Prior Diagnoses and Age of Diagnosis in Children Later Diagnosed with Autism

Autistic people are at increased likelihood for medical and psychiatric co-occurring conditions (Rosen et al., 2018), and more research is needed to understand the ways that this impacts diverse autistic populations, and age of autism diagnosis. Given the commonality of co-occurring diagnoses, Mandell et al. (2007) published a study using Medicaid data to examine diagnoses children received before autism. The data included children living in urban Philadelphia in the late 1990s, who were diagnosed with Autistic Disorder in the DSM-IV (APA, 1998). More than 56% of the children received a diagnosis other than autism on their first mental health visit, most commonly ADHD. Race impacted outcomes; Black children were five times more likely to receive a diagnosis of adjustment disorder compared to ADHD, and more than 2.5 times more likely to receive a diagnosis of Conduct Disorder (Mandell et al., 2007). Boys were also more likely than girls to receive conduct disorder. Possible explanations authors suggested included symptom presentation, parent report, or clinician bias (Mandell et al., 2007). Research with more recent data points that includes a more expansive examination of gender, and geographic region would improve the field’s understanding of these trends. This paper focuses on updated data (20 years after Mandell), and can capture new trends in race, gender, age, and criterion of spectrum diagnosis, as well as neurodevelopmental concerns such as attention deficit/hyperactivity disorder (ADHD) and oppositional defiant disorder (ODD), mood and anxiety diagnoses, adjustment disorders, intellectual disability, as these are commonly diagnosed in childhood and co-occurring with autism.

Significant research has investigated the symptomatic and genetic/ environmental risk factor overlap among ADHD, ODD, conduct disorder (CD), and autism (Geluk et al., 2012). Co-occurrence between autism and ADHD is estimated at approximately 30% (Rosen et al., 2018). Symptoms of inattention, such as lack of eye contact, or fine motor stereotypy (which can be viewed as fidgeting), can be particularly difficult to differentiate between the two disorders. Positive correlations among autism, ADHD, ODD, and CD symptoms have been found in both boys and girls (Kerekes et al., 2014).

Anxiety is considered the most common co-occurring diagnosis with autism (Lai et al., 2019). This may present as social anxiety, generalized anxiety, or may be environmentally determined (e.g., related to sensory concerns or routine disruptions). Little research exists on adjustment disorders, and their differentiation from healthy or typical reactions to stress are unclear (Casey & Bailey, 2011). Adjustment diagnoses are commonly used to secure referrals for other services, when clinicians are not comfortable making a diagnosis (Chomienne et al., 2011). Relatedly, researchers have also found support for a connection between autism and obsessive–compulsive disorder (OCD), particularly in restricted and repetitive behaviors and preferences for routines, which can be diagnostic across both of these labels (Jiujias et al., 2017; Pazuniak & Pekrul, 2020). There is also evidence for shared genetic and etiological origin (Meier et al., 2015).

An intersectional approach to research in autism elucidates the ways in which race, biological sex, geographical location, cognitive ability, and many other factors combine to impact the lives of diverse autistic people. Autistic Black children are diagnosed later, despite parents reporting concerns at similar times to their White counterparts (Constantino et al., 2020). The intersectional identity as both Black and disabled means research indicates increased risk of difficulty securing coordinated care, obtaining diagnosis, or feeling mistreated by providers (Bishop-Fitzpatrick & Kind, 2017; Broder-Fingert et al., 2020; Cascio et al., 2020). Furthermore, gold-standard measures for autism evaluation, have also been criticized for their lack of cultural relevance to diverse populations (Kim et al., 2013; (ADI-R; validation sample was 4% Black, 77% male); Pugliese et al., 2015 (ADOS-2; validation sample was 6.4% Black, 77% male); Ronkin et al., 2021).

Females are diagnosed with autism at lower rates than males, approximately 4:1; the ratio is reduced to 2:1 when looking at children with intellectual disability (Hull & Mandy, 2017). The field debates whether this discrepancy is due to a female protective effect, or whether girls are being under-diagnosed due to bias in diagnostic measures (Hull & Mandy, 2017; Kreiser & White, 2014). This study will look at biological sex as a factor in diagnosis and uses female or girl in reference to biological sex at birth.

