Co-occurring Autism Spectrum Disorder and Gender Dysphoria in Adolescents

This study is a secondary data analysis of a large electronic medical record (EMR) dataset formed by the PEDSnet learning health system network of 8 pediatric hospital institutions across the United States (Children’s Hospital Colorado, Children’s Hospital of Philadelphia, Cincinnati Children’s Hospital Medical Center, Nationwide Children’s Hospital, Nemours Children’s Health, Seattle Children’s Hospital, Stanford Children’s Health, Ann and Robert H. Lurie Children’s Hospital of Chicago).¹⁸  Patients were included in the analysis if they (1) were 9 to 18 years old and (2) had ≥2 inpatient or outpatient encounters at a PEDSnet member institution between 2009 and the data extraction date (March 2022), with at least 1 completed visit in the 18 months before the data extraction date. Variables included age on the data extraction date, sex, ethnicity, race, and health insurance type category (private, public, or other) as documented in the patient’s EMR, median household income quintile (first = lowest, fifth = highest), which was identified by matching the 2020 American Community Survey 5-year estimates of median household income to zip code data extracted from the EMR, and the PEDSnet institution. EMR-reported sex represented the sex assigned at birth for most patients and was used to be consistent with previous PEDSnet research.⁵  The “other” health insurance category included any insurance type that was not listed as private or public (e.g., self-pay, charity). Analyses included descriptive statistics and adjusted logistic regression models, with the outcome being any presence of a GD diagnosis and the main predictor being any presence of an ASD diagnosis (see Supplemental Table 7 for included diagnosis codes). The main regression analysis included the full sample of youth and was adjusted for age, EMR-reported sex, household income quintile, and health insurance type, and modeled PEDSnet institution as a random effect to account for clustering at the health system level. Marginal effects were used to identify the predicted probability of receiving a GD diagnosis among those with and without an ASD diagnosis when holding all other covariates constant. We then repeated this analysis with an interaction term for each demographic characteristic to identify specific differences in the association between ASD and GD diagnoses among demographic groups and again calculated marginal effects to assess the magnitude of the differences in the probability of a GD diagnosis among youth with an ASD diagnosis by demographic characteristics (eg, Hispanic vs non-Hispanic, public vs private insurance). The Bonferroni correction was used in these demographic analyses to account for the family-wise error rate (Type I) when conducting multiple comparisons, with the P value set at .05.

All study procedures were reviewed and determined to be not research involving human subjects by the Seattle Children’s Institutional Review Board.

Among the 919 898 patients meeting inclusion criteria, the mean age was 13.6 years (SD = 2.6). In the full sample, 50.8% were reported as male in the EMR, 15.9% identified as Hispanic ethnicity, 55.2% identified as White race, 40.3% used private insurance, and 32.5% used public insurance. Overall, 4.4% (n = 40 249) had an ASD diagnosis alone, 0.5% (n = 4925) had a GD diagnosis alone, and 0.05% (n = 464) had both ASD and GD diagnoses (Table 1). GD diagnosis was more prevalent among youth with an ASD diagnosis compared with youth without an ASD diagnosis (1.1% vs 0.6%). In addition, youth with cooccurring ASD/GD were older (15.4 vs 13.6 years), more likely to be reported as female in the EMR (62.7% vs 49.2%), identify as White (78.0% vs 55.2%), and use private insurance (48.7% vs 40.3%), and less likely to identify as Asian (2.2% vs 4.9%) or Black (2.4% vs 16.9%) compared with all other patients in the sample.

