Changes over Time in ASD Prevalence and Characteristics Among Children Aged 3-17 Years, National Health Interview Survey, United States, 1997-2011

Background:  Previous studies suggest a dramatic rise in the estimated prevalence of autism spectrum disorder (ASD) in the past two decades. Further information is needed on whether there have been concomitant changes in the characteristics of children identified with ASD, in order to better understand patterns of identification and potential risk factors among sub-groups of children.

Objectives:  To estimate trends in prevalence of ASD among US children aged 3-17 years in 1997-2011 and to describe differences in associations between ASD and selected sociodemographic and birth characteristics in early (1997-2003) versus late (2004-2010) time periods.

Methods:  The National Health Interview Survey (NHIS) is an ongoing national survey representing the civilian, non-institutionalized population of the United States. We included children aged 3-17 years in NHIS 1997-2011. A child was considered to have autism if a parent/caregiver answered affirmatively to the question: “Has a doctor or health professional ever told you that [CHILD] had autism?” For the 2011 survey, the term “autism spectrum disorder” replaced “autism”. Logistic regression determined associations between autism or ASD and selected characteristics including sex, race/ethnicity, age, maternal education, family income-to-poverty ratio, health insurance, and birth weight. The family income-to-poverty ratio was derived using multiple-imputation. Analyses accounted for the complex survey design, and data were weighted to provide nationally-representative estimates.

Results:

The estimated prevalence of autism increased from 0.8/1,000 children in 1997 to 10/1,000 in 2010; the prevalence of ASD in 2011 was 12/1,000. For all subgroups examined except black children, prevalence increased more than two-fold between the 1997-2003 and 2004-2010 time periods; in black children, the prevalence increase of 50% did not reach statistical significance.

In both 1997-2003 and 2004-2010, the odds of ASD were significantly higher among: males (versus females, 1997-2003 aOR:3.8(2.4-6.1), 2004-2010 aOR:3.9(2.9-5.3)) and children insured by Medicaid (versus privately insured, 1997-2003 aOR:2.1(1.1-4.1), 2004-2010 aOR:2.0(1.4-2.9)). Although findings were not statistically significant in  1997-2003, in both time periods the odds of ASD were higher among children whose mothers had a college degree (versus no  college degree,1997-2003 aOR: 1.6(0.9-2.5), 2004-2010 aOR: 1.5(1.1-2.0)) and children with birth weight >4 kilograms (versus 2.5-4 kilograms, 1997-2003 aOR: 1.2(0.7-2.0), 2004-2010 aOR:1.4(1.1-1.9)). For 2004-2010 only, the odds of ASD were significantly lower among black children (versus white, 1997-2003 aOR: 1.1(0.6-1.8), 2004-2010 aOR:0.5(0.3-0.7)) and Hispanic children (versus white, 1997-2003 aOR: 0.8(0.5-1.2), 2004-2010 aOR:0.4(0.3-0.6)).       

Because the autism question changed in 2011, we did not combine these data with the 2004-2010 grouping; nonetheless, preliminary assessment indicates a similar pattern of results for 2011 and 2004-2010.

Conclusions:  The prevalence of parent-identified ASD increased more than ten-fold overall from 1997-2010. Similar increases were observed for all sub-groups except black children, who had a smaller increase, possibly indicating a lag in identification.  While many characteristics associated with ASD were similar for the 1997-2003 and 2004-2010 time periods, in the later time period only there was a significant differential between children who were white and children who were either black or Hispanic. These data have potential implications for better understanding which population sub-groups might be under-identified and thus, under-served.