A.I. - Will it reduce our salaries?

[coroner]

I just heard an interview with Mustafa Suleyman, CEO and co-founder of one of the world's leading AI companies. He also published a best-selling book on AI The Coming Wave: Technology, Power, and the Twenty-first Century's Greatest Dilemma

Link to the podcast: Mustafa Suleyman interview (@30:47)

This is what he said regarding AI and its impact on healthcare:
"AI models are performing an incredible role being able to detect disease at human level performance across the board. The cost of producing a diagnosis and a proposed treatment pathway is about to plummet to zero marginal cost. It is going to be essentially free in the next decade to provide a world-class diagnosis. That is going to transform healthcare for billions of people."
Granted, he is an AI guy, so he will be hyping it up because he's got a book to sell and a company dependent on it.

I started a prior thread on AI 4 years ago. Instead of bumping the old one (see link below), I thought I'd make a new one because AI has advanced since then and anyone who hasn't been living under a rock has heard the growing media frenzy.
https://forums.studentdoctor.net/threads/mainstream-media-a-i-is-better-than-humans-to-dx-images-in-pathology.1365161/

But, based on his statements above, that made me wonder. The general sentiment from prior threads on AI is that as physicians, we won't become totally obsolete: at least not in the short term (who knows what will happen in the next 25-50 years). And AI will be more of an enhancement to medical providers rather than a replacement.

But what about reimbursement? Even if what he says that, "The cost of producing a diagnosis and and proposed treatment pathway is about to plummet to zero marginal cost." is partially true, that could potentially lead to reduced reimbursement for many medical specialties and is concerning. My retirement track might have been prolonged...either that, or shortened as I will have to look for more lucrative side hustles...

---

Members don't see this ad.

 

[caffeinegirl]

The quote you mentioned is hype. The assumption of zero marginal cost is far-fetched to say the least.
AI will make the job more efficient, and less prone to mistakes (that is of course if it is developed properly with validated ground truth training).
In pathology, what this has already translated to is screening of slides by AI with preliminary reports already generated for pathologist review. It's a combination of having a resident/fellow preview and put in a prelim diagnosis of the case (although the AI is better diagnosticallly) and have the attending sign it out.
This already makes it quicker to review and sign out a case and makes it quicker to filter out easy/hard cases. What that means financially is that you will get paid less/case if it already has AI review. But your overall take home salary may be the same - you'll just be signing out more cases in the same amount of time (this assumes that the practice has increased volume available to sign out).
There are tons of AI-assist tools already in practice in pathology - for breast, prostate, and skin - with more on the way.
Of course, I only touched upon the diagnostic/clinical side. For research, the potential is tremendous - especially for prognostic and targeted medicine.

 

[caffeinegirl]

Touching on reimbursement - there are already codes for digital pathology in the works. But that $$ will go to the lab/dig path provider. I expect that pathologist salary can be either $$/slide/CPT (contract style), or employed with a base salary + volume bonus, or a PC-only arrangement.
I don't think the PC reimbursement component (not your take home pay) will be reduced in the short-term due to AI (although there could be arguments for it).
I don't think you have to look at side-hustles yet @coroner - but those with tech savvy and digital pathology experience will have a leg-up during this transition. Currently the set up cost is still high - so only those high volume labs will be taking the plunge. Once consolidation is "complete" in the field, this will become more widespread.

 

[shikimate]

Can AI gross mastectomies and placentas? As far as I know, you gotta hire people to do that and those people cost 100K+/yr. Talk about "zero marginal cost", "essentially free"...
Does AI know when to order recuts? Submit additional sections?
Also forensics will never go away. That's why I continue to sharpen my autopsy skills, even though I'd never do forensics full time.
I also wonder what AI will do with dusty slides? Dandruff staining for PanKer?

 

[Med Director New England]

No matter how good AI is / gets it will need human supervision - so we got that going for us.

If we truly have a looming “shortage” of pathologists in the near future - AI will push us back to par by efficiency gains in a post AI apocalyptic medical system.

In the near future digital might put some solo paths out of business.

 

[caffeinegirl]

Agreed- AI in general will decrease physician workhours/FTE. As for AI in pathology - it will not replace the technical component (grossing and histology) but actually adds the step of slide scanning and QC. Its value lies in the professional component - slide interpretation and case report generation. A good algorithm will be able to determine slide artifacts (such as dust), and also triage cases that may need additional sections/stains. You still need a pathologist to "run the show." You'll just need less pathologist workhours. And let's not forget that the majority of pathologists in this scenario do not need to be onsite at the lab and can be fully remote. That's a huge change.

 

[hurrycane]

That statement sounds like something VCs would like to hear to pump free money into the startup 😁

 

[Zarniwoop]

So we become like cytotechs, who put a stamp of approval on machine screened slides?
And cytotechs become ______ ?

Why have AI outperform us at our 100+ year old process? Wouldn't it be better if the entire nature of processing and examining tissue changed?

