CDDS Autism Zombie
(click to make larger)I have already written extensively about the California Department of Disability Services autism data. Its getting to hard to imagine that there is something to still say.
But at least 4 times a year, I find myself writing something about the latest CDDS report. This is because the CDDS issues a quarterly report and a group of individuals who advocate a vaccine etiology of autism analyze the data in manner that inevitably supports their theory. This happens every quarter. This happens no matter how many times we criticize the use of the CDDS data because it is not supposed to be used as epidemiologic data, as the CDDS themselves clearly state. Just like Ouija Boards and astrology, there are always a few who resurrect this undead creature, and send it growling off towards public consumption.
What is more, this zombie seems impervious to attack. If the data indicate a continued rise, this is proof that vaccines cause autism and even though much of the thimerosal has been removed, there is still some being used in the flu vaccines and this is the likely culprit. But, if the numbers drop, this is also proof that the vaccines cause autism, because the sudden drop can be correlated to the not quite complete removal of thimerosal.
I should say that some take a more reasonable approach. They accept the criticism that these data are not meant to be used for epidemiology. However they then argue that it is the best that can be had. And that if one sees trends here, even if the data are flawed, this is still indicative of an actual effect. This argument is actually somewhat widespread.
But the argument is still flawed. If one refers to the graph, it should be clear that the natural statistical variation allows for an occasional decrease; even one extending for a few quarters before an uptake is noted. That is the nature of statistics; one is going to see an occasional slight downswing. Moreover, the fact that the CDDS data are so regular does not make them reliable or valid.
The CDDS uses a form version from 1986. It has the following categories.
0 = 'None''
What is more, this zombie seems impervious to attack. If the data indicate a continued rise, this is proof that vaccines cause autism and even though much of the thimerosal has been removed, there is still some being used in the flu vaccines and this is the likely culprit. But, if the numbers drop, this is also proof that the vaccines cause autism, because the sudden drop can be correlated to the not quite complete removal of thimerosal.
I should say that some take a more reasonable approach. They accept the criticism that these data are not meant to be used for epidemiology. However they then argue that it is the best that can be had. And that if one sees trends here, even if the data are flawed, this is still indicative of an actual effect. This argument is actually somewhat widespread.
But the argument is still flawed. If one refers to the graph, it should be clear that the natural statistical variation allows for an occasional decrease; even one extending for a few quarters before an uptake is noted. That is the nature of statistics; one is going to see an occasional slight downswing. Moreover, the fact that the CDDS data are so regular does not make them reliable or valid.
The CDDS uses a form version from 1986. It has the following categories.
0 = 'None''
1 = 'Full Syndrome''
2 = 'Residual State''
9 = 'Suspected, not Diagnosed
The form counts the 1’s and 2’s. These categories are not part of DSM-IV-R. They are older categories based on earlier ideas. Thus, it can not be correctly said that CDDS offers services only to Autistic Disorder as defined by the DSM-IV. Full syndrome includes both PDD-NOS and Autistic Disorder.
The form counts the 1’s and 2’s. These categories are not part of DSM-IV-R. They are older categories based on earlier ideas. Thus, it can not be correctly said that CDDS offers services only to Autistic Disorder as defined by the DSM-IV. Full syndrome includes both PDD-NOS and Autistic Disorder.
An Autistic Spectrum Disorders Best Practice Guidelines For Screening, Diagnosis and Assessment statewide standards was held in 2002. A revision to this practice that conforms to the DSM-IV and ICD-9 will not be implemented until 2008.
These data are not good measures of Autistic Disorder, as they have been advertised by some persons, almost unceasingly. Moreover, there are no reliability checks conducted to make sure the State diagnosticians are conforming to recommended practices. This makes them an unreliable and invalid, source of autism data. For these reasons any attempt to use these data for purposes of epidemiology is ill-founded.
These data are not good measures of Autistic Disorder, as they have been advertised by some persons, almost unceasingly. Moreover, there are no reliability checks conducted to make sure the State diagnosticians are conforming to recommended practices. This makes them an unreliable and invalid, source of autism data. For these reasons any attempt to use these data for purposes of epidemiology is ill-founded.
posted by Interverbal at 7:22 PM
4 Comments:
Just wondering why you say:
"Full syndrome includes both PDD-NOS and Autistic Disorder."
My children are PDD-NOS because they do meet most but not all of the DSM-IV criteria. At the same time they are receiving services and treatments for autism.
So from what you are saying they should not be classified as having autism in collected statistics?
Just curious.
Hi MJ,
Totally fair question.
The CDDS have been advertised by Mr. Rollens and others in way that led us and others to believe that the CDDS only counted Autistic Disorder. And more or less always counted Autistic Disorder.
The CDDS does not always count Autistic Disorder (for any number of reasons) and they include children with PDD-NOS.
This is an adequate way to help get services to kids who need it, but it is a just an awful way to try to do get validity especially according to the current DSM standards.
It can be summed up like this: The CDDS is using an outdated, 21 year old reference criteria, that uses outdated terms that no one else still uses. Rollens and company are misleading people when they talk about "full syndrome autism", without explaining what that means in this case.
I understand that the California is using an outdated model. What I have trouble understanding is why exact adherence to the DSM-IV criteria is worth noting. Since the criteria is almost solely behavior based you are going to have a substantial difference in diagnosis from PDD-NOS to autism from doctor to doctor.
From what I understand even the "better" tests such as the ADOS have the same bias.
So until there is an objective standard to measure against isn't a broader definition better than a narrower one?
Hi MJ,
"What I have trouble understanding is why exact adherence to the DSM-IV criteria is worth noting."
Because to make a claim about "full syndrome autism", but have the criteria be different from the rest of the world, will end up misleading the rest of the world if they try to use such data for epidemiology.
A fundemental rule of statistics is the requirement that we measure the same thing, in the same way, everytime.
So CDDS can have a different defintion from the DSM-IV and ICD-10, but others can not jump in and try to compare their version of full syndrome autism, to someone else's definition and say "Yep, autism is going down/up these days."
"Since the criteria is almost solely behavior based you are going to have a substantial difference in diagnosis from PDD-NOS to autism from doctor to doctor."
The criteria is solely behavior based, there are no none behavioral criteria. It is logical then to expect that we see some variety in doctor to doctor and area to area. But what is remarkable is how similar prevalence rates are when researchers use the same instruments and methods.
In the research we see a clear differentiation between PDD-NOS and Autistic Disorder in terms of scores on a number of measures such as IQ. These categories are not as porous as some would suspect. To place them together changes the scores and nature of what we are looking at.
"From what I understand even the "better" tests such as the ADOS have the same bias."
Depends on the test, it is not true for all of them. Most of the better tests came about in the early 90's just before or after the advent of DSM-IV. But any test, no matter how helpful, no matter if it is the ADOS, or GARS, or CARS, or ADIR, is enough by itself to diagnose a child. It is not enough to get a ADOS score, then diagnose a child based on that.
"So until there is an objective standard to measure against isn't a broader definition better than a narrower one?"
No, it is not even really broader. The CDDS have mentioned that some kids labeled with Autistic Disorder, don't qualify as full syndrome for one reason or the other. The CDDS, defintion is not broader.
But even if it was really broader, we still should not comapre apples and oranges by comparing the CDDS based data to epidemiological data based on the DSM-IV or ICD-10.
Post a Comment
<< Home