Flutterby™! : Autism rates? 2005-12-29 03:05:08.517+00

Autism rates?

2005-12-29 03:05:08.517+00 by Dan Lyke 7 comments

In response to a message my mom forwarded which claimed that autism rates were sharply rising in the U.S., I went off and did a quick search on said rates. What I found was the usual assortment of "yes it is/no it isn't" back and forth. On the one hand, the CDC points out that "Autism was added as a special education exceptionality in 1991" and there are no clear signs that the increase of diagnosis indicates an increase in incidence, others show the sorts of exponential trend-lines that seem a little steeper than just a constant number of new diagnoses per year that one would expect to level off after the various age groups in the sample categories are filled.

What I don't see immediately is whether the incidence of developmental disability is rising. What one would expect to see, if autism is just a new re-classification inside developmental disability, is that the overall rates are remaining the same. On the other hand, I'm also aware that schools are able to get more funding if they can classify kids as developmentally disabled, so I'd guess that that number is rising sharply too.

Statistics is hard.

I'm also wondering if anyone has informed opinion on the EPA & DuPont PFOA settlements. While I'm seeing a lot of fingers pointed, I'm not sure I've found anyone who's willing to come forward with evidence for PFOA and cancer in humans (and there are several things that cause cancer in rats and not mice, so cancer in rats isn't necessarily an indicator), aside from compounds that result from overheating and burning.

In the past different behaviors were simply called that - now its a disease... the goverment and school systems have too much time on their hands IMO.

Yeah, looking at that second graph, especially with their plot of the percent of the second derivative, it sure looks like the funding mechanisms got changed around roughly 1990 and people have been learning how to take advantage of 'em for a while.

Not that I blame them, that's a good portion of what running a school is about: Find ways to get things classfied such that your school gets the most funds possible. But it does speak to some of the larger failure of government that we need to be dealing with.

But what happens is that the children are damaged. They are marked forever as 'autisic' or worse and can never escape that marking in our database age. Terrible and destructive - money for destroying the children. Money for nothing chicks for free. Oh thats off topic... but I am sure the autistic mind would like that thought and be scolded for it when its perfectly healthy in my autistic mind.

One difficulty is that we have conditions that are defined by what I call Chinese-menu criteria. For example, Asperger's Syndrome is defined in the DSM as people having "at least 2 of the following characteristics....." and then you pick and choose from a list. We don't diagnose appendicitis or lymphoma this way, which just goes to show the general muddle of the field.

I have two nephews who seem to lie on that broad spectrum that starts at Asperger's and end with extreme autism. In some ways they are amazing, but their strange relationship to other people makes them somewhat less functional in the normal world. Yes, it can be terrible to be labeled early in life, but frankly if you spent 10 minutes with my nephews you'd know something was wrong, even if you never met them before. Their sister got married the other week, and the party afterwards with music and dancing and general happy commotion nearly made them hysterical with over stimulation. I have fears that they will never get the help they need to find some functional independent niche in the wider society.

#Comment Re:demographics and population based studies made: 2005-12-29 20:54:08.193072+00 by: m

I spent my career in a variety of Public Health projects, and I will aver that the proper application of statistical analysis to population based studies is indeed a bitch.

The highest quality data generally available in Public Health is considered to be birth and death records. The basic info on the front end certificate data is generally correct, because the involved relatives tend to notice when their newborn son is classified as a female. But the back end data that is used for statistical work does not even have the parents or bereaved checking over the base demographics.

Even though this data is supposed to be fully massaged and reasonability checked, it was always problematic. Data required by law was frequently missing, illegible or simply impossible.

Other data was even worse. I was associate with setting up a risk analysis system for HIV infection in the mid '80s. Data collectors and transcribers were highly motivated. Some 6% of the patients tested were HIV positive, even though they claimed no known risk factors. This was quite surprising for testing that was done anonymously, where individuals would be expected to be reasonably honest. The data was echo checked a third time between test requests and existing code identified data, without finding significant errors. It turned out that the clinic transcription process, where the HIV counselors filled out the test requests and risk history, was being done from their interview logs.

After all data was re-correlated with the logs, the 6% positive rate for no known risk factors fell to less than 0.05%. The HIV counselors, though highly motivated, had modified the protocol for their data transcription process, and gross blunder resulted.

The moral of this, and many similar stories, is that data is never as good as the owners or users think it is. And usually, its a hell of lot worse.

This does not even include the post edit analysis of the data, which is entire trip by itself. Let me just say that as wonderful as PC based analysis systems are, they are all too often abused by individuals who know no more than that a larger correlation coefficient is more impressive than a small one.

And the biggest problem, is at the start with the project design. Too many projects end up in a place where they couldn't prove the sun comes up in the morning, because the initial design is so poor.

This is not to say that such work does not have meaning. Frequently it is all there is, and something is better than nothing. Meta-studies where the results of a group of other studies are compared, can be useful. But they too often fall prey to the ills that wreck so many smaller studies. Recently, a couple of studies of studies have been done. Gross estimates of half the general scientific studies were not reproducible, and for medical studies the results were worse -- only a third could be reproduced. So, how much can you believe the studies of studies?

That is science. The really interesting stuff often comes in finding out why labs a, b and c do not get the same result as d and e. And when the conditions controlling the phenomena are understood, then the results do become reproducible. Then there is a useful result. The brainy part, is knowing the difference.

And, FWIW, there was some play in the press last year about sharply increased rates of autism amongst techies in the Valley; the surmise was "two geeks with Asperger's find love..."