Sunday, March 30, 2014

Autism Rising : The Emperor's New Clothes

It's that time of the year again - winter is finally starting to fade, spring is in the air, and Autism Awareness month is right around the corner.  Which means that it is time to play one of my favorite games - rationalize about the latest massive increase in the rate of autism.

In case you haven't heard, the CDC just this week released the latest in its series of reports on autism prevalence.  This new report which looked at children who were eight in 2010 found that overall 1 in 68 children had a form of autism - 1 in 42 boys and 1 in 189 girls.  

Since a picture is worth a thousand words, here is what the CDC's autism prevalence figures have looked like over the past 10 years.


As you can see the autism prevalence figures from just this one source have more than doubled in just ten years.  I have to wonder how much higher the rate is going to have to go before public health officials and the medical community start taking the problem seriously.  What does the number have to be before we stop trying to explain away the increase and start acknowledging that the increase just might be because more people are developing autism?

The CDC's explanation this year is a mixed bag of more children with typical intelligence being recognized, "better awareness", and, somewhat strangely, saying that even more awareness is needed in certain areas.  Entirely missing from the explanation or interpretation is anything that suggests there might be a very real problem.

How can it be that after ten years of study the CDC can't figure out if the increase is real?  How can they say in their 2000 report when the new, unexpectedly high rate was 6.7 per 1,000 -
Studies of ASD prevalence in the United States during the 1990s have identified rates of 2.0 - 7.0 per 1,000 children, a greater-than-tenfold increase from rates of 0.1 - 0.4 per 1,000 children identified during the 1980s. [...] However, because the number of persons identified for services is dependent on multiple factors (e.g., changing eligibility criteria, increased awareness, and changes in service availability), aggregate data might underestimate prevalence and should be used with caution in examining population-based trends.
And now, ten years later, when the rate is 14.7 per 1,000, when there are more than 2 children with a diagnosis of autism for every one back then, they say -
The global prevalence of autism has increased twentyfold to thirtyfold since the earliest epidemiologic studies were conducted in the late 1960s and early 1970s. At that time, prevalence estimates from European studies were one in 2,500 children in the population, and by the 2000s prevalence estimates from large surveys were 1%–2% of all children. Although the underlying reasons for the apparent prevalence changes are difficult to study empirically, select studies suggest that much of the recent prevalence increase is likely attributable to extrinsic factors such as improved awareness and recognition and changes in diagnostic practice or service availability.
How can the expected rate of autism go from 1 per 1,000 to 6.7 per 1,000 to 14.7 per 1,000 and the explanation stay the same?

To preempt any snarky comments, yes, there are some external social factors that can explain away some of the increase without their being an actual increase in the number of people with autism.  But to suggest that there were 14 people with autism in the past for every 1 that was identified but no one noticed due to social issues is just absurd.  It is very hard to miss a large number of children who have functional language issues, don't socialize well, have restricted interests, repetitive behaviors, sensory issues, and the tendency to meltdown on a regular basis.

The other part of the CDC's explanation, the idea that increase is bring driven by children with more typical intelligence, is complex one that would require its own post. But, suffice it to say, like last year's "later diagnosis" theory it doesn't hold up under close scrutiny.  The increasing IQ numbers could easily be attributable to the increasing availability and use of early intervention services such as speech therapy and ABA.  If you start increasing a child's receptive and expressive communication ability and teach them how to better self-regulate then, everything else being equal, they will be score better on an IQ test.

Besides which, since these numbers are in line with last year's NSCH increase but are based on one static age, how can you reconcile the current findings with last years?  Especially since the median age of diagnosis did not change in the current CDC report from the one before?  Oh wait, you try to do it like so -
Comparison of findings from the ADDM Network 2010 surveillance year with results from the CDC 2011–2012 National Survey of Children's Health (NSCH) on parent-reported ASD prevalence (5) revealed some similarities. Collectively, these two studies used three complementary data sources: health (ADDM), education (ADDM), and parent-report (NSCH). The report based on NSCH data estimated ASD prevalence of 2.0% among children aged 6–17 years in 2011–2012. Like the ADDM Network, NSCH also found a large increase in ASD prevalence compared with its previous estimate, which was based on 2007 data. The NSCH attributed this increase to children who received diagnoses at an older age, with a greater proportion judged as having mild (less severe) ASD according to parent report. Although not synonymous with ASD without co-occurring intellectual impairment, the increased number of children with milder ASD diagnosed at an older age in the NSCH study parallels the increasing percentage of children with normal intellectual ability and ASD identified in the ADDM Network.
In other words, the NSCH report attributed the increase to increased number of milder, more intelligent children being diagnosed later in life (i.e. after age 8) and the current CDC report says that is comparable to milder, more intelligent children being diagnosed earlier in life (i.e before the age of 8)?  Even though the NSCH data is more recent than the CDC data?

To leave you on a happy note, the most recent data from the CDC is already quite stale because it is talking about children who were born in 2002.  It doesn't tell us what the rate of autism is in children who are being born today.  Given that the rate has gone up considerably every time we look at a later birth year I think it is safe to assume that the rate for children born today is much higher than 1 in 68.

If you want to get an idea of what the numbers might look like for children born after 2002 you can look at the NSCH numbers.  But  the NSCH data contains a lot of extra noise that makes it hard to compare to the CDC numbers.  However, if you take the data from the last three NSCH data sets, pull out the data on the children who were eight at the time, and chart it with the CDC numbers by birth years you get a somewhat clearer picture -


The numbers aren't exactly comparable because of differences in how they are derived.  The NSCH data is based on parental report and might be overstating the rate.  But the NSCH data also covers a much larger part of the country than the CDC numbers do so might be more representative of the country as a whole.

In any event, I think it is interesting that the two disparate data sets are showing the same trend.  If you measure the same thing in two different ways and come up with a similar answer then that tells you your answer is probably correct.

If the trend in the two data sets has continued to current day then the autism number for children born today will be about 1 in 25.  Maybe in twelve years when the CDC releases data for children born this year we will be treated to explanations about how our awareness is so much better than it was today.  Because, like ten years ago today, we would finally be seeing the half of the autism population that we can't see today.