Effects of Intersectionality Along the Pathway to Diagnosis for Autistic Children With and Without Co-occurring Attention Deficit Hyperactivity Disorder in a Nationally-Representative Sample

Abstract

Children with complex behavioral profiles (e.g., ASD + ADHD) may experience delays in obtaining a final diagnosis. Low-resource or underrepresented groups may be at even greater risk for delayed diagnosis. We assessed the effect of sociodemographic factors, symptom complexity and co-occurring conditions, and identifier of first symptoms on diagnostic trajectories among children aged 3–17 years diagnosed with ASD (n = 52) or ASD + ADHD (n = 352) from a nationally-representative sample. Race/ethnicity and gender disparities were evident in both groups. Race, symptom complexity, and co-occuring conditions predicted age of final diagnosis and wait time between first concern and final diagnosis, both of which were staggeringly high. Results suggest a complex influence of sociodemographic factors on the diagnostic pathway, and risk of health disparities as a function of intersectionality.

Introduction

Autism Spectrum Disorder (ASD) is diagnosed in White Non-Hispanic (WNH) children more than any other race or ethnic group (Christensen et al., 2016), and in males more than females (1 in 151 girls relative to 1 in 37 boys; Hiller et al., 2015). These differences are likely more reflective of social and cultural biases than biological prevalence (Loomes et al., 2017; Magaña et al., 2013), as well as health disparities in access to early detection and intervention (Magaña et al., 2013; Mandell, et al., 2009). Disparities in healthcare access and quality can lead to an initial misdiagnosis, and in turn, a delay in receiving the correct diagnosis. For example, African–American children are two times more likely to receive a diagnosis of conduct disorder, and over five times more likely to receive a diagnosis of adjustment disorder, before their ASD diagnosis relative to White children (Mandell et al., 2007). This problem reflects providers’ tendency to overlook parents’ developmental or socioemotional concerns about their children and focus on so-called “problem behaviors”, which can impede the process of early identification for neuro developmental conditions with socioemotional core features, like ASD.

Hispanic/Latinx families have also reported difficulty receiving a correct ASD diagnosis due to limited access to care, low English-language proficiency, socioeconomic status, and cultural stigma regarding disabilities (Zuckerman et al., 2014). As a result, Hispanic/Latinx individuals are likely underrepresented in the autistic community; indeed, one study found a significantly lower enrollment of students diagnosed with ASD in predominantly Hispanic/Latinx school districts relative to non-Hispanic/Latinx-serving districts (Palmer et al., 2010), and another found that Hispanic/Latinx families were more likely to receive a misdiagnosis of a language disorder first (Mandell et al., 2009).

This prior work suggests that the autistic community is vulnerable to a multitude of potential health disparities that may impede or delay a correct diagnosis. Disparities in pediatric care stem primarily from a history of institutional racism, implicit biases among providers, and disproportionate barriers to accessing care among underrepresented groups (for review, see Johnson, 2020), rather than to family or cultural knowledge or awareness.

The Role of Primary and Specialty Care in Diagnosis

There are many factors that are associated with delayed, missed, or incorrect diagnosis, such as low socioeconomic status, symptom severity, and level of parental education (Fountain et al., 2011, Jo et al., 2015). But perhaps the most universal factor is the logistical difficulty of coordinating the assessment and diagnostic process. In many cases of suspected ASD, the complex presentation of symptoms requires consultation with clinicians from multiple lines of service (Goin-Kochel et al., 2006). Families may also need to change providers within a single line of service depending on scheduling, payer status (e.g., private insurance coverage versus government healthcare benefits), and a clinician’s level of expertise with ASD. Highlighting this complexity, one study of 495 parents of children diagnosed with ASD found that the average diagnosis required a family to interact with 4–5 different clinicians (Dyches et al., 2004).

Even when primary care physicians serve as a medical home for continuity of care across a patient’s childhood, minority families still struggle to access subspecialty doctors with developmental, pediatric, and mental health expertise (Flores & Tomany-Koran, 2008; Elster et al., 2003; Nguyen et al., 2007). Particularly for families with limited resources, multiple jobs, and low education, coordinating care presents a significant challenge. The consequences of a multi-step diagnostic process are myriad, including (but not limited to): increased wait times from first concern to diagnosis, difficulty with coordination of care, problems with payer status/eligibility, and a feeling of disconnection between patients and providers.

Attention Deficit Hyperactivity Disorder (ADHD)

Primary care physicians, educators, early intervention providers, and family members are positioned to engage in surveillance for co-occurrence of other developmental disorders among individuals diagnosed with ASD. However, symptom overlap between autism and related developmental disorders–in particular, Attention Deficit Hyperactivity Disorder (ADHD)–can create confusion in the assessment and diagnostic process (APA, 2013). ADHD, a neurodevelopmental condition characterized by difficulty maintaining focus or paying attention, is classified into one of three sub-types: Predominantly Inattentive Presentation, Predominantly Hyperactive–Impulsive Presentation, or Combined Presentation (APA, 2013). The Predominantly Inattentive Presentation is marked by difficulty focusing on tasks such as completing self-care activities (e.g., dressing) or school work, becoming easily bored or distracted during play, and difficulty following instructions and listening. In contrast, the Predominantly Hyperactive–Impulsive Presentation is marked by difficulty sitting still for long periods of time and controlling impulsive behavior. The Combined Presentation type is marked by co-occurrence of these two symptom profiles. While the core symptoms of ADHD can create downstream social challenges, social-communication difficulties are not a diagnostic feature in ADHD as in ASD.

An ADHD diagnosis is typically initiated in the context of primary care or educational settings (AAP, 2011), and is assigned based on a clinician’s assessment of the child’s symptoms relative to DSM-5 criteria. In many cases, psychiatrists, developmental pediatri2011cians, and neurologists may be required to aid in diagnosis and treatment management. In the US, 11% (> 1:10) of children receive an ADHD diagnosis (Visser et al., 2014). One study with a nationally-representative sample of children followed longitudinally from kindergarten to eighth grade found that, compared to WNH children, children from underrepresented groups were less likely to be diagnosed with ADHD (Morgan et al., 2013). Specifically, Hispanic children were 56% less likely, African-American children were 36% less likely, and children from Other racial and ethnic groups were 48% less likely to be diagnosed with ADHD than WNH children, and children from underrepresented groups were also significantly less likely to be using medication. This study also found that boys were two times as likely to be diagnosed with ADHD compared to girls.