Families of autistic children living in rural areas report less access to specialized providers, intervention options, and longer waitlists (Antezana et al., 2017; Mello et al., 2016; Zhang et al., 2017). To manage distance-to-care burdens, rural schools can become a primary resource for families, however, this can mean delays in services (Antezana et al., 2017). Living in more rural areas is also related to higher rates of intellectual disability co-occurrence (Palmer et al., 2010). Socioeconomic status can interact with race and rurality to impact families’ abilities to access evaluations, coordinated care, educational supports, early intervention services, and support through adulthood. Research suggests that higher SES is associated with higher rates of autism, and research participation in autism research studies likely due to diagnostic resources (rather than true prevalence differences; Cascio et al., 2021; Durkin et al., 2017).

Delayed diagnoses of autism reveal inequities in access to diagnostic evaluations and healthcare supports. This study examined age at autism diagnosis as well as previous diagnoses received by autistic children enrolled in Missouri Medicaid. The first aim of this study was to examine how age of diagnosis of autism from 2 to 10 years old in Missouri differs based on sociodemographic factors. It was hypothesized that Black children would have a delayed diagnostic age compared to White children, girls would have a delayed diagnostic age compared to boys, children living in rural areas would have a delayed diagnostic age compared to those living in an urban setting, and number of prior diagnoses would be associated with later diagnosis.

The second aim was to examine prior diagnoses assigned to children who later receive autism diagnoses and whether race, sex, or geographic location impact those diagnoses. The diagnostic categories were chosen based on common childhood and developmental diagnoses, common autism comorbidities, and capture a sampling that are diagnosed across genders. It was hypothesized Black children would be more likely to receive prior diagnoses of conduct disorders compared to the larger population, girls would be more likely to receive anxiety- and obsessive–compulsive (OCD) related disorders compared to boys, and children in rural regions would be more likely to receive diagnoses of ADHD and adjustment disorders compared to children in urban settings. The authors would stratify by race, geography, and sex to look at differences in prior diagnoses as an exploratory hypothesis.

Participants were Medicaid enrollees with DOBs between Jan 2005 and Nov 2018 and who received an autism diagnosis between Sept 2015 and Feb 2020 (to prevent skew in the data due to the Covid-19 pandemic). In late 2014, Medicaid converted from ICD-9 to ICD-10; to keep the diagnosis as controlled as possible, we only included children diagnosed in ICD-10 system (F84.0 Autistic Disorder). Children with claims data for at least one year prior to autism diagnosis were included in analyses (See Fig. 1 for accounting of data cleaning procedures; n = 13,850). Only White and Black identified participants were included in the analyses of the study (n = 12,241), as Missouri Medicaid did not provide ethnicity to researchers, and other racial subgroups samples were too small to draw conclusions.

This study explored how multiple elements of a child’s identity (race, sex, disability, and geographical region) may impact their access to and experience within the medical system. Missouri Medicaid covers adults and children who are categorized as low income (i.e., for a household of four, income before taxes below $35,245 qualifies), and those with a disability. Medicaid eligibility was included in initial data, and families that qualified because of a pre-existing disability were excluded.

Data security procedures and protocol were approved by University of Missouri-St. Louis Institutional Review Board. Participants were autistic children diagnosed in the Missouri Medicaid system, and de-identified data was provided to the researchers. To maintain confidentiality, age in weeks at autism diagnosis and prior diagnoses was tabulated before data was received by authors.

The study was pre-registered at Open Science Framework (Registration DOI Retracted for Blinding). Preliminary analyses checked t-tests, correlations, and ANOVA testing; whether age at diagnosis was related to each of the predicted dummy-coded variables (sex [1 = male], geography [1 = urban], race [1 = White], disability [1 = had Medicaid code of intellectual disability pre-autism diagnosis]).