Mixed logistic regression revealed 3 times greater adjusted odds (aOR) of GD diagnosis among youth with an ASD diagnosis compared with the odds among youth without an ASD diagnosis (95% confidence interval [CI]: 2.72–3.31; Table 2), and an average predictive margin (probability) of a GD diagnosis of 0.019 (95% CI: 0.013–0.024) among those with an ASD diagnosis compared with 0.006 (95% CI: 0.005–0.008) among those without an ASD diagnosis (z = 6.53, P <.001). Results of the interaction models of GD (Table 3) indicated a significant interaction between ASD and EMR-reported sex for females compared with males (aOR =1.77, 95% CI: 1.45–2.16) and between ASD and insurance type for those using public insurance compared with private insurance (0.65, 95% CI: 0.52–0.81). Post-hoc analyses comparing the magnitude of the differences in effects using predictive margins and pairwise comparisons further indicated additive interactions (see Tables 4–6 and Fig 1) such that, among those with an ASD diagnosis, youth whose EMR-reported sex was female were more likely to have a GD diagnosis compared with youth whose EMR-reported sex was male (predictive margin [probability]: 0.034 vs 0.007, z = 6.53, P <.001), Asian youth were less likely to have a GD diagnosis compared with non-Asian youth (0.008 vs 0.019, z = −3.94, P <.001), Black youth were less likely to have a GD diagnosis compared with non-Black youth (0.004 vs 0.020, z = −6.14, P <.001), White youth were more likely to have a GD diagnosis compared with youth of color (0.023 vs 0.012, z = 4.92, P <.001), and youth who used private insurance were more likely to have a GD diagnosis compared with youth who used public insurance (0.027 vs 0.012, z = 5.20, P <.001). No significant differences among median household income quintiles emerged (Table 6).

The results of this study support those of previous research with clinical samples indicating significant cooccurrence rates of ASD and GD in the US adolescent population.^1–5  We also identified significant demographic differences in these cooccurrence rates by EMR-reported sex, race, and insurance type, indicating potential disparities in ASD and GD diagnosis and service use. Overall, these results can be used to inform the development and implementation of strategies to reduce these disparities and improve health outcomes for youth with cooccurring ASD and GD.

Our results supported our first hypothesis that youth with an ASD diagnosis would be significantly more likely to have a GD diagnosis compared with those without an ASD diagnosis, and are also consistent with previous research indicating a significant cooccurrence of ASD and GD in pediatric populations.^1–6  However, the prevalence of GD among youth with ASD across the 8 institutions in our study was notably lower than that previously identified in pediatric neuropsychology clinic populations (1.1% vs 5.4%).⁴  Such differences may be attributable to various factors, including smaller samples used in previous studies, or because our study used a broader population of youth seen in hospital systems compared with previous research conducted in specialty clinics in which youth are actively seeking care. In addition, much of the previous research has focused on traits of ASD or gender dysphoria/diversity which could overestimate prevalence, whereas we relied on diagnosis codes, which likely leads to underestimates. Nevertheless, these results add further evidence for enhanced screening in both ASD and gender clinic settings, as well as specialized programs and support services for youth with both ASD and GD and their families.

Our second hypothesis focused on demographic differences in the cooccurrence of ASD and GD diagnoses was partially supported. Regarding sex, we found that youth with an ASD diagnosis whose EMR-reported sex was female were more likely to have a GD diagnosis compared with youth with an ASD diagnosis whose EMR-reported sex was male. Although the existing literature is mixed on whether there are differences by sex in ASD/GD cooccurrence,¹⁹  research reveals that ASD diagnoses are more common among cisgender males than cisgender females, which has been attributed to diagnostic criteria that are biased toward male-typical behaviors.⁸  Consequently, this could suggest that youth who were assigned female at birth but exhibit male-typical behavior²⁰  and/or identify as male²¹  may be more likely to be diagnosed with ASD. Given research indicating that individuals assigned female at birth tend to be undiagnosed, misdiagnosed, and be diagnosed with ASD later,⁹  further research focused on the order of and age at which youth with cooccurring ASD/GD receive their ASD and GD diagnoses is warranted to further understand these differences by sex or gender.

Regarding ethnicity/race, results revealed no significant differences in cooccurring diagnoses between Hispanic and non-Hispanic, nor between multiracial and single-race youth. However, we did observe that youth of color with ASD were significantly less likely than White youth to have cooccurring diagnoses, and this was particularly the case for Black and Asian youth. Although little research has focused on ethnic and racial differences in cooccurring ASD and GD because of small sample sizes, our results are consistent with demographic differences and disparities in access to ASD and gender-affirming care services independently.^{7,10,12,14,17}  Importantly, both ASD and GD can impact youth and/or caregiver comfort in seeking health care services and provider perceptions of these youth, which can be further exacerbated by cultural norms, implicit biases, and structural racism in the health care setting.²²  Therefore, systems may need to be designed to address these biases to effectively engage youth of color with cooccurring ASD and GD and their families to meet their unique health care and support needs.²³ 

