As devoted as our healthcare practice is to analytics that produce critical insights, a new study came out about Artificial Intelligence (AI) and physician prescribing habits that makes me believe that we are headed in a vaguely dystopian direction. I always have Kurt Vonnegut’s voice in the back of my mind warning that technology can be both magnificent and enormously destructive. Such is the case with physicians spending less time with patients and more time with data and computer models.
The latest study in this realm, coming out of Indiana University, claims to validate predictive computer models that help physicians prescribe the optimal course of medicine for their patients vs. the physician alone. The idea is simple – that doctors should rely heavily on data rather than their own “intuition” in order to make better and more “economical” prescribing decisions. On its face, the rationale behind the prescribing study makes perfect sense. Patient outcome studies are, by and large, reliable data from which one can make informed decisions.
And I understand that most (as the article states) aren’t calling for data models to “replace” a physician’s judgment. But the concept of physicians relying too heavily on data to make critical treatment decisions is troublesome, as physicians’ “intuition” and personal experience is exactly what you are paying for when you seek a certain course of therapy.
In the field of psychiatry, for example, talk therapy is woven into the fabric of a patient’s diagnosis and treatment cascade. Although psychiatrists are medical doctors, and sustain an astounding base of biological and neurological knowledge, their practice is as much of an art form as it is a scientific endeavor. The physician’s intuition is critical, and they must rely on their experience, gut, and focus on the individual case. Human interaction and emotional connection are irreplaceable in this field, and will be decisive factors in prescribing the right medical treatment – whether it’s a drug, device or procedure.
In the case of an internist, we may see more of a use for the computer model – especially if he/she is isolated and doesn’t see the sheer number of patients the way urban doctors do. And older internists that are set in their ways, and not familiar with the latest and greatest therapeutic options and procedures in their profession, would certainly elevate their patients’ standard of care by looking closely at this data.
But in the words of a close friend who is a partner at a NY-based gastroenterology practice, “I would never let data like this be a significant factor in how I treat my patients. If a computer model told me that a 25 year old male with rectal bleeding didn’t need a colonoscopy, and the patient’s verbal description of the bleed was slightly outside of the norm, I would go ahead and scope them anyway. I can tell by a simple conversation what the best course of action is. And I have saved more than a few lives based on my own intuition, and when needed, the judgment of my partners, colleagues and medical staff.”
Another related reason why this computer model is troubling is that physicians are spending significantly less time with patients these days due to managed care and poor insurance reimbursement. Physicians, especially primary care physicians, need to see more patients per hour in order to make a decent wage. Many of these docs would have no choice but to embrace dispensing with time- intensive consultations with patients, speaking with colleagues and reading literature to determine the right course of action. A quick data search on medication may very well do the trick.
So, if this approach is widely embraced at physician offices and large hospital systems, a logical outcome is that face-time with patients will further erode, and deeper physician evaluation of a patient and their medical history will be eliminated. Intangibles that emerge from these human conversations are at the heart and soul of diagnosis.
Finally, I fear that increasing economic pressures on our healthcare system is moving physicians further away from trying better, more innovative therapies toward less expensive options that may actually do more harm than good. This “one-size fits all” data approach seems like a recipe for stagnation in a system that needs to support new approaches to medicine.
On first blush, the Indiana University study – and others like it – make me think that the technology pendulum is swinging a bit too far to the right of what common sense would dictate in the healthcare industry.
Do you think this potential shift in physician prescription behavior will significantly impact how pharmaceutical companies and public relations practitioners communicate with decision makers?
Connect with Michael: