Tuesday, May 31, 2011

2 Wolfram Alpha On Lyme Disease Diagnosis

So, I'm not sure how familiar my readers are with Wolfram Alpha. If you've ever heard of the Mathematica software package, these are the guys that created it.

Wolfram Alpha is an online tool that uses computational algorithms and data modeling to output objective information on a variety of topics.

In other words: It takes statistics and facts from numerous sources and creates a database on given topics on the fly.

(To learn more about what Wolfram Alpha's online tool does, go here: http://www.wolframalpha.com/tour1.html)

You can go to their homepage and type any term into their search engine and it will build a page based on that term and nothing but that term. Instead of a list of different link results like you see on Google, you get one page with many different kinds of information on that term or topic.

So for fun, I typed "Lyme disease" into their homepage.

The link for that query is this: http://www.wolframalpha.com/input/?i=Lyme+disease

Now, as part of the output for that query, I received the following table of data:


This is one of those moments where my head is breaking, and I can see where they are drawing their data from in the notes below the table - so I need to follow up and see if I can look at the original source for this data.

But just based on what I'm looking at right now, the words "Oh my god, where are they getting these numbers?" escaped my lips.  

According to the footnote, the data is not CDC based, but is based on actual patient visits to US healthcare providers between 2006-2007. I have to wonder why Wolfram Alpha is mining data from this source and not more recent NAMCS/NHAMCS statistics or another source, but it is interesting to see something that is not CDC for a change.

The data in this table does and does not reflect what I've observed online in the Lyme disease patient community.

For one thing, the chart states far more men than women contract Lyme disease and are diagnosed for it than women.  Yet the number of women I see online who discuss Lyme disease outweighs the number of men a great deal.

After reading Polly Murray's book, The Widening Circle, why women are more active online participants and more activist in general has been made more clear to me: the children. If you are sick with Lyme disease, that's bad enough - but if your child is sick, that's worse, and mothers seem to be more likely than fathers to reach out for answers and support online to help not only themselves but their children. (Sorry dads, I don't like the dissing either, and I'm sure a lot of you are active in your child's health - but statistically speaking more women do end up dealing with their kids' health issues for whatever reason.)

This reaching out and forming support groups began with women way before the internet - the internet is just an extension of what happens in real life.

(By 1992 there were over 100 patient support groups for Lyme disease - which kinda chips away at this idea that Lyme disease is a disease spread by reading the internet. What was the internet in 1992? Usenet?*)

But in this table, it's saying far more men get Lyme disease, and I see fewer of them online.

Okay, so another thing about this table,  just so the reader is clear on what they're reading, it doesn't take 22 doctors for one woman to achieve a diagnosis for Lyme disease here, even though I've heard so many stories about this phenomenon - no, what the table means is that for every 22 office visits a doctor sees, one of them will be a woman that gets diagnosed with Lyme disease. 

That one in six visits for men seems really high to me - too high, if you ask me - but in reading the fine print these are estimates and estimates which are weighted for US demographics (how? in what manner?).

I think what's striking about this one year of data is the extrapolation of just how many patients were diagnosed with Lyme disease over the course of 131,748 doctor visits. 

86,700 people in one year... is that closer to reality than the CDC reported 40,000 or so?

I really don't know how much stock to put in this data at all. I feel like writing Wolfram and asking them what gives. How was the estimation made and weighed and why are they drawing their data from NAMCS/NHAMCS - and if so, can they get more recent data or is this it? 

I'm thinking it would be more accurate a head count to get diagnosis numbers from doctors rather than the CDC, because not all cases are reported - but I don't know how reliable these numbers are or how the estimation was computed.

Update:

Just to add to this, check out this other table generated from the same data set:


Interesting.

Now look at the ratio for the number of  men who are diagnosed with Lyme disease to the number of women.  And now look at the total number of estimated cases. 

The only factor changed for initial data output was from "primary diagnosis at visit" to "any diagnosis at visit". What exactly does this mean?

