Prevalence of autism among adults in Canada: results from a simulation modelling study

Abstract

Objective

To estimate the prevalence of autism among adults living in Canada.

Design

A Monte Carlo simulation modelling approach was employed. Input parameters included adult population estimates and mortality rates; autism population all-cause mortality risk ratios; and autism prevalence estimates derived from child and youth data due to the lack of adult data. This approach was executed through 10 000 simulations, with each iteration generating a distinct data scenario. Prevalence estimates were reported as the mean with the 2.5th and 97.5th percentiles, corresponding to a 95% simulation interval (SI).

Setting

Where possible, Canadian data sources were used, including the 2019 Canadian Health Survey on Children and Youth and Statistics Canada mortality rates and population estimates.

Primary outcome measure

National prevalence estimates of autistic adults living in private dwellings in Canada, with variations in prevalence by sex at birth and province/territory considered.

Results

The findings suggest the prevalence of autism among adults in Canada to be 1.8% (95% SI 1.6%, 2.0%). National prevalence estimates by sex at birth were 0.7% (95% SI 0.6%, 0.9%) for females and 2.9% (95% SI 2.6%, 3.2%) for males. Provincial/territorial estimates ranged from 0.7% in Saskatchewan (95% SI 0.3%, 1.3%) to 3.6% in New Brunswick (95% SI 2.4%, 5.1%).

Conclusions

The limited availability of data on autistic adults constrains our ability to fully understand and address their unique needs. In this study, autism prevalence was estimated based on diagnosed cases, which excludes individuals without a formal diagnosis. Additionally, other factors such as data availability and methodological assumptions may influence the modelling of prevalence estimates. As a result, our findings should be interpreted within the context of these limitations. Nevertheless, this study provides a valuable reference point for understanding autism prevalence among adults in Canada.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Prevalence estimates of autism among adults in Canada were derived and reported by sex at birth and by province/territory.

• A dynamic Monte Carlo simulation approach was used, allowing for seamless updates as new data becomes available.

• Due to the limited data on autistic adults, multiple data sources were employed, which may introduce unique methodological challenges.

• Prevalence estimates are based on diagnosed cases, likely underestimating the true prevalence of autism, particularly among groups less likely to receive a diagnosis, such as females.

• Given the dearth of Canadian data on the risk of mortality among autistic people, we relied on pooled all-cause mortality rate ratios from a systematic review of non-Canadian studies, which may not accurately reflect the mortality burden of autistic people in Canada.

Introduction

Autism, also known as autism spectrum disorder, is a lifelong neurodevelopmental condition. Autistic people may communicate and connect with others differently, have sensory processing differences or focus intensely on particular interests or activities.^{1 2} Autistic people may also have other physical, intellectual, learning or mental health conditions, which can introduce further complexities and challenges.²,⁵ In 2019, approximately 1 in 50 (or 2.0%) children and youth (<18 years of age) in Canada were estimated to have an autism diagnosis.⁴ Of these, just over half (53.7%) were diagnosed before age 5.

Autism data are mainly focused on children, as autism tends to be diagnosed during the developmental period.^{4 6 7} However, autism is a lifelong condition, and many autistic adults continue to need ongoing support while experiencing disparities in employment, social relationships and overall well-being.^{8 9} The derivation of estimates in adults is hindered by a national and global scarcity of population-based studies that quantify the number of autistic adults.⁸¹⁰,¹³ This lack of data limits our understanding of this population’s needs, experiences and the broader implications for public health and policy decisions.

We sought to address this data gap, specifically the need for national and provincial/territorial estimates of diagnosed adult autism prevalence in Canada. We used a simulation modelling approach inspired by the work of Dietz and colleagues.¹⁴ Dietz et al used US national data on children to derive autism prevalence, adjusting for higher mortality rates among autistic people and extrapolating this estimate to national and state adult populations. Their study estimated that 2.2% (95% simulation interval (SI) 2.0%, 2.5%) of US adults 18–84 years old have autism, with the estimated prevalence among adult males being four times higher than that among adult females: 3.6%, (95% SI 3.1%, 4.0%) vs 0.9% (95% SI 0.6%, 1.1%), respectively. Moreover, there were noteworthy variations in prevalence at the state level.

