Monday, June 15, 2009

Notes on autism severity and the DSM-V

I was asked briefly to comment on the notion of "severity" of autism as currently proposed for the DSM-V. Here are some far from complete notes (some from earlier writing or work of mine) I organized in response:

Currently, “severity” of autism most often refers to the attempt to quantify the obviousness of autistic traits and abilities. The more obvious these traits and abilities are judged to be, and therefore the more atypical a person is judged to be, the more “severe” autism is considered to be.

Being “more severe” (having more obvious autistic traits and abilities, or being more obviously atypical) is widely presumed to be worse than being “less severe” (having less obvious autistic traits and abilities or being less obviously atypical), while being not autistic at all (no obvious autistic traits and abilities) is considered an “optimal outcome.”

However, there is no current consensus on what neurocognitive differences underlie observable autistic atypicalities and therefore fundamentally define what makes a person autistic rather than nonautistic. At this fundamental level (what autism is), there is currently no consensual way to judge who is “more autistic” or “less autistic” (which would be construed as “more severely autistic” and “less severely autistic”) or whether this would be good, bad or irrelevant.

Currently there is/are no consensual instrument(s) agreed to measure autism “severity.” The major diagnostic instruments (ADOS, ADI-R) do not work like scales, though they have often and misleadingly been misapplied this way. A recent preliminary attempt to transform the ADOS into a scale of “severity” has just been published (Gotham et al., 2009).

On popular measures of “severity” (the CARS, the ADI-R improperly used as a scale, the SRS), an autistic who has one or more very high apparent abilities (that is, has a highly uneven profile of apparent abilities) will score as more "severe" than an autistic whose apparent abilities are uniformly low.

The large proportion of autistics recently reported to have exceptional skills (Howlin et al., 2009) would be judged as less “severe” if they lost those skills. Exceptional skills are considered to increase “severity” (that is, atypicality) in autism, not decrease it. This would also likely be true of autistic cognitive strengths which have been revealed in the peer-reviewed literature in comparisons between the performance of autistics and nonautistics on a wide range of tasks (I have started compiling an inventory of these, which was the subject of an IMFAR 2009 poster). Indeed, at IMFAR 2009, the loss of exceptional autistic abilities (Eigsti et al., 2009; from Deborah Fein’s group) was reported as evidence for “optimal outcomes” in autism.

In the same direction, we have found that autistics who scored higher on pieces of the ADI-R (and would therefore commonly be assumed to be "more severe") had a wide range of better visual skills than autistics who scored lower on these measures (Caron et al., 2006).

Perhaps unsurprisingly then, autism “severity” (defined and measured in various ways) has been a poor predictor of outcomes in autism (Howlin, 2005). Here are two examples from papers reporting very good outcomes in autism:

It cannot be emphasized strongly enough that… it would have been impossible for anyone to predict this outcome. (Kanner et al., 1972)

… early history explained little of the variance in outcome. Indeed, the good and poor outcome groups differed little with respect to early impairments in social responsiveness, deviant language, and bizarre behaviors. (Szatmari et al.,1989)
Currently the best adult outcomes reported in the literature still belong to autistics who as children met the narrowest, strictest, and presumed-to-be most “severe” autism diagnostic criteria ever devised (Kanner et al., 1972; Szatmari et al., 1989; Farley et al., 2009). Asperger (1944/1991) described an individual whose outcome was outstanding (he was a successful academic) as “grossly autistic” with “impossible behavior,” etc.

Some of the most obvious (that is, most “severe”) and successful autistics have been and continue to be individuals regarded as autistic savants. Indeed, the “obviousness” of these individuals, the fact that their atypicalities are so obvious (which would be judged as so “severe”), has been successfully exploited in the autism literature as a means of investigating the fundamental atypicalities which make autistics autistic.

