5 ms·

I would argue that "professional judgement" should be taken out of the equation as much as possible. What would we lose by requiring any prescribed intervention

I would argue that "professional judgement" should be taken out of the equation as much as possible. What would we lose by requiring any prescribed intervention to go through a rigorously-defined flowchart/checklist? A lot of doctors think they know better than the consensus best practice. They're almost universally wrong. Off-label prescription? Mostly horse crap, unless it's part of a serious clinical trial. Unfortunately, it would be damn near impossible to get most doctors to give up any of their powers of discretion. It's all too tied up with the prestige of the role, so anything that challenges it is seen as a personal attack. For example, see how doctors in general react to pretty much any technology that automates part of their job, like reading an ECG. You can prove that the machine does a better job than the human, and they'll still reject it.

code_biologist 3y ago
This “flow chart” can exist now on the the back of well prompted large language models. Not that there isn’t potential for huge headaches for all parties (hallucination around medical topics will be really dangerous) but doctors should be prepared for far more sophisticated self diagnosis coming their way. I think the safest use might just be for lay users to get a third opinion and sanity check doctors.
bigfudge 3y ago
I think someone more knowledgeable than me should reply to this, but my understanding is that most medicine is still too complex, with too little data to reliably make machine judgement. We don’t have the training data. The comment is also uneccessarily confrontational. There are plenty of areas in medicine where checklist and sops have been widely adopted.
- jeffparsons 3y ago
  > [...] my understanding is that most medicine is still too complex, with too little data to reliably make machine judgement. Indeed! I should have been clearer: I'm not suggesting any kind of AI / ML stuff — rather, only that where there is an established and uncontroversial best practice process for, e.g., diagnosing and treating X (which there often is), then I would hope for there to be a requirement to make a record of the path a doctor took to the conclusion, including a record any deviations like not ordering some standard test that most doctors would before starting a patient on Y. It could literally be a printed flowchart with check boxes on it and room for comments. The goal is to make the cowboys out there a little more accountable (incentivise doing things by the book), and to stop treating vague notions of "professional judgement" (a common protest against doing things by the book, I hear) as sufficient reason for ignoring the consensus best practice or simply forgetting important steps. > The comment is also uneccessarily confrontational. Fair call. I suppose ranting and raving doesn't really help anything. But I sincerely believe that a significant part of the problem is the defensiveness that tends to arise when a person is told (or it is implied): you are fallible. This tool will help to catch your mistakes, and make you more effective. I'm sure this happens in every industry, and I see working with that as 90÷ of the challenge. > There are plenty of areas in medicine where checklist and sops have been widely adopted. Yep, I agree wholeheartedly. And it saves lives. Hospitals in particular have made huge improvements to patient care/survival in this way, and continue to do so every year with new initiatives. I just think there are still a lot of, uh, exciting opportunities for future excellence. :)

code_biologist 3y ago

This “flow chart” can exist now on the the back of well prompted large language models. Not that there isn’t potential for huge headaches for all parties (hallucination around medical topics will be really dangerous) but doctors should be prepared for far more sophisticated self diagnosis coming their way. I think the safest use might just be for lay users to get a third opinion and sanity check doctors.

bigfudge 3y ago

I think someone more knowledgeable than me should reply to this, but my understanding is that most medicine is still too complex, with too little data to reliably make machine judgement. We don’t have the training data. The comment is also uneccessarily confrontational. There are plenty of areas in medicine where checklist and sops have been widely adopted.

jeffparsons 3y ago
> [...] my understanding is that most medicine is still too complex, with too little data to reliably make machine judgement. Indeed! I should have been clearer: I'm not suggesting any kind of AI / ML stuff — rather, only that where there is an established and uncontroversial best practice process for, e.g., diagnosing and treating X (which there often is), then I would hope for there to be a requirement to make a record of the path a doctor took to the conclusion, including a record any deviations like not ordering some standard test that most doctors would before starting a patient on Y. It could literally be a printed flowchart with check boxes on it and room for comments. The goal is to make the cowboys out there a little more accountable (incentivise doing things by the book), and to stop treating vague notions of "professional judgement" (a common protest against doing things by the book, I hear) as sufficient reason for ignoring the consensus best practice or simply forgetting important steps. > The comment is also uneccessarily confrontational. Fair call. I suppose ranting and raving doesn't really help anything. But I sincerely believe that a significant part of the problem is the defensiveness that tends to arise when a person is told (or it is implied): you are fallible. This tool will help to catch your mistakes, and make you more effective. I'm sure this happens in every industry, and I see working with that as 90÷ of the challenge. > There are plenty of areas in medicine where checklist and sops have been widely adopted. Yep, I agree wholeheartedly. And it saves lives. Hospitals in particular have made huge improvements to patient care/survival in this way, and continue to do so every year with new initiatives. I just think there are still a lot of, uh, exciting opportunities for future excellence. :)