Mirroring Dietz et al methodology, our analysis aimed to implement a modelling strategy using data on children/youth to estimate the prevalence of diagnosed autism among adults in Canada ≥18 years of age, focusing on variations by sex at birth and province/territory. This contribution will help to advance population health research and surveillance by providing initial baseline estimates and addressing an important knowledge gap.

Methods

Data sources and input parameters

Data sources and respective input parameters are outlined in table 1.

Table 1. Input parameters used to simulate autism prevalence estimates.

Input	Time frame	Source/data available	Description and limitations
Autism population all-cause mortality risk ratios	2010–2016	Catalá-López et al (2022)17:Pooled all-cause mortality risk ratio among males: 2.09 (1.50, 2.92);Pooled all-cause mortality risk ratio among females: 4.87 (3.07, 7.73)	Pooled all-cause mortality risk ratios from a systematic review, comparing the risk of all-cause mortality rates among autistic individuals to the general population by sex Included studies were heterogeneous in terms of characteristics (eg, age), design, settings and outcomes* Pooled all-cause mortality risk ratios were based on eight studies from the US, Sweden, Finland and Denmark†. No Canadian studies were included in the systematic review Of these eight studies reporting mortality risk ratios, seven adjusted for age‡
Mortality rates	2019	Statistics Canada mortality rates19 by province/territory, age and sex	Population-level data by province/territory, age and sex (for population estimates and mortality rates) In 2019, the estimated Canadian population was 37 601 230 Population sizes were 18 918 505 for females and 18 682 725 for males Population-level estimates are available for the general adult population only; no autism-specific estimates were available.
Population estimates	2019	Statistics Canada population estimates18 by province/territory, age and sex
Autism prevalence(ages 1–17)	2019	Canadian Health Survey on Children and Youth (CHSCY)15 prevalence estimates	Prevalence by province/territory, age and sex A cross-sectional, complex design survey that collects information on the health and well-being of children/youth using both web-based and paper and pencil methodologies Children/youth were identified as having autism if the person most knowledgeable (PMK) reported ‘Yes’ to the question: “Has(child/youth) been diagnosed with any of the following long-term conditions? Autism spectrum disorder, also known as autism, autistic disorder, Asperger’s disorder or pervasive developmental disorder” Long-term conditions, as defined by CHSCY, last 6 months or more and have been diagnosed by a health professional Excluded from the survey were children and youth living on First Nation reserves and other Aboriginal settlements in the provinces, and children and youth living in foster homes and institutions Given that the PMK reports this data, there is potential for under-reporting or over-reporting of autism.

^*Studies were eligible for inclusion if they were observational, on persons with autism according to standard operationalised diagnostic criteria and reported mortality risk ratios comparing individuals with autism to the general population/those without autism, or provided sufficient data allowing for their derivation.

^†The ages of these study cohorts are described in table 1 of Catalá-López et al¹⁷ Briefly: one study included participants who were mostly ≥4 years at first evaluation; another study reported a mean age of 7.5 years at admission and 43.4 years at follow-up; one study assessed participants in early childhood at admission, with a mean age of 33.2 years at follow-up; another had a mean age of 10.8 years at diagnosis and 35.8 years at follow-up; another reported a mean age of 19.8 years at diagnosis; another included participants mostly >9 years at diagnosis, with a median age of 19 years at death; one study had a mean age of 33.1 years at evaluation; and the eighth study reported a mean age of 8 years at diagnosis.