The assumption that having more obvious autistic traits and abilities is bad (“more severe”) has resulted in prejudicial judgments in the literature. An autistic boy (who has accomplished more as a child and young adult than most people, autistic or otherwise) with a full-scale IQ of more than 100 and a strong performance on numerous tests is judged to be “low-functioning,” because he is so obviously atypical. That is, he is judged to be very “severely” autistic (Bonneh et al., 2008; Belmonte et al., 2009), and this must solely be a very bad thing. In Aldred et al. (2004), preschool autistic children are falsely judged to be “low-functioning” or “high-functioning” solely on the basis of presumed autism “severity” falsely drawn from ADOS scores. It is being falsely and prejudicially assumed that an autistic whose traits and abilities are more obvious must function poorly.

The proposed DSM-V changes will enhance societal prejudices against autistic traits and abilities. These prejudices already serve to prevent autistics whose traits and abilities are more obvious from being considered able to learn, to communicate, to make decisions about our lives, to walk around freely, to be employed, etc. Under the proposed DSM-V changes, autistics who are highly and obviously atypical would be even more harshly judged, without there being anything approaching a sufficient empirical rationale for this. Autistics whose communication and social interaction is highly atypical, or who most strongly pursue their focused abilities and interests (something which has widely been acknowledged leads to good outcomes, but has largely not interested researchers; Howlin, 2005), will now be at risk of being regarded as “most severe.” This is at a time when the term “severe autism” is popularly equated with low expectations, hopelessness, dread and horror.

Also, in my experience, the term “severe autism” is commonly used as a justification for why autistic people not only can but must be treated in ways that would be instantly recognized as unacceptable in any other population. This includes lowering or discarding very basic standards of science and ethics, a pervasive practice now applied to the entire autistic population. I am referring to the basic standards of science and ethics that automatically protect and benefit nonautistics, and without which they could not proceed safely in society, much less have good outcomes.

The proposed DSM-V wording raises numerous specific concerns; I will mention two. One is that it enshrines the false and harmful assumption, an assumption showing a striking absence of reciprocity on the part of the DSM-V committee, that if autistics communicate or interact in atypical and possibly very adaptive ways, then this communication or interaction must be defective or wrong, or must not exist at all. Another concern is that by using words such as “rituals,” the DSM-V is perpetuating the misconstrual of autistics’ focused abilities and interests (and motor mannerisms). This misconstrual led to the demonstrably false and harmful assumption that autistics’ atypicalities in this area amount to “obsessions” and must be treatable in the same manner as OCD (see recent failed citalopram RCT, King et al., 2009; a failed fluoxetine RCT has been announced by Neuropharm and Autism Speaks, but not published yet).

In these and other ways the proposed DSM-V autism criteria define possibly very adaptive but atypical autistic traits and abilities as being more “severe” (worse or more defective) the more they are obvious. This leads to the assumption that such traits and abilities should ideally be suppressed and eliminated.


Aldred, C., Green, J., & Adams, C. (2004). A new social communication intervention for children with autism: pilot randomised controlled treatment study suggesting effectiveness. Journal of Child Psychology and Psychiatry, 45, 1420-1430.

Asperger, H. (1944/1991). 'Autistic psychopathology' in childhood (Frith, U., Trans.). In: Frith, U. (Ed.), Autism and Asperger Syndrome. Cambridge University Press, Cambridge, UK, pp. 37-92.

Belmonte, M.K., Bonneh, Y.S., Adini, Y., Iversen. P.E., Akshoomoff, N.A., Kenet, T., Moore, C.I., Simon, H.J., Houde, J.F., & Merzenich, M.M. (2009). Autism overflows with syntheses. Neuropsychology Review, 19, 273-4.

Bonneh, Y.S., Belmonte, M.K., Pei, F., Iversen, P.E., Kenet, T., Akshoomoff, N., Adini, Y., Simon, H.J., Moore, C.I., Houde, J.F., & Merzenich, M.M. (2008). Cross-modal extinction in a boy with severely autistic behaviour and high verbal intelligence. Cognitive Neuropsychology, 25, 635-52.

