Autism Prevalence Mashup

Perhaps the most important unresolved question today about autism is whether it is becoming more common.  Are we in the midst of an epidemic of autism or are social forces just making it appear that way.  The answer to this question has the potential to reshape how we think about autism and how we should deal with it.

If the rate of autism is not increasing then our focus should be on providing services and finding new ways to help people with autism live more productive lives.  But, if the rate of autism is increasing, then our immediate concern should be figuring out why and putting a stop to it.

While I don't think there is yet a clear answer to the overall question, I think it is safe to say that the number of children with a label of autism is increasing each year.  The problem is trying to figure out if this increase is real or not.

Since I think the question is important, I decided to take some administrative prevalence data from studies that I was reading and put them together to see what they looked like.  The papers are both about autism prevalence in Montreal, Canada and both use yearly administrative data from school boards to estimate the prevalence of autism.

There are several problems with what I am going to talk about below, so let me get those out of the way.  First, the general disclaimers about administrative prevalence apply - so yes, I know that there are problems with using this type of data.  Second, the papers used two different school boards so there is a problem with combining the data like I am doing.  But, since both of the school boards are in the Montreal area and both used the same definition of autism, I think the comparison has some relevance.

So, without further ado, I give you the first chart.  The years on the bottom are the birth year of the children.  Keep in mind that the age of the children is going from the youngest on being on the right (2002) and the oldest being on the left (1987).

Although, the birth years aren't really exact birth years but rather an approximation based on what grade the child was in.  So if a child was held back a grade or two as might be common for children with special needs (if that even happens in Montreal, I don't know), then the ages could be skewed.

The blue line represents the children from the first study and the green line represents the children from the second.  I indicated the point at which the children on both lines would have been eight years old.  That can be important because that is the age where we can assume that most children with autism have been recognized and given the appropriate label.

The red line is a three period moving average of both lines and represents the overall trend of the numbers.

(Click for a larger version)

I think there are a few notable things on this chart.

First, I think the overall trend is clear.  Younger children are more likely to have a label of autism than older ones and the number of children with an autism label is growing by about ten percent per year.

Second, notice that the prevalence goes up as the children get younger up until age eight when it starts to fall? This is what I was talking about above where younger children might not yet have a diagnosis, so we can assume that the numbers before age eight are lower than they should be.

Third, if you look at where the blue and green lines overlap, you might notice that the rate of autism in the children from the second study is about half of that from the earlier study.  For example, in 1993 the children in the earlier study were about ten years old and had a rate of about 71 per 10,000 but in the second study the children would have been 15 and had a rate of about 48 per 10,000.

As I have commented on before, it looks like there might be a substantial group of children who have a diagnosis of autism when they are younger but lose it as they get older.  This is the third independent data set where I have seen this relationship and I am starting to wonder about it.  If I am not mistaken, this rate is similar to the recovery rate that has been historically demonstrated for ABA.

Moving on, another nice thing about these two studies is that they gave a breakdown between the different types of autism.  One of the theories about the increasing rate of autism is that, as time goes on, there are more milder cases being recognized and that is partially responsible for the increase.  But, as you can see on the chart below, that didn't happen here.

(Click for a larger version)

Even though the average prevalence increased between the two studies, the breakdown of the types of autism is essentially unchanged.  The majority of autism cases here are PDD-NOS, followed by classic autism, and finally Aspergers bring up the rear.  And, as I talked about before, CDD is extremely rare with only 1 child having that label.

Take away from this what you will, I just found it to be interesting.  Or maybe this is a sign that I need to get a better hobby...


References

1. Fombonne, Eric, Rita Zakarian, Andrew Bennett, Linyan Meng, and Diane McLean-Heywood. 2006. “Pervasive developmental disorders in Montreal, Quebec, Canada: prevalence and links with immunizations.” Pediatrics 118:e139-50.  doi: 10.1542/peds.2005-2993

2. Lazoff, Tamara, Lihong Zhong, Tania Piperni, and Eric Fombonne. 2010. “Prevalence of pervasive developmental disorders among children at the english montreal school board.” Canadian journal of psychiatry. Revue canadienne de psychiatrie 55:715-20. Link