Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

inflecthealth.medium.com

Post reply on HN
Print view View original

291–300 of 368 posts

Re: I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

#291

Post by salawat »

SoftTalker wrote: post #141
operatingthetan wrote: post #22

Earlier quoted context omitted.

>is bad at understanding the atypical. To be fair I've found most doctors require a lot of convincing if your problem atypical as well.

As they should. If you hear hooves, think horses not unicorns.

Zebras, deer, moose, and giraffe, bulls of many types, all exist.

I'd say if you hear hoofbeats, contemplate what hooved animal would be appropriate to the situation and try to eliminate the easy ones. Lord knows, you might find a Zebra escaped from the zoo, or a bull from someone's field from time to time.

In short, life is messy, and taking any aphorism to the death will lead to just that.

Re: I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

#292

Post by VeninVidiaVicii »

PostOnce wrote: post #246
MrPatan wrote: post #227

One complaint he has is "It doesn't know to ask the right questions". Well, the prompt was to give diagnoses, not questions. Ask GPT for the follow up questions first, then the diagnoses. This is fascinating in that, because now the machine speaks human, we subconsciously ascribe human agency to it. This guy was instictively treating it like a colleague, who would naturally ask follow up questions unprompted. But you…

Other commenters take offense at your comment "doctors are done" and point to doctors doing physical things. GPT can already write robot control and machine vision code (poorly). The real questions are harder. For example, what about patients who lie or are too stupid to answer correctly? They can still be helped by a real human doctor, but GPT would be clueless. Another angle: we didn't have radiation sickness, HIV,…

> For example, what about patients who lie or are too stupid to answer correctly?

I can only speak from my work in the emergency department, but this is a leading cause of misdiagnosis among (human) doctors.

In fact, a lot of diagnosis relies on the patient’s story, and if they lie, you can’t feasibly perform every single test on them to figure out what’s actually wrong or why.

Re: I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

#293

Post by aix1 »

jeffrallen wrote: post #225

I found this quote really interesting: > this is not an uncommon scenario; about 8% of pregnancies discovered in the ER are of women who report that they’re not sexually active. We have so much work to do as a society to get honest about our bodies. Hoping my children do better; they are already getting better education than my wife did.

I don't think this is about education. It's more about religion, sex crime etc.

Re: I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

#294

Post by apwheele »

MrPatan wrote: post #220
apwheele wrote: post #218

Earlier quoted context omitted.

I think people are more impressed with the "well-described explanation", which is a (mostly bullshit) interpolation from prior text on the internet. If you just focus on the "feed a narrative and get out a probability of a particular diagnosis" -- the ability to do that with other machine learning models has already existed. If you give me a sample of 100k ER text records and final diagnoses, I would be pretty confid…

What percent of human doctors diagnoses is also well-decribed bullshit, though?

Totally get the point -- in the ER it is a mix of narrative and tests, so hopefully not so many. But in general practice, probably more BS than any of us would like.

Part of the difficulty with healthcare (that I myself don't have an easy answer to), is that you have very asymmetric error costs. So chest pain, even if heart attack is low probability, you should be risk adverse in many scenarios and still do due diligence to eliminate that as a possibility.

Not saying that smart systems are not possible, just still skeptical that level of complexity is possible on the backbones of LLMs. The smart sounding narratives I am concerned are mostly red herrings though in terms of outcomes we care about, accuracy/making right decisions/etc.

Re: I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

#295

Post by aix1 »

tedunangst wrote: post #18
chimeracoder wrote: post #11

> So after my regular clinical shifts in the emergency department the other week, I anonymized my History of Present Illness notes for 35 to 40 patients — basically, my detailed medical narrative of each person’s medical history, and the symptoms that brought them to the emergency department — and fed them into ChatGPT. It's quite shocking that the doctor would openly admit to violating HIPAA in such a brazen way. HI…

And yet medical journals are filled with articles with sufficient detail that other doctors can even learn to make diagnoses from reading them.

FWIW, case reports are governed by a different set of rules and are not directly comparable to what's described in this article:

https://hipaa.yale.edu/sites/default/files/files/Case%20Repo...

