Thursday, November 11, 2010

The Hazards of Self-Diagnosis

Flickr picture by The Doctr
A few weeks ago, I talked about why it wasn't the best idea to give yourself the Autism Quotient and act based on the results of the test.  The Autism Quotient is not meant to be a diagnostic test and, even if you get a high score on the test, you are still fairly unlikely to have autism.

But, for the sake of discussion, lets say that you do have a history of having some autism-like symptoms and decide to take the test.

If you get a high score on the test you might believe that, based the test result and your life experiences, it is likely that you have a form of autism.  After all, you are an intelligent, mature person who has carefully done research and made honest self-appraisals.  And the AQ test bears out your idea, so what could be wrong?

Well, to start off with, you might have schizophrenia, not autism.  According to a study1 published earlier this week (open access, go read it) -
In clinical practice, differential diagnosis of high-functioning autistic spectrum disorder (ASD) and schizophrenia (SCH) is difficult but important. It is especially difficult when adult patients with ASD have psychotic symptoms as a result of maladjustment to their circumstances. Diagnosis of ASD requires a knowledge of early developmental history, but sometimes that is difficult to clearly ascertain when the patient is an adult. If such patients were not diagnosed as having ASD during their childhood, we cannot distinguish their symptoms from the positive symptoms of SCH. Similarly, when ASD patients are in social withdrawal or in an autistic state, it is difficult to distinguish their state from the negative symptoms of SCH. Thus, a reliable measurement for differentiating the two disorders is needed.
This study looked specifically at the Autism Quotient (AQ) test and how well it could distinguish between high-functioning autism and schizophrenia in adults.  The researchers gave the AQ to 51 adults of normal intelligence who had either autism or pdd-nos and 46 adults who had schizophrenia.

The main result was not too surprising.  The autism group, as a whole, scored higher on the AQ than the schizophrenia group. But - and this is the important bit - the AQ misclassified 8 (17%) out of the 46 patients who had schizophrenia as having autism.  As the researchers say in their discussion section, "high AQ patients with SCH cannot be distinguished from ASD by using only the scores of the total AQ and its subscales".

Let me put this another way.

The AQ can't reliably tell the difference between schizophrenia and high functioning autism.  Trained clinicians - people who do this everyday as part of their job - have a hard time as well.  So what do you think the chances are that an untrained person, researching the condition on their own, is going to be able to tell the difference?

Now, before you think this doesn't matter, consider that schizophrenia is known to be more common in adults that autism is.  The current figure for schizophrenia is about 1 in 90, or about 1.1% of adults over the age of 18.  The rate for autism in adults is not well understood but is, by most accounts, much lower than 1%.  I think it would be safe to say that, even if we were to be highly optimistic about how many adults have autism, there are going to be 3 adults with schizophrenia for every 1 with autism.

So, assume that you had a group of 1,000 adults where you had 4 adults (1 in 250) that had autism and 11 that had schizophrenia, and that you gave them the AQ.  Based on what we know about the AQ, we would expect about 25 of the group to score in the autism range.  Out of that group, most (20) would be misidentified, 3 would have autism, and 2 would have schizophrenia.

Or in simple terms, if the AQ gives you a high score and it is picking up on something that is really there, you have almost a 40% chance of having schizophrenia instead of autism.  That little fact might be important as schizophrenia is more easily treated than autism is.

All of this is just one small example of how a self-diagnosis could go wrong.  There are many other ways that the process can go wrong.  The bottom line is, if you think you have a problem, don't self-diagnosis.  Go talk to a professional who knows what they are doing.

1. Naito, Kenichi, Yusuke Matsui, Kiyoshi Maeda, and Kiwamu Tanaka. 2010. “Evaluation of the Validity of the Autism Spectrum Quotient (AQ) in Differentiating High-Functioning Autistic Spectrum Disorder from Schizophrenia.” The Kobe journal of medical sciences 56:E116-24.