As in the autistic community, racial differences in prevalence may reflect disparities in access to care rather than biological differences, in part due to the payer status of many White Non-Hispanic (WNH) families who have private insurance that enables them to access specialty care (Siegel et al., 2015). For over a decade, studies have clearly demonstrated disparities ADHD diagnosis for children of color (e.g., Mandell et al., 2007), but still these issues remain. More epidemiologic surveillance studies are needed to evaluate whether the biological prevalence of ADHD is equal across all racial/ethnic groups, or if racial/ethnic differences in reported prevalence rates are solely reflective of disparities in surveillance, assessment, and care. It is particularly important that these types of studies not rely on retrospective analysis of school- or parent-reported diagnosis, but instead, engage in equivalence trials involving systematic prospective surveillance, screening, and diagnosis.

Co-occurring Diagnoses of ASD + ADHD

According to the 2007 National Survey of Children’s Health, 67% of children with ADHD had at least one additional reported mental health or developmental condition (Larson et al., 2011). Another study reported that 70% of autistic children meet the criteria for a diagnosis of at least one psychiatric comorbidity (Abdallah et al., 2011). For example, depressive disorders are 4 times more likely among autistic individuals relative to the general population (for review, see Hudson et al., 2019). There is some overlap in the symptom profiles of ADHD and ASD, including attention processing, performance monitoring, face processing and sensory processing (for review, see Lau-Zhu et al., 2019). There is also mounting evidence demonstrating overlap of heritable genes and spontaneous mutations that may contribute to the manifestation of both disorders (Visser Rommelse et al., 2016; Kim et al., 2017; Martin et al., 2014), as well as similar abnormalities in brain structure and function (for review, see Rommelse et al., 2011).

Prior to 2013, the Diagnostic and Statistical Manual for Mental Disorders did not permit co-diagnosis of ASD + ADHD (APA, 2000). Following extensive research, the plausibility of an ASD + ADHD diagnosis was acknowledged and accommodated in the DSM-5 (APA, 2012). Because the allowance for co-diagnosis is relatively new, there is a lack of established best-practice guidelines specific to the combined presentation of ASD + ADHD. Variability in diagnostic and evaluation approaches to determining whether a co-occurring diagnosis of ASD + ADHD is appropriate likely contributes to the wide range of estimated symptom overlap found in the literature, with an estimated 28–87% of autistic children exhibiting ADHD symptoms (Mansour et al., 2017). This wide range may also be credited to discrepancies in sample size, age, ethnicity, or gender ratio (Amr et al., 2012; Mahajan et al., 2012; Pondé et al., 2010; Simonoff et al., 2008; Sinzig et al., 2009).

Intervention Approaches

Prior studies of national samples have highlighted the fact that pharmacological interventions, alone or in combination with behavioral supports or therapies, are a highly prevalent approach to management of ADHD symptoms (e.g., Visser et al., 2015; Danielson et al., 2018). For example, Visser et al. (2015) found that 74% of children with ADHD in their sample received pharmacological intervention. Only 41% of the children receiving pharmacological intervention also received behavioral supports or therapies. This result suggests that while pharmacological interventions are not the only approach to management of ADHD symptoms, they are used in the majority of cases. Preschool-aged children with ADHD and children were less likely to have received pharmacological intervention than older children, but more likely to receive behavioral therapy. Children of racial and ethnic backgrounds other than WNH were less likely to have received pharmacological intervention and more likely to receive behavioral intervention compared to their WNH peers. A later study by Danielson et al. (2018) reported similar trends, and also found that children with ADHD and a current co-occurring condition were more likely to receive pharmacological intervention than those without a co-occurring condition, and that girls were more likely than boys to receive behavioral treatment. These results suggest differences in the treatment approaches used based on age, identities, and co-occurring diagnoses, highlighting the potential risk for disparities.

Pharmacological intervention for ADHD largely consists of the use of stimulants (e.g., methylphenidate) (AAP, 2011). Stimulants are also often used as an off-label pharmacological intervention for symptoms of ASD, given the overlap between these two disorders with respect to executive function skills targeted by stimulants (Barnard-Brak et al., 2016; Karalunas et al., 2018). Pharmacological interventions for ASD and ASD + ADHD specifically target inattention and hyperactivity, stereotypical and repetitive behaviors, aggression, and self-injurious behavior (McDougle et al., 2008).

In contrast, while behavioral interventions (e.g., Applied Behavior Analysis; ABA) are not as commonly used to manage symptoms of ADHD, they are the predominant approach for ASD (Antshel et al., 2013). ABA is widely regarded by the clinical community as having the strongest empirical support for efficacy, particularly in the context of early intervention, though few randomized controlled trials have been conducted to fully explore this approach (Reichow et al., 2012). Notably, public outcry against ABA from the self-advocate community has gained traction in recent years, and emerging literature calls into question whether it should remain the dominant approach to intervention (Leaf et al., 2021). Additional evidence-based interventions are commonly implemented in educational settings, including speech therapy, occupational therapy, cognitive therapies, social skills groups, and physical therapy. Others have less empirical support, but are commonly-reported by parents as complementary therapies used to augment their child’s educational or medical intervention plan (e.g., music therapy, sensory integration therapy; Goin-Kochel et al., 2007).

