Developmental functioning and symptom severity influence age of diagnosis in Canadian preschool children with autism

Abstract

Background

Early diagnosis of autism spectrum disorder (ASD) is essential in most Canadian jurisdictions to access interventions that improve long-term child outcomes. Our main objective was to identify factors associated with timing of ASD diagnosis in five provinces across Canada.

Methods

Factors influencing age of diagnosis were assessed in the analyses of an inception cohort of children diagnosed with ASD between ages 2 and 5 years. We examined bivariate associations and using a series of multiple variable regression models, evaluated the unique contributions of developmental functioning, ASD symptoms and demographic variables. Children with known genetic abnormalities, or severe sensory or motor impairments interfering with assessment were excluded.

Results

Participants were 421 children (84.6% boys). The mean age of diagnosis was 38.2 months (SD=8.7), an average of 19 months after parents identified initial concerns. Factors associated with later diagnosis included more advanced language and cognitive skills, and higher levels of restricted repetitive behaviour symptoms. Child sex and family demographics were not associated with age of diagnosis. In regression analyses, language and cognitive skills accounted for 6.8% of variance in age of diagnosis and ASD symptoms contributed an additional 5.5%. Provincial site accounted for 4.0% of variance in age of diagnosis, independent of developmental skills and ASD symptoms.

Interpretation

Diagnosis of ASD occurred, on average, 19 months after parents’ initial concerns. Language and cognitive skills, symptom severity and provincial site accounted for variation in age of ASD diagnosis in this Canadian cohort. Variable presentation across the developmental continuum must be considered in planning assessment services to ensure timely ASD diagnosis so that outcomes can be improved. Policy and practice leadership is also needed to reduce interprovincial variability.

Autism spectrum disorder (ASD) affects communication, social interaction and behavioural development. Early intensive intervention is critical, improving outcomes for children with ASD as young as 18 to 24 months (1,2). Despite efforts to increase awareness of early signs (3,4), and some evidence of a recent trend toward earlier age of ASD diagnosis (hereafter, ‘AoD’) (5), recent epidemiologic studies in the USA (6–9) and UK (10) suggest that the mean AoD remains at 4 to 5 years. Thus, many children are not diagnosed before entering the school system (9–11), despite concerns being identified by their parents in the first 2 years of life (12,13). In a recent systematic review of 42 studies, milder ASD symptom severity (14–17), lower socioeconomic status (8,9,18–20) and initial parental concerns less specific to ASD (e.g., behavioural problems) (15,16) were most consistently associated with later AoD (5). Other child (e.g., sex, cognitive level), family (e.g., ethnicity) and community/health system (e.g., urban versus rural residence) were less consistently related to AoD. Conflicting findings were attributed to methodological differences (e.g., clinic versus registry-based studies), cohort effects (changes in clinical profiles and referral patterns over time) and differences in health systems across countries (5).

To our knowledge, five studies of AoD have been conducted in Canada. Two were from the National Epidemiologic Database for the Study of Autism in Canada (NEDSAC), which involved surveillance for ASD diagnoses through a variable combination of educational, health and service agency records at sites across Canada (21,22). In two studies, perinatal registries were used to establish provincial birth cohorts in Alberta (23) and Nova Scotia (24), then linked to physician billing records, a method that was validated against expert clinical diagnosis in the Nova Scotia cohort (25). In the fifth study, rates of ASD were reported for children living in Newfoundland, based on clinical records from the regional diagnostic centre (26). The mean or median AoD in these Canadian studies ranged from 39 to 65 months, possibly reflecting the wide age periods for ascertainment and data sources (see Table 1). Factors associated with timing of diagnosis also varied, mirroring findings among international studies (5), with limited data reported on the influence of child characteristics other than basic demographics. The sole study among these five that was based on clinical records rather than population registry data did not report factors associated with AoD (26).

Table 1.

