Eighth Grade in Children Diagnosed With ADHD in the Fifth Grade
Discussion
We investigated racial/ethnic disparities in ADHD diagnosis in a large, nationally representative cohort of schoolchildren as they attended kindergarten to eighth grade. Results indicated that racial/ethnic minority children are much less likely than otherwise identical white children to receive an ADHD diagnosis. Children at lower risk include those displaying greater academic achievement, those with more frequent learning-related classroom behaviors, and those without health insurance. Boys, children frequently engaging in externalizing problem behaviors, and those raised by an older mother are more likely to receive a diagnosis of ADHD. Being raised by an English-speaking parent also increased the likelihood of diagnosis. In addition, and again after extensive statistical control, children diagnosed with ADHD were much less likely to use prescription medication for the disorder if they were Hispanic, African American, or of other races/ethnicities.
Previous research has reported racial/ethnic disparities in ADHD diagnosis and treatment, particularly for African Americans.9 However, whether and to what extent these disparities also occur among Hispanics and children of other races/ethnicities has not been systematically examined.37,38 Our study extends previous work by establishing that children who are African-American, Hispanic, or of other races/ethnicities are less likely than whites to be diagnosed with ADHD. These racial/ethnic disparities in ADHD diagnosis occur as early as kindergarten and continue at least until the end of middle school. Racial/ethnic minorities are also less likely to be using medication to treat the disorder by the end of elementary and middle school. These racial/ethnic disparities are not attributable to many potential confounding factors that also strongly predict children’s receipt of an ADHD diagnosis or medication use.
Our analyses relied on multi-item parent response to identify children as being diagnosed with ADHD. Although previous work has repeatedly found that parents are valid and reliable reporters of their children’s ADHD symptomology, diagnosis, and treatment response,26–31 and has relied on the same type of parent survey items used here,7 we were unable to independently verify to what extent children identified by their parents also met formal ADHD diagnostic criteria. Although we controlled for many potential confounding factors, including changing levels of behavioral functioning and academic achievement, it is possible that unmeasured variables may have contributed to the lower risk we attribute to children’s status as racial/ethnic minorities. Our study analyzed a single cohort of children, whose experiences (eg, health insurance coverage) may have resulted from period-specific events and conditions. Consistent with previous work,7,9,13 we interpreted the directionality of the disparities as indicating underdiagnosis for racial/ethnic minorities. Another possibility is that children who are white are comparatively overdiagnosed. Additional research is needed to determine the extent to which such under- and overdiagnosis occurs. Our study examines over-time diagnosis until the end of eighth grade, which constituted the final survey wave of the ECLS-K. We may have observed other diagnosis patterns had the ECLS-K’s data collection continued into high school.
Our study provides clinic- and school-based professionals with risk estimates that are more rigorously derived than those currently available, particularly regarding over-time diagnosis rates for ADHD. Underdiagnosis for African Americans, Hispanics, and children of other races/ethnicities may occur because clinicians are disproportionately responsive to white parents who are more likely to solicit ADHD diagnosis and treatment of their children.39,40 Clinical practice and policy may need to be redirected to ensure that children from minority families are appropriately evaluated, diagnosed, and treated for ADHD.
Our study should inform efforts to reduce racial/ethnic disparities in ADHD diagnosis and pharmacologic treatment. We found that disparities in ADHD diagnosis and treatment occur for children who are Hispanic and of other races/ethnicities as well as for African Americans. To date, clinicians and policy-makers have had very little information about disparities in diagnoses and treatment of non-African American minority groups. Medical and school-based professionals should ensure that their efforts to reduce racial/ethnic disparities in ADHD diagnosis and treatment also extend to groups other than African Americans. Our findings provide additional support to calls for increasing solicitations by pediatricians, school psychologists, teachers, and other clinicians of concerns by minority parents for their children’s learning and behavior,15,41 ensuring sensitivity to differing cultural values about disability during well-child visits and referrals,42 and encouraging symptom recognition and help-seeking behaviors by these parents,43 and suggest that these efforts by clinicians should be made throughout minority children’s early life course.44
Footnotes
- Accepted April 12, 2013.
- Address correspondence to Paul L. Morgan, PhD, 211 Cedar Building, University Park, PA 16802. E-mail: paul.l.morgan{at}gmail.com
Dr Morgan conceptualized the study research questions and design, collaborated on the analyses and their interpretation, drafted the initial manuscript, and approved the final manuscript as submitted; Dr Staff also conceptualized the study’s research questions and design, conducted the study’s initial analyses, reviewed and revised the manuscript for important content, and approved the final manuscript as submitted; Drs Hillemeier and Farkas collaborated in the study’s research questions and design, made substantive intellectual contributions to the data analysis and interpretation, revised the manuscript for important content, and approved the final version to be published; and Mr Maczuga conducted the study’s analyses with Dr Staff, revised the manuscript for important content, and approved the final version to be published.
The content of this publication does not necessarily reflect the views or policies of the US Department of Education.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Supported by grant R324A120331 from the National Center for Special Education Research, Institute of Education Sciences, US Department of Education; and infrastructure support provided by NIH/NICHD grant R24HD041025-11. Funded by the National Institutes of Health.
References
- Copyright © 2013 by the American Academy of Pediatrics
In this issue
