A new pilot program in Utah lets patients with chronic conditions such as hypertension, diabetes and thyroid disorders skip seeing a doctor and get prescriptions renewed by an artificial intelligence system that analyzes their prescription history before issuing a refill.
I understand the impulse behind this decision — I really do. As an emergency room physician, I see the problem this program is trying to solve walk through my doors every single day.
As an emergency room physician, I see the problem this program is trying to solve walk through my doors every single day.
Last week, a woman in her 60s showed up at my ER at 11 p.m., not because she’d had a sudden emergency but because she’d run out of her blood pressure medication days earlier. She wouldn’t have been able to see her primary care doctor for weeks and couldn’t get anyone to approve the prescription over the phone.
So she endured a four-hour wait at the ER and was charged a copay many times higher than what she’d have paid for a regular office visit, just to acquire a prescription she’d been taking successfully for five years.
What happened to her happens constantly.
Diabetic patients run out of insulin. Patients with chronic pain have to wait for an appointment. Patients with anxiety or depression miss doses because they can’t get in to see their psychiatrist. These aren’t people trying to game the system, they’re people caught in a health care access crisis. They have jobs they can’t miss to make a midday appointment three weeks from now, and they have conditions that can’t wait.
But even though Utah’s AI prescribing program is trying to solve a legitimate problem, it eliminates the doctor entirely from certain clinical decisions. That’s not solving the health care access problem; it’s giving up on the solution.
Introducing AI prescribers at this moment in American health care strikes me as profoundly tone-deaf. It takes the one element of health care that’s supposed to be personal, the physician-patient relationship, and replaces it with an algorithm. It tells patients that their care can be automated, that their concerns can be addressed by a chatbot, that the human element of medicine is optional.
Medicine has always been more than matching symptoms to treatments, it’s a relationship built on trust, continuity and the understanding that someone is looking out for your well-being as a whole person, not just managing your diseases as isolated data points. When we reduce health care to a series of transactions that can be automated, we lose something essential.
When we reduce healthcare to a series of transactions that can be automated, we lose something essential.
At the same time, patients have good reason to feel that when they do see a doctor they’re being rushed. That’s because primary care in the U.S. works on a reimbursement model that pays doctors to see as many patients as possible in the shortest length of time. Doctors have to fill their schedules to keep their practices financially stable.
But that’s not the only problem. For every hour a doctor spends with patients, there’s about two hours spent on paperwork. This is a big driver of the health care access problem. This administrative workload for doctors includes managing prior authorizations for medications and procedures, completing insurance paperwork and appeals, and writing electronic health record notes. These notes have changed from brief clinical summaries to lengthy legal documents to meet billing needs. On top of that, there is a constant flow of messages from patients, labs to check and referrals to manage.
AI is helping solve some of these problems.
AI ambient scribes can listen to patient visits and automatically generate clinical notes, freeing doctors from typing while trying to maintain eye contact and build rapport with their patients. AI systems are being developed to manage prior authorizations, and patient summary tools can condense months of medical records into brief overviews, helping doctors quickly understand complex patients.
Such applications of AI allow doctors to see more patients, spend more meaningful time with them and practice at the top of their license. This is how technology should work in medicine. It shouldn’t take doctors out of the equation.
Here’s what we lose when we do that: clinical judgment, physical examination and context. That woman who came to my ER for a blood pressure refill? When I examined her, I found significant swelling in her legs that she’d been ignoring. A further workup revealed that she had signs of heart failure. That means a simple refill would have been dangerous. She needed evaluation by cardiology and an inpatient stay for congestive heart failure.
An AI system looking at her refill request would have seen a patient on blood pressure medication requesting a refill. I saw a patient who needed something else.