Previous studies of age of ASD diagnosis and associated factors in Canada

Authors	Sample type	Age range for ascertainment of ASD cases	Case definition ASD	N	Mean age at diagnosis (months)	Older age for…
Ouellette-Kuntz et al. (22)	Four province registry (varies across sites: school records, clinic records, other service agency records)	Diagnosis to age 17 years; minimum age 18 months to ‘school-age’, depending on service provider	Clinical diagnosis	769	39.0–55.0a,b	Southeastern Ontario (vs. Newfoundland and Labrador) PDDNOS, Asperger’s (vs. autistic disorder)
Burstyn et al. (23)	Provincial perinatal registry (all live births 1998–2004) in Alberta	Maximum age of diagnosis = 10 years (ascertainment to March 31, 2008)	physician billing records (ASD- related code)	1,138	44.2	Aboriginal parent (vs. other ethnicity) Less recent birth cohort
Coo et al. (21)	Service-based surveillance system at six sites across Canada	Diagnosis to age 15 years; minimum age 18 months to ‘school-age’, depending on service provider	Chart review/ clinical diagnosis	2,180	43.3–64.7a,b	PDDNOS, Asperger’s (vs. autistic disorder) rural (vs. urban) residence no sibling with ASD
Frenette et al. (24)	Provincial perinatal registry (all Nova Scotia live births 1990–2002)	Minimum age 2 years; follow-up ranged age 3–15 years	Physician billing records (ASD-related codes)	884	55.2b	First born County of residence Comorbid ADHD
Pelly et al. (26)	Regional assessment centre serving all children living in Avalon Peninsula, Newfoundland	Age < 15 years	Clinical diagnosis	272	46.1b	Not reported

ADHD Attention Deficit Hyperactivity Disorder, ASD Autism spectrum disorder, PDDNOS Pervasive Developmental Disorder—Not Otherwise Specified

^aAcross multiple regions.

^bMedian age of diagnosis (mean not reported).

The objective of this study was therefore to assess AoD and associated factors in an inception cohort of newly diagnosed preschool children with ASD drawn from five Canadian provinces.

METHODS

Participants and setting

Participants were drawn from a Canadian longitudinal study (‘Pathways in ASD’), which includes sites in Vancouver, Edmonton, Hamilton, Montreal and Halifax. Inclusion criteria were: (a) within 4 months of receiving clinical diagnosis of ASD, confirmed by the Autism Diagnostic Observation Schedule (27), the Autism Diagnostic Interview-Revised (28) and expert rating of DSM-IV-TR criteria (29) and (b) chronological age between 24 and 60 months at the time of diagnosis. Children were excluded if they had (a) neuromotor disorder interfering with study assessments; (b) known genetic abnormalities or (c) severe vision or hearing impairment. Participants were recruited between 2005 and 2011. To ensure independence of observations, only one child per family participated. The study was approved by each site’s Research Ethics Board.

Measures

Demographics

The parent most knowledgeable (generally the mother) completed a questionnaire covering ethnicity, years of schooling, household income, employment and marital status.

Autism diagnostic observation schedule (ADOS)

The ADOS uses standardized activities to elicit communication, social interaction and repetitive behaviour relevant to ASD. Inter-rater reliability is excellent (27). The ADOS consists of four modules for individuals of differing language levels, the first three of which were used to assess participants in this study. The scoring algorithm is organized into two domains, social affect (SA) and restricted repetitive behavior (RRB) (30). For comparability across modules, a severity score for each domain and overall symptoms was calculated (31,32).

Autism Diagnostic Interview: Revised (ADI-R)

The ADI-R is a standardized semi-structured parent interview used in the differential diagnosis of ASD (28). It consists of three major domains: (a) language and communication, (b) reciprocal social interaction and (c) RRB. The ADI-R discriminates well between autism and other forms of developmental disability. Inter-rater reliability is excellent (33). The ADI-R includes an item that queries at what age parents first had developmental concerns about their child. Both the ADOS and ADI-R were administered by trained and research-reliable examiners.

Vineland Adaptive Behavior Scale Second Edition (VABS-II)

The VABS-II is a semistructured interview administered to parents or caregivers to assess functioning from birth to 18 years in the domains of communication, daily living skills, socialization and motor skills (34), which generate an Adaptive Behavior Composite (ABC). The VABS-II has been shown to have adequate reliability and validity (34).

