Maternal and Child Health Journal

, Volume 16, Supplement 1, pp 151–157

Parent-Reported Prevalence of Autism Spectrum Disorders in US-Born Children: An Assessment of Changes within Birth Cohorts from the 2003 to the 2007 National Survey of Children’s Health

Authors

    • National Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention
  • Catherine Rice
    • National Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention
  • Marshalyn Yeargin-Allsopp
    • National Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention
  • Coleen A. Boyle
    • National Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention
  • Michael D. Kogan
    • Maternal and Child Health Bureau, Health Resources and Services Administration
  • Carolyn Drews
    • Department of Epidemiology, Rollins School of Public HealthEmory University
  • Owen Devine
    • National Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention
Article

DOI: 10.1007/s10995-012-1004-0

Cite this article as:
Schieve, L.A., Rice, C., Yeargin-Allsopp, M. et al. Matern Child Health J (2012) 16: 151. doi:10.1007/s10995-012-1004-0

Abstract

The prevalence of autism spectrum disorders (ASD) from the 2007 National Survey of Children’s Health (NSCH) was twice the 2003 NSCH estimate for autism. From each NSCH, we selected children born in the US from 1990 to 2000. We estimated autism prevalence within each 1-year birth cohort to hold genetic and non-genetic prenatal factors constant. Prevalence differences across surveys thus reflect survey measurement changes and/or external identification effects. In 2003, parents were asked whether their child was ever diagnosed with autism. In 2007, parents were asked whether their child was ever diagnosed with an ASD and whether s/he currently had an ASD. For the 1997–2000 birth cohorts (children aged 3–6 years in 2003 and 7–10 years in 2007), relative increases between 2003 autism estimates and 2007 ASD estimates were 200–600 %. For the 1990–1996 birth cohorts (children aged 7–13 years in 2003) increases were lower; nonetheless, differences between 2003 estimates and 2007 “ever ASD” estimates were >100 % for 6 cohorts and differences between 2003 estimates and 2007 “current ASD” estimates were >80 % for 3 cohorts. The magnitude of most birth cohort-specific differences suggests continuing diagnosis of children in the community played a sizable role in the 2003–2007 ASD prevalence increase. While some increase was expected for 1997–2000 cohorts, because some children have later diagnoses coinciding with school entry, increases were also observed for children ages ≥7 years in 2003. Given past ASD subtype studies, the 2003 “autism” question might have missed a modest amount (≤33 %) of ASDs other than autistic disorder.

Keywords

AutismPrevalenceBirth population

Introduction

Autism spectrum disorders (ASDs), which include autistic disorder, Asperger disorder and pervasive developmental disorder not otherwise specified (PDD-NOS), are characterized by impairments in social interactions and communication and restricted, repetitive and stereotyped behaviors and interests [1]. ASD prevalence increased markedly in the United States and other developed countries over the past several decades [217]. Recent studies indicate that in the United States, ASD prevalence has continued to increase and has reached ~1 % [18, 19].

Nine of ten sites participating in the Autism and Developmental Disabilities Monitoring (ADDM) Network in the 2002 and 2006 surveillance years reported an increase in ASD prevalence during that 4-year period; the average increase was 57 % (a change from 6.0 to 9.4 cases/1,000 children) [19]. ADDM tracks ASDs in 8-year old children residing in select U.S. populations through periodic reviews of special education and specialist health records. Reports from the 2003 and 2007 phases of the National Survey of Children’s Health (NSCH), also suggest a large ASD increase during the 4-year interval between surveys, from 5.5 to 11.0 cases/1,000 U.S. children aged 3–17 years [11, 18].

Three general underlying explanations should be considered as potentially contributing to the observed prevalence increases: (1) a change in how prevalence was measured by surveys and surveillance systems (internal measurement effects); (2) changes in ASD awareness, screening tests, access to diagnostic services, diagnostic criteria, and special education placements that all lead to a greater identification of children with ASDs in the community (external population identification effects); and (3) changes in either ASD genetic susceptibility among persons having children within a population or non-genetic risk factors (etiologic genetic and/or environmental effects).

