Autism epidemiology in Hong Kong children and youths aged 6–17: Implications on autism screening and sex differences in the community

Abstract

Epidemiological studies on autism lack representation from Asia. We estimated the prevalence of autism among children and youths in Hong Kong using a two-stage approach. In addition, we evaluated the psychometric properties of the screening instrument and explored sex differences within an epidemiological context. A random school-based sample of 5,865 children and youths were screened with the Autism Spectrum Quotient-10 (AQ-10). Then, a subsample of 317 participants underwent the Autism Diagnostic Interview-Revised assessment. Prevalence was estimated by applying positive and negative predictive values (PPV/NPV) of AQ-10 derived from the subsample to the entire cohort. None of the screened negative participants had autism, resulting in an NPV of 100%. Discrepant PPVs were noted for males (20.4%) and females (5.20%). The estimated prevalence was 2.57% using sex-specific PPVs. Explorative analysis on AQ-10 Positive participants without the diagnosis (i.e. ‘false positives’) showed significantly elevated autistic symptoms. The prevalence of autism in Hong Kong is comparable to the recent estimates in Western countries, which poses a significant public health challenge. Despite the high false-positive rates, AQ-10 remains valuable for excluding autism and identifying those with autistic symptoms. Furthermore, community-based studies are crucial to address sex differences in autism expression.

Lay abstract

Although studies have found that autism is becoming more common, little is known whether this is true in Asian countries. This study looked into how many children and teenagers in Hong Kong might have autism. We first screened 5,865 school-aged children and youths with the Autism Spectrum Quotient-10 (AQ-10). Then, we conducted in-depth interviews with 317 of them to assess for autism. We found that around 2.57% of children and youths aged 6–17 years in Hong Kong might have autism. This number is similar to that of the Western countries. Furthermore, we also discovered that boys who were screened positive on the AQ-10 were more likely to have autism than girls. This could be because autism is more common in boys, and the AQ-10 might be better at spotting autism in boys than in girls. In the future, studies will need to find an optimal way of detecting autism in the community, considering how autism may present differently in boys and girls.

Keywords

Asian autism epidemiology prevalence screening sex differences

Introduction

Historically, autism was once considered a rare condition predominantly affecting males and accompanied by intellectual disability. However, with the reconceptualization of the condition as a spectrum and increased awareness, the condition is currently not as rare as previously considered. Indeed, a concurrent rising trend in the prevalence of autism reported worldwide has been speculated to reflect an actual increase (Zeidan et al., 2022). Epidemiological studies serve as a first step in delineating factors contributing to the prevalence of autism.

Since the revision of the DSM-5 diagnostic framework in 2013, the median prevalence of autism reported by studies across the globe was 1%, and the prevalence reported varies in the range of 0.01%–4.36% (Zeidan et al., 2022). However, existing epidemiological studies on autism were largely based on Caucasian populations, with an underrepresentation of those from Africa and Asia. To comprehensively understand the global variations in autism prevalence, regional studies across diverse populations based on standardized case-identification approaches and contemporary diagnostic criteria are essential (Hussain et al., 2023). A recent study conducted on 10,000 children aged 6–10 years in three metropolitan areas in China revealed a prevalence rate of around 1%, which was at the lower end of the global range (Sun et al., 2019). Hong Kong Special Administrative Region (Hong Kong SAR) of China is a unique community among all Chinese metropolitan regions, where a Westernized socioeconomic infrastructure is docked within a culturally-ethnically Chinese-predominant society. Before 2008, the prevalence of autism in Hong Kong was estimated to be 0.161% for youth under 15 years of age, using the DSM-4 diagnostic criteria and a clinic-registry approach (Wong & Hui, 2008).

Other than the regional differences, methods of case identification can also contribute to variations in prevalence estimates. For instance, registry-based approaches that use single-stage case identification could underestimate the true prevalence owing to the presence of undiagnosed cases in the community. On the other hand, community-based studies might capture the ‘unregistered’ cases. Recognition of cases outside the registry is especially relevant to the contemporary conceptualization of autism, which includes individuals with optimal levels of functioning and do not require active support from the medical or educational systems (Harstad et al., 2023).

Given the complexity of autism diagnosis, large-scale community-based epidemiological studies, including the present study, often adopt a multi-stage design with screening followed by confirmation of diagnosis. The accuracy of the screening and diagnostic instruments is therefore of paramount importance. While studies often adopted screening instruments with established sensitivity and specificity, their positive and negative predictive values (PPV and NPV) were seldom examined when applied at the general population level. Psychometric studies often utilize a case-control design with a balanced proportion of cases and non-cases. The PPV and NPV yielded from such studies would not be applicable to the general population, as these values are heavily influenced by case prevalence, and the prevalence of most disorders is much lower in the general population. Furthermore, the differences in the instrument’s psychometric properties between males and females warrant greater scrutiny. As differences in phenotypic presentation between autistic males and females have been described (Hull et al., 2020), the properties of the instruments theoretically would be expected to behave differently between sexes. Although cumulative data thus far have failed to support sex-specific scoring and algorithms for commonly used instruments, the existing studies were limited by ascertainment bias. This bias stemmed from diagnoses being established by the same diagnostic tools under examination, leading to a circularity of argument in the absence of alternative diagnostic benchmarks such as multi-source validated measures (Kaat et al., 2021). Therefore, a community-based study, such as the present study, provides a rare opportunity to examine the psychometric properties of instruments in real-life situations where the whole spectrum of neurodiversity and its intersection with sex are present. For example, a recent study showed that there were sex differences in the psychometric properties of the Modified Checklist for Autism in Toddlers (M-CHAT-R/F) when applied to a community sample of toddlers (Eldeeb et al., 2023).

The present study reports a systematic investigation of the prevalence of autism in Hong Kong among children and adolescents, using the DSM-5 diagnostic criteria and a two-stage approach on a random epidemiological sample of the entire territory. Furthermore, the psychometric properties and sex differences of our screening instrument, namely the Autism Spectrum Quotient-10 (AQ-10), were examined at a community level.

Methods

The present study was part of a territory-wide epidemiological study of psychiatric disorders among children and adolescents aged 6–17 years in Hong Kong (Chan et al., 2025). The study recruited a random school-based sample for systematic evaluation of the presence of more than 30 DSM-5-defined psychiatric disorders with the Diagnostic Interview Schedule for Children Version 5 (DISC-5) (Shaffer et al., 2000). As autism was not included in the DISC-5, a two-stage approach was applied to the epidemiological sample for prevalence estimation.

School-based random sampling

The Hong Kong SAR Government provides 12 years of free education for children and adolescents from age 6 to 18 years, and up to 96.6% of our targeted population (i.e. all children and adolescents aged 6–17 years in Hong Kong) are attending schools. Furthermore, clinical psychologists at the Child Assessment Service in Hong Kong, a governmental service that provides territory-wide assessment services for all children, would conduct intelligence quotient (IQ) tests for those suspected of intellectual disability. Only children with a diagnosis of intellectual disability would be eligible for enrolment into special education schools. Given the context of the educational system in Hong Kong, a territory-wide school-based sampling provides the best available portal to systematically enrol a representative community sample. To encompass all main regions and types of schools in Hong Kong, a stratified random sampling approach was adopted. For the first stratum, a complete list of 1,291 schools was obtained from the Education Bureau of the Hong Kong SAR Government, with schools classified by levels (primary or secondary), funding types (private or subsidized), geographical districts (the New Territories East and West, Kowloon or Hong Kong Island) and curriculum (special education schools for intellectual disability or mainstream schools for normal intelligence). A random number was generated for each school under each category, and school administrators under each category were approached by the order of the random number list. With consent at the school level, the next stratum of sampling of students was performed by generating a random number list of anonymised students for every class in each grade corresponding to the actual headcounts of registered students. Parents or a main caregiver was then approached to enrol their children to participate in the study according to the order of appearance on the list. Those who were unapproachable or declined were replaced by the next student on the list. A sampling ratio of one in seven in each class was adopted to reach the target sample size of 6,000. This sample size was determined for the whole psychiatric epidemiological study with the software Epi Info by the US Centers for Disease Control and Prevention (CDC). It showed that a sample size of 6,000 could give a margin of error for disorders prevailing at 5% with a 95% confidence interval (CI) ± 0.55% (i.e. an 11% margin of error in either direction), given an estimated population size of 680,000 for children and adolescents in the age range 6–17 years in Hong Kong. While only 1.3% of our targeted population was studying in special schools (Census and Statistics Department, Hong Kong SAR Government, 2022), the special school group was oversampled to >300 to provide an adequate sample size for estimation of psychiatric disorders in DISC-5. The parents or caregivers of all participating students provided written informed consent, and the study was conducted in accordance with the Declaration of Helsinki.

Screening and identifying autism in two stages

In the first stage of the epidemiological study, the parent or caregiver of all participants completed the Autism Spectrum Quotient-10 Child Version (AQ-10-Child, for children up to 11 years old) or Adolescent Version (AQ-10-Adol, for adolescents between aged 12–17 years old) alongside the DISC-5 assessment and questionnaires on socioeconomic status, family background and the child’s medical history and school adjustment. The AQ-10, a screening instrument for autism spectrum conditions, comprises 10 items rated on a four-point Likert-type scale and scored binarily (Allison et al., 2012). The brevity of the AQ-10, compared to other available screening instruments for autism, was critical for the feasibility of the epidemiological study given the extensive research procedures for parents or caregivers. Both AQ-10-Child and AQ-10-Adol were validated in Hong Kong, with established local cutoff scores of five (sensitivity = 0.82 and specificity = 0.87) and six (sensitivity = 0.80 and specificity = 0.87), respectively. Compared to the original version in the United Kingdom, a lower AQ-10-Child cutoff score of five was adopted, as the local validation study showed that the original score of six reduced sensitivity to 0.65 (Leung et al., 2023).

The second stage was conducted simultaneously with the first stage to meet the study’s timeline. Participants who completed the AQ-10 and DISC-5 were stratified by the AQ-10 result (AQ-10 Positive and AQ-10 Negative) and school level (primary and secondary). Consecutive participants from each group were invited for the second-stage confirmation of autism with the Autism Diagnostic Interview-Revised (ADI-R) (Lord et al., 1994) until the target sample size was reached, ensuring balanced representation across groups. The ADI-R was administered by psychiatrists with experience in child and adolescent mental health or by research staff majoring in psychology and with experience in conducting diagnostic assessments of psychiatric conditions for children and adolescents in research settings. All ADI-R interviewers were blinded to the AQ-10 score. A clinical psychologist qualified as an ADI-R trainer, and the co-author (PWLL) provided the training of the ADI-R for all research interviewers. The inter-rater reliability at the item level of the ADI-R was ascertained with Cohen’s Kappa > 0.7 for all research interviewers. Regular research meetings were conducted by the first author and co-authors (OWHW, KYCL and PWLL, who were clinicians in child and adolescent mental health with more than 10 years of experience in assessing and treating autism) to ensure accurate scoring of each item in the ADI-R through reviewing the written responses verbatim and discussion with the interviewers. We adopted a scoring algorithm of the Chinese version of the ADI-R that aligns with the latest DSM-5 criteria, with its psychometric properties established in Hong Kong. In brief, the items in the ADI-R were mapped to the subdomains of the DSM-5 criteria: socio-emotional reciprocity (A1), nonverbal communicative behaviours (A2), social relationships (A3), motor movements (B1), adherence to routines/resistance to change (B2), highly restricted, fixated interests (B3) or hyper- or hypo-reactivity to sensory input (B4). A research diagnosis of autism is made when all three of the A subdomains and at least two of the four B subdomains score at or above the cutoff scores criteria (Lai et al., 2022). To determine the sample size needed for the second diagnostic stage by ADI-R, a Cochran equation that took into account the sensitivity and specificity of the AQ-10 and the sampling ratio was used (Cochran, 1977), which yielded a minimum requirement of 300 participants. Each participating family received an honorarium of Hong Kong Dollars 300 upon completion of the research procedures. There was no community member involved in the design and conduct of the present study.

Sociodemographic characteristics of the study population

To assess the sociodemographic characteristics of the study population, parents or caregivers were asked about the highest level of education attained and the monthly household income. Education levels were categorized as primary or less, secondary, post-secondary or tertiary. Household income was classified into three groups: low (first to second decile), middle (third to eighth decile) and high (ninth to tenth decile) based on the 2021 Population Census of the Hong Kong SAR (Census and Statistics Department, Hong Kong SAR Government, 2022).

Prevalence estimation and statistical analysis

Based on the second stage’s ADI-R results, the sex-specific PPV and NPV of the AQ-10 were computed and compared given the known phenotypic diversity in autism between males and females as mentioned earlier. They were then applied to the number of AQ-10 Positive and AQ-10 Negative males and females in the entire cohort to estimate the overall prevalence. Referencing the work of Alshaban et al. (2019), we calculated the 95% CI of the prevalence estimations using binomial simulations with 10,000 iterations, where n = sample population and p = proportion of AQ-10 Positive/Negative. Between-group characteristics were compared with t-tests or the Mann–Whitney U test for continuous data, and chi-square tests for categorical data. All statistical analyses were performed with Jamovi version 2.3.28 and R version 4.3.2.

Results

Study population characteristics

Of all the schools approached, 72.8% (N = 126) consented to participate in the study. Reasons for rejection at the school level included administrative difficulties and insufficient manpower to support the study. Then, the average response rate at the participant level was 71.7%. Among those who did not respond, around 87.1% of the parents or caregivers declined due to study procedures being incompatible with their schedules or lack of interest after receiving an explanation of the study details, while approximately 12.9% could not be reached by the research team. The study population included in this report comprised 5,865 participants with AQ-10 scores, consisting of 5,552 children and adolescents with normal IQ from mainstream schools and 313 children and adolescents with intellectual disabilities from special education schools aged 6–17 years. The majority (86.2%) of the study population was of Chinese ethnicity, 12.1% was non-Chinese Asians, 0.3% was African, 0.2% was White and the rest (1.2%) was of mixed ethnicities. These participants completed the DISC-5 interview and provided valid responses for the AQ-10 screening. There were more males than females (3,410 vs 2,455), and significantly more males (24.6%) were screened positive on the AQ-10 than females (13.2%) (χ² = 116, p ⩽ 0.001). Otherwise, age and other sociodemographic characteristics were comparable between the AQ-10 Positive and AQ-10 Negative groups. The detailed breakdown of the study population, categorized by school level and type according to their AQ-10 results, is presented in Table 1.

Table 1.

Demographics of participants and the comparisons between AQ-10 Positive and AQ-10 Negative groups.

	All participants		AQ-10 Positive		AQ-10 Negative		Statistics
	Mean/n	SD/%	Mean/n	SD/%	Mean/n	SD/%	U/χ²	df	p
n	5865		1164	19.85	4701	80.15	-
Age	10.9	3.15	10.85	3.17	10.9	3.15	2.70E + 06		0.544
Primary
Mainstream	3303	56.31	616	18.65	2687	81.35	-
Special school	178	3.03	117	65.73	61	34.27	-
Secondary
Mainstream	2249	38.35	348	15.47	1901	84.53	-
Special school	135	2.30	83	61.48	52	38.52	-
Sex
Male	3410	58.14	839	72.08	2571	54.69	116	1	<0.001***
Female	2455	41.86	325	27.92	2130	45.31	116	1	<0.001***
AQ-10 score	3.11	2.18	6.49	1.27	2.28	1.42	43,727		<0.001***
Household income
Low	498	8.53	99	8.54	399	8.53	0.286	2	0.867
Middle	3973	68.07	782	67.47	3191	68.21
High	1366	23.40	278	23.99	1088	23.26
Highest parental education attained
Primary or less	107	1.83	15	1.30	92	1.97	5.02	3	0.171
Secondary	2723	46.67	545	47.06	2178	46.57
Post-secondary	631	10.81	111	9.59	520	11.12
Tertiary	2374	40.69	487	42.06	1887	40.35

AQ-10: Autism Spectrum Quotient-10.

***

p < 0.001.

Second-stage ADI-R for confirmation of DSM-5 autism

A total of 728 participants from both AQ-10 Positive and AQ-10 Negative groups from primary and secondary schools were approached for the ADI-R, of which 154 (21.2%) could not be reached, 257 (35.3%) rejected the invitation and 317 (43.5%) agreed and completed the diagnostic interview. Age, sex and AQ-10 scores were comparable between those who participated and those who rejected the ADI-R assessment. However, families that participated in the ADI-R had a lower household income than those that did not participate in the ADI-R (8.54% vs 5.12%, χ² = 9.56, p = 0.008) (Table 2). The sociodemographic characteristics of these 317 participants also resembled those of the entire study population (Table S1 in Supplemental Materials). Among these 317 participants, 159 were from primary schools, while 158 were from secondary schools. There were 140 AQ-10 Positive participants (102 males and 38 females) and 177 AQ-10 Negative participants (91 males and 86 females). The sex distribution within the AQ-10 Positive and AQ-10 Negative groups in the subsample resembled that of the entire study population. Finally, 23 participants (21 males and 2 females) were diagnosed with autism by ADI-R in the AQ-10 Positive group, while none in the AQ-10 Negative group received the diagnosis (Figure 1).

Table 2.

Comparisons between participants not participating and completing the ADI-R.

	Not participating ADI-R		Completed ADI-R
	Mean/n	SD/%	Mean/n	SD/%	U/χ²	df	p
n	411	56.46	317	43.54
Age	11.47	3.11	11.26	3.12	62,606		0.365
Sex
Male	248	60.34	194	61.20	0.0552	1	0.814
Female	163	39.66	123	38.80	0.0552	1	0.814
AQ-10
Score	4.18	2.54	4.04	2.548	62,889	0.42
Positive	190	46.23	141	44.48	0.221	1	0.638
Negative	221	53.77	176	55.52	0.221	1	0.638
School level
Primary	214	52.07	159	50.16	0.261	1	0.609
Secondary	197	47.93	158	49.84	0.261	1	0.609
Household income
Low	21	5.12	27	8.54	9.56	2	0.008**
Middle	262	63.90	220	69.62
High	127	30.98	69	21.84
Highest parental education
Primary or less	3	0.73	1	0.32	0.984	3	0.805
Secondary	193	47.07	157	49.68
Post-secondary	44	10.73	32	10.13
Tertiary	170	41.46	126	39.87

ADI-R: Autism Diagnostic Interview-Revised; AQ-10: Autism Spectrum Quotient-10.

p < 0.01.

Effect of non-participation on prevalence estimation

Given the differences in household income among families that did and did not participate in the ADI-R assessment, we hypothesized that a latent motivational factor existed to influence the family’s willingness to participate and examined whether this factor might bias the probability of identifying autism. We referenced the work by Kim et al. (2011) to model this latent variable, where the participant’s sex, AQ-10 scores and family income class were used to predict participation versus non-participation in the ADI-R assessment in a logistic regression. The regression showed that only family income (high vs low) significantly predicted lower odds of participating in the ADI-R assessment, with an odds ratio (OR) of 0.41 (95% CI: 0.22–0.78, p = 0.007), while sex and AQ-10 score were not significant predictors (p = 0.564 and 0.294). From the regression, a probability score of participating in the ADI-R assessment was generated to serve as a surrogate of the motivational factor for each participant ever invited. Then, within the subsample of the second stage’s ADI-R assessment, another logistic regression was used to predict autism versus non-autism with the probability score, the participant’s age, sex, the highest parental education level and the AQ-10 score as predictors. In this logistic regression, only the AQ-10 score predicted autism with an OR of 1.95 (95% CI: 1.47–2.59, p < 0.001), suggesting that the willingness to participate and other sociodemographic characteristics were not associated with the probability of identifying autism (Table S2 in Supplemental Materials).