Merrill-Palmer-Revised Scales of Development (M-P-R)

The M-P-R is a measure of intellectual development for children aged 2 to 78 months. The developmental index (DI) is comprised of cognitive, receptive language and fine motor scales. For this study, an M-P-R developmental quotient (MP-R DQ) was calculated as DI age equivalent divided by age, ×100.

Preschool Language Scale–4th edition (PLS-4)

The PLS–4 is a language test for children to age 83 months. It provides standard scores for total language (TL), auditory comprehension and expressive communication and has excellent test–retest reliability (36). The TL standard score was used in this study.

Child Behavior Checklist (CBCL 1.5–5)

The CBCL/1.5–5 is a measure of externalizing and internalizing behaviour problems in typically developing preschool children (37), and validated in children with ASD (38). The CBCL generates T-scores (mean = 50, SD=10) based on a normative sample of children aged 1.5 to 5 years.

Statistical approach

The data were analyzed using SPSS version 23.0.2 (39). AoD was determined from the clinical record, and the age of first concern, in a standardized manner, from the ADI-R. We examined site differences in AoD using analysis of variance and post-hoc testing using Tukey honestly significant difference test. We then examined bivariate correlations (Pearson r for continuous variables and Spearman ρ for categorical variables) to assess the association of each measure with AoD. Last, to assess associations between AoD and broader domains (i.e., developmental level, ASD symptoms, socio-demographic factors, as well as site), we ran a series of hierarchical regression models, including predictors with significant bivariate correlation with AoD (significance set at P<0.01 to adjust for multiple comparisons). Linear regressions were performed using Mplus (version 7.4) (40) which incorporates a full information maximum likelihood approach for handling missing data.

RESULTS

Sample characteristics

Participants were 421 children with ASD (356 boys and 65 girls) across five sites (Vancouver [21.9%] Edmonton [16.6%], Hamilton [16.2%], Montreal [31.8%] and Halifax [13.5%]). Efforts were made to recruit all newly diagnosed children with ASD at each site, with a mean participation rate of 60%. Descriptive statistics for study measures are presented in Table 2 and stratified by site in Appendix 1. No child was excluded due to severity of cognitive nor motor impairment. One child was excluded due to a diagnosis of Rett Syndrome subsequent to enrolment.

Table 2.

Participant descriptive statistics

		N	%
Sex
Male		356	84.6
Female		65	15.4
Site
Halifax		57	13.5
Montreal		134	31.8
Hamilton		68	16.2
Vancouver		92	21.9
Edmonton		70	16.6
Marital status
Married/common law		339	80.5
Other		47	11.2
Parent most knowledgeable years of schooling
13+ years		215	51.1
<13 years		167	39.7
Household Income
>$80,000		143	34.0
<$80,000		237	56.3
Ethnic/cultural heritage
Caucasian		286	67.9
Other		101	24.0
	n	Mean	Standard Deviation	Minimum	Maximum
Child’s age at diagnosis (months)	421	38.21	8.71	19.17	59.57
Age at time of consent (months)	421	39.87	9.00	21.14	61.15
VABS-2 Adaptive Behaviour Composite	399	72.76	10.13	49	101
PLS-4 Total Language standard score	386	65.18	19.17	50	136
M-P-R DQ (MP-R DAE/age ×100)	388	56.00	21.00	5	145
ADI T1 Social domain total	405	17.26	5.14	3	29
ADI T1 Communication domain total	405	11.97	3.37	0	23
ADI T1 Restricted Repetitive Behaviour domain total	405	5.14	2.29	0	12
ADOS T1 Overall Severity Metric	406	7.57	1.70	2	10
ADOS T1 Restricted Repetitive Behaviors Severity Metric	406	7.86	1.74	1	10
ADOS T1 Social Affect Severity Metric	406	7.39	1.83	2	10
CBCL Externalizing problems T-score	365	56.15	10.40	28	89
CBCL Internalizing problems T-score	365	60.24	9.28	37	85

Age of diagnosis by site

Mean AoD was 38.2 months (SD=8.7 months), on average 19.0 months (SD=8.2 months) after parents reported that they first recognized concerns. There was a modest but significant difference in AoD across sites (F(4,416)=4.31; P=0.002). Post-hoc analysis showed that at two sites, Vancouver and Halifax, children were significantly older at diagnosis (by 4.5 and 5.0 months, respectively) than the Montreal site, where mean AoD was 36.2 months. Edmonton and Hamilton AoD (respective means of 38.1 and 37.7 months) did not differ from that of Montreal.