Because the NSCH covers children over a wide age range (3–17 years), prevalence estimates from 2003 and 2007 can be compared within subgroups based on both child’s age at survey and child’s birth year (i.e. birth cohorts). Comparison of population prevalence within birth cohorts is particularly informative as it holds both genetic and non-genetic prenatal and early postnatal etiologic exposures constant. Accumulating evidence supports the prenatal period as the likely key exposure time for non-genetic risk factors for ASD [20, 21]. Thus, prevalence change observed within a given birth cohort is most likely explained by either internal measurement effects or external population identification effects.

We estimated birth-cohort specific ASD prevalence and prevalence change between the two NSCH surveys for children born between 1990 and 2000 (and thus aged 3–13 years in 2003 and 7–17 years in 2007). Although these were independent cross-sectional surveys, they were both nationally representative of the U.S. population of children. Additionally, we limited our analysis to U.S.-born children to eliminate potential differences from in-migration. Because the ASD/autism question was revised between surveys, we are unable to separate internal from external identification effects. Nonetheless, this study provides information relevant to the most recent NSCH and ADDM 4-year ASD trends and examines the impact of composite internal and external identification effects.

Methods

Sample Selection

The 2003 and 2007 NSCH were nationally representative random-digit-dial telephone surveys on the health of children <18 years of age. The respondent was a parent or guardian residing in the household. Details of the survey methodology were published previously [22, 23]. Altogether, 102,353 children were included in the 2003 NSCH and 91,642 children were included in the 2007 NSCH. Estimates from both surveys are weighted to reflect the U.S. non-institutionalized population of children in the respective survey years; samples for both NSCH cycles were independently drawn.

From each NSCH dataset, we initially selected all children who were: U.S.-born (95 and 96 % of children in 2003 and 2007 surveys, respectively); aged 3–17 years; and not missing data on autism (2003) or ASD (2007) (<1 % missing for each survey). This selection included 80,937 and 74,575 children from 2003 and 2007 surveys, respectively. We conducted descriptive analyses to compare prevalence estimates by child age at survey from these samples.

The remaining analyses were organized according to child’s birth year, and thus, were necessarily limited to birth cohorts represented in both surveys, 1990 through 2000. These final samples included 56,565 children who were U.S.-born in 1990–2000 and aged 3–13 years at 2003 survey administration and 56,728 children who were U.S.-born in 1990–2000 and aged 7–17 years at 2007 survey administration.

Autism Ascertainment

In the 2003 NSCH, autism was ascertained from the question: “Has a doctor or health professional ever told you that [CHILD’S NAME] has autism?” There were no follow-up questions on the specific autism spectrum disorder or on whether the child currently had the condition. In the 2007 NSCH, ASD was ascertained with two questions: “Were you ever told by a doctor or other healthcare provider that [CHILD’s NAME] had autism, Asperger’s disorder, pervasive developmental disorder, or other autism spectrum disorder?” and if yes, “Does [CHILD’s NAME] currently have autism or ASD?”

Data Analysis

For the 2003 NSCH survey sample, we calculated point prevalence estimates for each year of child’s age based on the single autism question. For the 2007 NSCH sample, we calculated age-specific prevalence estimates in two ways: based only on an affirmative response to the single “ever told your child has an ASD” question and based on affirmative responses to both ASD questions, thus indicating the child “currently has an ASD”. While we believe it is most appropriate to report population-based prevalence of ASD using the more conservative “current ASD” estimate, our objective was to compare estimates from the 2003 and 2007 surveys. We thus examined both prevalence measures from the 2007 NSCH as the 2007 “ever ASD” question is most closely aligned with how autism was ascertained in the 2003 survey. Moreover, there is a large difference between the two 2007 estimates; ~40 % of children classified as “ever ASD” were not additionally classified as “current ASD” [18].

