Two Fatal Technical Flaws in the DSM-5 Definition of Autism

A poorly written DSM criteria set is useless. The essential prerequisite to an accurate diagnosis is that different clinicians can agree on whether it is present or absent in a given individual. If the definition lacks precision, different people will interpret it in their own different and idiosyncratic ways.

The DSM-5 definition of Autistic Spectrum Disorder has two fatal technical flaws that make it impossible to interpret and use reliably.

A truncated version of its Criterion A reads as follows:

"A. Persistent deficits in social communication and social interaction across multiple contexts, as manifested by the following...

1. Deficits in social-emotional reciprocity;

2. Deficits in nonverbal communicative behaviors used for social interaction;

3. Deficits in developing, maintaining, and understanding relationships."

The DSM 5 examples offered for each of these three items are vague enough to overlap into normality, but I wouldn't have made a big fuss about this.

The really fatal flaw here is that no instructions are given as to whether one item, two items, or all three items must be present to make the diagnosis of Autism Spectrum Disorder. The diagnosis will vary dramatically from rater to rater, institution to institution, and place to place depending on which of these three different possible convention is chosen. It will be even more impossible than it is now to determine rates of autism and why they shift so much over time.

The second fatal flaw comes in the following statement attached to the end of the criteria set for Autism Spectrum Disorder:

"Note: Individuals with a well established DSM-IV diagnosis of autistic disorder, Asperger's disorder, or pervasive developmental disorder not otherwise specified should be given the diagnosis of autism spectrum disorder. Individuals who have marked deficits in social communication, but whose symptoms do not otherwise meet criteria for autism spectrum disorder, should be evaluated for social communication disorder. "

This throws wide open to raters the choice of using DSM-IV criteria or DSM-5 criteria depending on their personal preferences.

DSM-5 has essentially made it clinician's choice how to define and diagnose Autism Spectrum Disorder. Some will require one item from criterion A; others two; yet others three; and some will chuck DSM-5 altogether and use the very different definitions that are contained in DSM-IV.

The diagnosis of Autism is already badly muddled. There has been a forty-fold increase in rates in just 20 years. Some of this is due to the introduction of Asperger's in DSM-IV, some to improved case finding and reduced stigma, but a significant portion comes from loose and inaccurate diagnosis.

DSM-5 turns the current confusion into a complete Babel. The impossibly vague and confusing DSM-5 definition of Autism Spectrum Disorder is essentially useless for clinical or research purposes and is not a trustworthy guide for determining school services.

How could this possibly happen? Sadly, the inevitability of this kind of blooper was predictable and predicted. Criterion writing is a rare skill. Despite years of practice, I am pretty bad at it and I have known only a handful of people who could really do it well. The DSM-5 work groups had little supervision and were given far too much freedom to write criteria their own way. The result is an amateurish botch.

These egregious mistakes require immediate correction. Too much rides on the diagnosis of autism to let it ride on each clinician's random preferences.

Allen Frances is a professor emeritus at Duke University and was the chairman of the DSM-IV task force.