^‡Summarised in online supplemental etable 8 of Catalá-López et al.¹⁷

We used multiple sources to derive required estimates for analyses, including:

• Canadian Health Survey on Children and Youth (CHSCY)¹⁵ 2019 autism prevalence estimates among children/youth 1–17 years. The CHSCY is a national, cross-sectional survey administered by Statistics Canada that collects health data on children/youth living in private dwellings. The 2019 survey achieved a response rate of 52.1%. Excluded from the survey’s coverage are children and youth living on First Nation reserves and other Aboriginal settlements in the provinces, and children and youth living in foster homes and institutions. Data were collected through electronic questionnaires and phone interviews, spanning 11 February 2019 to 2 August 2019. The primary survey respondent was the person most knowledgeable in the household, typically a parent. Autism was assessed through the following question: “Has [child/youth] been diagnosed with any of the following long-term conditions? Autism spectrum disorder, also known as autism, autistic disorder, Asperger’s disorder, or pervasive developmental disorder”. The full unweighted sample size was 39 951, with 819 children and youth having been diagnosed with autism. Prevalence estimates were weighted to represent children/youth 1–17 years of age, and the bootstrap method was used to calculate 95% CIs. Additional information about CHSCY is available through Statistics Canada.^{15 16}

• Pooled all-cause mortality rate ratios among males and females with autism as compared with the general population from a systematic review by Catalá-López et al.¹⁷ These rate ratios were based on studies from the USA, Sweden, Finland and Denmark. No national Canadian studies were available.

• Statistics Canada 2019 population¹⁸ and mortality¹⁹ estimates by province/territory, age and sex at birth.

Modelling approach

We applied a Monte Carlo simulation modelling approach to estimate the prevalence of autism among adults. We first derived the prevalence of diagnosed autism among children/youth (1–17 years of age) based on the 2019 CHSCY (2.0%, 95% CI 1.8%, 2.2%).¹⁵ For plots comparing the proportions of children/youth with autism living in Canada and age at the time of survey completion, and the frequencies of children/youth with autism living in Canada and age at the time of diagnosis based on data from the 2019 CHSCY, see online supplemental figure S1A, respectively.

Anticipating that prevalence estimates may be unreliable for groups (province/territory and sex at birth) with less than ten respondents (eg, for Yukon, Northwest Territories, Nunavut), we assumed case numbers in these groups to approximate the national average. We proposed that the SE in these regions would equate to 50% of the prevalence of autism, signifying a high uncertainty and a large coefficient of variation of 50%. See online supplemental box S1 for a summary of modelling assumptions.

Subsequently, we adjusted prevalence estimates using the pooled all-cause mortality rate ratios (males – 2.09; 95% CI 1.50, 2.92; females – 4.87, 95% CI 3.07, 7.73) among autistic adults, compared with the general population.¹⁷ We assumed this relative mortality rate to remain constant among all age groups. Finally, to account for differences in the overall population distribution, we adjusted prevalence using population estimates for province/territory, age (in 1-year intervals, 18–89 years) and sex at birth.¹⁸ Algorithms corresponding to these data steps and adapted from Dietz et al¹⁴ are displayed in online supplemental figure S2.

We applied Monte Carlo simulation²⁰ to estimate adjusted autism prevalence rates, stratified by province/territory and sex at birth. We recalibrated the all-cause mortality rate ratio in each iteration by drawing a value from a normal distribution characterised by specific mean and SE values. This method allowed us to simulate a feasible degree of stochastic variability. For every province/territory-sex at birth subgroup, the initial computation of autism prevalence was based on a presumed beta distribution, with alpha and beta parameters derived from the empirical prevalence and its associated SE.

Next, within each age category (in 1-year intervals, 18–89 years of age) of a given province/territory-sex at birth subgroup, we derived (a) the autism mortality rate, calculated as the product of the group’s all-cause mortality rate ratio; (b) the prevalence count of autism cases, using the group’s population and autism prevalence; (c) a recalibrated autism prevalence to incorporate the anticipated probable impact of autism survival rate on prevalence and (d) the revised count of autism cases within the population, achieved by multiplying the group’s population with the adjusted autism prevalence. This process was reiterated across all age groups (1-year intervals, ages 18–89) within each province/territory and sex at birth. We amalgamated the outcomes from each simulation, encompassing the computed metrics and the associated province/territory, sex at birth, age group and population data. This approach was executed through 10 000 simulations, with each iteration yielding a distinct data scenario. We summarised prevalence estimates as the mean with the 2.5th and 97.5th percentiles, corresponding to prevalence estimates with a 95% SI.²¹ We used Python V.3.10 (Python Software Foundation) and R V.4.3 (R Foundation for Statistical Computing) for analyses. We conducted and reported this study according to STRESS-DES (Strengthening the Reporting of Empirical Simulation Studies Discrete-event Simulation) guidelines²² (online supplemental table S1).