Aim 1 used a blocked linear regression with sociodemographics and number of prior diagnoses as predictors and age of autism diagnosis as the outcome. For aim 2, the a priori plan was to use a single multinomial logistic regression with the specific prior diagnosis as the outcome, with geography, race, and sex as predictor variables. A large proportion of children had a high diagnostic load (19.1%), and the authors did not want to force children into one diagnostic category, but rather capture their multiple diagnoses. Thus, the plan was switched to running six individual logistic regressions, with post-hoc explorations of multiple overlapping diagnoses. The outcomes in logistic regressions were the presence (1) or absence (0) of the following: ADHD diagnosis, conduct related diagnosis, adjustment and related diagnosis, anxiety and related diagnosis, mood and related diagnosis, and intellectual disability diagnosis. The Benjamini–Hochberg False Discovery Rate (Benjamini-Hochberg, 1995) correction was used to protect against multiple comparisons inflation of type 1 error, adjusting for the 22 regression findings between aims 1 and 2. In exploratory analyses, trends within race, sex, and geographical locale were assessed based on patterns in the descriptive analyses. Chi-square distribution testing with a Benjamini–Hochberg False Discovery Rate corrected significant difference testing was used to assess post hoc analyses.

Tables 1 and 2 show descriptive statistics of the sample (n = 12,241). The average age of diagnosis was 362.35 weeks (6.9 years), the median was 355 weeks (6.8 years), and the mode was 156 weeks (3.0 years). White children were diagnosed older than Black children (7.05 years old vs. 6.55; p < 0.001), and among White participants, no significant differences in average age of diagnosis were found between urban and rural populations (t(10,232) = − 1.53, p = 0.126, 95% CI [− 11.08, 1.36]; Cohen’s d = − 0.030), nor between male and female (t(10,240) = 1.21, p = 0.225, 95% CI [− 2.86, 12.17]; Cohen’s d = − 0.029). Among Black participants, no significant differences in average age of diagnosis were found between urban and rural (t(1,996) = 0.05, p = 0.959, 95% CI [− 22.099, 23.28]; Cohen’s d = 0.004). There were differences based on sex, t(1996) = 2.082, p = 0.037, 95% CI [1.04,34.89]; Cohen’s d = 0.115, such that Black girls were diagnosed significantly younger than Black boys. See Fig. 2, and Supplemental Table 1, for intersectional sociodemographic comparisons.

Post-hoc and exploratory analyses were initiated to investigate findings in cases that violated apriori predictions (Fig. 3). Urban White females and males appear similar with prior diagnostic rates of 36.34% and 36.55%, but look different than Urban Black populations, where males had a prior diagnostic rate of 27.86% and females had a prior diagnostic rate of 20.63%. Both White males and females had significantly more prior diagnoses than their counterparts, Black males, X² (1) = 34.99, p < 0.001, phi = 0.081, and Black females, X² (1) = 29.83, p < 0.001, phi = 0.14, respectively.

The differences in conduct disorders were more closely examined in the urban subsample because they violated apriori hypotheses. Rates of conduct disorders in Black and White children in urban samples were significantly different, X² (1) = 32.59, p < 0.001, phi = 0.093, a small effect size, with White children having higher rates of diagnosis than Black children. In the urban male sample, racial differences in diagnostic rates of conduct disorder were significant, X² (1) = 22.96, p < 0.001, phi = 0.065, a small effect size, with White males being diagnosed at significantly higher rates than Black males. The differences in ADHD were similarly examined because they impacted a large segment of the participants and were contrary to hypotheses, with White males having significantly higher rates than Black males.

Having no prior diagnoses before the autism diagnosis was associated with being Black and living in an urban area. More prior diagnoses were associated with later autism diagnosis. Increased diagnostic load is likely an indicator of a complex presentation (Casanova et al., 2020). It could also be related to access to medical care, with more diagnoses reflecting more visits to doctor’s offices, and referrals or medical orders in the healthcare system. Race was a significant predictor of being diagnosed with every prior diagnosis, aside from intellectual disability, likely indicating more diagnostically complex autistic Black children are not included in this sample as they are not yet diagnosed with autism, hence the under-representation by racial group noted earlier. This might reflect how providers view children, as well as beliefs families hold about child development (i.e., less diagnostic-seeking behaviors at early signs, more tolerance for variability in child development).

