Ramblings of an old Doc

 

 

Sometimes you don’t feel well. You have some symptoms and don’t know what to do about them, or maybe your friend told you he/she isn’t feeling well. In the latter case, and if you aren’t a Doctor, best to tell him/her to talk with one…. this isn’t something you try to do at home without really knowing what you’re doing. Sometimes you just want an idea of what might be wrong and what to do as initial self care. In that framework, if you are legally responsible for yourself, o.k. If not, just use it to try and outguess “House”.

The best advice about diagnosing health problems given to medical students is, “Don’t try to diagnose and treat yourself or your relatives.”

If this is for what you think might be non-emergent (yes Jafo, it is proper usage) symptoms or you don’t know and it’s for yourself, you can try out Symcat.

I took some screenies of “Headache” and the subsequent screens,  and hope they’re legible. It’s interesting (for me, anyway) to watch it calculate odds and likelihoods as you add or modify symptoms. What interested me was the degree of sophistication the algorithms demonstrated and if they matched my thinking. So if you’re going to test it, on one of the screens there’s a place to check off a “just trying it out”, and you should do that.

“The smart webapp walks you through what a medical professional might ask in a quick background check, from what your symptoms are and how long you've been experiencing them to how old you are and your gender, and family history of diseases. Symcat then uses clinical data (aggregate patient health records) to predict what could be wrong and give you advice.”

- http://www.theatlantic.com/health/archive/2012/03/data-driven-diagnosis-a-google-for-your-health-symptoms/254506/ 

Well somewhat, but that will probably improve.

I found it pretty accurate over a range of symptoms. I tested the more critical symptoms first and found very good to excellent correlation to what I would have recommended. I also tested less critical problems like sore throat, cough, etc. You can choose the default “%” or the “Likelihood” of diagnosis in the right column.

While accurate, I found it not to be spontaneous and it didn’t ask the '”right questions” at the “right time” for the problems of lesser acuity. It also tended to send you to the E.R. That might have been because they are still building the “Things I can do to treat myself” option. What I found to be good was the “Find a Doctor near you” option.

Again, please use this webapp with discretion. It isn’t perfect, and it isn’t “Watson” (the IBM supercomputer slated to help with very tough diagnostic problems). Even Doctors with decades of experience and the expert advice of specialists and sub-specialists are stumped at times. I think you’re better off with a caring person, but you can tell me that’s a bias, and you’ll be correct.

Please use common sense: Don’t automatically take its results as Gospel. Always err (if you are erring) on the side of “this could be something more serious. I should call and talk to a health Professional.” And remember you are biased and will have either the tendency to minimize your symptoms, or exaggerate them depending on too many factors to enumerate but which should be taken into consideration by a Professional who knows you, your cultural and personal factors.

It’s an art as well as a science.

That’s the strength, but also the Achille’s heel of these high powered statistically driven apps. It’s true that these statistics yield more accurate probabilities and likelihoods true for large patient populations and can be very helpful in judging treatment efficacies or diagnostic probabilities. The degree to which they can be applicable to an individual is the sticky point. As time goes on and more information is fed into these systems, they will give more refined answers. This is a great webapp being developed by two medical students at Johns Hopkins. They are deserving of praise.

A Doctor does it quite differently, though. He uses all his senses simultaneously, and integrates his impressions and experience very differently than a computer does and also has a “radar” which defies any sort of description except it rarely fails.

There is however,  no “Delphic Oracle”. There will always be uncertainty since there are new diseases occurring and being “discovered”, and old ones which are constantly evolving. This isn’t any more a static system than the world in which it occurs.

That’s why I’ll keep looking avidly at what’s being done, but keep in mind the limitations of both men and machines.

Sources:

http://www.theatlantic.com/health/archive/2012/03/data-driven-diagnosis-a-google-for-your-health-symptoms/254506/

http://lifehacker.com/5895878/symcat-diagnoses-your-health-symptoms-like-a-doctor-would-then-tells-you-what-to-do

edit: Have added "Medical" to the title. You're right, Philly.


Comments
on Mar 24, 2012

I thought this was going to be about computer problems.............boy, was I surprised. 

There is a commercial on TV that shows a Doctor using a jack hammer working on a street and he hits a water main.  The announcer than asks 'Do you want your Doctor doing your job?'..........Then why are you doing his?  I find it's best to leave the folks who are skilled and knowledgeable at their job to do it.