The Computer Doctor
So it will be a little more time before computers take over our jobs as doctors. A new study found that “online sites that are symptom checkers, those tools that ask for information and suggest a diagnosis, are accurate only about half of the time”. Whoops. Now, I know some people think that the artificial intelligence will get catch up but I disagree. There is something more to being human. There are nuances that only a doctor, who is not rushing, can pick up. Read any Malcolm Gladwell book. This doesn’t mean they won’t keep trying, though. Less docs mean more money for hospital, administrators and insurers. They will see this 50% accuracy as progress. Heck, “the diagnosis accuracy rate for physicians is (only) 85 to 90 percent” so at what point do they say they are close enough? That is the only thing they care about. Find that number and full steam ahead with computer doctors!
We have these amazing myths about numerology, just amazing.
A myth is a story built on a set of axioms that teaches an authoritative meaning about something big. It becomes a myth when certainty about its truth is held with a passion out of proportion to the know-ability of its foundations, its axioms. It cannot be a story, but must be a beloved story.
The first chapter of Genesis is a story. Those fundamentalists who hold that it is literally and precisely true make it into a myth – they hold that truth based of faith. In spite of the assumption by some that “myth” is a contemptuous term, it only means a story based upon faith, not bedrock reason. Our culture is defined to some degree by the myths that we hold, and our fervency of belief; and the malignity of our myths.
Much of our current numerology phase is based upon myth that numbers *ARE* the most precise way of describing the things they are attached to – and moreover, the number is the fundamental core existence of that thing in its entirety. In quantum physics, things are waveforms and no more or less. Why not insist that all things are merely a measurable point in Hilbert space, a set of numbers, a tensor? Diabetes – a diagonalizable matrix, with a trace no less! That’s all it really is!! And the treatment – just the eigenvector(s) of the diabetes matrix, with A1c as eigenvalues of the transformation! Sure sounds smart to me!
But what rubbish! An entire field of mathematics, statistics/probability, studies the validity of measurements and what assertions can be drawn from them. The sheer massiveness of a set of number does not imply that they have the least smidgen of meaning.
We talk about using a machine to make a “diagnosis,” and another machine to grade its “accuracy,” based on intervening trivialities like workup of the diagnosis. The feedback loop is one machine’s ability to grade another.
When I take a history, I not only ask about the symptoms, but the impact on the life of the patient. Dyspareunia, for instance, is said to have a prevalence of (to affect) 10% of women and an unknown number of men during their lives. Achieving the diagnosis or finding its origins means nothing by itself – how much does it impact the life of the patient? One can confidently say after October 1, 2015, – “That’s a F52.6 (nonorganic dyspareunia) , good buddy!” or “An N94.1, or an F52.8, can’t tell.” Of course, I’m referring to ICD-9 625.0, to speak in today’s lingo.
Our myth is that those cryptic letters and numbers have some sort of meaning in human health. They are semiotic code values for a machine to associate with something in a language that is standardized across machines. “F52.6 is a true member of the set of disease assignments for person SmithJoe3239.” I may assert that the sentence is nonsensical – you may insist that I do so because I am a technophobe or some other sort of troglodyte.
All this crap has nothing do to with one human healing another – a mission that has existed as long as humans have had empathy and reason to speculate, that if others can get sick and suffer and die, so can I. Our latest myth to escape that reality involves skipping down the big Data Yellow Brick Road, and going to see the Digital Wizard. WTF, why not?
We’ve had it for years. Only now, the patients plug in the symptoms rather than describe them over the phone to an administrative functionary who then plugs them into a grid. One less person on the payroll.
Speaking of Malcolm Gladwell, there is an interesting interview with him on Medscape this week.
One thing computers can’t do is have a conversation with patients. I spent about 20″ yesterday with a gentleman reassuring him he didn’t have residuals from a concussion 4 months ago, but was simply overloading his brain with work issues. A computer can’t do that. A computer can’t listen to a lady who is frustrated with the system while her husband is undergoing chemo and a workup for syncope.
If you read Gladwell’s interview with Eric Topol on Medscape, he makes the point about what his 85 year old mother wants in a physician, and it ain’ t about a lot of technology, it’s about listening and trust.
Sorry, I just don’t trust my computer.
Eric Topol is an idiot.