Psychometric properties of AQ-10 in a community sample and sex differences

As none of the AQ-10 Negative participants received a diagnosis of autism in the second-stage ADI-R assessment, using the validated cutoffs of the AQ-10 resulted in an NPV of 100% for both males and females. The PPV of the AQ-10 for both males and females together was 16.31%. When stratified by sex, the PPV of the AQ-10 was significantly lower (χ² = 4.65, p = 0.03) in females (5.2%) than in males (20.4%) (Table 3). Further stratification by school level did not yield a significant difference between the PPVs in primary schools and secondary schools for males (χ² = 0.969, p = 0.33) and females (χ² = 0, p = 1). With the marked difference in PPV by sex, we accordingly estimated the prevalence of autism using sex-specific PPVs.

Table 3.

Results of the diagnostic confirmation of autism by the ADI-R.

		Autism	Non-autistic	PPV	NPV	FPR
Overall (male & female)	AQ-10 Positive	23	118	16.31	100	83.70
Overall (male & female)	AQ-10 Negative	0	176	16.31	100	83.70
Male only	AQ-10 Positive	21	82	20.39	100	79.61
Male only	AQ-10 Negative	0	91	20.39	100	79.61
Female only	AQ-10 Positive	2	36	5.26	100	94.74
Female only	AQ-10 Negative	0	85	5.26	100	94.74

ADI-R: Autism Diagnostic Interview-Revised; AQ-10: Autism Spectrum Quotient-10; PPV: positive predictive value; NPV: negative predictive value; FPR: false-positive rate; FNR: false negative rate.

Estimation of autism prevalence

To address the unbalanced sex distribution and enhance representativeness, we reconstituted the study population based on sex, school level, intelligence and household income according to the census data in Hong Kong before estimating the prevalence (Census and Statistics Department, Hong Kong SAR Government, 2022). This was done by (1) dividing the study population into 24 strata (2 sexes × 2 school levels × 2 intelligence groups × 3 household income levels), (2) computing the rate of AQ-10 Positive and AQ-10 Negative in each stratum (Table S3 in Supplemental Materials), (3) based on the sample size of 5,865 and the Census data, the distribution of the population in each stratum was projected (Table S4 in Supplemental Materials) and (4) using the projected number of the population in each stratum multiplied by the corresponding AQ-10 Positive/Negative rate to calculate the number of AQ-10 Positive and AQ-10 Negative in each stratum. In the reconstituted sample, there were 3,020 males and 2,845 females. Proportionally, there were 651 AQ-10 Positive males and 343 AQ-10 Positive females (Table S5 in Supplemental Materials). Applying the sex-specific PPVs, the overall prevalence of autism was 2.57% (95% CI: 2.23%–2.91%), with a male-to-female (M:F) ratio of 7.39:1. We also applied the original DSM-4 algorithm of the ADI-R to determine whether the prevalence would be significantly different. While there was no change in the classification of autism in females, there were two less males being classified as having autism, resulting in a decrease of the PPV of AQ-10 to 18.4%. Replicating the aforementioned procedures yielded a prevalence of 2.39% (95% CI: 2.08%–2.71%) using the DSM-4 algorithm.

Explorative analysis of autistic symptoms in AQ-10 Positive males and females without research diagnosis of autism

Given the low prevalence of autism in the population, the false positive rates (FPRs) of the AQ-10 in our ADI-R subsample for males and females were 79.6% and 94.7%, respectively (Table 3), indicating a large number of false positives by the AQ-10, that is, those who were AQ-10 Positive but did not meet the criteria of the ADI-R DSM-5 algorithm (n = 118, with 82 males and 36 females). Since autism is conceptualized as a continuum from individuals who are functionally impaired because of autistic symptoms to those with an absence of the symptoms (De Groot & Van Strien, 2017), it is interesting to explore the symptoms of autism in the AQ-10 False-positive participants in contrast to the AQ-10 Negative participants (n = 176, with 91 males and 85 females). Using multivariate analysis of covariance (MANCOVA), log-transformed diagnostic algorithm subdomain scores of the ADI-R (ever or abnormal at age 4–5 years, with A1–3 reflecting the deficits in the social communication and interaction domain and B1–4 representing the restricted and repetitive behaviour domain) were compared between the two groups with age as a covariate. An interaction term of ‘Group * Sex’ was also added to the model to explore the differential symptoms between males and females. Compared to the AQ-10 Negative group, the AQ-10 False-positive group had significantly higher ratings on all seven symptoms, suggesting that symptoms of autism were present in the group despite not reaching the research diagnosis. Notably, while female sex was associated with lower ratings on the symptoms of A2 (deficits in nonverbal communicative behaviours), A3 (deficits in developing, maintaining and understanding relationships) and B1 (stereotyped/repetitive speech, motor, use of objects), significant ‘Group * Sex’ interactions were observed in A2 and B1 (Table 4 and Figure 2). Post hoc examination revealed that the AQ-10 False-positive females scored comparably to the AQ-10 Negative females on these two symptom domains (A2: t = 1.83, df = 289, p_Tukey = 0.260; B1: t = 0.789, df = 289, p_Tukey = 0.859).

Table 4.

Summary of MANCOVA on the ADI-R subdomain scores between the AQ-10 False-positive and AQ-10 Negative groups and their interaction with sex.