For the birth cohort analyses, we compared the 2003 “ever autism” prevalence estimate for each birth year to the corresponding 2007 “ever ASD”, and “current ASD” estimates. We calculated relative differences for each comparison and determined statistical significance using z tests for the difference in proportions from independent samples with the assumption of unequal variances.

We also assessed the distribution of several demographic characteristics among autism cases reported in 2003 and both “ever” and “current” ASD cases reported in 2007. These analyses were stratified into three broad birth cohort groups: 1990–1992; 1993–1996; and 1997–2000. Assessment of individual birth years was precluded by small sample sizes. We assessed the statistical significance of differences between 2003 and 2007 distributions using z tests.

All estimates were weighted to reflect the US non-institutionalized population of children. Standard errors used in statistical tests were adjusted to account for the complex sample design with SUDAAN software (Research Triangle Institute, Research Triangle Park, NC).

Human subjects review was not required for this study since this was a secondary analysis of de-identified datasets.

Results

Among children aged 4 years or older, there was a greater than twofold difference in age-specific prevalence estimates of diagnosed ASD (“ever ASD”) reported in 2007 than diagnosed autism in 2003 (Fig. 1). The highest age-specific autism estimates from the 2003 survey were observed for children ages 7 and 8 years (9.9 and 9.5 cases/1,000 children) and estimates ranged from 6.9 to 9.9 cases/1,000 among children ages 7–10 years; lower estimates were observed for children aged 3–6 years (range 3–5 cases/1,000) and 11–17 years (range 4–6 cases/1,000). A peak age was not observable among children in the 2007 survey. Although we observed year-to-year fluctuations, estimates of “ever ASD” were all quite high for children ages 5–13 years (range 18–29 cases/1,000). While 2007 “current ASD” estimates were lower (range 12–17 cases/1,000) than 2007 “ever ASD” estimates, the general pattern by child age was similar.
https://static-content.springer.com/image/art%3A10.1007%2Fs10995-012-1004-0/MediaObjects/10995_2012_1004_Fig1_HTML.gif
Fig. 1

Point prevalence estimates for autism in 2003 and 2007 by child age at survey, U.S. born children 3–17 years of age at survey, 2003 and 2007 National Survey of Children’s Health

Among children who were U.S.-born in 1990–2000, the estimated overall ASD/autism prevalence was substantially higher in 2007 than 2003 regardless of whether the 2007 estimate included children reported to have ever had ASD or children only reported to currently have ASD. We observed differences within most birth-cohort strata. Among the youngest children, those born from 1997 to 2000, the estimated autism prevalence in 2003, when they were ages 3–6 years, was dramatically lower than the ASD prevalence estimates in 2007, when they were 7–10 years of age (Table 1). The relative increase was between 300 and 600 % when the 2003 estimates were compared to 2007 “ever ASD” estimates and between 200 and 400 % when compared to 2007 “current ASD” estimates.
Table 1

Autism estimates within birth cohorts, U.S. born children 3–17 years of age at survey, 2003 and 2007 National Survey of Children’s Health

Birth year

2003 Survey

2007 Survey

% Difference

Child age

“Ever” autism prevalence (/1,000)

SE

Child age

“Ever” ASD prevalence (/1,000)

SE

“Current” ASD prevalence (/1,000)

