The people working on DSM 5 have difficulty getting their story straight. First they said they didn't care what impact DSM 5 would have on prevalence rates of the different mental disorders. Their indifference was so profound that the expensive DSM 5 field trials omitted the simple and obvious step of systematically comparing what rates would be using DSM 5 versus DSM IV.
Last week, the New York Times ran a front page story reporting that DSM 5 changes would dramatically reduce rates of autism. This produced an uproar in the autism community and instigated a petition against DSM 5. A more narrowly defined diagnosis, it was feared, would likely result in much reduced school and mental health services.
The DSM 5 leadership was forced to change its tune, responding to the impassioned criticism in a Medscape piece reassuringly titled "Concern Over Changes to Autism Criteria Unfounded, Says APA."
DSM 5 claims to be making only small changes that will not really effect the prevalence of autism:
We've just moved from the idea that you need 2 criteria from column A, 1 from column B, 1 from column C, and 1 additional criterion that could land in any of those categories to one requiring all 3 manifestations of impairments and social communication and interaction.
And according to the APA:
...field testing has not indicated that there will be any change in the number of patients receiving care for autism spectrum disorders in treatment centers -- just more accurate diagnoses that can lead to more focused treatment.
This makes absolutely no sense. The DSM 5 assertion of rate neutrality is, just on the face of it, completely impossible. A simple comparison of how DSM IV and DSM 5 criteria are written makes apparent that DSM 5 has to be much more restrictive.
The criteria set for Autistic Spectrum Disorder starts with two components: An 'A' criterion whose stem reads "persistent deficits in social communication and social interaction across contexts" and a 'B' criterion defined as "restricted repetitive patterns of behavior, interests or activities". DSM-IV Asperger's also starts with two components: An "A" component whose stem reads "qualitative impairment in social interaction" and a "B" component whose stem reads "restricted repetitive and stereotyped patterns of behavior, interests, and activities". So, superficially at least, they would seem to be equivalent.
But there is a huge difference in the number of items needed to satisfy each of these criteria in DSM 5 vs DSM IV. In DSM-5, the 'A' criterion is met only if ALL THREE of its problems are present (i.e., problems in social-emotional reciprocity, in use of non-verbal communicative behaviors, and in developing relationships). In contrast, DSM-IV Asperger's has four problems listed in its 'A' criterion and Only Two Of Four are required. So DSM 5 requires a 100% hit rate for A criteria problems; while DSM IV requires only 50%. The B criterion works the same way. DSM-IV Asperger's requires only one item be present, whereas DSM-5 requires two.
Here is the math. According to Dr. Brian Reichow of the Yale Child Study Center, there are 2,688 possible combinations of ways to get a diagnosis of autism using the DSM-IV rules. In stark contrast, there are only 6 possible combinations using the new DSM 5 method. Based on logic and math, one would assume that DSM IV autism is definitely broader in its reach and more heterogeneous in its composition. The much narrower DSM 5 definition would (as claimed by APA) indeed be much more specific. But APA can't have it both ways -- the flip side of specificity has to be lower rates of prevalence.
The extent of narrowing will vary in different real life situations depending on the mix of severities. The exclusion of current cases would be less an issue in tertiary care settings dealing with the severely impaired -- because they more easily meet all three of the DSM 5 'A' and two 'B' criterion items. But the rates of Asperger's should drop precipitously under DSM 5 rules.
And that is precisely what Dr. Volkmar and his colleagues at Yale found in their empirical, head to head comparison of DSM IV and DSM 5. Only 45% of those who met any DSM IV diagnosis of autism (ie of Autistic Disorder; Asperger's; or Pervasive Developmental Disorder, not otherwise specified) met the more demanding standards of DSM 5. Twenty-five percent of those who get the diagnosis of Autism in DSM IV would lose it in DSM 5; 75% of those with DSM IV Asperger's are undiagnosed in DSM 5; and 80% of PDD-NOS. While this study is admittedly preliminary, I would give it much more weight than any of the empty DSM 5 reassurances.
DSM 5 derived its proposed criteria set from unpublished, not yet peer reviewed, tertiary care data that may not generalize at all well to the wide range of settings in which the diagnosis of autism must be made. This is clearly an inadequate foundation on which to develop a new definition of autism.
And it is false reassurance to appeal to an un-posted field trial that was indifferent to prevalence and made the stunning mistake of not doing a head-to-head comparison of DSM IV vs DSM 5. Moreover, the field trials were conducted in non-representative academic settings where the impact on rates is likely to be very different than it will be in the real world.
My previous blog expressed the concern that rates of autism have been inflated because of the loose DSM IV definition and also the way it is used because of a too-close linkage to school services. I suggested the need for more precise diagnosis, but also a reduced authority for DSM in school service requirements. A kid who needs help shouldn't lose it just because psychiatric diagnosis in this area remains so imprecise. I would welcome a more accurate DSM 5 definition of autism -- but the method of deriving the new DSM 5 criteria is suspect and its claim to be rate neutral seems simply absurd.
Allen Frances is a professor emeritus at Duke University and was the chairman of the DSM-IV task force.