Lou Gehrig's Disease - Motor Neuron Disease - Amyotrophic Lateral Sclerosis
Thought it had been cured by now? Still no known cause. Still no cure. Still quickly fatal. Still outrageous.

Monday, October 1, 2018

He Can Tell You How Many of Every Species There Are in the Great Lakes

Captain Reggie Fisher knows he can.

You see, he had a smart method for counting every Walleye that was really slick.  He went out in three of our five Great Lakes -- Michigan, Erie, and Ontario.  He caught Walleye, tagged them, and tossed them back in, over and over and over and over.  He recaught some with tags and some new ones.  Captain R kept records and did that long enough that a model (ok, he made some assumptions and picked among models) could statistically tell us all how many Walleye are in the Great Lakes.

He came up with a number that statistically tells us that there are 10,725 Walleye in all the Great Lakes.

But wait...
  • Do fish that were tagged and thrown back behave differently than the ones that escaped his nets?  
  • How do we know that his sample reflects the Walleye in Lakes Superior and Huron where he didn't go fishing?  Maybe they are different because it's so much colder up there.  
  • Are there Walleye that swim deep and always escape the nets?  
  • His number seems awfully small.  How do we do a common-sense check on his results?  Perhaps we should measure it against some known data?  
This seems like a case where we need to be a little skeptical of the answer that the computer chunks out.  Does it pass the common-sense test?

And now Captain R says he has the technique be the definitive source of populations of every species in the Great Lakes.

Wait a minute!

Silly story, you say?  I wish it were.

Recapture (repetitive sampling, marking, and resampling) is an established way to figure out how many beans are in the jar or how many fish are in a lake.  It gets a lot more complicated when there are multiple jars and multiple lakes and diverse beans and fish.

Here's the problem --

Prevalence (the number of people with ALS) per 100,000 as a rate is hard for me to picture, but something really bothers me when the prevalence rate increased 32% from 2002 to 2004. That's a lot!  Would that not have caused a national ALS panic in 2004?

I multiplied the prevalence rate by the U.S. population and here are the total number of people with ALS they say we had in 2002, 2003, 2004.  That seems small.  Maybe we need to be skeptical of the answer that their computer chunked out?

Then I pulled out the MND Death Certificate data from that we can get online in the CDC's Wonder database, and it makes those counts of people with ALS in those years really look weird.

Did over half the people with ALS in 2002 die in 2002?  Does such a small but severely rising population of people with ALS make sense with the fairly flat death certificate data?

In the recapture paper, you can see the bold pitch for being the purveyor of patient counts for all diseases based on the wonderful statistical results they got with ALS.  Really? Just because the CDC published this paper doesn't mean that we shouldn't be asking questions and challenging the concept of fishing through Medicare, Medicaid, and VA files as being the end-all solution to determining patient populations.

The only reason that file-fishing technique made any sense for ALS when the Registry was proposed was that people with ALS had a shortened wait for Medicare and it seemed like most would be captured there.  We have found a myriad of problems in the technique in the decade since.  Some people with ALS still able to work delay the Medicare decision.  Some people with ALS find better coverage in private plans to which they may be entitled and never enter Medicare or Medicaid or VA.  Some people with ALS opt into Medicare Advantage plans and are not included in the Medicare files.

We have every right and obligation to challenge the results as not passing the common-sense test.  It's interesting, sure, but is it right?

And to suggest that with this, the CDC has the end-all technique that we should use for other diseases (that don't even have the possible greater populations in Medicare or VA files) is suggesting that we're pretty gullible taxpayers who are willing to settle for whatever we get.

We're not.