SE

2003 versus 2007 “ever” ASD

2003 versus 2007 “current” ASD

1990

13

4.2

1.6

17

10.1

1.7

6.5

1.3

140.5*

54.8

1991

12

5.9

1.5

16

16.4

3.4

11.2

3.1

178.0*

89.8**

1992

11

5.9

1.5

15

13.1

2.3

6.9

1.6

122.0*

16.9

1993

10

8.7

2.3

14

11.2

2.3

6.8

1.5

28.7

−21.8

1994

9

6.9

1.6

13

27.5

7.5

13.6

4.9

298.6*

97.1**

1995

8

9.5

3.3

12

19.9

4.6

17.4

4.5

109.5**

83.2**

1996

7

9.9

3.3

11

26.0

8.4

11.3

5.2

162.6**

14.1

1997

6

3.3

0.8

10

22.3

5.2

16.0

4.8

575.8*

384.8*

1998

5

4.1

1.0

9

21.7

5.1

12.9

2.5

429.3*

214.6*

1999

4

5.0

1.3

8

21.4

5.8

15.9

5.5

328.0*

218.0**

2000

3

2.9

1.0

7

18.9

3.5

13.3

3.0

551.7*

358.6*

Total 1990–2000

3–13

6.0

0.6

7–17

18.9

1.5

12.0

1.1

215.0*

100.0*

* P < 0.05; ** P < 0.10 one-tailed z test

Among children born between 1990 and 1996, the 2003 prevalence estimates (at ages 7–13) were also notably lower than the 2007 estimates (at ages 11–17). The relative differences were lower than those observed for children in the later birth cohorts, but were still sizable. When we compared the 2003 estimate with the 2007 “ever ASD” estimate, the relative difference was between 100 and 200 % for five of the seven 1990–1996 birth cohorts, was much higher than this range (299 %) for one birth cohort, and was much lower (29 %) for the remaining birth cohort. When we compared the 2003 estimate to the 2007 “current ASD” estimate, differences remained, although the magnitude was reduced. For six birth cohorts, the 2007 estimate for “current ASD” exceeded the 2003 estimate of “ever autism” and for three of these cohorts, relative differences were greater than 80 %.

We observed comparable patterns of results as those presented after sample restriction to males and non-Hispanic white (NHW) children (data not shown). Small sample sizes precluded us from conducting additional subgroup analyses.

Given the prevalence differences, we assessed whether the demographic profile of children reported to have autism/ASD in 2003 versus 2007 varied. Among the youngest children (1997–2000 birth cohorts), 2007 “ever” ASD cases were less likely than 2003 “ever” autism cases to be NHW and more likely to come from households in which the highest education level was ≤high school. We observed similar findings for the oldest children (1990–1992 birth cohorts) but not for children born from 1993 to 1996. In addition to differences in autism cases between surveys, for two of the three birth cohort groups examined, the proportion of non-Hispanic black (NHB) children was notably lower among 2007 “current” ASD cases than 2007 “ever” ASD cases (Table 2).
Table 2

Percentage distribution of demographic factors among children classified as having autism or ASD in the 2003 and 2007 National Surveys for Children’s Health according to year of birth