Being White was associated with having an anxiety disorder, adjustment disorder, and mood disorder. Broder-Fingert et al. (2013) found racial differences in treatment of autistic children, specifically, Black children had fewer appointments and evaluations from psychiatrists. While autism is considered primarily biological in nature (Parellada et al., 2014), anxiety and mood disorders feature a combination of environmental and biological precedents (Schiele & Domshke, 2018). One explanation is that Black families are a protective factor for neurodiverse children, indeed, many protective factors have been hypothesized through cultural and family structures including resilience, social support, feelings of efficacy, extended family network support and more (Liu et al., 2017; Taylor et al., 2015). Again, the under-representation of Black children in the sample complicates interpretations of this finding.

Similar rates of ADHD co-diagnosis (24%) as Rosen and colleagues were found (30%; 2018). In the ADHD model, children who were White, males, and living rurally were all more likely to be diagnosed with ADHD. Significant research supports racial disparities in diagnosis of ADHD, wherein Black children are diagnosed less despite similar symptoms (Winders-Davis et al., 2021). ADHD is commonly diagnosed by pediatricians or family practitioners, who children living rurally have more access to (Committee on Quality Improvement, Subcommittee on Attention-Deficit/Hyperactivity Disorder, 2000). Similarly, being white and living rurally were both related to being diagnosed with conduct disorder. This could be because school providers are more comfortable offering this diagnosis (then reported to family practitioners, entering the medical record to support prescriptions or referrals). Boys and girls did not have different rates of diagnosis, in line with prior research which indicates rates of conduct disorders are similar across genders (Rowe et al., 2010).

The relationships among race, sex, and geographic region in ADHD and conduct disorders warranted post-hoc analyses. When looking at urban males, White boys had significantly more diagnoses than Black boys, and the pattern held with urban females. Prior research has found disparities in ADHD, with Black children having significantly lower rates of diagnosis and medication (Coker et al., 2016; Shi et al., 2021; Winders Davis et al., 2021). While biological/genetic indicators may cause disparities between male and female prevalence (Merikangas & Almasy, 2020), racial disparities are hypothesized to be influenced by environmental factors such as provider bias or structural racism (Moody, 2016; Winders Davis et al., 2021). This could include parent beliefs about symptoms, cultural influences on co-occurrence (Slobodin & Masalha, 2020), or cultural insensitivity in diagnostic processes and racism in schools (Kang & Harvey, 2020; Moody, 2016). These same factors may have influenced the under-representation of Black children in this Medicaid sample.

The data on anxiety, mood, and adjustment disorders have many similarities. Our study found lower rates of anxiety co-occurrence (around 10%) than is established in the literature (Lai et al., 2019; Rosen et al., 2018). In the model of anxiety disorders, sex was not a significant predictor, so apriori hypotheses were not supported. That being said, Lai et al. (2011), also found similar rates of anxiety, depression, and OCD symptoms in male and female autistic participants, so there is some precedent in the literature for these findings.

Mandell et al. (2007), had very different findings. The prior study was conducted over 25 years ago, in an urban center on the east coast, with no analysis of geographical influences and limited gender analysis. It categorized children in the prior diagnosis they had received the most, rather than across multiple diagnostic categories, and found that Black children were more likely to be diagnosed with another condition before autism. There are a few reasons why our findings may not match the earlier data. First, autism diagnostic rates in general, and in particular among people of color, have increased dramatically (Nevison & Zahorodny, 2019). The average age of diagnosis has gotten younger, while the DSM-5 put autism on a spectrum, meaning that a broader group of people fall under the umbrella (APA, 2013). Changes to the healthcare system mean that more people have healthcare in the US as compared to the 1990s, allowing for more access to services. This paper’s significant updates to prior findings highlight the importance of replication in the field of clinical psychological research (Fletcher, 2021).

Intellectual disability diagnostic rates in this data (3.7%) were much lower than they are hypothesized to be, which is around 36% (Maenner et al., 2023). Children may be carrying ‘developmental delay’, or other similar terms, in medical records, and will later be diagnosed with intellectual disability. Furthermore, children who were brought on Medicaid with a diagnosis of intellectual disability may have been cut during data cleaning procedures which excluded participants who came on Medicaid due to a disability.