Gaps in Current Knowledge

Disparities along the diagnostic and care pathway for autistic individuals have historically received little attention. A PubMed search (“disparit*[Title/Abstract] AND autis*[Title/Abstract]”) conducted in September 2020 yielded only 266 peer-reviewed papers and review articles available in the literature, with none appearing prior to 2002 and over half published in the last 5 years. Several studies have established disparities in age of diagnosis and access to early intervention among underrepresented groups (Durkin, et al., 2010; Ennis-Cole et al., 2013; Magaña et al., 2013; Nguyen et al., 2016), and a recent review of general healthcare access and quality concluded that some autistic individuals were at greater risk for socioeconomic-, racial, and gender-based disparities (Bishop-Fitzpatrick & Kind, 2017). Most recently, Hall and colleagues (Hall et al., 2020) found that autistic individuals who identified as LGBTQ + experienced disparities in care such as barriers to accessing appropriate care providers and obtaining prescriptions compared to straight, cisgender autistic individuals. In fact, nearly 40% of LGBTQ + autistic respondents reported being refused care, relative to 20% of straight, cisgender autistic respondents. Moreover, LGBTQ + autistic individuals were more likely to self-report poor mental and physical health than straight, cisgender autistic individuals.

The intersection of income and race/ethnicity also impacts early identification of neuro developmental conditons such as ASD and ADHD (Zuckerman et al., 2018). Zuckerman and colleagues found that low-income parents from racial/ethnic minority groups were less familiar with early signs or general characteristics of neurodevelopmental conditions compared to white non-Hispanic parents. They were also less-likely to report knowing a friend or family member with a neurodevelopmental condition, decreasing their opportunities for information-sharing.

These two studies highlight the importance of considering how the intersection of various identities may interact to produce different disparities in health care and health knowledge, particularly with respect to the stages of surveillance and diagnosis, where families’ knowledge of early signs and access to appropriate, responsive, and timely care is most critical. However, current work does not specifically address the potential for multiplicative effects of disparities for individuals from underrepresented or vulnerable communities who also have multiple psychological or psychiatric diagnoses (e.g., ASD + ADHD). This gap in the literature motivated the current study.

In this study, we examined intersectionality (Crenshaw, 1989), or the way that different social identities may interact with and influence one another to produce different lived experiences and outcomes. We examined both (1) interactions between different social identities and demographic characteristics (e.g., gender, race, poverty level), and (2) the interaction between these identities and diagnostic complexity (e.g., ASD vs. ASD + ADHD).

Belonging to two or more groups vulnerable to health disparities identities may place individuals with complex diagnoses at greater risk for delays in care (Alegria et al., 2011; Slopen et al., 2016). In addition to documented disparities in health knowledge and health care at the intersection of race/ethnicity and a diagnosis of ASD or ADHD (e.g., Zuckerman et al., 2016), recent studies have also noted disparities in healthcare access and quality for LGBTQ + autistic individuals (Hall et al., 2020) compared to straight, cisgender autistic individuals. This area of work highlights the potentially-detrimental effects of intersectionality at multiple stages along the pathway from surveillance to intervention. The overarching goal of this study was to understand the effect of intersectionality among diagnostic and sociodemographic variables for individuals diagnosed with ASD versus ASD with co-occurring Attention Deficit Hyperactivity Disorder (ASD + ADHD). Given the high incidence of ASD and ASD + ADHD, an analysis of potential health disparities and diagnostic factors is crucial for efficient diagnosis and effective treatments. Our specific aims were to:

1. Measure differences in aspects of the diagnostic pathway (e.g., age of first concern, age of final diagnosis, wait time) by diagnostic group (ASD, ASD + ADHD), controlling for sociodemographic factors (race, sex, poverty level);

2. Measure differences in aspects of the diagnostic pathway (e.g., age of first concern, age of final diagnosis, wait time) by diagnostic group (ASD, ASD + ADHD), controlling for various other factors including co-occurring conditions (current and ever) and type of intervention used; and

3. Combining the knowledge from Aims 1 and 2, build predictive models for the diagnostic pathway (age of first concern, age of final diagnosis, wait time), controlling for sociodemographic and other variables.

Method

Sample and Delivery

This study used data from the ADHD module of the 2014 National Survey of the Diagnosis and Treatment of ADHD and Tourette Syndrome (NS-DATA), which is a follow-back survey of a nationally drawn subgroup of respondent households from the 2011–2012 US National Survey of Children’s Health (NSCH). This survey was sponsored by the Centers for Disease Control and Prevention’s National Center on Birth Defects and Developmental Disabilities and National Center for Health Statistics. A complete description of this survey (including each item and corresponding response options) and sample population has been published elsewhere (CDC, 2015; Visser et al., 2015). Briefly, NS-DATA was a follow-up survey of respondent parents who reported that their child had ever received an ADHD diagnosis by a doctor or healthcare provider on the 2011–2012 NSCH. The response rate for NS-DATA was 47%; when combined with the 23% response rate from the 2011–2012 NSCH, the final NS-DATA response rate was 11%. There were 2966 completed interviews overall for the ADHD module of NS-DATA; however, the analyses for this study were restricted to the sample of parents with children who were currently diagnosed with ASD, based on parental report at the time of NS-DATA (n = 404).

Variables

There were three outcome variables of interest. Age of first concern, in months, was measured through the question “How old was [the child] when [you or another parent or guardian/someone at [the child]’s school or daycare/a doctor or healthcare professional not at [the child’s school/someone else] was first concerned with [the child]’s behavior, attention, or performance?”. This item was not specific to concerns for ASD or ADHD. Age of final diagnosis, also in months, was measured via the question “How old was [the child] when you were first told by a doctor or other health care provider that [he/she] had ADHD?”. Wait-time was calculated as the number of months between first concern and final diagnosis.

The primary classifier of interest was current diagnostic status of child (co-occurring ASD and ADHD or only ASD). Other potential predictors were classified according to the following broad constructs: demographic characteristics, intervention variables and co-occurring conditions. Within the demographic construct, respondent-reported variables included: respondent’s relationship to child, sex of child, first healthcare provider type to diagnose condition, race and ethnicity, marital status, and age. Under the construct of interventions, variables included therapies involving behavioral/cognitive type, social skills, nutrition and other. Within the co-occurring conditions construct, the primary classifying variable was the child’s current diagnostic status, specifically with respect to ASD and ADHD (ASD, ASD + ADHD). In addition, the child may have also had other current or past diagnoses including significant behavioral problems, developmental disability, intellectual disability, anxiety, or depression.

