Autism Data Is All Over the Place

According to the Centers for Disease Control (CDC) website, autism is least prevalent in White non-Hispanic populations and most prevalent in adolescents of Asian/Pacific Islander background—just slightly above Hispanic.

Another CDC paper details how it was highest amongst Hispanic individuals and lowest amongst non-hispanic Whites.

Yet a National Institutes of Health (NIH) study listed non-Hispanic Black adolescents as having the highest prevalence and Hispanic individuals having the least.

In a 2018 report to Congress from Health and Human Services (HHS), it notes that White adolescents were more likely to be identified with autism spectrum disorder (ASD) than Black, but it was completely different in the 2014 analysis. Asian adolescents—the most prevalent according to the CDC—are barely mentioned in the report.

An analysis of 2022 National Survey of Children’s Health (NSCH) data also seems to show Black adolescents having the highest prevalence (4.5 percent) with Asians having the least (2.9 percent).

But the CDC’s National Health Interview Survey (NHIS) data also shows the condition to be distinctly associated with American Indian and Alaska Native resources (AIAN) adolescents—5.2 percent. Hispanics in that survey only had an autism rate of 2.9 percent. Both of those surveys show Asian adolescents having the lowest rates for autism, rather than the highest as shown on the CDC website.

Health survey data isn’t always reflective of the full population, and there are various biases and errors in any survey analysis. Each one may also have differing definitions of what constitutes an autism diagnosis. NCSH data is self-reported while Department of Education data comes from the schools’ designation.

Yet some of the disparities are so great as to questions how useful the datasets truly are.

For example, data from the Department of Education’s Individuals with Disabilities Education Act (IDEA) tells a very different story to that of CDC data. IDEA data is only based on those enrolled in school and within the disabilities program, so it won’t capture the complete picture of the whole autistic adolescent population. But it paints a picture as to how much the data can vary from survey to survey.

And in the IDEA data, Asian and Native American students particularly stand out for having autism rates five to six times that of other race/ethnic groups.

Wide Discrepancy in State Rates As Well

It’s not just issues with race and ethnicity. The datasets appear to diverge substantially when comparing state-by-state data. The states with the highest autism rates in the NSCH data are sometimes not even close to those in the Department of Education data.

The two datasets come from very different surveys, so there is bound to be some error and variation between the two, but their differences appear more than slight and show little correlation (Pearson: .35045).

Some states show similarities. For example, Massachusetts and Maine rank at the top in both datasets, which also aligns with a CDC 2017 report on the subject. But then there’s a more recent 2020 report that says California ranks highest, but in the NSCH and Department of Education data it is ranked 14th and 5th respectively.

Various websites detail Florida as having the highest rate, and that aligns with the NSCH data, but in the Department of Education data it is ranked 15th.

NSCH data uses complex weighting methodology to adjust for some of its sampling issues.

NCSH State Sampling Biases

In particular, the NCSH survey data appears to skew heavily towards certain states and less towards others that might alter the outcome. The more samples there are in one state relative to the population, the more accurate the data is likely to be.

For example, while heavily populated states like New York and California have the most data points, other states stick out for having a much higher sampling rate relative to their population.

Ideally, sampling rates for a state should parallel the population size. Larger states should have larger sample sizes. But for the NCSH data, that is not always the case.

Wyoming stands out for having the ninth most samples in the survey data, yet it is one of the least populated states with only 132,424 adolescents based on 2021 population estimates. A similar issue exists for Vermont, Alaska, North Dakota, and the District of Columbia—relatively large amounts of sample data for small populations.

Texas has about the same amount of sample data as Vermont despite it being a much larger and much more populous state. The same goes for Florida and Illinois.

Sampling Issue In Flawed Vaccine-Autism Paper

Issues related to sampling errors and autism also surrounded some of the accusations of a vaccine-autism link. Andrew Wakefield is the disgraced former physician best known for fraudulently promoting the theory that childhood vaccinations can lead to autism, but another scientist, Brian Hooker, also espoused that belief in a paper indicating a connection between autism and African American children.

Hooker’s research was based off of data from a prior study that had withheld certain data showing a high autism prevalence in African American children—double that of other races.

But the reason the data was withheld from the original paper was because of the low sample population—there were few vaccinated African American children in the cohort to make a reasonable implication—making it highly susceptible to bias.

Hooker’s paper would be retracted and the original author would eventually pen an apology for leaving out the data in the first place.

Other Biases and Error in Health Data

Health reporting data is fraught with many issues, such as the CDC’s “checkbox” issue with maternal mortality—wherein maternal deaths were heavily over-reported, including for elderly women with no likelihood of ever being pregnant. It regularly included deaths completely unrelated to giving birth like accidents and drug overdoses. The result made the United States appear to have the worst maternal mortality rates in the developed world.

Such reporting biases and errors are not confined to the U.S. Australia is sometimes reported as having the highest autism rates in the world, but according to an Australian Financial Review article, the high rates are likely driven by the country’s National Disability Insurance Scheme (NDIS)—the government disability insurance program. Reporting of autism in Australia exploded as benefits from national disability insurance grew.

Autism rates have exploded across the world in recent years, and there is ongoing debate if the high rates are a reflection of changing definitions, increased reporting, an actually growing prevalence, or some combination of all three.