Re: I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

#296

Post by aix1 »

giraffe_lady wrote: post #283
aix1 wrote: post #268

Earlier quoted context omitted.

There's also an implicit assumption that his ground-truth diagnoses are 100% correct.

It's briefly but explicitly stated: "the “right” diagnosis — or at least the diagnosis that I believed to be right after complete evaluation and testing"

Point taken, it's an explicit assumption (but still an assumption).

Re: I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

#297

Post by iso1631 »

ChuckNorris89 wrote: post #256
iso1631 wrote: post #255

Earlier quoted context omitted.

Not when you're a doctor in the US dealing with the shortages of privitised car where you have a limited amount of patients who can afford your time

How does the US doctor get impacted by that? Your cheap shot jab at an counter argument makes no sense. You probably mean the patience are dealing with the shortages, not the doctors. Also, I'm only commenting on the EU system as that's the one I'm dealing with right now. Why do you feels the need to start a holy war on how the US system is worse? I have no expertise with the US system so I'm avoiding any comparisons…

The US doctor has an incentive to overmedicalise the patients as he'll make more money by upselling stuff they don't need. An AI assisting the diagnosis is detrimental to that as the AI will be doing the upselling and getting the kickbacks from those "ask your doctor about" advertisers.

A doctor could choose to see 20 people a day and charge $100 each, or see 10 people a day and charges $200 each. Indeed perhaps there's an incentive to see fewer people as they'll pay the same hourly rate but cost less to treat as they're just paying $250 an hour to talk about last night's game.

Reality is though that 28% of US people have to wait more than a day to see a doctor, far higher than Germany, UK, Sweden and Norway.

https://worldpopulationreview.com/country-rankings/health-ca...

> A common misconception in the U.S. is that countries with universal health care have much longer wait times. However, data from nations with universal coverage, coupled with historical data from coverage expansion in the United States, show that patients in other nations often have similar or shorter wait times.

So perhaps AI could help with the disfunctional american system afterall.

Re: I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

#298

Post by bityard »

lechacker wrote: post #278
MrPatan wrote: post #227

One complaint he has is "It doesn't know to ask the right questions". Well, the prompt was to give diagnoses, not questions. Ask GPT for the follow up questions first, then the diagnoses. This is fascinating in that, because now the machine speaks human, we subconsciously ascribe human agency to it. This guy was instictively treating it like a colleague, who would naturally ask follow up questions unprompted. But you…

There's 0 (zero) chance I'd use a robot doctor over a (qualified, capable) human doctor. None.

Finding a capable doctor is not easy in the first place. There are a surprising number of primary care physicians and even specialists that are really just terrible at their job, at least that has been my experience.

Re: I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

#299

Post by rscho »

weard_beard wrote: post #271
ale42 wrote: post #229

Earlier quoted context omitted.

> Doctors are so done. I would be curious to see the outcome if the patients entered the symptom descriptions themselves...

Anonymously without fear of being judged by a human who might be required to report dangerous or illegal situations... How many more people might be saved if they had to anonymously tell a computer the list of drugs they use regularly? Or to use the author's example, how many ectopic pregnancies might be resolved when the patient can freely admit they were raped by a family member?

If you think patients more easily release such info to a robot than to a human, I think you are mistaken. Trust is a large part of medicine, and many (most?) laypeople would fear breach of confidentiality. Irrational, yes perhaps but still...

Re: I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients

#300

Post by IIAOPSW »

MrPatan wrote: post #227

One complaint he has is "It doesn't know to ask the right questions". Well, the prompt was to give diagnoses, not questions. Ask GPT for the follow up questions first, then the diagnoses. This is fascinating in that, because now the machine speaks human, we subconsciously ascribe human agency to it. This guy was instictively treating it like a colleague, who would naturally ask follow up questions unprompted. But you…

I disagree. I've seen it ask questions unprompted. It happened in a jailbreak I stumbled into by accident. The LLM could do it in principle. OpenAI restricts it in unrpoductive ways.

https://i.imgur.com/WIzVXMm.jpeg

Post reply on HN