HealthGrades Buys WrongDiagnosis –but what are they really getting?

HealthGrades, which produces quality reports on doctors and hospitals, just paid $6.2 million to purchase Adviware’s website WrongDiagnosis.com along with another web property. I never heard of WrongDiagnosis but the name sounded intriguing and I took a look. There are some interesting diagnosis checklists –explaining what questions a doctor is likely to ask when a patient presents with certain complaints– and what’s behind the questions. There are some other areas of the site to check your symptoms or to view popularly misdiagnosed diseases.

The site is too busy for my liking and somewhat overloaded with pharma DTC ads and Google AdSense ads. The approach makes sense as a revenue model –show people the applicable drugs for the diseases they think they have.  I’d be willing to put up with all the clutter if there were useful self-diagnosis tools, but what’s there isn’t terribly sophisticated.

I asked SimulConsult CEO Michael Segal, MD, PhD for his assessment. Here’s what he told me:

They have what is effectively a huge table of findings in diseases.  There is no information about incidence of diseases or frequency of findings in various diseases, and thus there is no realistic sensitivity or specificity.  As a result, they have only a rudimentary way of suggesting other findings useful to check, based only on the number of diseases having each finding in their database.  Also, findings are treated the same regardless of the age of onset and there is no way to enter absence of a finding.

This reminds me of the database that the March of Dimes did 25 years ago.  When I did a rotation in clinical genetics with the expressed interest of writing a diagnostic decision support program the geneticist told me that it was a nice aspiration but unworkable, and showed me the March of Dimes database as proof.  I then got him to explain to me in detail how he does a diagnosis, which formed the basis for our software.  Years later, when I showed our software to him, he said, “This works.”  I told him that of course it works since it was doing what he taught me to do.

I tried entering the case that we use for our simple demo.  It took an excruciatingly long time.  I first did “Cloudy vision and Kyphosis and Large head,” which gave exactly one answer, but it was the wrong diagnosis (!).  If I’d chosen “Cornea/lens symptoms” instead of “cloudy vision” I’d have gotten the correct answer, but I wasn’t able to find that term until I looked inside the disease that I knew to be the correct diagnosis and figured out what I needed.  In contrast, our software lists the correct diagnosis as #2 after one finding and #1 after 2 findings.  WrongDiagnosis gave me 137 unranked diagnoses after one finding and 21 unranked diagnoses after two findings.

Maybe HealthGrades will improve the site over time. There’s certainly plenty of opportunity to do so.

October 16, 2008

Leave a Reply

Your email address will not be published. Required fields are marked *