How Changing a Questionnaire Nearly Doubled America's Autism Rate
About 1.25 percent of kids in the U.S. had a diagnosis of autism spectrum disorder from 2011 to 2013, according to the National Health Interview Survey. In 2014, it was 2.24 percent — a colossal move, in statistical terms.
That doesn't mean an extra 600,000 kids developed autism last year. The difference is explained by a change in the order of questions and other adjustments in the 2014 survey, the Centers for Disease Control reports today. The statistical hiccup is a lesson in the difficulty of measuring health in the general population. It's especially hard for conditions like autism — a developmental disorder marked by difficulty in communicating and socializing — that have had shifting diagnostic criteria.
The federal government has been using the National Health Interview Survey, intended to be nationally representative of U.S. households, since 1957. Surveyors visit people at home and ask questions about household members of all ages. Parents answer for their children. In recent years, the survey has questioned families of about 11,000 kids age 3 to 17 on autism and other developmental delays.
In 2011 through 2013, surveyors asked parents whether doctors had ever told them their child had an intellectual disability, and then asked about any other developmental delays. Then they provided a list of 10 conditions, including autism alongside diabetes, arthritis, and others, and asked the parents to tell them which, if any, their child had.
In 2014, rather than including autism on a list of other conditions, the survey inserted a question specifically about autism spectrum disorders. It came after the question about intellectual disabilities and before the question about other developmental delays. That apparently led some parents who might have previously chosen "other developmental delay" to indicate an autism diagnosis instead.
There were big swings in the responses about autism and other developmental delays after the survey changed. But when researchers looked at the total number of people who responded yes to either question, there was very little difference:
The more direct autism question "is likely to receive more attention and more thoughtful responses," authors from the CDC's National Center for Health Statistics write. So it's probably capturing diagnoses the earlier surveys missed.
"True year-to-year changes of the magnitude observed are unusual," the authors note, "and require abrupt or dramatic changes in the risk factors acting on the population."