Birth year and demographic factor

2003 Survey year “ever” autism cases

2007 Survey year “ever” ASD cases

2007 Survey year “current” ASD cases

1990–1992 Births

N = 90

N = 302

N = 181

% with factor

% with factor

% with factor

Race/ethnicity

 Hispanic

2.5

10.1*

12.4*

 Non-Hispanic White

71.1

60.5

63.5

 Non-Hispanic Black

24.3

20.3

19.5

 Non-Hispanic other/multi-racial

2.1

9.1

4.7

<200 % Federal poverty limit

33.4

41.5

43.5

Highest household education high school or less

10.9

36.8*

31.2**

Female

24.2

35.1

35.9

1993–1996 Births

N = 159

N = 369

N = 247

% with factor

% with factor

% with factor

Race/ethnicity

 Hispanic

13.8

13.4

13.5

 Non-Hispanic White

65.7

68.7

78.0

 Non-Hispanic Black

18.0

13.4

4.0**,++

 Non-Hispanic other/multi-racial

2.5

4.6

4.4

<200 % Federal poverty limit

52.5

35.7

24.8*

Highest household education high school or less

34.6

22.4

16.8*

Female

17.3

17.2

18.0

1997–2000 Births

N = 104

N = 371

N = 253

% with factor

% with factor

% with factor

Race/ethnicity

 Hispanic

10.4

15.1

19.5

 Non-Hispanic White

75.5

59.1*

65.0

 Non-Hispanic Black

9.0

19.1

7.9++

 Non-Hispanic other/multi-racial

5.1

6.8

7.5

<200 % Federal poverty limit

39.8

54.0

51.6

Highest household education high school or less

25.6

42.6**

39.3

Female

20.0

20.2

16.7

P < 0.05; ** P < 0.10 two-tailed z test comparing 2007 estimate to 2003 estimate

++ P < 0.10 two-tailed z test comparing 2007 “current ASD” estimate to 2007 “ever ASD” estimate

Discussion

We previously reported nationally-representative autism and ASD prevalence estimates from the 2003 NSCH [11] and 2007 NSCH [18] and noted a twofold difference. Here, we demonstrate the difference remains evident when either child age at survey or child’s birth year is held constant.

Our observation that in 2003, peak ASD prevalence rates occurred at 7–10 years of age is consistent with a 1996 population-based study in metropolitan Atlanta that documented peak prevalence at age 8 years [7]. Recent studies indicate that ASD diagnosis age is decreasing for U.S. children. Studies from the ADDM Network indicate the median age at earliest ASD diagnosis was 5.7 years among children born in 1994 [24] and 4.4 years among children born in 1998 [19]. Among children with full-spectrum autism served at a California regional developmental disabilities center, 12 % more children who were born in 1996 than 1990 were diagnosed by age 5 years [16]. Moreover, the mean age at entry into this service system decreased from 6.9 to 3.3 years among children born in 1987 and 1994, respectively [5]. Past studies do not completely overlap with birth cohorts examined here. However, our finding that by 2007, ASD prevalence among children 5 years of age was comparable to estimates for older children is suggestive of a continued decrease in diagnosis age.

In strata based on birth year rather than child age at survey, prevalence was also consistently higher in 2007 than 2003, even for cohorts at the peak prevalence age (7 years or more) in 2003. Holding birth year constant eliminates potential prevalence variation that might occur across birth cohorts from the same target population because of year-to-year differences in ASD genetic susceptibility among persons having children. Also, differences observed within a given birth cohort are very unlikely due to secular changes in environmental triggers that affect children prenatally. Additionally, because we defined our study population as US-born children aged 3–13 years in 2003, differences within birth cohort are unlikely due to secular changes in those environmental triggers that might impact children post-natally up to age 3 years. Prevalence differences within birth cohorts are thus explained by either measurement differences or additional identification and classification of children between the two surveys.

We lacked data to discern specific factors contributing to the birth-cohort trends. Clearly, there was a measurement change. The survey question verbiage was expanded in 2007 to explicitly include all ASDs, rather than using the less specific term “autism”. A recent study of 8-year old children classified as having ASDs in Metropolitan Atlanta, showed that 66 % of cases had documentation of autistic disorder in health records or were receiving special education services under the specific “autism” eligibility category; the remainder had documentation of Asperger disorder (5 %), PDD-NOS or ASD-NOS (12 %), or did not have documentation of an ASD diagnosis/education placement, but had documentation of behaviors consistent with ASD diagnostic criteria (17 %) [25]. Thus, we might expect that use of the non-specific “autism” terminology in the earlier NSCH resulted in one-third of the “true” ASD cases being missed. However, the differences between 2003 “ever autism” and 2007 “ever ASD” estimates were much greater than 33 %. Moreover, it is unlikely that all ASD cases beyond autistic disorder were missed in 2003. In fact, the 2003 NSCH “autism” estimate for children aged 7–9 years is slightly higher than the 2002 ADDM “ASD” rate for children aged 8 years [14]. Thus, an affirmative response to the 2003 NSCH autism question likely included some children with Asperger’s Disorder and PDD-NOS.