Respondents affirmed whether the child was of Hispanic, Latino, or Spanish origin. Respondents were permitted to refuse or respond “Don’t Know” for this item. The interviewer then read a list of categories to describe the child’s race (White/Caucasian, Black/African–American, American Indian/Native American, Alaska Native, Asian, Native Hawaiian, Pacific Islander), and the respondent selected as many as they felt were applicable. Respondents were permitted to refuse or respond “Don’t Know” or “Other” for this item. Prior to analysis, we concatenated the race and ethnicity variables, and as a result of small frequencies in some cells, we collapsed race/ethnicity into four categories: White non-Hispanic, Black non-Hispanic, Hispanic, and Other. Respondents reported their household income and size. Prior to analysis, we recoded the household income and size variables into 4 categories according to federal poverty level guidelines: at or below 100% PL, above 100% to at or below 200% PL, above 200% to at or below 400% PL, above 400% PL. Response options were “Yes”, “No”, “Don’t Know”, or “Refused” for items pertaining to diagnoses, co-occurring conditions, and intervention types. Response options were “Male” or “Female”, “Don’t Know”, or “Refused” for sex. The interviewer read a list of response options for marital status, and respondents were asked to choose one (Married, Separated, Divorced, Widowed, Never Married). Respondents were permitted to refuse or respond “Don’t Know” for this item. Free responses were solicited for the respondent’s relationship to the child, medications, and provider types. For parsimony, we have not included the specific wording of each item here, since they have been previously published elsewhere (CDC, 2015).

Multivariable regression analysis models included terms for sex, respondent’s relationship to child (family, other), race (White non-Hispanic, Black non-Hispanic, Hispanic and Other), poverty level (at or below 100% PL, above 100% to at or below 200% PL, above 200% to at or below 400% PL, above 400% PL), relationship of identifier of first concern to child (family, other), first health care provider type to diagnose (occupational, physical, and speech therapists, pediatricians and general health care providers, psychologists) and all potential current and ever co-occurring conditions as well as therapies.

Data Analysis

Initial analyses involved assessing differences between children in the ASD versus ASD + ADHD groups in terms of all variables of interest. Descriptive statistics included weighted means and standard errors for continuous variables and weighted percentages for categorical variables, using the final raked ADHD weight. It is common practice to weight models when dealing with large national survey data, in order to address potential issues related to nonresponse and representative sampling. We followed the guidelines published by the CDC for use of data from the National Children’s Health Survey (CDC, 2012).

Weighted linear regressions were used to obtain p-values related to the outcome variables of interest and their association with child’s current diagnostic status. Similarly, weighted chi-squares were used to assess associations between categorical variables and current diagnostic status. Multivariable weighted regression analysis was conducted for each of the outcome variables, and various models were considered. The following steps were taken to arrive at a final model for each outcome. After a potential predictor was entered into the model, we assessed whether it was significant and/or whether it contributed to better model fit. If it fared well on either/both counts, it was retained in the model and we added the next potential predictor and repeated the process. At the end of this process, we arrived at a model with main effects which were important with respect to the outcome, based on the above two criteria. After that, we considered all possible two-way interaction terms among these variables in the model to test for effect modification. Those interaction terms were retained that were significant and/or contributed to better model fit. Goodness-of-fit of models was assessed using the R-square statistic. For each of the above analyses, weighting was done using the final sampling weights as calculated by CDC, which utilized post stratification weighting. Statistical analysis was conducted using the SAS-callable SUDAAN software.

Results

Sample Characteristics

Descriptive statistics for the two diagnostic groups with respect to outcomes, demographic factors, intervention variables, and co-occurring conditions are presented in Table 1. The sample consisted of 404 children (weighted count = 631,075) who were primarily WNH (65%) and male (80%), and about a quarter of the sample was at or below 100% PL (24%). The average age of first concern in the sample was 40.36 months (SE = 2.32) while that of final diagnosis was 65.78 months (SE = 2.38), resulting in an average wait-time between first concern and final diagnosis of 25.54 months (SE = 1.75). There were no statistical differences between the diagnostic groups with regard to these variables. A significantly higher proportion of the ADHD + ASD diagnostic group had a previous diagnosis related to behavioral problems compared to the ASD diagnostic group (48.4% vs 14.7% respectively, p = 0.005). A higher proportion of the ADHD + ASD group reported having intellectual disability compared to the ASD group, although not statistically significant (34.4% vs 14.8% respectively, p = 0.0502). A significantly higher proportion of the ADHD + ASD group also reported depression (23%) compared to the ASD group (6%), p = 0.01. The two diagnostic groups were not different with respect to the other remaining variables.

Table 1.

Comparisons between ASD and ASD + ADHD by demographic characteristics, intervention variables and co-occurring conditions