Caron, M.-J., Mottron, L., Berthiaume, C., & Dawson, M. (2006). Cognitive mechanisms, specificity and neural underpinnings of visuo-spatial peaks in autism. Brain, 129, 1789-802.

Farley, M.A., McMahon, W.M., Fombonne, E., Jenson, W.R., Miller, J., Gardner, M., Block, H., Pingree, C.B., Ritvo, E.R., Ritvo, R.A., & Coon, H. (2009). Twenty-year outcome for individuals with autism and average or near-average cognitive abilities. Autism Research, 2, 109-118.

Gotham, K., Pickles, A., & Lord, C. (2009). Standardizing ADOS scores for a measure of severity in autism spectrum disorders. Journal of Autism and Developmental Disorders, 39, 693-705.

Howlin, P. (2005). Outcomes in autism spectrum disorders. In: Volkmar, F.R., Paul, R., Klin, A.,Cohen, D. (Eds.), Handbook of Autism and Pervasive Developmental Disorders (pp. 201-220). Hoboken, NJ: Wiley.

Howlin, P., Goode, S., Hutton, J., & Rutter, M. (2009). Savant skills in autism: psychometric approaches and parental reports. Philosophical Transactions of the Royal Society B: Biological Sciences, 364, 1359-1367.

Kanner, L., Rodriguez, A., & Ashenden, B. (1972). How far can autistic children go in matters of social adaptation? Journal of Autism and Childhood Schizophrenia, 2, 9-33.

King, B.H., Hollander, E., Sikich, L., McCracken, J.T., Scahill, L., Bregman, J.D., Donnelly, C.L., Anagnostou, E., Dukes, K., Sullivan, L., Hirtz, D., Wagner, A., & Ritz, L. (2009). Lack of efficacy of citalopram in children with autism spectrum disorders and high levels of repetitive behavior: citalopram ineffective in children with autism. Archives of General Psychiatry, 66, 583-590.

Szatmari, P., Bartolucci, G., Bremner, R., Bond, S., & Rich, S. (1989). A follow-up of high-functioning autistic children. Journal of Autism and Developmental Disorders, 19, 213-225.


Jannalou said...

One of my concerns with the proposed severity scale is that it divides things into two categories. So what if someone is "severe" in one category and "mild" in the other? Do they just even it out and say the person is "moderate" - even though that's inaccurate?

I'm glad it's just proposed, and I hope they decide on some other system. This one just isn't going to work the way they think it will.

codeman38 said...

What Jannalou brings up is something I've also been wondering about. One of my biggest personal issues is that, from an overall standpoint, my skills are ridiculously uneven-- above the norm in some respects, well below the norm in others.

And that's the case even in things that people think would be related. For instance, I speak quite clearly and coherently, and thus appear extremely "high-functioning" in that respect-- but my comprehension of spoken language has been dodgy for decades. My visual processing is great for spotting typographical errors, but poor for judging the speed and position of cars on the road. And so on...

Michelle Dawson said...

Just for clarity, "level of functioning" and "autism severity" have been wrongly confused or conflated. In the original post I provide examples of this from the literature.

But (setting aside their validity just for a second) the two terms refer to different dimensions. They have in common that there is no current consensus on how to measure either.

Anonymous said...

There's already the "functioning" dimension, measured by IQ as you note. This one is predictive of outcome.

There's no evidence that symptom severity, on the other hand, is predictive of outcome (e.g. Szatmari et al. 1989) as you are also aware.

So what would be the useful purpose of introducing this new "severity" dimension? If they want to divide ASD in the DSM-V, divisions in terms of IQ have more support in research.

(Certainly, the Asperger's/Autism separation never really had much support in research.)

A dimension that measures how much ASD affects academics and employment might have more real-world usefulness than one that describes "disturbances" in communication and such.

Michelle Dawson said...

