Exactly how many ways are there to get an autism diagnosis?

There’s a saying in autism circles that I really like: “If you’ve met one person with autism, you’ve met one person with autism”. I reckon I’ve met a few hundred in my time, and I can vouch for the fact that people with autism are a pretty mixed bunch.

This has a big influence on the way I think about autism and, increasingly, the way I conduct autism research. Rather than just trying to determine whether or not X (whatever X is) happens to be impaired in autism, my colleagues and I look at our group of participants with autism and try and work out which individuals have problems with X and what makes them different from the individuals with autism that don’t have a problem. This seems an eminently sensible approach to us. One that might actually tell us some useful stuff. But it doesn't always go down very well.

I’ve given a few talks on the issue, which have been really well received. But then again, I’ve mostly been speaking to teachers and paediatricians – people who actually work with lots of different people with autism every day – or to researchers who don’t do autism research and so have a fairly neutral perspective.

But when we try and publish this research, we invariably come up against reviewers conditioned to think in terms of autism as a big box of amorphous abnormality; reviewers who consider that any study is "fatally flawed" and "theoretically irrelevant" if it doesn't finish up with a conclusion that X is impaired / normal / enhanced (delete as applicable) in autism. Not all reviewers, I hasten to add. But enough to make it very difficult for journal editors to accept the papers.

In my talks, I usually start off with a couple of slides that show the current diagnostic criteria for autistic disorder according to the clinical psychologist’s bible, DSM IV. In a nutshell, there are 12 different things that the person making the diagnosis should look for – 12 different boxes that can be ticked or not. To get a diagnosis, you need at least 6 ticks. It’s not quite as simple as that though. The 12 boxes are divided into three categories, headed social interaction; communication impairments; and repetitive behaviours / restricted interests. You need at least two ticks in the first category and at least one tick in the other category. If you don’t, then you don’t have autism.

I know what you’re thinking but them’s the rules.

I then point out the fact that, in principle, this means that you can have two different people with the same diagnosis despite ticking six completely different boxes. And I illustrate this with two imaginary kids who have nothing in common other than their diagnostic label.

Anyway, today I was busily trying to write a paper related to these issues and I got to wondering, just how many different ways are there to get an autism diagnosis?

So this is what I did. I made a spreadsheet in Excel which had 12 columns, corresponding to the 12 different boxes that can be ticked. I then filled each of the rows below with a different combination of ticks and crosses (actually 1s and 0s).

Altogether there were 4096 possible combinations of ticks and crosses, going from all crosses to all ticks with everything in between (mathematically inclined readers will spot that 4096 is 2 to the power 12).

Then I added a 13th column that showed for each row whether the particular combination of ticks and crosses would get you a DSM IV diagnosis of autistic disorder. Finally I added up the number of rows with a positive diagnosis.

The answer… drum roll… is 2027. That’s right. Two thousand and twenty seven different ways of getting an autism diagnosis.

Now, you can argue that some of the combinations might not actually exist in reality. And you might well be right. But that’s not really the point.

The point is that a lot of autism research begins with the premise that even though there are 2027 different ways of getting an autism diagnosis, being in the autism 'gang' makes you (a) fundamentally similar to all the other gang members and (b) fundamentally different to everyone that doesn't make the grade. That’s a big, big assumption. And I for one am not willing to make it.


Notes:

I've posted the Excel sheet here. Feel free to download it, check my sums, and tell me if I've done something stupid!!


Update (31/01/12):

The always excellent Autism and Oughtisms pointed out that in DSM 5, the number of combinations will drop from 2027 to 11. However, that's mainly because the 8 boxes under social and communication are being replaced with a single box called "Deficits in social-emotional reciprocity". Although this is mandatory (you'll need a tick in this box to get a diagnosis), it's incredibly broad. "Deficits" can range “from abnormal social approach and failure of normal back and forth conversation through reduced sharing of interests, emotions, and affect and response to total lack of initiation of social interaction”.

Then Allen Frances in the Huffington Post somehow concludes that the number of combinations will go down from 2688 to 6. I have no idea where he gets these numbers from. He also completely misses the point that the number of combinations is only meaningful if the items themselves are the same.

Finally, writing in Scientific American, Ferris Jabr got some astronomer dude to write a computer program for him to work out the combinations. He gets 2027 and 11 too. And computers are always right. Right? Jabr argues that the 2027 is meaningless because cluster analysis suggests there are far fewer combinations that actually exist. He's right of course. But it doesn't change the fact that, regardless of whether we're looking at DSM IV or DSM 5 criteria, the heterogeneity within autism is massive, and it's a bit bonkers for us all to keep pretending that it isn't there.