Outcome variables	Overall (n = 404)	ASD + ADHD (n = 352)	ASD (n = 52)	p-value
	n (M, SE)	n (M, SE)	n (M, SE)
Age of first concern (months)	401 (40.36, 2.32)	349 (40.94, 2.57)	52 (35.77, 3.89)	0.27
Age of final diagnosis (months)	399 (65.78, 2.38)	349 (66.37, 2.61)	50 (61.01, 4.93)	0.34
Wait time (months)	397 (25.54, 1.75)	347 (25.55, 1.87)	50 (25.40, 4.84)	0.98
Demographic characteristics	n (%, SE)	n (%, SE)	n (%, SE)	p-value
Respondent relationship to child				0.14
Family	365 (91.77, 2.47)	318 (91.15, 2.77)	47 (96.69,1.92)
Other	37 (8.23, 2.47)	32 (8.85, 2.77)	5 (3.31, 1.92)
Sex				0.09
Male	319 (79.78, 3.36)	272 (78.29, 3.69)	47 (91.69, 6.18)
Female	85 (20.22, 3.36)	80 (21.71, 3.69)	5 (8.31, 6.18)
Diagnosing provider type				0.18
Occup./phys./speech	213 (57.17, 4.69)	183 (55.32, 5.08)	30 (72.03, 9.97)
Pediatr./gen. health care	4 (0.82, 0.50)	4 (0.93, 0.56)	0
Psychologist	163 (42.01, 4.69)	146 (43.75, 5.07)	17 (27.97, 9.97)
Race				0.34
White Non-Hispanic	299 (65.09, 4.53)	260 (65.24, 4.85)	39 (63.90, 12.79)
Black Non-Hispanic	29 (13.88, 3.21)	24 (14.07, 3.51)	5 (12.38, 7.15)
Hispanic	33 (14.39, 3.73)	28 (13.41, 3.82)	5 (22.06, 13.23)
Other Non-Hispanic	40 (6.64, 2.09)	37 (7.28, 2.34)	3 (1.67, 1.38)
Poverty level				0.18
< 100% PL	59 (24.08, 4.29)	53 (25.99, 4.70)	6 (8.00, 5.07)
100%−199% PL	74 (19.13, 3.72)	66 (19.87, 4.09)	8 (12.86, 5.78)
200%−399% PL	135 (26.45, 3.66)	119 (25.63, 3.88)	16 (33.29, 11.15)
400% + PL	120 (30.35, 4.31)	103 (28.51, 4.48)	17 (45.85, 13.02)
(Ever) Co-occurring conditions
Behavioral disability				0.005
Yes	142 (44.65, 4.65)	129 (48.44, 4.97)	13 (14.66, 6.98)
No	260 (55.35, 4.65)	221 (51.56, 4.97)	39 (85.34, 6.98)
Developmental disability				0.81
Yes	245 (67.43, 4.24)	219 (67.73, 4.62)	26 (65.08, 10.08)
No	154 (32.57, 4.24)	130 (32.27, 4.62)	24 (34.92, 10.08)
Intellectual disability				0.05
Yes	120 (32.14, 4.34)	109 (34.41, 4.76)	11 (14.73, 7.14)
No	278 (67.86, 4.34)	238 (65.59, 4.76)	40 (85.27, 7.14)
Anxiety				0.12
Yes	145 (25.29, 3.56)	135 (26.76, 3.91)	10 (13.59, 6.62)
No	259 (74.71, 3.56)	217 (73.24, 3.91)	42 (86.41, 6.63)
Depression				0.01
Yes	103 (21.13, 3.62)	95 (23.08, 4.04)	8 (6.01, 2.92)
No	298 (78.87, 3.62)	254 (76.92, 4.04)	44 (93.99, 2.92)
(Current) Co-occurring conditions
Behavioral disability				0.42
Yes	84 (94.24, 5.19)	81 (94.41, 5.23)	3 (73.21, 25.07)
No	4 (5.76, 5.19)	3 (5.59, 5.23)	1 (26.79, 25.07)
Developmental disability				0.34
Yes	213 (99.66, 0.20)	194 (99.78, 0.16)	19 (98.19, 1.75)
No	5 (0.34, 0.20)	3 (0.22, 0.16)	2 (1.81, 1.75)
Intellectual disability				1.00
Yes	110 (91.37, 5.53)	101 (91.37, 5.83)	9 (91.36, 8.37)
No	9 (8.63, 5.53)	7 (8.63, 5.83)	2 (8.64, 8.37)
Anxiety				0.91
Yes	121 (81.69, 6.87)	21 (18.44, 7.25)	1 (16.37, 15.73)
No	22 (18.31, 6.87)	112 (81.56, 7.25)	9 (83.63, 15.73)
Depression				0.57
Yes	78 (84.29, 7.79)	72 (84.79, 8.01)	6 (69.16, 24.02)
No	23 (15.71, 7.79)	22 (15.21, 8.01)	1 (30.84, 24.02)
(Current) therapy type
Behavioral/cognitive				0.16
Yes	246 (99.71, 0.19)	220 (99.69, 0.21)	26 (100, 0)
No	3 (0.29, 0.19)	3 (0.31, 0.21)	0
Social skills				0.14
Yes	222 (73.92, 4.24)	198 (76.16, 4.47)	24 (56.01, 12.95)
No	97 (26.08, 4.24)	81 (23.84, 4.47)	16 (43.99, 12.95)
Nutrition				0.99
Yes	76 (73.63, 6.41)	65 (73.65, 6.73)	11 (73.41, 20.70)
No	41 (26.37, 6.41)	38 (26.35, 6.73)	3 (26.59, 20.70)
Other				0.91
Yes	42 (66.65, 10.62)	39 (66.49, 11.14)	3 (69.53, 24.59)
No	23 (33.35, 10.62)	21 (33.51, 11.14)	2 (30.47, 24.59)

Age of First Concern

Results from a weighted ANOVA for age of first concern are presented in Table 2. Having ever experienced behavioral problems was significantly associated with an earlier age of first concern, whereas ever experiencing depression was associated with a later age of first concern. Specifically, first concern was expected to be about 11 months later in children who were ever depressed compared to those who were never depressed, controlling for all other variables in the model (p = 0.03). On the other hand, first concern was expected to be about 14 months earlier for those children who ever had significant behavioral problems, compared to those who did not (p = < 0.001).

Table 2.

Results of weighted multiple linear regression models for three outcome variables of interest