Multivariate tests
		Pillai’s trace	F	df1	df2	p
Group		0.06	16.99	7	283	< 0.001***
Sex		0.05	2.34	7	283	0.025*
Group * Sex		0.03	1.32	7	283	0.241
Univariate tests
	Dependent variable	SS	df	MS	F	p
Group	A1	27.22	1	27.224	44.1953	< 0.001***
	A2	2.35	1	2.346	33.0856	< 0.001***
	A3	7.19	1	7.194	77.2126	< 0.001***
	B1	0.87	1	0.865	17.9771	< 0.001***
	B2	0.54	1	0.540	20.5089	< 0.001***
	B3	0.49	1	0.487	17.1578	< 0.001***
	B4	0.42	1E + 00	0.422	18.0543	< 0.001***
Sex	A1	2.13	1	2.132	3.46158	0.064
	A2	0.46	1	0.461	6.50186	0.011*
	A3	1.09	1	1.087	11.6706	<0.001***
	B1	0.29	1	0.289	6.01213	0.015*
	B2	0.03	1	0.027	1.03245	0.31
	B3	0.04	1	0.039	1.37189	0.242
	B4	0.04	1E + 00	0.042	1.79788	0.181
Group * Sex	A1	0.33	1	0.33373	0.54	0.462
	A2	0.31	1	0.30854	4.35	0.038*
	A3	0.05	1	0.0536	0.58	0.449
	B1	0.21	1	0.21326	4.43	0.036*
	B2	0.03	1	0.02649	1.01	0.317
	B3	0.01	1	0.00529	0.19	0.666
	B4	0.00	1	1E-04	0.00	0.948
Residuals	A1	178.02	289	0.61599
	A2	20.49	289	0.0709
	A3	26.93	289	0.09317
	B1	13.91	289	0.04812
	B2	7.61	289	0.02634
	B3	8.20	289	0.02836
	B4	6.75	289	0.02335

Results were controlled for age. ADI-R: Autism Diagnostic Interview-Revised; AQ-10: Autism Spectrum Quotient-10; A1: socio-emotional reciprocity; A2: nonverbal communicative behaviours; A3: social relationships; B1: motor movements; B2: adherence to routines/resistance to change; B3: highly restricted, fixated interests; B4: hyper- or hypo-reactivity to sensory input.

p < 0.05; ***p < 0.001.

Discussion

Prevalence of autism among children and youth in Hong Kong

This is the first epidemiological study that applied DSM-5 diagnostic criteria to investigate the prevalence of autism in Hong Kong, a metropolitan city with unique internationality among the major cities in China and Asia. The study population was based on a representative random sample, covering a broad range of school-aged children and youth across varying socioeconomic statuses and intelligence levels. Its strengths include case identification at the population level, use of locally validated autism measures, a reasonably large sample size, an adequate number of screened children recruited for case determination, careful examination of potential bias in non-participation at the diagnostic stage and the use of sex-specific PPVs in prevalence estimation.

The prevalence of autism in Hong Kong, estimated at 2.57% in the present study, aligns with contemporary estimates from other Asian countries, such as 2.64% in South Korea (Kim et al., 2011) and 3.22% in Japan (Saito et al., 2020). This figure reflects a marked increase from Hong Kong’s earlier estimate of 0.16% in 2008 (Wong & Hui, 2008), consistent with trends observed in a recent meta-analysis, which reported significantly lower prevalence estimates in earlier years before 2008 than those of recent studies in China (Wang et al., 2018). Similarly, the U.S. national surveillance by the CDC reported a steady rise from 0.67% in 2000 to 2.76% in 2020 (Maenner et al., 2024). Several factors may contribute to this global apparent increase in the prevalence of autism. The conceptualization of autism in terms of a spectrum and changes in the diagnostic criteria over the past two decades is a primary consideration (Rosen et al., 2021). However, methodological aspects of the present study, including assessment tools, study design (two-stage approach), setting (community-based) and scope (encompassing the entire territory and wider age range) of our study and possible higher recognition of autism features among contemporary parents, may have underpinned the apparent rise in prevalence in Hong Kong. Given the substantial methodological disparities across existing studies, it is difficult to disentangle the extent to which the differences in prevalence estimates are due to methodological variations, including case definition, or an actual increase. Yet, one large-scale population study found that changes in case definition from DSM-4 pervasive developmental disorder to DSM-5 autism spectrum disorder did not bring any substantive changes in autism prevalence estimates in the same group of community children (Kim et al., 2014). In other words, changes in case definition in recent decades may not be as influential in determining prevalence as suspected. The present study also found an insignificant change in the prevalence estimate from using the earlier DSM-4 scoring algorithm (2.39%) to that of the recent DSM-5 (2.57%). Hence, the possibility of an actual increase in autism prevalence should not be readily dismissed, where standardized methodologies of case identification across studies and time are needed. Simultaneously, the escalating global autism prevalence from recent surveys should raise significant concerns from a public health standpoint given the resulting surge in demand for resources for early detection and treatment. Research has shown that early intervention for children with autism can offer considerable benefits to them and save a significant amount of costs needed for subsequent educational accommodation. For example, early intervention was estimated to save USD 208,500 per child across 18 years of education in Texas (Chasson et al., 2007), while USD 250,000 per child would be saved in annual special education costs with early intervention for children with autism in North Carolina (Cooper, 2022).

When adopting the locally derived lower cutoff score of five for the AQ-10-Child following the local validation study (Leung et al., 2023), the screening instrument AQ-10 demonstrated a 100% NPV in both males and females in this population-based study. This is relevant to the precision of our prevalence estimation given that binarizing the AQ-10 into positive and negative for projection in the whole study population would fail to consider the dimensional positive correlation between the AQ-10 score and the probability of an autism diagnosis, as shown in our logistic regression (Table S2 in Supplemental Materials). Of note, 4 out of 17 males with autism diagnosed by the ADI-R had an AQ-10-Child score of five. Hence, adopting the original higher AQ-10 cutoff of six derived in the UK for our ADI-R participant selection could potentially result in an NPV of less than 100%, which may inflate the prevalence estimation owing to the high proportion of AQ-10 Negative participants in the population.

Utility of the AQ-10 as a screening instrument for autism in the community

The AQ-10 was developed as a brief screening tool to identify ‘red-flag’ symptoms that prompt comprehensive diagnostic assessment for autism (Allison et al., 2012). Thus, the 100% NPV found in the present study supports the instrument’s effectiveness in excluding autism when utilized as a universal screening measure in the community. Importantly, this property remained consistent across sexes, affirming that the established cutoff can effectively distinguish individuals unlikely to have autism for both males and females.