In response to Joseph, there is no current consensus on how to accurately assess autistics' intelligence, or at what age this might become predictive.

When measured intelligence has been predictive, and it has not consistently been, it is unclear what accounts for this.

Studies claimed to show that measured intelligence predicts outcomes include the series of papers by Rutter and colleagues, spanning 1967-1970, where about one-third of the children were deemed untestable and half the entire sample received less than two years of education.

In the typical population, measured intelligence assessed early in childhood (with the kinds of instruments also used to assess autistics) does not predict intelligence later in life. See Mackintosh (1998) for a review.

Anonymous said...

Are there outcome studies where IQ has not been predictive of outcome?

Sure, there are many ways they could measure IQ, but however they do it in different studies, I think it's typically predictive of outcome statistically.

I agree it's not necessarily clear why IQ is predictive of outcome. I've looked at the raw data from Szatmari et al. (1989), for example. IQ here predicts outcome, but the lower-half of autistics ranked by outcome (as listed in a table in the paper) all attended special education. The upper-half presumably attended mainstream education. If you stratify the groups by special education vs. other, the IQ trends within each group practically go away.

There's also a non-significant trend of age vs. IQ in Szatmari et al. (1989).

Finally, there might have been some subjectivity in how Szatmari et al. ranked autistics by outcome.

Anemone said...

My impression based on my reading is that an IQ above 50 predicts a better outcome than an IQ below 50, but that how much IQ is above 50 doesn't predict anything.

I can see that until people are actually clear on what autism is (and isn't), it's pretty much impossible to determine severity. Severity of what? I would be considered very high functioning by a lot of standards, because I can "pass", at least in the short term. But I can't hold down a job, which counts for a lot more in my book.

Angel said...

Beautifully written. I greatly appreciate your blog. Kudos to you!

Michelle Dawson said...

In response to Joseph, in Szatmari et al. (1989), individual-level data re early history IQ are not reported. There is only a mean and SD reported for the group.

While the S-B is mentioned as a test that was used, other tests may also have been used. The age at which children were originally tested is also not reported, though you can narrow it down to age 6 or younger. The authors decided to take the latest score reported within that age range.

Early IQ in this study was correlated with later VABS composite scores, but these also are not reported individually.

There is also the problem of which tests are used to determine early "IQ." This may not be an IQ score but, for example, a Vineland score.

In order to provide evidence that IQ in childhood is related to later outcomes, researchers have worked hard to locate various thresholds post-hoc. But looking at the data in more detail is more informative. For example, both Howlin et al. (2004) and Farley et al. (2009) found that neither PIQ nor VIQ predicted outcomes.

Sebastian said...

A dimensional approach to ASD in the diagnostic criteria may have advantages over a categorical diagnosis of atypical autism / Asperger / Autism, including how 'severe' ASD is seen and treated in society (and inclusive to subclinical autism such as non-verbal learning disorder):

By researching a dimensional measure of ASD it might be established that indeed everyone could be regarded as somewhat autistic. Regarding autism as an extreme of normal that every person shares to some degree may contribute to understanding and more respectful treatment whereby the artificial construct of an "cut-off point" which divides people in "us" vs. "them" does not facilitate understanding.

How about regarding atypical sensory perception as a core symptom of autism whereby 'severity' would be measured by hypersensitivity to sensory stimulus e.g. sensory perception thresholds that have been shown to be lower in all sensory channels measured so far for autistic people. The inclusion of sensory perception as a core feature of ASD may help reconceptualize autism from a behavioral defined phenomena to an information processing phenomena and given the present conceptualization for autism, improve how autistic people are regarded in society.

People could be diagnosed or discovered to be X% autistic instead of a priori deviant. Categorical diagnostic categories such as ASD may however be a matter of public health policy and not necessarily a science. That a "cut-off point" exists between autistics and nonautistics can be questioned.

farmwifetwo said...

