AI Better Than ID Physicians by 2030

I’ve almost decided to give up blogging and content creation. What is the point when AI can write a better blog than I can?

However, my audience consists mainly of physicians, and as a retired physician, I often ask other physicians when they think AI will replace them. “Oh, it won’t happen to my subspecialty,” I frequently hear.

So I asked AI, and here is its reply:

As a physician, you’re right to be curious about how AI will impact different medical subspecialties. The integration of AI into healthcare is rapidly evolving and will likely affect various specialties to different degrees. Here’s an overview of how AI may impact different medical fields and which ones may be most at risk:

Subspecialties Most At Risk

Interestingly, AI hones in on image-based specialties such as radiology, pathology, and dermatology.

I believe radiology has been called out as a dying specialty for the last 20-30 years. Of course, there are now shortages and impressive salaries.

AI quickly points out that it won’t replace these specialists but rather “augment their work, improving efficiency and accuracy.” That doesn’t sound right to me. When you can snap a picture of a lesion in your primary care clinic and have AI make the diagnosis, you don’t need to refer to dermatology (except to have that lesion excised if it is bigger than those silly punch biopsies that bleed like stink).

However, with slides and films, someone has to sign off on the diagnosis.

So, is the future a physician managing physician extenders and AI? It just might be.

Subspecialties Least at Risk

On the other hand, the three least affected specialties are surgery, emergency medicine, and oncology. The first two are chaotic and hands-on, which is unsuitable for current AI. However, oncology is almost entirely protocol-driven (in my limited experience). Care management, side effects, education, and counseling can all be done by AI instead of an oncology practice. Once AI can do the procedures and surgeries, we need orderlies to gurney the patients to the right machine and get them on the table. But that is decades in the future. Maybe.

The benign conclusion is that AI will augment rather than replace physicians.

I’m a retired Infectious Diseases physician. Honestly, in a few years, AI will be able to do everything I can and do it better. And it won’t be upset when the ED calls in the middle of the night asking for the dose of azithromycin in a 2-year-old.

How AI Will Replace Infectious Diseases Physicians

As a case study, I asked AI about how at-risk ID physicians are.

Since it didn’t want to offend me by calling me worthless in a few years, it again says that AI will augment and enhance the work. Here are the key points:

Decision Support

AI thinks it will help ID physicians make decisions. Nope, it will help primary care docs and hospitalists make decisions, obviating the need for an ID consult. After all, I can do 90% of my consults on the phone with access to the primary data. If AI has access to the primary data and the flow of clinical information, it will probably offer you this decision support before you even know you need it. Time to change the vanco/zosyn to ceftriaxone. Or Keflex x 5 days upon discharge. It will remind you it is time to send a C. diff test, or it may just order it for you. Who will collect the sample? Not AI.

Diagnostic Assistance

AI can read a blood smear for malaria and is an asset in gram stain interpretation. It can then tell you what antimicrobial to use and determine the course of treatment, possible side effects, and counseling needed.

Medical Records

As an ID physician, not infrequently, I felt I was the only one looking over the old records. Complicated patients often require more records review than time spent on the H & P. Soon, AI will have a live, updated, and ongoing summary of overall care and can parse individual issues. The computer will automatically keep and update the problem list and order and cancel unnecessary tests (like the daily labs in-house). The problem is ununified medical records full of faxes and scans from OSH.

Limitations

Of course, AI currently has limitations. Apparently, its antibiotic recommendations are appropriate in only 36% of cases, and some are harmful. In my experience, that is about the same accuracy as antibiotics provided in outpatient and, perhaps, inpatient settings. Antibiotic prescription is as much art as science, as many older physicians don’t bother understanding the new science and use decades-old approaches to antibiotic treatment.

When will AI replace your friendly neighborhood physician?

So, when will AI replace physicians? Not soon, but it is inevitable that workflows will change. Physicians will manage teams and oversee AI.

AI won’t disrupt the profession; it will replace it. Eventually, meanwhile, given the looming physician shortage, AI may help plug a few holes.

ID doctors, your days are numbered. There is nothing AI can’t do that I could as an ID physician, and there is a lot it can do much better.

All the more reasons to focus on Financial Independence. Instead of FU money, you will need insurance to give you time to adapt when AI takes your job. As it steals RVUs, physicians must find new ways to provide value in a world where cognitive skills are no longer enough.

AI and your professional societies won’t tell you this, but FI is the best way to ensure that you will be fine even once AI agents replace you.