In the original study of the AQ-10, the PPV was 0.86 for the AQ-10-Adol and 0.94 for the AQ-10-Child among 1,000 autistic and 3,000 non-autistic individuals. As PPV is dependent on both the prevalence of the condition in a particular sample and the psychometric properties of the instrument, it is not surprising that the PPVs of the AQ-10 were only low to modest in our current study, as autism is relatively rare in the community setting (i.e. 2.57% compared with the original study’s 25% in its case-control sample; Allison et al., 2012). Compared with the Social Communication Questionnaire (SCQ), another commonly used screening instrument for autism, a low PPV of 8.6% was also noted when the SCQ is applied in a population-based prevalence study with a similar two-stage design as the present study. This low PPV was considered acceptable for a relatively rare condition like autism in the general population (Alshaban et al., 2019). Furthermore, with autism being less prevalent among females, the PPV was even lower at only 5.26%. A lower female PPV was similarly reported in a recent study that applied the M-CHAT in a community sample of toddlers (Eldeeb et al., 2023). On the surface, this could question the efficiency of using the AQ-10 to identify youth with autism, which would result in a significant number of false positives. However, our explorative analysis showed that the participants classified as ‘false positive’, despite not reaching the research diagnosis criteria, still exhibited elevated levels of autistic symptoms. Individuals with broad autism phenotypes or autistic traits are not without clinical needs and should not be overlooked (Kurtz et al., 2023; Yule et al., 2021). Hence, the AQ-10 is still useful in identifying individuals with elevated autistic symptoms, particularly in scenarios where other related clinical conditions (e.g. attention-deficit/hyperactivity disorder, anxiety and depression) are present. This can be crucial for formulating a comprehensive management approach that considers the role of autistic symptoms within the broader clinical context.

Sex-specific symptom expression may moderate the detection of autism in the community

While the inherent sex bias of autism could account for the lower PPVs observed in females, the projected M:F ratio in the present study at 7.47:1 exceeded ratios reported in other contemporary studies, including a meta-analysis that reported a pooled ratio of 4.20:1 (Loomes et al., 2017). Although the reason behind this higher ratio in our study is unclear, we considered possibilities from the study’s methodological approaches and the sexual dimorphism of autistic symptom presentation. Historically, the majority of the assessment and diagnostic instruments for autism were developed based on the male prototype (Lai et al., 2023), yet recent studies have identified quantitative and qualitative differences in autistic symptom expression between males and females, such as restricted and repetitive behaviours (Hull et al., 2020) and impairment of eye gaze to social stimuli (Harrop et al., 2019). In our explorative analysis, the symptom load in nonverbal communication and stereotyped/repetitive behaviours of the AQ-10 False-positive females was significantly less than that of the AQ-10 False-positive males and yet comparable to that of the AQ-10 Negative females. This could suggest that the ADI-R may not be sensitive enough to detect these symptoms in females, or that the manifestation of these symptoms in the broad autism phenotype of females could be qualitatively different from that of males. A recent study showed that the sensitivity of the standardized diagnostic instrument, Autism Diagnostic Observation Schedule (ADOS), alone in detecting known diagnoses of autism is lower in females than in males (D’Mello et al., 2022). Collectively, this prompts a need for optimisation of diagnostic practices for females within research settings. Such investigations should involve a representative community cohort spanning the autism spectrum to mitigate ascertainment biases that limit current studies (Kaat et al., 2021).

Limitations

This study has several limitations. First, the female population was relatively underrepresented in the second-stage ADI-R assessment, which may have affected the precision of PPV estimation in females. Second, the confirmation of an autism diagnosis relied solely on the ADI-R, without additional clinical evaluation and observational assessments, such as the ADOS. This may have reduced the accuracy of case detection and confirmation based on clinical best-estimate. Third, although the AQ-10 is the only validated screening instrument available in Hong Kong, its local validation study did not include autistic individuals with intellectual disabilities. Thus, there is less confidence in the accuracy of prevalence estimation within this subgroup. Finally, although we statistically examined the effect of non-participation on the probability of an autism diagnosis, the response rate for the ADI-R was only moderate. The lower participation of high-income families may be attributed to their heavier work schedules and the monetary incentives offered being relatively less appealing to them. The lower representation of high-income families in the second-stage ADI-R may still have unknown effects on our prevalence estimates.

Conclusion

This is the first territory-wide epidemiological study on the prevalence of autism in Hong Kong using the DSM-5 diagnostic framework and a two-stage diagnostic approach in a representative population sample. The prevalence of autism was estimated to be 2.57%, which was in line with the recent global trend of a rise in autism prevalence. This poses a significant public health challenge, demanding increased resources for early detection and treatment. The AQ-10 has proven to be useful in excluding autism and identifying autistic symptoms in the community using the locally validated age-specific cutoffs for both males and females. However, it remains obscure yet intriguing how plausible sexual dimorphism might impact the diagnosis and detection rate of autism in epidemiological studies. Future studies should re-examine the psychometric properties of diagnostic instruments for autism when applied in a community setting.

Supplemental Material

sj-docx-1-aut-10.1177_13623613251360269 – Supplemental material for Autism epidemiology in Hong Kong children and youths aged 6–17: Implications on autism screening and sex differences in the community

Supplemental material, sj-docx-1-aut-10.1177_13623613251360269 for Autism epidemiology in Hong Kong children and youths aged 6–17: Implications on autism screening and sex differences in the community by Oscar WH Wong, Sandra SM Chan, Steven WH Chau, Winnie CW Chu, Carol SW Ho, Stephy WS Ho, Se Fong Hung, Samara Hussain, Kelly YC Lai, Angela MW Lam, Holly HL Lo, Karen KY Ma, Suk Ling Ma, Flora YM Mo, Pak Chung Sham, Caroline KS Shea, Suzanne HW So, Kelvin KF Tsoi and Patrick WL Leung in Autism

Footnotes

Acknowledgements

The authors would like to thank the team of Work Heart research assistants, along with Dr. Brian Or, Dr. Jason Wong and Dr. Iris Ma for their efforts in the recruitment of participants and data collection for this study.

ORCID iDs

Oscar WH Wong

Kelly YC Lai

Karen KY Ma

Suk Ling Ma

Patrick WL Leung

Ethical Considerations

The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. The authors assert that all procedures contributing to this work comply with the ethical standards of the Joint CUHK-NTEC CREC (Ref: 2018.497).

Consent to Participate

Written informed consent was obtained from the parents or legal guardian of the participants.

Author Contributions

Oscar WH Wong: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Writing – original draft; Writing – review & editing.

Sandra SM Chan: Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Supervision; Writing – original draft; Writing – review & editing.

Steven WH Chau: Investigation; Methodology; Writing – review & editing.

Winnie CW Chu: Funding acquisition; Project administration; Writing – review & editing.

Carol SW Ho: Investigation; Project administration; Writing – review & editing.

Stephy WS Ho: Data curation; Methodology; Project administration; Writing – review & editing.

Se Fong Hung: Conceptualization; Data curation; Funding acquisition; Investigation; Methodology; Project administration; Supervision; Writing – review & editing.