For analyses focused on socioeconomic status, we did not observe differences by median household income quintile but did find that youth with an ASD diagnosis who used private insurance were more likely to have a GD diagnosis compared with those who used public insurance. This could be related to the correlation between income and insurance coverage.²⁴  These findings could also be related to variations in insurance coverage based on health care needs and diagnoses and may be particularly relevant among youth with ASD who qualify for public insurance (e.g., Medicaid) but live in a state in which gender-affirming care is not explicitly covered by public insurance plans.²⁵  These results reveal the need to design systems that address biases and highlight the importance of insurance coverage for services associated with these diagnoses to ensure that youth with cooccurring ASD and GD receive the care they need.

This study should be interpreted within the context of several limitations. First, although using the large PEDSnet dataset afforded us the opportunity to conduct further subgroup analyses that have not been possible in past research, we were still limited in our ability to understand the associations between these 2 relatively rare diagnoses among smaller racial subgroups. In addition, our inclusion criterion requiring patients to have been seen at a member institution at least once in the 18 months before March 2022 may have introduced further bias because of disparities in access to health care during the COVID-19 pandemic.²⁶  Future research should, therefore, continue to use large, representative datasets to better understand potential disparities in diagnoses and access to care, and to design appropriate screening and intervention services for these groups of youth.

Relatedly, it is important to acknowledge that many factors can influence whether patients receive timely and accurate diagnoses of ASD and GD, all of which may exacerbate health disparities. First, it is likely that there are patients in our sample who were misclassified because of undiagnosed ASD and/or GD, leading to an underestimate of the true prevalence. This could be related to the fact that PEDSnet diagnosis data are extracted from billing codes and problem lists, but do not pull information from the patient’s past medical history or clinician notes. Moreover, there could be misclassification among those whose diagnoses do not appear in administrative data within these health care systems, whether due to receiving care elsewhere, inadequate documentation, or intentional decisions not to record diagnoses that may lead to stigma or discrimination.

In addition, our results may be affected by the quality and accuracy of demographic information in the EMR. For example, we used the sex reported in the EMR for our analyses to be consistent with previous PEDSnet research,⁵  which is likely the sex assigned at birth for most patients; however, there are some patients for whom this could have been changed to reflect their affirmed gender identity. More work is needed to appropriately classify individuals in this dataset. Similarly, youth who identify as gender-diverse but have not received a GD diagnosis, whether because they are not seeking gender-affirming medical care or because of difficulties in accessing such care, would have been excluded from our analysis. This may not only lead to an underestimate of gender diversity in our sample but also could mask further disparities in access to gender-affirming care. Similar concerns exist for the ethnicity and race data in the EMR, as a large body of research has revealed that ethnicity and race data are rarely self-reported and often inaccurate in the EMR, particularly for Hispanic, Asian, and Indigenous populations.²⁷  Thus, future research using EMR data should prioritize the use of self-reported gender identity and ethnicity/race information to better understand how intersectional identities and inequities related to racism, sexism, and transphobia can influence care access and health outcomes.

These results represent an important step in understanding the intersectionality of identities among youth with cooccurring ASD and GD, and further highlight the need for and design of services that address compounding disparities in access to diagnostic services and treatment among youth with cooccurring ASD and GD and other marginalized identities. In summary, our study adds to the growing body of research on cooccurring ASD and GD and provides important evidence to support additional research, improved screening for ASD and gender diversity, and care integration and coordination in specialty clinics and in underserved communities.

The research reported in this publication was conducted by using PEDSnet, a National Pediatric Learning Health System, and includes data from the following PEDSnet institutions: Children’s Hospital Colorado, Children’s Hospital of Philadelphia, Cincinnati Children’s Hospital Medical Center, Nationwide Children’s Hospital, Nemours Children’s Health, Seattle Children’s Hospital, Stanford Children’s Health, and the Ann and Robert H. Lurie Children’s Hospital of Chicago. Thank you to the PEDSnet Data Coordinating Center for providing us with these data. Thank you to Daksha Ranade and Victoria Soucek for their assistance in accessing and preparing the data for analysis.

aOR

adjusted odds ratio

ASD

autism spectrum disorder

CI

confidence interval

EMR

electronic medical record

GD

gender dysphoria

SES

socioeconomic status