An interesting and difficult question. FWIW... and my opinion only: (since Michelle hates opinions....)

1. The IQ scale is a waste of time because it does not tell you how someone learns and where their strengths and weaknesses are.

Which is why I lobbied hard and one child has had, the other will have full psychometric and language testing. It gave us that answer for the elder and hopefully will for the younger next year. If it doesn't for the younger we'll try something else.

2. What is currently labelled as autistic behaviours such as flapping, spinning, rocking, self-injuring occur in others who's had brain injuries or strokes. Therefore, cannot be claimed to be "autistic behaviours" and measurable.
Actually more and more I'm not convinced a diagnosis of autism actually exists in it's current format.

It's a group of symptoms that may include, but is not limited to, everything from regression as a child, self-injuring, self-stimulating behaviours, developmental delays, brain damage, OCD sensory and visual processing issues. Then there's the how much of each and how does it impact one's ability to be independant.

The best answer I have is that instead of trying to measure only IQ or psychometry or language... testing also needs to be done on self awareness, ethics (realizing when someone is lying to you or using you), banking and other independant living skills.

And the independant living skills need to be rated the highest because it doesn't matter how book smart you are or what your diagnosis is.... if you can't figure out how to make a meal, do a load of laundry, pay your bills, balance your cheque book or know when someones scamming you out of your $$$... You need someone to help and the system needs to know and help you get/pay for that help.

It's the best idea I have.


Anonymous said...

In response to Joseph, in Szatmari et al. (1989), individual-level data re early history IQ are not reported. There is only a mean and SD reported for the group.

Table II lists individual IQ scores. These are adult scores, evidently. The test was the WAIS-R.

Michelle Dawson said...

In response to Joseph, I had written, "Early IQ in this study was correlated with later VABS composite scores."

Sorry, that is not true (I knew that...). The correlation is between VABS at follow up (not reported individually) and IQ at follow-up (reported individually).

There was no correlation reported between early history IQ (not reported individually, as I wrote) and any outcome measure.

So Szatmari et al. (1989) provides no evidence that IQ in childhood is a predictor of later outcomes in autism. I knew that too, sorry.

Michelle Dawson said...

In response to farmwife2, I am inexpressibly grateful that I don't live in a part of the world where balancing a chequebook or being able to make a meal (etc.) is prerequisite for being at large without supervision.

And in farmwife2's model, should those who pay for quackery ("know when someones scamming you") be considered incapable of living "independently"?

In my experience, pretty much all human beings, including all children (autistic or not), need "help." The vast majority of people are extremely dependent on services and assistance provided by other people.

Michelle Dawson said...

In response to Sebastian, there is some evidence that the autism/Asperger's distinction is sometimes useful. Whether this will stand up over time is unclear. Right now we don't know any better. Making the distinction in research, at least in some areas, is cautious.

I agree that autism is about atypical information processing and analysis. But there is evidence that this atypicality is not reducible to "hypersensitive" typical processing (allowing for a continuum along this one dimension) which so far as I can tell is what is being proposed here.

Larry Arnold PhD FRSA said...

like it or not your comparisons actually are divisive against us anankastic folk.

OCD is OCD and is not any worse or any better than anything else.

It is just another form of neurodiversity and the way in which some of these DSM propositions are set up the two will merge into an amorphous soup of indistinction that these idiotes inelucidate in there dissolution into pulp fiction.

Michelle you talk some sense but it is all inflected with a great deal of predjudice and misunderstanding as to what the neurodiverse world beyond autism is.

Autism is nothing special and is only part of what is going on in this DSM revision, only part of the world in which difference is devalued and pathologised.

Michelle Dawson said...

Mr Rex wrote: "OCD... is not any worse or any better than anything else."

Please show me where I wrote that it is.

Roger Kulp said...

I find it interesting,that the proposed revisions,make no mention of developmental delay,which I have always thought ought to be among the top criteria for any sort of autism diagnosis.Basic stuff like toileting,walking,catching a ball.