 

[WEBB PINKERTON]

So we become like cytotechs, who put a stamp of approval on machine screened slides?
And cytotechs become ______ ?

Why have AI outperform us at our 100+ year old process? Wouldn't it be better if the entire nature of processing and examining tissue changed?

Will the amount of glass being pushed in the future be greatly reduced? I am just not real bullish on all this AI slide screening crap. Seems antiquated for what it likely coming.

You know what you need to be doing to prepare.....

 

[Med Director New England]

AI / algo’s are no more a threat to pathology than any other field.

In the scenario the OP described - developing the treatment plan using AI with all the various inputs of information (ie the oncologist job) in my view is much easier than using AI to become the eye’s & brain of the pathologist.

Also digital has to become widespread before AI will make a major impact on Histopath. We still have 5-15 years of the status quo.

But I do think it is a good idea to be in a group that is an early adopter. Academia and large labs will be at an advantage.

 

[caffeinegirl]

So we become like cytotechs, who put a stamp of approval on machine screened slides?
And cytotechs become ______ ?

Why have AI outperform us at our 100+ year old process? Wouldn't it be better if the entire nature of processing and examining tissue changed?

The process of processing and examining tissue is already changing - predicting treatment response, targeted treatment, and prognosis based on AI-detected morphological features is already happening. The classification of multiple neoplasms based on molecular expression is already happening. It's just a matter of time.

 

[caffeinegirl]

Paige Unveils Game-Changing AI That Revolutionizes Cancer Detection Across Multiple Tissue Types

Paige, a global leader in end-to-end digital pathology solutions and clinical AI applications, today announced the release of a groundbreaking product

www.businesswire.com

Just the tip of the iceberg

 

[mike-sheree]

Certainly sounds like the days of the morphologist are sunsetting 😢

 

[WEBB PINKERTON]

I still cant get over how quickly our oncologists adopted guardant360. Felt like it came out of nowhere. Definitely has had an impact on our work, less repeats for sure.

Tests for Patients with Advanced Cancer - GuardantHealth

Guardant Health offers cutting-edge tests for patients with advanced cancer, delivering insights to guide treatment decisions.

guardanthealth.com

My truffle pig said he is scared cause he heard about AI devices finding truffles. No one wants to be obsolete.

 

[coroner]

Also digital has to become widespread before AI will make a major impact on Histopath. We still have 5-15 years of the status quo.

Sounds plausible. Like most advancements in medicine whether in pathology, surgery, oncology, etc. the changes come first at the biggest hospitals and academic/research institutions. Community hospitals are last. As mentioned, the major effect AI would have would be as an adjunct (not replacement yet) to the PC of histopathology. So, after AI is implemented in the major academic institution via large-scale digitizing/scanning of images, the next step would be integrating them with the AI platform for diagnosis. Smaller/community hospitals would eventually get consolidated and funneled into a larger system that uses AI as workload becomes more efficient.

So we become like cytotechs, who put a stamp of approval on machine screened slides?
And cytotechs become ______ ?
Why have AI outperform us at our 100+ year old process? Wouldn't it be better if the entire nature of processing and examining tissue changed?

The process of processing and examining tissue is already changing - predicting treatment response, targeted treatment, and prognosis based on AI-detected morphological features is already happening. The classification of multiple neoplasms based on molecular expression is already happening. It's just a matter of time.

Acquisition and processing of tissue (TC) is manual labor requires actual manpower, so obviously AI could not replace that. But, who knows, maybe in 50-100 years we'll get an AI robot that can do that too. After all, there was a medical droid that delivered Luke & Leia Skywalker in Star Wars: Revenge of the Sith.

Also forensics will never go away. That's why I continue to sharpen my autopsy skills, even though I'd never do forensics full time.

That's true...somebody's gotta grind the meat! But, see example above of AI obstetrics...it could happen in forensics. We'll probably dead by then...

 

[Zarniwoop]

I still cant get over how quickly our oncologists adopted guardant360. Felt like it came out of nowhere. Definitely has had an impact on our work, less repeats for sure.

Tests for Patients with Advanced Cancer - GuardantHealth

Guardant Health offers cutting-edge tests for patients with advanced cancer, delivering insights to guide treatment decisions.

guardanthealth.com

My truffle pig said he is scared cause he heard about AI devices finding truffles. No one wants to be obsolete.

That pig of yours better start finding a new income stream (tattoo parlor, tea shop?). His reckoning may be sooner than he expects. Computers Are Learning to Smell

 

[KeratinPearls]

We are still in the first innings of AI and Pathology.

 

[sah]

So we become like cytotechs, who put a stamp of approval on machine screened slides?
And cytotechs become ______ ?

Why have AI outperform us at our 100+ year old process? Wouldn't it be better if the entire nature of processing and examining tissue changed?

The entire process may change. Grossing and glass slides replaced by AI melding radiologic images/lab markers/Nucleic Acid analysis/probes…