Something Killed 56,000 Doctor Jobs — and It Wasn’t AI

Self-Diagnosis is Already Replacing Doctors, and AI will Accelerate This

Discussion of AI and healthcare tends to get wrapped around the issue of “will AI replace doctors?” The problem with this all-or-nothing formulation is that it’s easy to refute and impossible to prove—and it tends to get doctors angry, besides. So let’s start with something more manageable:

In fact, it would be very surprising if the answer was no, since much simpler technologies are already replacing doctors without anyone even noticing. Things like like over-the-counter drugs.

Over-the-Counter Means Out-of-the-Clinic

16 years ago, one of the business headlines in the New York Times read “CLARITIN TO SELL OVER THE COUNTER”, and the article noted that this decision by the FDA would “will bring substantial savings for the uninsured and allow all patients to obtain the drug without a trip to the doctor.”

One day Claritin required a visit to an allergist, and the next day . . . it just didn’t. There’s no question that the allergist was “replaced” in that equation, but what exactly replaced her?

No doctor required

No doctor required

Very likely it was “Dr. Google”: the ability of most people nowadays to search on the internet to find out which allergy medicine to take. This clearly wasn’t AI, but it was a new combination of information organization (i.e. Google), a literate population, and a drug that had very few side effects.

The same thing has happened to more than 700 drugs over the last two decades, per the FDA, including:

Nicorette gum (nicotine), Children’s Advil and Children’s Motrin (ibuprofen), Orudis KT and Actron (ketoprofen), and Aleve (naproxen sodium) for pain relief and fever reduction; Femstat 3 (butoconazole nitrate) for vaginal yeast infection; Pepcid AC (famotidine), Tagamet HB (cimetidine), Zantac 75 (ranitidine hydrochloride), Axid AR (nizatidine), and Prilosec OTC (omeprazole magnesium) for heartburn; Rogaine (minoxidil) for hair growth; and Claritin (loratadine), the first non-sedating antihistamine.

Add to that list the anti-overdose drug Narcan (naloxone), newly available at the time of this writing in 46 states without a prescription.

Missing: 56,000 Practitioners

So there’s a big precedent for allowing patients to make decisions — including regarding medication use — that once required special expertise.

In some cases a crisis pushed the drug to over-the-counter (OTC), as with Narcan. In some cases, issues of insurance revenue and pharmaceutical sales played the biggest role, and in others simply the recognition that the upside outweighed the downside.

But in every case, the elimination of the need for a prescription has reduced the revenue going into the healthcare system, and reduced the need for doctors.

In fact, a 2012 study by Booz (done even before many of the drugs mentioned above were made OTC) concluded that

If OTC medicines did not exist, an additional 56,000 medical practitioners would need to work full-time to accommodate the increase in office visits by consumers seeking prescriptions for self-treatable conditions.

You read that right: nearly sixty-thousand medical practitioner jobs have already been eliminated in the United States, not by the coming specter of artificial intelligence or robot doctors, or even by minute clinics, but by the over-the-counter medication and a re-definition of which diseases are “self-treatable.”.

Oto = Oh No!

Since the ability of the average person to self-diagnosis and self-treat for a wide variety of ailments is well-established — as is the resulting decrease in the number of doctors required — it would make sense that people with access to better and more easy-to-use diagnostic tools would be able to do even more of this. And this is where AI will accelerate the diagnose-your-own-illness phenomenon.

In my speciality, pediatrics, ear infections account for about 20 million physician visits per year, putting them right up there with allergies as a common pediatric complaint (and my own practice experience certainly confirms this).

If a child comes into a clinic complaining of ear pain, or other symptoms that suggest an ear infection, the standard diagnosis is made by otoscopy: examination of the ear with a lighted magnifying scope.

Enter Oto: an attachment for your iPhone which lets you take photographs of your child’s eardrum, similar to what a doctor would see when looking through the otoscope. These photos can then be viewed by your doctor, or by doctors provided by CellScope, and a diagnosis made and prescription written (if needed) without the need for a physician visit.

The current Oto approach requires there to be a doctor to interpret the images — but medical image recognition is one of the hottest areas of research in artificial intelligence. It doesn’t take much imagination to recognize that if an algorithm can be trained to distinguish ear infection images from normal eardrums that would eliminate the need for the large part of 20,000,000 doctor visits per year for ear infection in the United States alone: the equivalent of several thousand full-time pediatricians.

The Frog in the Hot Water

It’s famously easy for even large changes to happen so slowly that they aren’t noticed. I have a feeling that AI in healthcare will be like this (just like the unnoticed changes from OTC medications). And not as many doctors will be needed, so over time not as many will apply to medical school…

From the patient or parent side, one day you’ll be using that Oto device and the app will just read the image without having to send it to a doctor (update: my sources at the company confirm that it is working to incorporate AI image recognition). And you’ll be happy because it will be faster.

And if it’s proven that the algorithm is at least as accurate in diagnosing ear infections as a pediatrician, how long before Oto is able to send a prescription straight to the pharmacy?

Or, similarly, your child will be coughing but an AI app on your phone will listen to the cough and diagnose pneumonia and send the prescription.

Or you’ll take a photo of that rash on your feverish child and tell you it’s nothing — or tell you to immediately go to the emergency department.

And then one day you’ll be talking to your grandchildren about how when you were a child you used to go to the doctor’s office when you were sick. And to the bank to get paper money, and to the music store to get physical music disks . . . and how strange that seems today.