Sensitivity analyses

We conducted sensitivity analyses using the same Monte Carlo simulation modelling approach to ascertain the impacts of potential female underreporting and historical variation in autism prevalence. Two alternative scenarios were examined: (a) an upward adjustment of the prevalence of diagnosed autism among females aged 1–17 years in the 2019 CHSCY by a factor of 4/3²³ and (b) downwards adjustments to the prevalence of diagnosed autism among all children and youth aged 1–17 years, as reported in the 2019 CHSCY, by factors of 2 and 4.²⁴,²⁶ In relation to scenario (a), Canadian reports estimate an overall male-to-female autism prevalence ratio of approximately 4:1.²⁴,²⁶ However, a systematic review from 2017 suggests that the true ratio may be closer to 3:1, indicating potential underreporting or underdiagnosis among females.²³ To account for this, a correction factor of 4/3 was applied to female prevalence estimates. With regard to (b), reductions in overall autism prevalence were implemented to explore the effects of historically lower diagnosis rates, which have been reported in prior Canadian studies²⁴,²⁶

Patient and public involvement

Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Results

Results for autism prevalence are presented in figure 1 and table 2 (national and provincial/territorial estimates) and table 3 (national and provincial estimates among adults by sex at birth).

Figure 1. Prevalence of autism among adults with 95% simulation intervals (SI) by Canadian jurisdiction and sex at birth. N.L., Newfoundland; N.W.T, Northwest Territories; P.E.I., Prince Edward Islands.

Table 2. National and provincial/territorial autism prevalence estimates among adults in Canada.

	Prevalence (95% SI)*
Canada	1.8% (1.6%, 2.0%)
Province/territory	-
Alberta	1.9% (1.3%, 2.4%)
British Columbia	2.0% (1.5%, 2.5%)
Manitoba	1.5% (0.9%, 2.3%)
New Brunswick	3.6% (2.4%, 5.1%)
Newfoundland and Labrador	2.0% (1.2%, 3.0%)
Northwest Territories	1.9% (0.7%, 3.7%)
Nova Scotia	1.3% (0.7%, 2.1%)
Nunavut	1.9% (0.7%, 3.7%)
Ontario	1.9% (1.7%, 2.1%)
Prince Edward Island	3.2% (2.3%, 4.4%)
Quebec	1.5% (1.1%, 2.0%)
Saskatchewan	0.7% (0.3%, 1.3%)
Yukon	1.8% (0.7%, 3.7%)

^*Derived using simulation strategies proposed by Dietz et al,¹⁴ see Methods section of the manuscript and Figure S2online supplemental figure S2.

SI, simulation interval.

Table 3. National and provincial/territorial autism prevalence estimates among adults in Canada by sex at birth.

