By the Numbers: Autism Is Not a Math Problem

At a meeting of the Icelandic Medical Association last week, Yale University child psychologist Fred Volkmar gave a presentation on how the American Psychiatric Association (APA) is changing the definition of autism. In his talk, Volkmar came to a startling conclusion: more than half of the people who meet the existing criteria for autism would not meet the APA’s new definition of autism and, therefore, may not receive state educational and medical services.

The APA defines autism in a reference guide for clinicians called the Diagnostic and Statistical Manual for Mental Disorders (DSM). The newest version of the manual, the DSM-5, is slated for publication in May 2013.

In Iceland, Volkmar presented data from an unpublished preliminary analysis of 372 high-functioning autistic children and adults with IQs above 70. He plans to publish a broader analysis later this year. On a key PowerPoint slide that Volkmar shared with Scientific American, he notes that there are 2,688 ways to get a diagnosis of autistic disorder in DSM-IV, but only six ways to get a diagnosis of autism spectrum disorder in DSM-5. Although intriguing at first glance, it turns out that both these numbers are slightly wrong—and that they are pretty much useless when comparing the DSM-IV and DSM-5. You cannot reduce autism to a math problem.

Hubble Fellow Joshua Peek of Columbia University was asked to code a computer program that would calculate the total possible ways to get a diagnosis of autistic disorder in DSM-IV and the total possible ways to get a diagnosis of autism spectrum disorder in DSM-5. You can do the math by hand, too, if you like: It all comes down to factorials. The DSM-IV criteria are a set of 12 items in three groups from which you must choose 6, with at least two items from group one and at least one item each from groups two and three. The DSM-5 criteria are a set of seven items in two groups from which you must choose five, including all three items in group one and at least two of the four items in group two. Peek's program crunched the numbers: there are 2027 different ways to be diagnosed with autism in DSM-IV and 11 ways to be diagnosed with autism in DSM-5.

One might think that those statistics make it absurdly easy to qualify for a diagnosis of autism in DSM-IV and incredibly difficult to meet the criteria for autism in DSM-5, but those numbers alone don't tell you anything unless you understand how common each symptom of autism is in the general population. Symptoms of autism are not randomly distributed throughout the population and the symptoms do not cluster together in random combinations. Research in the past decade has shown that some symptoms appear together much more often than others. In fact, that is one of the main reasons that the APA has consolidated the DSM-IV criteria for autism into fewer, denser and more accurate criteria in the DSM-5. The idea is that the DSM-IV criteria allowed for too many possible combinations, many of which rarely occur; the DSM-5 criteria, in contrast, better reflect the most common combinations of symptoms.

Specifically, the APA has merged two distinct groups of symptoms from the DSM-IV—deficits in communication and deficits in social interaction—into one group in the DSM-5 because someone with autism almost always has both kinds of symptoms.

Most psychiatrists applaud this consolidation because, as several recently published studies have shown, the new criteria are more precise: they rarely assign autism to people who do not have it. However, the DSM-5 criteria may be a little too strict, ignoring some autistic people with milder symptoms. Two recently published studies suggest an easy fix: if the new criteria require that patients meet one fewer symptom—four out of seven instead of five out of seven—high-functioning autistic people will not be excluded.