Independent Variables and Effects	Model 1: age of first concern (months)			Model 2: age of final diagnosis (months)			Model 3: wait time (months)
	Beta Coeff	SE Beta	p-value (B = 0)	Beta Coeff	SE Beta	p-value (B = 0)	Beta Coeff	SE Beta	p-value (B = 0)
Intercept	85.06	14.79	< .001	97.17	11.33	< .001	− 0.86	13.03	0.95
Diagnostic group
ASD	− 51.47	15.10	< .001	− 44.98	14.80	< .001	6.81	11.21	0.54
ASD + ADHD	< .001	< .001		< .001	< .001		< .001	< .001
Race/Ethnicity
White Non-Hispanic	< .001	< .001		< .001	< .001		< .001	< .001
Black Non-Hispanic	− 7.49	6.05	0.22	− 10.49	5.34	0.05	− 21.97	7.03	< .001
Hispanic	4.65	6.75	0.49	− 0.31	6.00	0.96	1.12	5.61	0.84
Other	− 6.52	8.84	0.46	4.87	8.50	0.57	24.79	16.48	0.13
Poverty Level
< 100%PL				< .001	< .001
100%−199% PL				− 1.50	5.59	0.79
200%−399% PL				0.08	4.89	0.99
400% + PL				8.29	5.31	0.12
Diagnostic Group X Race/Ethnicity Interaction
ASD, WNH	< .001	< .001
ASD, BNH	25.83	9.35	0.01
ASD, Hispanic	4.64	9.78	0.64
ASD, Other	24.7	11.06	0.03
ASD + ADHD, WNH	< .001	< .001
ASD + ADHD, BNH	< .001	< .001
ASD + ADHD, Hispanic	< .001	< .001
ASD + ADHD, Other	< .001	< .001
Relationship of first concerned person to child
Family	− 33.95	13.45	0.01	− 27.98	9.89	< .001	3.88	6.80	0.57
Other	< .001	< .001		< .001	< .001		< .001	< .001
Diagnostic group, relationship
ASD, Family	40.69	15.31	0.01	39.01	15.14	0.01
ASD, Other	< .001	< .001		< .001	< .001
ASD + ADHD, Family	< .001	< .001		< .001	< .001
ASD + ADHD, Other	< .001	< .001		< .001	< .001
Behavioral problems
None	< .001	< .001		< .001	< .001		< .001	< .001
Current or Prior	− 14.42	4.04	< .001	− 18.07	4.18	< .001	− 4.91	6.74	0.47
Developmental disability
None	< .001	< .001		< .001	< .001
Current or prior	− 4.1	4.36	0.35	− 7.05	4.18	0.09
Intellectual disability
None	< .001	< .001		< .001	< .001
Current or prior	7.85	5.17	0.13	7.41	5.47	0.18
Anxiety
None				< .001	< .001		< .001	< .001
Current or prior				9.04	3.95	0.02	− 0.52	5.68	0.93
Depression
None	< .001	< .001		< .001	< .001		< .001	< .001
Current or prior	10.96	4.95	0.03	13.59	4.52	< .001	19.43	5.48	< .001
Intervention
None	< .001	< .001					< .001	< .001
Intervention	− 8.66	5.64	0.12				15.8	5.81	0.01

Because wait time is calculated as a difference score between age of first concern and age of diagnosis, and because the effect of poverty level on age of diagnosis was nonsignificant in the second model, we did not include poverty level in the third model

There was also a significant interaction between race and diagnostic status. As seen in Fig. 1, the mean age of first concern was later for the ADHD + ASD group compared to the ASD group for White Non-Hispanics, Hispanics and Others. However, it was the opposite for Black Non-Hispanics. There was also a significant interaction between diagnostic status and relationship of the person who first raised concern about the child. When first concerns were reported by family members, the average age of the child was not different between the two diagnostic groups. However, when other persons outside the family reported first concerns, the mean age of children diagnosed with ASD was much earlier than the mean age of children diagnosed with ASD + ADHD.

Fig. 1.

Interaction plot between diagnostic status and race for age of first concern

We predicted that poverty level would not play a notable role in age of first concerns, because prior literature has demonstrated that low-income families often report concerns at the same time as higher-income families. Parents often report that these concerns are not taken seriously until much later, which could reflect an influence of poverty level on age of diagnosis. However, the effect of poverty level was nonsignificant. Because wait time is calculated as a difference score between age of first concern and age of diagnosis, and because the effect of poverty level on age of diagnosis was nonsignificant in the second model, we did not think it would be valuable to include poverty level in the third model.

Age of Final Diagnosis

Results from a weighted ANOVA for age of final diagnosis are presented in Table 2. A history of behavioral problems and ever having had diagnoses related to anxiety or depression were significant predictors of age of final diagnosis. In particular, holding all other predictors in the model constant, the expected age of final diagnosis for those who ever had significant behavioral problems was 18 months earlier than those who did not (p < 0.001). However, those who had ever been diagnosed with an anxiety disorder were expected to be diagnosed 9 months later compared to those who had not (p = 0.02). Similarly, those with a prior or current of depression were expected to be diagnosed about 13.5 months later compared to those who never had a diagnosis of depression (p = 0.003).

Black Non-Hispanics experienced a 10-month delay in final diagnosis compared to their White Non-Hispanic counterparts (p = 0.05). As seen in Fig. 2, there was also a significant interaction between diagnostic group and relationship of the person who first raised concern about the child. The mean age of final diagnosis was not different between the ASD and ASD + ADHD groups when first concerns were raised by family members. However, when other persons outside the family reported first concerns, the mean age of children diagnosed with ASD was much younger than the mean age of children diagnosed with ASD + ADHD at the time of final diagnosis.

Fig. 2.

Time from age of first concern to age of diagnosis varies by diagnostic group (ASD vs. ASD + ADHD) and by the relationship between a child and the person reporting first concerns

Wait Time

Results from a weighted ANOVA for wait-time between age of first concern and age of final diagnosis are presented in Table 2. Race, ever having therapeutic intervention, and having a prior or current diagnosis of depression were significant predictors of wait-time. In particular, holding all other predictors in the model constant, the expected wait-time for Black Non-Hispanics was lower compared to their White Non-Hispanic counterparts (p = 0.002); those who received any form of a therapeutic intervention were expected to have a higher wait-time of about 16 months compared to those who received none (p = 0.007); those who had a prior or current diagnosis of depression were expected to have an average wait-time of 19 months more compared to those who did not (p < 0.001).

Discussion

In this study, we assessed the impact of sociodemographic and diagnostic factors on the diagnostic trajectory of children diagnosed with ASD and ASD + ADHD. We included sociodemographic factors such as sex, race, and poverty level, and diagnostic factors such as health care provider utilization, identifier of first concerns, and additional comorbidities. Our results suggest a complex influence of sociodemographic factors on the diagnostic pathway, and a risk of health disparities in age of first concern, age of diagnosis, and wait time on children as a function of intersectionality.

Comparing Symptom Profiles

The ASD + ADHD and ASD groups differed in the incidence of behavioral problems, intellectual disability, and depression, with the ASD + ADHD group reporting higher incidence of each than the ASD-only group. Notably, risk for depression may be higher for complex cases with multiple diagnoses (23% in the ASD + ADHD group versus 6% in the ASD-only group). In all other aspects of symptomatology, the two groups were equivalent.

