Why computers can never replace physicians as diagnosticians


Category : Medical Rants

As I wrote recently, a good friend asked me that question – why not use computers to make difficult diagnoses?  Danielle Ofri's excellent post about the anchoring heuristic – Well: Falling Into the Diagnostic Trap – stimulated these comments:

Does it make sense to use a diagnostic software – one which has all the rules such as "If valium is given, B.P should go down else move to diagnosis 2" I think we are talking about finite and specific rules so it should be quite possible to create such a system and thus be rid of all sorts of biases of the human brain.


Wouldn't a computer have figured this out?

The answer is no.  The challenge of diagnosis starts with collecting and attending to the data.  Computers will never be able to take a proper history, because good history requires interactive skills, imagination, and attention to body language.  Taking a good history is not a one time process, but something that we do over time.  As new data arrive, we go back to the patient and ask more questions.

Daily I do "pre-rounds" when the interns and residents (along with the students) present new patients and discuss old patients prior to visiting each patient.  Sometimes stories make complete sense and jive with the physical exam and laboratory tests.  Sometimes the presentation leaves me confused.  Like many physicians before me, I then often say, I am confused and when in doubt "let's go to the bedside".

Computers are only as good as the data we enter.  If we are clueless likely the problem is the data input – either into our brains or a computer.  When we lock into a diagnosis, likely we slow down collecting more data, taking more history, repeating the physical exam.  We have to train ourselves to keep thinking and remain skeptical about our diagnoses.  Computers cannot do that for us.  Computers cannot touch the patient.  They cannot interpret the little clues that make us search more carefully.

Comments (3)

Agreed.  There is something to be said for clinical intuition.  Working in the ER, I have seen multiple cases where a patient presents with a seemingly straightforward medical complaint but on exam and through interacting with the patinet I can tell they are more ill than they appear.  Computers may be good reference tools but they should not replace the patient-provider face to face interaction. 

i completely disagree. as time goes on, we will see much better technology that will eventually fully replace physicians whether it is surgery or diagnostics.

Partly agreed. If you take example of EEG device and hardly injured patient which can not answer on questions, all what you have is brain activity. There are very good progress of developing algorithms (very complicated) to decide what type of injury are in patient case. I think that computer can better read that brain activity then doctor itself.

Post a comment