	Sex at birth	Prevalence (95% SI)*
Canada	Female	0.7% (0.6%, 0.9%)
Canada	Male	2.9% (2.6%, 3.2%)
Province/territory	-	-
Alberta	Female	0.8% (0.4%, 1.4%)
Alberta	Male	2.9% (2.0%, 3.9%)
British Columbia	Female	0.7% (0.4%, 1.1%)
British Columbia	Male	3.3% (2.3%, 4.4%)
Manitoba	Female	0.4% (0.1%, 0.8%)
Manitoba	Male	2.7% (1.5%, 4.2%)
New Brunswick	Female	0.9% (0.3%, 2.0%)
New Brunswick	Male	6.4% (4.1%, 9.2%)
Newfoundland and Labrador	Female	0.3% (0.1%, 0.7%)
Newfoundland and Labrador	Male	3.7% (2.0%, 5.8%)
Northwest Territories	Female	0.7% (0.2%, 1.5%)
Northwest Territories	Male	2.9% (0.8%, 6.5%)
Nova Scotia	Female	0.5% (0.1%, 1.4%)
Nova Scotia	Male	2.2% (1.1%, 3.5%)
Nunavut	Female	0.7% (0.2%, 1.6%)
Nunavut	Male	3.0% (0.8%, 6.5%)
Ontario	Female	0.7% (0.5%, 0.9%)
Ontario	Male	3.1% (2.7%, 3.6%)
Prince Edward Island	Female	0.7% (0.2%, 1.5%)
Prince Edward Island	Male	5.9% (4.0%, 8.1%)
Quebec	Female	0.8% (0.4%, 1.3%)
Quebec	Male	2.3% (1.6%, 3.1%)
Saskatchewan	Female	0.5% (0.1%, 1.2%)
Saskatchewan	Male	1.0% (0.4%, 2.0%)
Yukon	Female	0.7% (0.2%, 1.5%)
Yukon	Male	2.9% (0.8%, 6.4%)

^*Derived using simulation strategies proposed by Dietz et al,¹⁴ see Methods section of the manuscript and Figure S2online supplemental figure S2.

SI, simulation interval.

We estimated the national prevalence of autism among adults in Canada to be 1.8% (95% SI 1.6%, 2.0%). Estimates by sex at birth were 2.9% (95% SI 2.6%, 3.2%) for males and 0.7% (95% SI 0.6%, 0.9%) for females, with prevalence among males approximately four times higher than females. We observed considerable variation between provincial/territorial estimates (table 2), ranging from 0.7% in Saskatchewan (95% SI 0.3%, 1.3%) to 3.6% in New Brunswick (95% SI 2.4%, 5.1%). Provincial/territorial prevalence estimates (table 3 and figure 1) among males and females were both highest in New Brunswick (males—6.4%, 95% SI 4.1, 9.2%; female—0.9%, 95% SI 0.3, 2.0%).

Results from sensitivity analyses are summarised in online supplemental tables S2 and S3. When adjusting for potential female underreporting (assuming a male-to-female prevalence ratio closer to 3:1 rather than 4:1), the revised national autism prevalence among adults in Canada was 1.9% (95% SI: 1.7%, 2.2%), with an adjusted prevalence for females of 1.0% (95% SI: 0.6%, 1.5%) (online supplemental table S2). On reducing prevalence estimates by a factor of 2, the national autism prevalence among adults was 0.9% (95% SI: 0.6%, 1.3%), with revised estimates of 1.5% (95% SI: 0.9%, 2.3%) for males and 0.4% (95% SI: 0.2%, 0.6%) for females (online supplemental table S3). When reducing prevalence estimates by a factor of 4, the national autism prevalence among adults was 0.5% (95% SI: 0.3%, 0.7%), with adjusted estimates of 0.7% (95% SI: 0.4%, 1.1%) for males and 0.2% (95% SI: 0.1%, 0.3%) for females (online supplemental table S3).

Discussion

We conducted a simulation analysis to approximate national and provincial estimates for autism prevalence among adults in Canada by sex. The prevalence estimates generated from this study are a first for Canada and address a critical data gap, illustrating the relevance of a simulation approach in such a context. Our national estimates resembled those of Dietz et al,¹⁴ showing a prevalence of 1.8% in the Canadian setting compared with 2.2% in the USA. In addition, the corresponding sex ratios in both studies indicated a four times higher prevalence in males than females. We also observed considerable variability in prevalence estimates among different provinces/territories. These estimates can help advance public health research and surveillance by offering a reference for estimating the prevalence of autism among adults in Canada.