Samara HUSSAIN: Data curation; Formal analysis; Investigation; Methodology; Project administration; Writing – review & editing.

Kelly YC Lai: Conceptualization; Data curation; Funding acquisition; Investigation; Methodology; Project administration; Writing – review & editing.

Angela MW Lam: Data curation; Formal analysis; Investigation; Writing – review & editing.

Holly HL Lo: Data curation; Investigation; Project administration; Writing – review & editing.

Karen KY Ma: Data curation; Formal analysis; Investigation; Writing – review & editing.

Suk Ling Ma: Conceptualization; Methodology; Writing – review & editing.

Flora YM Mo: Conceptualization; Methodology; Writing – review & editing.

Pak Chung Sham: Conceptualization; Funding acquisition; Methodology; Supervision; Writing – review & editing.

Caroline KS Shea: Methodology; Writing – review & editing.

Suzanne HW So: Conceptualization; Data curation; Funding acquisition; Investigation; Methodology; Writing – review & editing.

Kelvin KF Tsoi: Formal analysis; Methodology; Writing – review & editing.

Patrick WL Leung: Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Supervision; Writing – original draft; Writing – review & editing.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Phase I of the Commissioned Study on Mental Health Survey: Hong Kong Child and Adolescent Psychiatric Epidemiologic Survey: Age 6 to 17, Health and Medical Research Fund Commissioned Study, Food Health and Welfare Bureau (Ref: MHS-P1(Part 1)-CUHK).

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

OWHW has access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Supplemental material

Supplemental material for this article is available online.

References

Allison

Auyeung

Baron-Cohen

(2012). Toward brief ‘red flags’ for autism screening: The short autism spectrum quotient and the short quantitative checklist in 1,000 cases and 3,000 controls. Journal of the American Academy of Child & Adolescent Psychiatry, 51(2), 202–212.e7. https://doi.org/10.1016/J.JAAC.2011.11.003

Alshaban

Aldosari

Al-Shammari

El-Hag

Ghazal

Tolefat

Ali

Kamal

Aati

N. A.

Abeidah

Saad

A. H.

Dekair

Khasawneh

Al Ramsay

Fombonne

(2019). Prevalence and correlates of autism spectrum disorder in Qatar: A national study. Journal of Child Psychology and Psychiatry, 60(12), 1254–1268. https://doi.org/10.1111/JCPP.13066

Census and Statistics Department, Hong Kong SAR Government. (2022). https://www.censtatd.gov.hk/

Chan

S. S. M.

Wong

O. W. H.

Hussain

Tsoi

K. K. F.

K. K. Y.

Chau

S. W. H.

S. L.

Lai

K. Y. C.

Chu

W. C. W.

H. H. L.

S. W. S.

Leung

C. C.

Yiu

K. K. L.

S. H. W.

Sham

P. C.

Hung

S. F.

Leung

P. W. L.

(2025). Twelve-month prevalence of DSM-5 mental disorders and the psychosocial correlates- a child and adolescent psychiatric epidemiologic survey in Hong Kong SAR. The Lancet Regional Health – Western Pacific, 57, 101533. https://doi.org/10.1016/J.LANWPC.2025.101533

Chasson

G. S.

Harris

G. E.

Neely

W. J.

(2007). Cost comparison of early intensive behavioral intervention and special education for children with autism. Journal of Child and Family Studies, 16(3), 401–413. https://doi.org/10.1007/S10826-006-9094-1

Cochran

W. G.

(1977). Sampling techniques. John Wiley & Sons. https://www.wiley.com/en-us/Sampling+Techniques%2C+3rd+Edition-p-9780471162407

Cooper

(2022). Reducing special education costs by providing early intervention for autistic children. Behavioral Interventions, 37(2), 397–414. https://doi.org/10.1002/BIN.1839

De Groot

Van Strien

J. W

. (2017). Evidence for a broad autism phenotype. Advances in Neurodevelopmental Disorders, 1(3), 129–140. https://doi.org/10.1155/2011/545901

D’Mello

A. M.

Frosch

I. R.

C. E.

Cardinaux

A. L.

Gabrieli

J. D. E.

(2022). Exclusion of females in autism research: Empirical evidence for a ‘leaky’ recruitment-to-research pipeline. Autism Research, 15(10), 1929–1940. https://doi.org/10.1002/AUR.2795

10.

Eldeeb

S. Y.

Ludwig

N. N.

Wieckowski

A. T.

Dieckhaus

M. F. S.

Algur

Ryan

Dufek

Stahmer

Robins

D. L.

(2023). Sex differences in early autism screening using the modified checklist for autism in toddlers, revised, with follow-up (M-CHAT-R/F). Autism, 27(7), 2112–2123. https://doi.org/10.1177/13623613231154728

11.

Harrop

Jones

Zheng

Nowell

Schultz

Parish-Morris

(2019). Visual attention to faces in children with autism spectrum disorder: Are there sex differences? Molecular Autism, 10(1), 1–10. https://doi.org/10.1186/s13229-019-0276-2

12.

Harstad

Hanson

Brewster

S. J.

Depillis

Milliken

A. L.

Aberbach

Sideridis

Barbaresi

W. J.

(2023). Persistence of autism spectrum disorder from early childhood through school age. JAMA Pediatrics, 177(11), 1197–1205. https://doi.org/10.1001/JAMAPEDIATRICS.2023.4003

13.

Hull

Petrides

K. V.

Mandy

(2020). The female autism phenotype and camouflaging: A narrative review. Review Journal of Autism and Developmental Disorders, 7(4), 306–317. https://doi.org/10.1007/s40489-020-00197-9

14.

Hussain

John

J. R.

Dissanayake

Frost

Girdler

Karlov

Masi

Alach

Eapen

(2023). Sociocultural factors associated with detection of autism among culturally and linguistically diverse communities in Australia. BMC Pediatrics, 23(1), 1–11. https://doi.org/10.1186/S12887-023-04236-2

15.

Kaat

A. J.

Shui

A. M.

Ghods

S. S.

Farmer

C. A.

Esler

A. N.

Thurm

Georgiades

Kanne

S. M.

Lord

Kim

Y. S.

Bishop

S. L.

(2021). Sex differences in scores on standardized measures of autism symptoms: A multisite integrative data analysis. Journal of Child Psychology and Psychiatry, 62(1), 97–106. https://doi.org/10.1111/JCPP.13242

16.