Symptom complexity was a significant predictor of age of final diagnosis. Specifically, the presence of anxiety and depression led to a later age of diagnosis for both groups by around 1 year relative to those without such co-occurring psychiatric conditions. This may further reflect the difficulty with differential diagnostic processes when ASD symptoms overlap with or are masked by other conditions. Conversely, the presence of significant behavioral problems led to an earlier age of diagnosis for both the ASD and ASD + ADHD groups by an average of 18 months. Children with behavioral problems may present with more severe or disruptive symptoms, and may be fast-tracked through a diagnostic pipeline in order to expedite their access to services.

This further highlights the clinical difficulty of diagnosing ASD “pure” versus ASD “plus” (i.e., with other co-morbidities); if symptom profiles overlap to great degree, it is challenging for providers to justify differentiating between these two populations.

From First Concerns to Diagnosis

For most racial and ethnic category members, the ASD + ADHD group also reported a higher mean age of first concern than the ASD-only group. This finding is in alignment with common clinical anecdotes from families that social communication and behavioral problems are often viewed first by providers through the lens of an attention disorder, and then later assessed and re-diagnosed as also related to ASD.

The mean age of first concern was lower for Black non-Hispanic children in the ASD + ADHD group than the ASD group, and statistically equivalent to that of White non-Hispanic children diagnosed with ASD + ADHD. However, the most striking disparity was for Black non-Hispanic individuals diagnosed with ASD, whose mean age of first concern was 56.09 months, nearly 2 years later than other groups diagnosed with ASD. This finding highlights an alarming health disparity when considered in conjunction with prior work also reporting a later age of diagnosis for Black children diagnosed with ASD compared to White-non Hispanic children. This may reflect differences in underlying factors related to access to care that were not measured in this dataset, for example, the myriad barriers to screening, assessment, and diagnosis that families from underrepresented groups often face. These include feeling ignored or dismissed in caregiver-provider or caregiver-educator interactions, experiencing racial/ethnic biases at points along the diagnostic trajectory, and difficulty obtaining high-quality family-centered care (Dababnah et al., 2018).

The impact of institutional racism is that minority families are often not given the tools and care necessary to seek and pursue a diagnosis once they do report a concern. In our sample, Black non-Hispanic children received a final diagnosis an average of 10 months later than their White non-Hispanic counterparts, highlighting a serious health disparity that likely delayed the availability of services for these families. It is possible that families from underrepresented racial/ethnic groups have differing awareness of autism symptoms and cultural expectations of children’s behavior (Tek & Landa, 2012; Gourdine et al., 2011; Mandell et al., 2007; Donohue et al., 2019). However, this issue of misdiagnosis or delayed diagnosis crosses cultural boundaries with regard to other vulnerable groups, for example, autistic women and girls and members of the LGBTQ + community (Hall et al., 2020; Haney, 2016). As such, it is irresponsible to lay the burden of early identification at the feet of caregivers, who consistently report being dismissed or ignored by providers (Dababnah et al., 2018). More likely, this disparity stems from a long and ongoing history of barriers to care faced by underrepresented groups, as identified nearly a decade ago by the Institute of Medicine (Smedley et al., 2002). Policies and practices in our medical and educational systems must change to prevent misdiagnosis and drastically-delayed diagnoses for children from at-risk groups.

Surveillance practices differ widely among caregivers, educators, and clinicians, and can also lead to differences in the time required to obtain a final diagnosis. When family members were the ones to report first concerns, children in the ASD and ASD + ADHD groups were identified and diagnosed at similar ages. However, others outside the family reported first concerns about children with ASD at a much lower age than children diagnosed with ASD + ADHD, and therefore this subset of the ASD diagnostic group was also diagnosed earlier than their counterparts in the ASD + ADHD group. Notably, for the ASD group, mean age of first concern was reported earlier by others outside of the family than by family members by almost a year, while for the ASD + ADHD group, family members reported concerns over 30 months later. This gap closed to some degree between first concern and diagnosis, such that when others outside of the family raised first concerns, children with ASD + ADHD were only diagnosed around 20 months later than when family members raised first concerns.

Wait Time

Contrary to our hypotheses, wait time from first concern to diagnosis did not differ significantly as a function of diagnostic group alone (ASD versus ASD + ADHD). However, it is worth noting that the mean wait time for both groups was 2 years, a staggeringly long trajectory regardless of diagnostic classification. It may be that wait times are less influenced by the specific diagnosis than by limitations on time and resources from state-level agencies that often authorize, directly fund, or provide staff for diagnostic centers (e.g., schools, county agencies, etc.), or by other factors such as the source of first concerns, as described above.

Race was a significant predictor of wait time from first concern to diagnosis, though not in the expected direction. Black non-Hispanic families had a lower wait time than their White non-Hispanic counterparts. This does not, however, suggest a lack of disparity—remember that these children received a diagnosis almost a year later than White non-Hispanic counterparts. In fact, this delay has been documented to extend well beyond a year, with a recent study finding that the average age of diagnosis for Black children with ASD was around 65 months, a striking 42 months after parents reported initial concerns (Constantino, et al., 2020). This result is in alignment with prior work suggesting a legacy of biased concern among providers with respect to Black children, where a tendency to focus on externalizing behaviors (e.g., adjustment disorder, ADHD) has been shown to take precedence over identifying developmental or socioemotional concerns, resulting in misdiagnosis that must later be revised (Mandell et al., 2007). This bias may explain why delays were noted in the ASD-only group but not the ASD + ADHD group, where externalizing behaviors may have been more prominent, leading providers to refer children sooner for comprehensive assessments that resulted in their dual diagnosis. The ultimate negative impact of delayed identification likely outweighs the benefit of a shorter pipeline from concern to diagnosis, given the amount of time lost along the developmental trajectory without intervention.

Children with depression also had a protracted wait time from first concern to diagnosis, lasting nearly 2 years longer than those without depression. While depression was a predictor of earlier diagnosis, this protracted wait time likely leads to frustration for families awaiting access to services. Further work is needed to better understand the system-level influences on wait time, and the impact of protracted wait time on families’ quality of life, psychological and emotional well-being, and access to services.