Nor have I ever understood why regression is rarely figured into an autism diagnosis,or factored into the severity.Often autistics who regress are excluded from studies.

Why is this ?

Would severe or multiple learning disabilities,without a low IQ/ retardation be considered "intellectually disabled"?What about significant differences between verbal and nonverbal IQ ? How would this be classified?

It looks like this new subclinical/ normal classification would push many aspies off the spectrum.That would be fun to sit back and watch.

Michelle Dawson said...

Roger Kulp wrote: "Often autistics who regress are excluded from studies."

Can you give examples of this?

Intellectual disability so far requires measured intelligence below a specified threshold, as well as difficulties with "current adaptive functioning" in at least two areas.

There are hundreds of named nuerodevelopmental disabilities. The vast majority of them (including Down syndrome and Fragile X) do not have a DSM-IV category.

Sebastian said...

I tend to believe that the degree of monotropism would be an adequate meassure of how I define autistic traits if it could be meassured.

David N. Andrews M. Ed., C. P. S. E. said...

"There are hundreds of named neurodevelopmental disabilities. The vast majority of them (including Down syndrome and Fragile X) do not have a DSM-IV category."

Indeed, and for those who don't know (Michelle probably does), this is because they do not fall into the classification of what DSM calls 'mental disorder' (not a term I espouse)... learning difficulties are essentially secondary issues (not to say that they are not important, but that - diagnostically - they are not primary issues: this is listed primarily as a chromosomal abnormality - not elsewhere specified). ICD 10 lists a main category, Q90.X with subcategories as follows: Q90.0, Q90.1, Q90.2, Q90.9; these classifications are all based on the means by which the trisomy is thought to have occurred, which the exception of the last one - which is a 'not otherwise specified' category. Since ICD leads DSM, it is likely that this classification under chromosomals rather than developmentals is the reason for the non-existence of a classification for Down syndrome in DSM. I suspect that the same process applies in the case of Fragile-X.

"I tend to believe that the degree of monotropism would be an adequate measure of how I define autistic traits if it could be measured."

My italics... As for monotropism... to be a useful measure of how autistic traits are defined, something would have to link the biological contributors to autistic behaviour to those actual behaviours. Even with the advances made in behavioural genetics, we know that no behaviours are inherently 'autistic': the reason we think of them as being 'autistic' is the clustering together that defines the behavioural characteristics we see as the diagnostic criteria. But we may well see similar behavioural characteristics in other categories (and even in non-categorised behavioural characteristics that accompany, for example, chromosomal or other types of congenital diagnostic categories. That's the first problem with any trait approach combined with the second problem. There is then the third issue, which is the extent to which the term 'monotropism' can be operationalised, since this will affect the extent to which the cognitive term 'monotropic' can be linked to both biological aspects and the behavioural characteristics: poor operationalisability will result in poor theory about how a cognitive issue can mediate a behavioural characteristic from a biological contributor.

I'm not sure that the term, then, is that useful in defining 'autistic traits', especially since no traits can be considered specifically 'autistic'.

With regard to 'severity' issues, I don't think that IQ has much going for it as an indication of how severely autistic someone is, provided that severity is a useful concept in understanding autism anyway.

Sebastian said...

>> Exceptional skills are considered to increase “severity” (that is, atypicality) in autism, not decrease it. <<

From my perspective, following the monotropism hypothesis, it could be hypothized that the greater the discrepancy between abilities and inabilities of a person, the more autistic that person is. Autism could be characterized as an extreme uneven distribution of abilities and inabilities of a person. By that definition, an autistic person exhibiting untypical skill might be considered as "severely" autistic, meaning to be more autistic in the sense that the skill profile of this person would be more uneven than that of less autistic people. The monotropism hypothesis itself does not judge abilities; as I see it, monotropism hypothizes that abilities and interests are monotropically distributed.