Intersections of Autism and Sexual and Gender Diversity Among Young People in New Zealand

Study Design

This was a nationwide cohort study of participants aged 5–24 years who completed the 2023 NZ Census. These data were accessed using the Integrated Data Infrastructure (IDI), a secure research database managed by StatsNZ that links anonymised administrative and survey data across government agencies to enable population-level analysis of social, health, and economic outcomes (Milne et al., 2019).

Autism

Autism was identified using an established IDI-based case-identification method (Bowden et al., 2020) drawing on diagnosis information from three health datasets: hospitalisation (the National Minimum Dataset [NMDS]), specialist mental health (the Programme for the Integration of Mental Health Data [PRIMHD]), and disability support services data (Socrates). Individuals were identified as Autistic if a lifetime diagnosis code for Autism was present in any of the three datasets (see Appendix Table 1 for details).

Gender and Sexuality Indicators

The 2023 Census asked about gender (male, female, another gender), sex at birth (male, female), and sexuality (see Appendix Table 2 for the Census questions). Utilising responses to the sex at birth and gender questions, StatsNZ generated a categorical transgender status variable which included ‘cisgender male’, ‘cisgender female’, ‘transgender male’, ‘transgender female’, and ‘transgender other’ (see Appendix Table 3 for details of how transgender categories were derived). For this study, we collapsed these responses into a binary (transgender/cisgender) variable, with only this binary variable used in analysis. StatsNZ categorised responses to the sexuality question as ‘heterosexual’, ‘homosexual’, ‘bisexual’, or ‘sexual identity not elsewhere classified’, and the question was restricted to those aged 15 years or older; unlike the gender question, which was asked of everyone. These were also collapsed into a binary (heterosexual/non-heterosexual) variable. Both the specific responses to the sexuality question and the binary variable were used in the analyses. There were high levels of missing data for these variables including 11.8% for gender, 18.2% for transgender status, and 27.2% for sexuality. Some missing data were imputed by StatsNZ (Statistics New Zealand, 2024); however, for this study, only ‘self-reported’ 2023 Census data were used. For those aged under 15 years, parents/caregivers are responsible for ensuring forms are completed, and therefore, may have completed on the child’s behalf.

Sociodemographic Variables

Age (5–9, 10–14, 15–19, 20–24 years), ethnicity (Asian; European; Māori; Middle Eastern, Latin American African [MELAA]; Other, Pacific), deprivation level, and urban/rural profile of residence were all derived from 2023 Census data. Age was measured in years as at Census night; we chose to begin the analysis at age five, as this is consistent with the literature regarding the average age at which gender diversity is first expressed (Holt et al., 2016; Tollit et al., 2023). Ethnicity was measured using Level 1 Groupings using the total response format meaning individuals could identify with multiple ethnic groups. Deprivation level reflects area-level socioeconomic disadvantage and was derived from the New Zealand Index of Deprivation (NZDep) 2023, a composite measure based on census variables including income, employment, housing, education, and access to transport and communication. Meshblocks of usual residence were linked to NZDep scores, which were collapsed into quintiles, where one represents areas of the least deprivation, and five indicates areas of the highest deprivation. Similarly, meshblocks were linked to the 2023 StatsNZ urban/rural classification and measured as a binary urban/rural indicator.

Statistical Analysis

All analyses were undertaken using Stata/MP 19.5 within the IDI. Initially, a descriptive summary of the study population was presented, stratified by Autism status. This was followed by observed and standardised rates of Autism by gender, transgender status, and sexuality. Standardised Autism rates were estimated using logistic regression models that adjusted for age, ethnicity, deprivation, and urban/rural variables. Standardisation was then applied using the margins command to generate adjusted prevalence estimates for each group, holding the covariate distribution constant across the population.

Community Involvement

Our research recognises the importance of engaging meaningfully with both the Autistic and SGD communities, with a strong emphasis on collaboration and co-production of knowledge. The authorship team includes multiple contributors from the SGD community, from the Autistic community, and individuals who identify as both Autistic and SGD, ensuring these perspectives are directly represented throughout the research process.

Limitations

This method for identifying formally diagnosed Autistic people in the population remains unvalidated, so the extent to which it identified false positives and false negatives is unknown. Moreover, it relies on health service use data, which does not include diagnosis information from primary care or outpatient settings. Thus, it likely undercounts the true population of diagnosed Autistic people in NZ. It also does not include those who have not been able to access a formal diagnosis of any kind. However, this is an established method used extensively in the literature (Bowden et al., 2022; McLay et al., 2025; Mujoo et al., 2023; Schluter et al., 2025; Vu et al., 2024). In addition, the use of Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV)-era diagnostic codes may result in the inclusion of conditions that are not part of the current DSM-5 classification of Autism spectrum disorder. In particular, Rett syndrome, previously classified within Pervasive Developmental Disorder (PDD) categories but no longer considered part of Autism, may have been captured, introducing a potential source of misclassification.

Our analyses relied on 2023 Census data to identify key variables, including those relating to gender and sexuality. One limitation of this is that parents are responsible for ensuring census forms for under-15s are filled out; they are not required to fill out the form on their child’s behalf, but many do. Some children will have filled out their forms themselves, and some would have had parents do it for them. Thus, there is likely to be an undercount of the SGD population in cases such as those where a young person is not yet ‘out’ to their family. In addition, we chose to restrict this analysis to only ‘self-report’ responses in the 2023 Census to ensure optimal reliability of these variables, that is, we did not include those that were imputed by StatsNZ. However, by limiting to self-reported data, this increases the extent of missing data and may limit the generalisability of these findings to the whole NZ population. The 2023 Census was the first time sexuality and ‘another gender’ options had been included in a NZ Census. The lack of a historical time series for these data, and an absence of other sources of whole-population data for validation, means we know little about the accuracy of the data and how well they represent the SGD community in NZ.

The findings of this report show there is a need for future research on the intersections of Autism and SGD identities. This report serves as a starting point to show the prevalence of intersecting Autistic and SGD identities. Further research looking at the demographics of those who sit at this intersection is needed, as well as other research that focuses on, for example, the healthcare and educational needs and outcomes of this group. Our findings show transgender people and those with less-common sexual orientations (i.e. those in the ‘sexuality not elsewhere categorised’ group) were more likely to be Autistic than any other group we looked at. Future research that investigates the experiences of these groups specifically is encouraged to ensure equity of outcomes.