Interventions

For children in both the ASD and ASD + ADHD groups, respondents reported using behavioral and cognitive interventions almost universally (99–100%), nutrition interventions for a large number of children in both groups (73–74%), and other interventions (e.g., physical therapy, occupational therapy, speech therapy, etc.) for a large proportion of children in both groups (66–69%). We recommend exercising caution when evaluating the high instance of “other” therapies in this sample, however. This may result from grouping several less-commonly-used interventions (e.g., PT/OT/Speech) together, despite the fact that they each target a very different skill set. When these therapies are provided by schools, parents are often unaware of the precise frequency and duration of time the child spends with a therapist, which is commonly less than in an outpatient clinical setting. For example, a chld may only receive 30 min of occupational therapy per 6 week session during school, compared to a child receiving outpatient services for 45 min every week with great regularity. This same issue is applicable to physical and speech therapies, which often differ greatly between school-based and outpatient settings.

The high prevalence of social skills intervention in the ASD + ADHD group may be in part because difficulty with communication and social interaction is a core symptom of ASD. The CDC also recognizes difficulty with social interaction as a side effect of ADHD. Therefore, the high instance of this treatment for the ASD + ADHD group is relatively unsurprising.

Finally, there was a high prevalence of anticonvulsant, anxiolytic, mood stabilizer, and psychostimulant use in the ASD group. This is not especially surprising, given that these medications are routinely prescribed by doctors. However, these medications have previously been shown to have a higher instance of adverse side effects in autistic children in comparison to typically-developing children (Antshel et al., 2016; McPheeters et al., 2011). Therefore, their use should arguably be justified based on alignment with a particular profile of needs and characteristics, rather than applied as a first line of treatment for autism.

Limitations

Our analyses were limited by the variables collected as part of the DTAT instrument, the wording of items and response options in the instrument, and the sample available in the DTAT database. This resulted in a reduced ability to interpret some of the results, especially those related to interventions. Future studies would benefit from the addition of free-response items related to intervention, rather than specific prompts, which may have resulted in missing or miscategorized information that would help to contextualize children’s experiences. It is possible that respondents’ recall may have also impacted our results; however, prior work has demonstrated that recall for developmental milestones is generally accurate among caregivers of children with developmental disabilities, particularly for domains other than language onset (Majnemer & Rosenblatt, 1994). Furthermore, a small number of participants in specific race and ethnicity groups reduced our ability to thoroughly assess disparities, given low representation of some of the groups that might have been most vulnerable as a result of their intersectional identities (e.g., Black Hispanic children). We were able to partially overcome this limitation by using the weighting procedures recommended by the CDC, but future studies of disparities in the process of diagnosing ASD and ASD + ADHD must make a concerted effort to build more robust relationships with underrepresented communities in order to ensure appropriate representation in these types of databases. The age of the DTAT dataset may be a limitation, given that some changes in surveillance and diagnostic procedures may have occurred in recent years, particularly given modifications to the diagnostic criteria for ASD and guidelines regarding co-occurring conditions that were implemented in the DSM-5 (APA, 2013). Comparison of the results from this dataset and more recent iterations may shed light on what, if any, changes in these relationships may have occurred. Finally, but of great importance, there are cross-national and cross-cultural issues that warrant consideration beyond the scope of this study. We limited our discussion to predominantly studies conducted in the United States, due to the fact that the dataset used in the present study was a national sample from United States families. However, these issues are likely persistent in other countries as well, and childrens’ identities may result in different intersectional effects depending on the national and cultural context in which they are experienced.

Conclusions

Access to healthcare has been implicated as a potential determinant of both the timeliness and accuracy of diagnosis, and may contribute to underrepresentation clinically and in the research literature (Emerson et al., 2015; Flores & Tomany-Koran, 2008; Elster et al., 2003). Efforts must be made by clinicians in all lines of service, as well as researchers and community advocates, to ensure that all demographic groups are represented and have access to needed diagnostic and treatment services in a timely manner. Given the established positive impact of early intervention on functional outcomes (National Research Council, 2001), efficient diagnostic pathways are critical. With regard to policy and education systems, it is important to consider whether a one-time evaluation and diagnosis is appropriate for young children, or if parents’ concerns should be revisited multiple times across a particular window of time during development. Further research is necessary to determine what, if any, “safety net” practices could be implemented to ensure that children with co-occurring conditions are not misdiagnosed, and that children from vulnerable groups are not overlooked or treated differently as a result of institutionalized biases related to their social identity. These may include increasing representation of Black and Hispanic providers in the workforce, improving implicit bias training and bias reduction efforts within educational and healthcare systems, and including patient advocates/navigators in the diagnostic process who can act as liaisons to ensure that caregivers’ concerns are heard and acted upon in a timely manner.

Although the ASD and ASD + ADHD groups did not differ statistically in their age of first concern or age of diagnosis, both of these timepoints on the diagnostic pathway were markedly delayed. For example, a wait time of 108 months (the longest in the ASD + ADHD group) represents a significant amount of lost time during which intervention could have been applied. For mean age of final diagnosis, 65.78 months collapsed across both groups, is nearly 6 years of age–far later than the age at which autism can be reliably diagnosed (18 months; Guthrie et al., 2013), and even higher than the national average of 4 years (median age = 52 months, Baio et al., 2014). Children diagnosed after the age of 4 are less likely to receive evidence-based behavioral interventions and more likely to use psychotropic medications or complementary alternative medicine than children diagnosed early (Zuckerman et al., 2016). In other words, protracted wait times can lead families to seek non-evidence-based interventions to meet their needs in the absence of a diagnosis that allows them access to the standard of care.

Further research is critical to identify barriers to a timely and efficient diagnostic pathway for children who may meet criteria for ASD and ADHD–that is, the pathway that involves the fewest number of referrals and the shortest time to diagnosis. The symptom overlap between diagnostic criteria for ASD and ADHD, the potential presence of other behavioral and psychological symptoms, and the effect of intersectionality for patients from underrepresented groups all contribute to the risk for delays in first concern and diagnosis. Downstream, the consequence of these delays may be implementation of non-evidence-based interventions or perhaps no intervention, accommodation, or support at all, potentially impeding functional ability and quality of life.