Child and family characteristics related to age of diagnosis

Bivariate correlations between AoD and child and family characteristics are summarized in Table 3; AoD was significantly related to the child’s cognitive level based on the M-P-R DQ (r=0.231; P<0.001) and language level on the PLS-4 (r=0.256; P<0.001). For illustration, we compared AoD, dividing the sample into quartiles based on the M-P-R DQ, revealing overall differences among quartiles (F(3,384)=10.22; P<0.001). Children in the top quartile (DQ>66) had an older AoD (mean = 41.9 months) compared to the lower three quartiles, which did not differ statistically from one another (means 35.1 to 38.3 months). A similar pattern of findings was obtained when stratifying the sample using the PLS-4 TL score: those in the top quartile (TL>74) were diagnosed later than those in the lower three quartiles (mean 42.8 months versus means 34.2 to 38.6 months, respectively). AoD was also later as ADI-R RRB symptom severity increased (r=0.232; P<0.001). To further explore this finding, we stratified the ADI-R RRB symptoms into ‘Insistence on Sameness’ and ‘Repetitive and Sensory Motor Behavior’ components (41); AoD was positively related to both components (r=0.145, P=0.004 and r=0.193, P<0.001, respectively). AoD was not related to child sex, ADI-R social or communication symptoms, internalizing and externalizing behavioural problems, nor family demographic factors (ethnicity, income, parent education, lone- versus two-parent status).

Table 3.

Bivariate correlations between potential predictors and age at diagnosis for the full sample

Correlations
Continuous measures	N	Pearson r
VABS-2 Adaptive Behaviour Composite	399	−0.001
PLS-4 Total Language standard score	386	0.256*
M-P-R DQ (MP-R DAE /age)	388	0.231*
ADI-R Social domain total	405	−0.089
ADI-R Communication domain nonverbal/verbal total	405	0.125
ADI-R Restricted Repetitive Behaviors domain total	405	0.232*
ADOS Restricted Repetitive Behaviors Severity Metric	406	−0.078
ADOS Social Affect Severity Metric	406	0.040
CBCL Externalizing problems t-score	365	0.065
CBCL Internalizing problems t-score	365	0.085
Categorical measures	N	Spearman rho (ρ)
Site	421	0.006
Child’s sex	421	0.000
Current marital status	386	−0.029
Years of schooling parent most knowledgeable successfully completed	382	−0.072
Estimated household income	380	0.082
Parent most knowledgeable ethnic or cultural heritage	387	−0.049

*Correlation is significant at the 0.01 level (two-tailed).

Multivariable models of age of diagnosis

Hierarchical regression models were run to examine the independent contributions of factors associated with AoD based on bivariate correlations significant at the 0.01 level or less (Table 4). Child developmental characteristics (language and cognitive skills) accounted for 6.8% of total variation in AoD (P<0.001). ASD symptom severity added 5.5% variance explained (F-change(1,421)=21.16; P<0.001). Site accounted for 4.0% added variance in AoD (F-change(4,421)=4.88; P=0.001).

Table 4.

Regression models for factors predicting age of diagnosis

	Developmental measures			Plus measures of ASD symptoms			Plus provincial site
Parameter	B	SE	P-value	B	SE	P-value	B	SE	P-value
Intercept	38.227	0.41	<0.001	38.233	0.400	<0.001	40.822	1.348	<0.001
M-P-R DQ (M-P-R DAE/age)	0.062	0.033	0.060	0.074	0.032	0.022	0.089	0.033	0.009
PLS-4 total language standard score	0.058	0.036	0.111	0.036	0.036	0.319	0.022	0.036	0.542
ADI-R Restricted Repetitive Behaviours				0.816	0.179	<0.001	0.851	0.184	<0.001
Site*									0.001
[Site=1]							4.540	1.462	0.002
[Site=2]							−0.076	1.220	0.951
[Site=3]							0.492	1.396	0.725
[Site=4]							2.769	1.304	0.034
Adjusted R-squared	0.068			0.113			0.153

Bold values indicate P < 0.05.