  1. "The current figure for schizophrenia is about 1 in 90, or about 1.1% of adults over the age of 18. The rate for autism in adults is not well understood but is, by most accounts, much lower than 1%.

    *snark* that's not a very good foundation to build your argument on.

  2. Well, if you can give be a good estimate of what the rate of autism in adults is then I will use it.

    Or better yet, lets assume that autism is just as common in adults as in the youngest children - 1 in 110. You would still have about a 29% chance of having schizophrenia instead of having autism if the AQ was actually onto something.

    The point is that as long as schizophrenia is more common than (or about as common as) autism that the inability to distinguish between the two is a problem.

  3. Except of course that the condition that is known as Schizophrenia, (Bleuler has an awful lot to answer for) contains a host of other symptoms not covered for by the AQ test.

    Now me old pal you are either getting perverse or delusional, I would suspect the former, in that you would appear to be constructing an argument to accord with a weltanshaung that only admits of autism where autism fits your criteria, and where every body else must be "crazy" to think so.

    I suppose you did come across catch 22 at some point?

    Well what exactly does lie at the root of your "neuorosis" anyway? What is it you actually fear because I can detect a lot of fear in your postings even if you cannot see it yourself. The notion of a cognisant and competent and self recognising autist fills you with dread because it upsets your inner world order perhaps.

    Well come to me, I am sybok, I feel your pain :)

  4. Look it's quite simple. People with autism online today:

    30% diagnosed
    20% un-but genuine
    30% bap
    20% schiz-xyz

    MJ, the AQ isn't anymore than a piece of paper with some questions about Star Trek.

    Why not write your own version using your own definitions of autism and schizophrenia?

    Oh, and of science too. I think you'll find our one a little rigid for you.

    A non-Cartesian space might be useful to get the graphs right.

    A relativistic parameter also might help get those pesky little points where they should be. At least in 1896.

  5. Grrrr! L. Spelling The Weltanschauung

    Mind you did you do that diliberately to fool Google Translate?

  6. Author,

    Of course there is far more to schizophrenia than is covered on the AQ. For that matter, there is far more to autism than is covered on the AQ - the AQ is aimed squarely at high functioning adults with normal intelligence and that group is certainly a smaller subset of the general autism population.

    But that isn't really the point.

    The point is that a even a rational, intelligent person who understands the subject is going to get it wrong a lot of the time when they self-diagnose. Now imagine what is going to happen with the non-rational, less intelligent people who doesn't understand the subject when they do the same.

    How many people taking the AQ on facebook do you think fit into the second category?

    As for being either perverse or delusional, I would have to say both, depending on the day of the week. But I am certainly not trying to be delusional or perverse when it comes to a definition of autism nor am I trying to make up my own definition. I actually try pretty hard to stick to the accepted definition of the condition.

    I just don't subscribe to the idea that people who are socially awkward and technically oriented all have a form of autism. But who knows, maybe under the DSM V that group would be considered to have autism.

    As for my "neurosis", that one should be obvious, I have three children who suffer from moderate autism. Every and every day they have to struggle to overcome the limitations imposed by their autism and to be able to deal with the world around them.

    But, haven't you heard, some "cognisant and competent and self recognising autist" has just announced that autism isn't a disability after all and shouldn't be cured. They announced that the problem is parents like me who just don't accept or love our children.

    No, that isn't fear of the "autist" that you are sensing.

  7. Socrates,

    It is kind of ironic you bring up writing your own version of the AQ.

    Remember the Adult Psychiatric Morbidity Survey from the UK that found that 1% of the adult population had a form of autism? They made up their own version of the AQ for that survey and used that to screen for people with autism.

    More importantly, they used this modified AQ to help extrapolate from the 19 adults found all the way back to the general population and arrive at the 1% figure. If even the original AQ has a hard time telling the difference between schizophrenia and autism, what do you think that says for the modified one used in this survey?

    If you are interested, see the following links for some details -

    By the way, non-Cartesian coordinate systems are useful, but I find imaginary numbers are the key to any sort of advanced thinking. Without them, you would not have a computer to type on.