Update #2:

Well, isn't this interesting. I googled "NAMCS" and my first result was this:
http://www.cdc.gov/nchs/ahcd.htm

On this page, we learn:
"The National Ambulatory Medical Care Survey (NAMCS) is a national survey designed to meet the need for objective, reliable information about the provision and use of ambulatory medical care services in the United States. Findings are based on a sample of visits to non-federal employed office-based physicians who are primarily engaged in direct patient care. 
The National Hospital Ambulatory Medical Care Survey (NHAMCS) is designed to collect data on the utilization and provision of ambulatory care services in hospital emergency and outpatient departments. Findings are based on a national sample of visits to the emergency departments and outpatient departments of noninstitutional general and short-stay hospitals."
So wait. This IS data from a survey the CDC knows about?

What percentage of these diagnosed cases are included in the CDC Lyme disease reported cases?

Why aren't these diagnosed cases mentioned on the reporting page for Lyme disease as a separate category of cases -  even if doctors did not fill out an official report?

Or are they mentioned somewhere on the CDC Lyme disease pages and I've missed it? Anyone want to confirm this for me?

Update #3:

Pointed out by an anonymous commenter on the blog just now - from the same page above:


Wow,  let's give the women lots of Ativan and give the guys none?

After reading this, we're all going to need Ativan.

Where they hell are these numbers coming from, though? The footnote here is really not useful.

And I'm sorry, if I am feeling sick with Lyme disease, getting it up is the last thing that would be on my mind... that must refer to a preexisting drug regimen. Even then... ouch.

I really have to wonder about these numbers.

Update #4 (hopefully the last?):

Okay, so I think I've answered some of my own questions here:
http://www.cdc.gov/nchs/ahcd/ahcd_faq.htm

Q. How are the data used? 
A. NAMCS and NHAMCS data are used to statistically describe the patients that utilize physician services and hospital outpatient and emergency department services, the conditions most often treated, and the diagnostic and therapeutic services rendered, including medications prescribed. The data are used by public health policy makers, health services researchers, medical schools, physician associations, epidemiologists, and the print and broadcast media to describe and understand the changes that occur in medical care requirements and practices. The data are disseminated in the form of public health reports, journal articles, and microdata files. 
Q: Can the ambulatory medical care surveys be used to find out how many people have a certain diagnosis?
A: No. The ambulatory medical care surveys (NAMCS and NHAMCS) are not based on a sample of the population. They are based on a sample of visits rather than a sample of people. The data can be used to find out how many ambulatory care visits were made involving a certain diagnosis. To get an idea of utilization of ambulatory care in the population, the number of visits can be divided by the population of interest to get a rate of visits for a diagnosis of interest.
Okay, partially answered. I still have to wonder how public health policy makers, epidemiologists, etc. use this data, though... and how closely this sample of visits maps to reality.

I really don't think anyone has a true picture of how extensive Lyme disease is and how many people have persisting symptoms of Lyme disease pre- and post-antibiotic treatment, and it would be useful to know how many cases physicians actually diagnose and treat versus report to the CDC as those are different numbers.

Even then, these surveys are only representational of doctor caseloads over a one week period during a given year - they are not solid figures.

* This is not all there was, but there were only 26 web sites in the world at the end of 1992.

2 comments:

  1. Even more telling...
    Using your Wolfram Lyme link, scroll down further as view at what is prescribed at that visit. The male/ female breakdown for tetracyclines was 84%/43%. For Benzodiazepines it was 2.7%/55%. Hmmmmmm

    ReplyDelete
  2. Anonymous,

    OMG, are you kidding?

    I need to post that... hold on, let me look...

    ReplyDelete

You can use <b>bold</b>, <i>italics</i>, and <a href="url">link</a> for links.

The Camp Other Song Of The Month


Why is this posted? Just for fun!

Get this widget

Lyme Disease

Borrelia

Bacteria

Microbiology

Related Posts Plugin for WordPress, Blogger...