Kim

Y. S.

Fombonne

Koh

Y. J.

Kim

S. J.

Cheon

K. A.

Leventhal

B. L.

(2014). A comparison of DSM-IV pervasive developmental disorder and DSM-5 autism spectrum disorder prevalence in an epidemiologic sample. Journal of the American Academy of Child & Adolescent Psychiatry, 53(5), 500–508. https://doi.org/10.1016/J.JAAC.2013.12.021

17.

Kim

Y. S.

Leventhal

B. L.

Koh

Y. J.

Fombonne

Laska

Lim

E. C.

Cheon

K. A.

Kim

S. J.

Kim

Y. K.

Lee

H. K.

Song

D. H.

Grinker

R. R.

(2011). Prevalence of autism spectrum disorders in a total population sample. American Journal of Psychiatry, 168(9), 904–912. https://doi.org/10.1176/APPI.AJP.2011.10101532

18.

Kurtz

M. R.

Kana

R. K.

Rivera

D. L.

Newman

S. D.

(2023). The role of the broader autism phenotype in anxiety and depression in college-aged adults. Frontiers in Psychiatry, 14, Article 1187298. https://doi.org/10.3389/FPSYT.2023.1187298/BIBTEX

19.

Lai

M. C.

Amestoy

Bishop

Brown

H. M.

Giwa Onaiwu

Halladay

Harrop

Hotez

Huerta

Kelly

Miller

Nordahl

C. W.

Ratto

A. B.

Saulnier

Siper

P. M.

Sohl

Zwaigenbaum

Goldman

(2023). Improving autism identification and support for individuals assigned female at birth: Clinical suggestions and research priorities. The Lancet Child and Adolescent Health, 7(12), 897–908. https://doi.org/10.1016/S2352-4642(23)00221-3

20.

Lai

K. Y. C.

Yuen

E. C. W.

Hung

S. F.

Leung

P. W. L.

(2022). Autism Diagnostic Interview-Revised Within DSM-5 Framework: Test of Reliability and Validity in Chinese Children. Journal of Autism and Developmental Disorders, 52(4), 1807–1820. https://doi.org/10.1007/S10803-021-05079-5

21.

Leung

C. N. W.

Leung

C. S. Y.

Chan

R. W. S.

Leung

P. W. L.

(2023). Can the UK AQ-10 be applicable to Chinese samples with autism spectrum disorder in Hong Kong? Cross-cultural similarities and differences. Autism Research: Official Journal of the International Society for Autism Research, 16(2), 302–314. https://doi.org/10.1002/AUR.2847

22.

Loomes

Hull

Mandy

W. P. L.

(2017). What is the male-to-female ratio in autism spectrum disorder? A systematic review and meta-analysis. Journal of the American Academy of Child and Adolescent Psychiatry, 56(6), 466–474. https://doi.org/10.1016/j.jaac.2017.03.013

23.

Lord

Rutter

Le Couteur

(1994). Autism diagnostic interview-revised: A revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. Journal of Autism and Developmental Disorders, 24(5), 659–685. https://doi.org/10.1007/BF02172145

24.

Maenner

M. J.

Warren

Williams

A. R.

Amoakohene

Bakian

A. V.

Bilder

D. A.

Durkin

M. S.

Fitzgerald

R. T.

Furnier

S. M.

Hughes

M. M.

Ladd-Acosta

C. M.

McArthur

Pas

E. T.

Salinas

Vehorn

Williams

Esler

Grzybowski

Hall-Lande

Shaw

K. A.

(2024). Prevalence and characteristics of autism spectrum disorder among children aged 8 years – Autism and developmental disabilities monitoring network, 11 sites, United States, 2020. MMWR. Surveillance Summaries, 72(2), 1–14. https://doi.org/10.15585/MMWR.SS7202A1

25.

Rosen

N. E.

Lord

Volkmar

F. R.

(2021). The diagnosis of autism: From Kanner to DSM-III to DSM-5 and beyond. Journal of Autism and Developmental Disorders, 51(12), 4253–4270. https://doi.org/10.1007/S10803-021-04904-1

26.

Saito

Hirota

Sakamoto

Adachi

Takahashi

Osato-Kaneda

Kim

Y. S.

Leventhal

Shui

Kato

Nakamura

(2020). Prevalence and cumulative incidence of autism spectrum disorders and the patterns of co-occurring neurodevelopmental disorders in a total population sample of 5-year-old children. Molecular Autism, 11(1), 1–9. https://doi.org/10.1186/S13229-020-00342-5/TABLES/4

27.

Shaffer

Fisher

Lucas

C. P.

Dulcan

M. K.

Schwab-Stone

M. E.

(2000). NIMH diagnostic interview schedule for children version IV (NIMH DISC-IV): Description, differences from previous versions, and reliability of some common diagnoses. Journal of the American Academy of Child and Adolescent Psychiatry, 39(1), 28–38. https://doi.org/10.1097/00004583-200001000-00014

28.

Sun

Allison

Wei

Matthews

F. E.

Auyeung

Y. Y.

Griffiths

Zhang

Baron-Cohen

Brayne

(2019). Autism prevalence in China is comparable to Western prevalence. Molecular Autism, 10(1), 1–19. https://doi.org/10.1186/S13229-018-0246-0

29.

Wang

Bin Zhang

C. H.

Ungvari

G. S.

Cao

X. L.

J. P.

Hou

C. L.

Jia

F. J.

Xiang

Y. T.

(2018). The prevalence of autism spectrum disorders in China: A comprehensive meta-analysis. International Journal of Biological Sciences, 14(7), 717. https://doi.org/10.7150/IJBS.24063

30.

Wong

V. C. N.

Hui

S. L. H.

(2008). Epidemiological study of autism spectrum disorder in China. Journal of Child Neurology, 23(1), 67–72. https://doi.org/10.1177/0883073807308702

31.

Yule

Wanik

Holm

E. M.

Bruder

M. B.

Shanley

Sherman

C. Q.

Fitterman

Lerner

Marcello

Parenchuck

Roman-White

Ziff

(2021). Nutritional deficiency disease secondary to ARFID symptoms associated with autism and the broad autism phenotype: A qualitative systematic review of case reports and case series. Journal of the Academy of Nutrition and Dietetics, 121(3), 467–492. https://doi.org/10.1016/J.JAND.2020.10.017

32.

Zeidan

Fombonne

Scorah

Ibrahim

Durkin

M. S.

Saxena

Yusuf

Shih

Elsabbagh

(2022). Global prevalence of autism: A systematic review update. Autism Research, 15(5), 778–790. https://doi.org/10.1002/AUR.2696

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB