Estimating the Prevalence of Autism Spectrum Disorder in New South Wales, Australia: A Data Linkage Study of Three Routinely Collected Datasets

Using three routinely collected data sources in NSW, most cases of ASD were ascertained from disability services data, with children more likely to be male, younger at first contact, live outside major cities and from more disadvantaged areas compared with cases identified in hospital or ambulatory data alone. Based on these data sources, we estimated the prevalence of ASD to be 1.1% by age 6 and 1.3% by 12 years of age. Overall, ASD prevalence among 6-year-old children increased an average of 4.1% per year over the study period. Prevalence rates were 3–4 times higher among male children but increased faster among female children.

The estimated 1.3% ASD prevalence from this study is higher than the most recent published population-level estimate from Western Australia, 0.5% of children born 1993–2005 (Bourke et al., 2016). That estimate was based on data from the Intellectual Disability Exploring Answers (IDEA) database, which identifies cases of intellectual and developmental disabilities from receipt of disability services and educational supports in schools (Petterson et al., 2005). The estimate from the current study is consistent with a more recent study of data on new autism treatment plans from Medicare, Australia’s universal healthcare program, which reported a cumulative incidence of 1.1% in 4-year-olds (May & Williams, 2018). Medicare reimbursement for autism treatment plans covers those with an autism diagnosis as well as those diagnosed with other pervasive developmental disorders. However, the prevalence estimate from the current study is lower than the most recent Australian estimates based on parental surveys, including an ongoing cohort study in Victoria (2.6–4.4%) (May et al., 2020) and the national Survey of Disability, Ageing and Carers (2.8%) (Australian Bureau of Statistics, 2019). It is also lower than the 2.3% prevalence among 8-year-old children in 2018 reported by the ADDM network, the main active autism surveillance program in the US (Maenner et al., 2021).

Compared to all births in NSW, children identified with ASD were more likely to live in regional and more disadvantaged areas. A previous study in Western Australia found ASD was associated with higher socioeconomic status (Leonard et al., 2011), but the association between socioeconomic disadvantage and ASD is complicated with cohort studies in different populations reporting contradictory results (Delobel-Ayoub et al., 2015; Durkin et al., 2017; Kelly et al., 2019; Roman-Urrestarazu et al., 2021; Yu et al., 2021). In the current study, disability services data only captured individuals who were receiving publicly funded services. A 2016 survey of Australian healthcare providers reported substantial wait times for autism diagnostic assessments, particularly in the public sector, which had a median wait of 16 weeks from time of referral up to a maximum of two years (Taylor et al., 2016). A study of referral and assessment pathways at a tertiary hospital in Melbourne reported children waited an average of 9 months for an ASD assessment (Bernie et al., 2021). Individuals with more resources may choose to avoid the public system and receive services from private providers (Kayrouz et al., 2014). In addition, regional differences in prevalence estimates may be shaped by differences in access to and availability of services.

As expected, we found that ASD prevalence increased over the study period, however the 4.1% average annual increase is slower than previous studies in Australia (May et al., 2020; Nassar et al., 2009) and the United States (Nevison & Blaxill, 2017; Zablotsky et al., 2019; Baio et al., 2018; Maenner et al., 2021). For example, a recent study of two Australian birth cohorts born four years apart (2000 vs. 2004), reported a 69% increase in prevalence of ASD from 2.6 to 4.4% at age 12–13 (May et al., 2020). The authors suggest that the dramatic increase was likely caused by changes in diagnostic criteria and the introduction of new funding from the Helping Children with Autism (HCWA) program, for which only their younger cohort was eligible. In July 2008, the Australian government introduced the HCWA program, which provided funding for early intervention services to registered children with an ASD diagnosis aged six years or younger (Bent et al., 2015). To receive these resources, an ASD diagnosis may have been emphasized rather than other neurodevelopmental conditions or intellectual disability (Croen et al., 2002; Nassar et al., 2009; Shattuck, 2006). In contrast, we found the prevalence of ID in disability services data decreased substantially over the course of the study period. This was unexpected and this decrease has not been reported in prior studies from Western Australia (Bourke et al., 2016; Nassar et al., 2009). This decrease is likely due in part to structural changes in funding for disability services in Australia (May et al., 2018; Australian Institute of Health and Welfare, 2018).

Comparing the three data sources used in this study, there was only limited overlap in individuals identified with three-quarters of cases found in disability services data alone. Limited identification from hospital admission data is expected, given that, while children with ASD are more likely to have comorbidities and to be admitted to hospital in early childhood (Alexeeff et al., 2017; Atladóttir et al., 2010; Brooks et al., 2021), the inpatient environment is not a primary site of treatment for neurodevelopmental disorders. A previous linked data study in New South Wales reported only 16% of individuals identified with intellectual disability were identified in hospital admission data alone, with the rest identified in disability services data (Reppermund et al., 2017). Similarly, a study in Western Australia found that hospital admission data poorly identified individuals with intellectual disability when compared to disability services and education data (Bourke et al., 2018).

Limited identification from ambulatory mental health data is more surprising, given the high level of mental health comorbidities experienced by individuals with autism. However, there are multiple barriers for individuals with ASD to receive mental health services, including discomfort and limited knowledge about ASD among some mental health practitioners (Adams & Young, 2021). In addition, the dataset used in this study included only one diagnosis code per encounter and did not include visits with general practitioners and private specialists. A validation study in Canada reported low sensitivity (17%) when identifying ASD from mental health outpatient data, with most cases identified in physician billing data (Dodds et al., 2009). Based on these results, hospital admission data and ambulatory mental health contacts alone do not sufficiently capture cases of ASD in New South Wales. However, it is important to link the three sources together because the 13% of children identified outside of disability services may represent a distinct population. Older at first contact, these children may have less severe autism symptoms or may lack access to services. Future studies should investigate this population to identify opportunities for improving access to autism services.

This study has several strengths including use of population-level data from three different service types - inpatient, outpatient, and community services. To our knowledge, this is the first study to estimate ASD prevalence in NSW using data linkage, and, while not complete in their coverage, these datasets together likely represent the most comprehensive data currently available in NSW. Under-enumeration may be possible, either because ASD cases were never diagnosed, or because they only received services from other sources, such as private providers and school accommodations. This lack of a “gold standard” prevented calculation of sensitivity and specificity metrics for each data source.

Previous validation studies in Australia, (Bourke et al., 2018) North America, (Burke et al., 2014; Dodds et al., 2009) and the Nordic countries (Lampi et al., 2010; Lauritsen et al., 2010) found that diagnosis codes for ASD in hospital admission and ambulatory mental health records produce few false positives. This was true when using medical record abstraction (Burke et al., 2014; Lauritsen et al., 2010), diagnostic interviews (Dodds et al., 2009; Lampi et al., 2010), or registries (Bourke et al., 2018) as a gold standard. As such, cases identified in hospital admission and ambulatory mental health data in NSW are expected to be true cases. Identifying cases from health and disability data sources can also reduce bias and over-reporting that may result from parental reporting.

Based on three routinely collected data sources, 1.3% of children born in NSW developed ASD by age 12. Identifying cases of ASD is important for evaluating prevalence and the health and wellbeing of children in a population. In this study we found disability services data were the best single source for identifying children with ASD in New South Wales, while hospital admissions or ambulatory mental health data alone should not be used for this purpose. However, if possible, all three data sources should be linked together to ensure children diagnosed at an older age and children without access to services are included.