ASD Autism spectrum disorder; B Regression coefficient; SE Standard error of estimate of regression coefficient.

*Reference site is Edmonton; sites 1–4 are Halifax, Montreal, Hamilton and Vancouver. ‘Site’ indicates total variance accounted for by comparisons among the five sites.

DISCUSSION

This study examined how child and family factors influenced AoD among Canadian preschool children with ASD, leveraging an inception cohort in five provinces. Mean AoD was 38.6 months and was influenced by child language and cognitive skills, ASD symptom severity, and to a lesser extent, site. The child’s sex did not influence AoD, nor did family demographics including ethnicity, income, parent education and lone- versus two-parent status. Overall, the measured factors accounted for a relatively small proportion of overall variance in AoD.

Mean AoD in this sample was at the lower end of the range reported in Canadian studies over the past decade (21–24) and lower than the mean age of 4 to 5 years reported in recent US and UK studies (5–10). However, only children diagnosed before age 5 were included in the current study, whereas other studies have ascertained to 8 years (6,7) or older (23,24). Regardless, mean AoD was, on average, 19 months later than when parents of children with ASD reported that they first had developmental concerns, a prolonged gap similar to that reported in previous studies (17,42,43). This highlights the importance of improving care pathways leading to diagnosis. Arguably, AoD of 38.6 months is unacceptably high for children whose symptoms clearly met diagnostic criteria in the preschool period, particularly as evidence accumulates for diagnosis-specific interventions targeted to children as young as 18 to 24 months (44). Early evidence-based intervention improves skill development and functioning and reduces long-term societal costs associated with ASD-related disability (45,46).

How can earlier diagnosis of ASD be achieved? Universal or targeted screening is one option but the potential benefits of ASD screening are controversial. The US Task Force for Preventative Health Care concluded that evidence is lacking (specifically, clinical trial evidence of long-term benefits of screening asymptomatic children) (47), whereas the American Academy of Pediatrics recommends universal screening (3). The Canadian Task Force on Preventative Health Care recently recommended against universal screening for developmental delay in preschool children with no apparent signs, but did not address ASD screening (48). Given that parent-report questionnaires have been shown to identify ASD earlier (47) and more consistently (49,50) than open-ended inquiry regarding parental concerns, further consideration is needed regarding using such tools; for example, in primary care settings where developmental concerns may come to practitioners’ attention early.

In this Canadian study, AoD was influenced by children’s developmental functioning, ASD symptom severity, as well as provincial site. Children in the top quartile of cognitive functioning (DQ>67, roughly corresponding to those without intellectual disability) were diagnosed 5 months later than children in the lower three quartiles (who did not differ from one another), with similar findings by language levels (i.e., only those with the highest level of language skills were diagnosed later, not all verbal children with ASD). Conversely, in a recent study that included data from 18 European countries, minimally verbal and nonverbal children were diagnosed later with ASD than verbal children (51). Intellectual and language skills have been variably associated with AoD in other previous studies (5) most of which did not stratify by developmental levels nor examine non-linear relationships (e.g., whether children at either end of the continuum may be diagnosed later). The association between advanced skills and later diagnosis is consistent with parent-reported experience (11). Prospective studies of infants at familial risk of ASD have also found that children with more advanced language and cognitive development are diagnosed later than those with more severe presentations, even with intensified surveillance (52,53). Children with ASD at the highest levels of functioning are disadvantaged by the following predicament: their stronger skills predict a more positive response to early intervention (54) yet they tend be diagnosed later, so are less likely to access such treatment. More attention is needed to the early clinical manifestations of ASD in such children to avoid delayed detection and ensure that children with ASD at all levels of functioning have the opportunity to receive intervention in a timely fashion.