Beyond measurement effects, external factors such as increased population awareness and access to ASD diagnostic services were likely important contributors to the prevalence differences. Some prevalence increases were expected for children aged 3–6 years in 2003, because some children have later diagnoses coinciding with school entry. Also, for some children, recognition and follow-up of social and communication difficulties by health care providers might be delayed because they lack access to comprehensive care and/or have milder symptoms. Nonetheless, the 328–575 % prevalence increases observed for these cohorts were striking. Additionally, prevalence increases were also observed for children in earlier birth cohorts who were already aged 7 and older in 2003. Within both late and early birth cohorts, children classified as “ever ASD” cases in 2007 were more likely to be a racial/ethnic minority and have parents with less education than children classified as “ever autism” in 2003. This demographic shift in case profile is consistent with a hypothesis of diagnostic delay among children whose parents had less access to services and/or less ability to navigate complex healthcare and education systems serving children with developmental concerns.

We previously reported that 2007 prevalence estimates for “current ASD” were 40 % lower than those for “ever ASD” [18]. Here we demonstrate that this differential occurs within and across birth cohorts. While we lack data to assess specific reasons for the discrepancies, we found that for the 1993–2000 birth cohorts, the subset of children reported as “current ASD” was much less likely to be NHB than the total group reported as “ever ASD”. This again, might be related to differences in ASD service access and use between racial/ethnic groups. There was no question on “current autism” in the 2003 survey. Thus, we cannot assess whether the differential between “ever diagnosed” and “currently has condition” was similar in 2003 and 2007.

While numerous studies assessed autism/ASD prevalence trends in select populations of U.S. children [5, 6, 810, 19], few assessed changes within birth cohorts. Two studies that did present birth-cohort findings were based on special education data. Gurney et al. [6] reported increases in autism special education placements in Minnesota within birth cohorts. For example, among children born in 1991, prevalence was 2.1/1,000 at age 6, 3.5/1,000 at age 8, and 5.2/1,000 at age 10. Newschaffer et al. [8] present similar findings for U.S. children receiving special education services from 1992 to 2001. Both studies were limited by using existing administrative data without standardized diagnostic criteria. Additionally, some children with ASDs were likely missed because they were not receiving special education under the autism eligibility placement. Our study supports and expands upon previous studies by demonstrating that trends across and within birth cohorts have continued into more recent time periods. Moreover, these nationally-representative U.S. samples based on parent-reported surveys likely include a broader range of children identified as having ASDs.

Nonetheless, several limitations should be considered. In addition to the aforementioned interpretation difficulties created by the question verbiage change, data from both surveys were based on parent report without clinical validation. However, previous studies suggest high reliability for the autism question [11] and comparability between estimates from NSCH and ADDM [11, 14, 18, 19]. Stratification into single-year birth cohorts decreased estimate precision; thus, some fairly large differences between 2003 and 2007 estimates (80–90 %) had P values in the 0.05–0.10 range. Nonetheless, the pattern of results supports a true difference in estimates. All but one difference were in the same direction, and most were quite large, given the short (4-year) time period. We analyzed two independent samples drawn from the U.S. non-institutionalized population of children in the respective survey years. Thus, population changes could affect the findings. We limited the sample to U.S.-born children to eliminate in-migration effects. However, we could not address out-migration or child deaths. Finally, our results are subject to biases associated with landline telephone surveys, including non-coverage of households without landlines. These biases were minimized by incorporating non-response and non-coverage adjustments into sampling weights.

Currently, the peak prevalence age period for ASDs is in flux. Evidence suggests that ages at both first evaluation for developmental concerns related to ASDs and initial ASD diagnosis are decreasing. Nonetheless, with increased awareness, we may continue to see increases in later “catch-up” diagnoses in children missed at younger ages for various reasons related to lack of health care access or failure of recognition of ASD behaviors at young ages. This study suggests both phenomena are occurring and their impact on recent ASD prevalence estimates is likely sizable.

Copyright information

© Springer Science+Business Media, LLC (outside the USA)  2012