One of the most fertile fields for malpractice is the humble diagnosis.  Errors in diagnosis lead doctors down the wrong path and result in either no treatment for a serious condition or a delay in treatment.  The best available evidence is that fully one-third of malpractice cases result from either misdiagnosis or delayed diagnosis.  I am surprised it is not more than that.  I see it all the time in my practice.  The patient sent home from the emergency department with chest pains that the doctor did not think were cardiac in nature.  The spot on the x-ray that the radiologist failed to recognize as lung cancer.  The patient with gradually developing paralysis that no one in the hospital seemed to notice until it is too late.  I could go on and on.  Help may be on the way.

Physician and a Woman Patient, 1665 Painting by Jan Steen - Pixels

I have written about some ways in which artificial intelligence (AI) is already making an impact on the practice of medicine.  Computers have been shown thousands of images of x-rays of the lung with lung tumors present.  With the benefit of this training, computers are able to detect lung tumors years before human radiologists can see something on the image.  There are many other ways in which AI is affecting the delivery of medicine and reducing errors.

Computers can play a role in preventing errors involving prescription medication.  With a list of the patient’s current medications, which can be taken directly from the patient’s electronic medical record (EMR), the computer can cross-check them against whatever new medication the doctor may be considering.  Possible negative interactions can be recognized in advance and decisions can be made about what drugs to discontinue and what new drugs to add.  The computer can also offer suggestions as to what the best medication might be in a certain circumstance and it might be one the doctor has not considered.  The computer can also correlate the patient’s medical history with the intended medication and may identify problems before they occur.  For example, a patient with a bleeding history may be a poor candidate for a particular drug that has the side effect of increased bleeding.  Maybe there is another medication to address the patient’s need that does not increase the risk of bleeding.

Doctors are the product of many years of training.  They know so much about so many things and medical science is providing new things to know every day.  The problem is our human brain.  There is simply no way that we can marshal all of our knowledge at one time.  Furthermore, we are victims of our unconscious biases.  For example, recency bias can make a doctor think of a recent case he had which involved many of the same symptoms, to think he is seeing the same thing again, and to fail to consider other options.  Confirmation bias can cause a doctor to consider a potential diagnosis too soon and then look for evidence that her initial thought is correct.

AI has the ability to marshal all of the knowledge it has at its disposal and devote it to the diagnostic problem at hand.  Better still, AI does not suffer from cognitive biases, although it is no better than the programming it has been given.  AI can “think” of rare conditions that a human doctor might never consider but that just might be what is ailing the patient.  There is an old saying in medicine that when a doctor hears hoofbeats, she thinks of horses and not zebras.  This is fine if the patient has a relatively common illness but can be devastating if the patient has something rarely seen – a zebra illness.

Doctors are increasingly willing to use AI in their practices.  Fewer and fewer of them see it as a threat.  More and more of them recognize its potential to make them better doctors, who can deliver better care to their patients.

AI is already changing the ways in which we live and interact with our environment.  Medicine is no exception and the rate of change is only accelerating.

 

The post Doctors Need Help, Cue the Robots first appeared on Sandweg & Ager PC.