The sex ratio findings in this study are well substantiated^{4 5 12 14}; however, it is important to acknowledge that females tend to be underdiagnosed or diagnosed later, compared with males.^{723 27},²⁹ While biological factors may contribute to a lower autism risk in females,²⁸ diagnostic criteria have largely been shaped by traits observed in male populations, making it more challenging to identify autism in females.²⁹ To address these differences, Hull et al have proposed a female autism phenotype, which captures sex-specific variations in the manifestation of autistic behaviours.²⁷ This includes the tendency for females to present internalising problems, elevating the risk of co-occurring conditions such as anxiety, depression, self-harm and eating disorders.²⁷³⁰,³² These overlapping conditions can obscure autistic traits, making diagnosis more difficult. Additionally, females are more likely to engage in social camouflaging, making detection through standard diagnostic procedures more challenging and increasing their risk of mental health difficulties.³⁰ This underdiagnosis among female children and youth would correspond with a lower reporting of diagnosed autism among females in the CHSCY, which would have been perpetuated in our simulation results. To account for this, we conducted a sensitivity analysis to examine how higher-than-reported female autism prevalence estimates would influence our model outputs. Furthermore, the variation observed between provincial/territorial estimates needs to be contextualised within a broader understanding of regional variations concerning availability and access to autism diagnostic assessment services.⁸ Residents of rural and remote areas especially face increased susceptibility to jurisdictional differences in the availability and access to autism diagnostic assessments,^{8 33 34} highlighting a need for further research in this area.

Limitations

Several limitations warrant consideration. First, using multiple data sources can introduce unique methodological limitations, as outlined in table 1. Second, we assumed that the prevalence of diagnosed autism remained relatively stable over time, while in fact, the prevalence of autism in Canada has been increasing since 2003.³⁵ This can be attributed, at least in part, to greater public awareness and more comprehensive screening efforts. A key piece of evidence supporting the role of improved screening, diagnostic and reporting efforts is the growing number of individuals being diagnosed with autism in adulthood.³⁶,³⁹ Given that autism is a lifelong condition that is typically detected in early childhood, it is reasonable to infer that many of these adults would likely have been diagnosed earlier had the screening and diagnostic tools of previous decades been as advanced as they are today. Thus, our approach assumes that autism diagnosis rates among adults in Canada would be similar to those of children/youth if they had received the same level of diagnostic coverage. This approach does not account for potential environmental or gene–environment interactions over time, as these remain challenging to quantify. This assumption aligns with Dietz et al,¹⁴ who noted that differential exposure to risk factors is unlikely to drive substantial variations in prevalence across birth cohorts. As the limited evidence available indicates an increase in diagnosed cases over time in Canada, we conducted sensitivity analyses—reducing prevalence estimates by factors of 2 and 4—to explore potential variations in prevalence under scenarios of significant temporal variation in prevalence. Third, given the absence of Canadian data on the risk of mortality among autistic people compared with the general population, we used pooled all-cause mortality rate ratios from a systematic review based on studies from the USA, Sweden, Finland and Denmark,¹⁷ which may not reflect the actual mortality burden experienced by autistic people in Canada. Notably, pooled all-cause mortality rate ratios were higher among females than males, and it is uncertain whether these sex differences in mortality rates are generalisable to the Canadian population. The underlying reasons for these disparities are also unclear: do they reflect true differences in mortality risk, or are they influenced by the higher likelihood of underdiagnosis among females?⁴⁰ Additionally, the greater prevalence of co-occurring conditions and internalising behaviours in females may further contribute to elevated mortality risk.⁴¹,⁴⁷ We also assumed these rate ratios to be uniform across provinces/territories and age groups. This approach mirrored that undertaken by Dietz et al,¹⁴ who assumed a constant standardised mortality ratio across age groups and US states.

Conclusions

This study generated simulated data that can serve as a foundational point of reference for estimating the prevalence of autism among adults in Canada. Further, this study used innovative modelling techniques in the Canadian context, and our estimates can be updated or refined as new data becomes available. However, acknowledging the intrinsic limitations of simulated estimates, there is a continued need for data collection, public health surveillance and research efforts directly involving the autistic adult population.

Language statement

We endeavoured to use language to describe autism that is preferred by the autistic community. Hence, based on recommendations of the Autism Alliance of Canada,⁴⁸ this manuscript uses identity-first language when referring to autistic people.

Data availability statement

Data may be obtained from a third party and are not publicly available.