As well, higher levels of parent-reported ASD symptoms in the RRB domain (indexed by the ADI-R) were associated with later ASD diagnosis. Both components—‘Insistence on Sameness’ and ‘Repetitive and Sensory Motor Behaviour’—were associated with older AoD. Previous studies have generally reported a negative (or no) association between ASD symptom severity and AoD (5). Thus the positive associations in this study are surprising, albeit modest, and may relate in part to reduced RRB symptom reporting among the youngest participants in this preschool age sample. Family demographics including income and parent education did not influence AoD within this cohort, in contrast to US studies, where such factors are among those most consistently identified. This may suggest that Canada’s universal healthcare may buffer against sociodemographic factors that affect access to assessment elsewhere—particularly in the USA—with the caveat that research participation biases and the narrow age range of our sample may have masked potential disparities in AoD present in the broader community. That said, in previous Canadian studies that reported on the influence of such factors, household income (neighbourhood-level data) was not associated with AoD (21,22). There were also site differences in AoD of up to 5 months. Clearly, the ideal would be no inter-provincial variation for such a serious health issue as ASD that affects young children, and with lifelong implications.

Strengths of this study include a relatively large sample from across Canada, an inception cohort design with consistency in ascertainment across sites and in standardized data collection including child and family characteristics to assess as potential correlates of AoD, and a multivariable modeling approach. However, there are several limitations. The sample was drawn from families who had agreed to participate in a longitudinal research study, which may have restricted variation related to socio-demographic variables. Ethical considerations precluded data collection on nonparticipants, which would have allowed formal assessment of participants’ representativeness. Diagnosis was limited to below 5 years, so mean AoD is specific to this age group and therefore represents an underestimate relative to studies that have ascertained to older ages. The availability of longitudinal data from an extensive battery of standardized measures provided a unique opportunity to assess clinical factors related to AoD in Canada but we acknowledge that associations may have been weakened by the constrained age interval, and other factors may operate in older children. As well, participants were recruited from regional academic diagnostic centres, which were based in urban areas with focused expertise in ASD. Factors influencing AoD in rural and remote communities and/or urban areas with less ASD assessment capacity or expertise may well differ. Finally, health service factors, including those that influence time between referral and diagnostic assessment, likely contribute significantly to AoD. In particular, data on wait times from initial referral to confirmation of diagnosis, an important driver of age of diagnosis of ASD, were not collected on this longitudinal cohort. Indeed, a large proportion of variance in AoD remained unexplained with data available in the present study, emphasizing the need for further research. For example, it would be helpful to know how variation in diagnostic models and service eligibility across Canada influence care pathways (e.g., assessment rates, wait times, access to intervention) as well as families’ experiences.

CONCLUSION

Reducing AoD will require attention to early clinical manifestations across the developmental continuum, educational support for community-based healthcare providers, and enhancement of the capacity and efficiency of care pathways to ensure timely access to diagnostic assessment. Further national and inter-provincial/territorial collaboration aimed at improving early detection strategies—linked to early ASD treatment—is urgently needed, including for children with more advanced intellectual and language functioning. Indeed, we encourage policy-makers and practitioners to take action on reducing inter-provincial variation and advancing evidence-informed assessment practices that accelerate ASD diagnosis. Such efforts, while time- and resource-intensive, have the potential to enhance long-term outcomes and quality of life for the many young Canadians living with ASD, and their families.

Appendix 1. Descriptive statistics for measures at T1 by site (entries are N, mean and standard deviation)

Site		M-P-R DQ (MP-R DAE/age)×100	PLS-4 T1 total language standard score	ADOS T1 Overall Severity metric	ADOS T1 Social Affect Severity metric	ADOS T1 Restricted Repetitive Behaviours Severity metric
1	N	51	43	54	54	54
	Mean	59	69.28	7.20	7.35	7.15
	Standard deviation	18	17.41	1.93	2.06	2.09
2	N	132	128	133	133	133
	Mean	56	64.31	7.49	7.67	7.21
	Standard deviation	18	17.92	1.81	1.82	1.60
3	N	66	67	66	66	66
	Mean	51	61.42	7.79	7.24	8.58
	Standard deviation	19	17.01	1.68	1.93	1.76
4	N	77	84	89	89	89
	Mean	54	65.56	7.89	7.52	8.40
	Standard deviation	18	18.60	1.45	1.71	1.36
5	N	62	64	64	64	64
	Mean	62	67.61	7.41	6.81	8.31
	Standard deviation	30	24.54	1.57	1.59	1.46