Comment by paulmd | Hacker News Reader

Comment by paulmd | Hacker News Reader

He's absolutely right, and people don't understand the level of "skilled work" that will be replaced by computers.

I write clinical decision support tools, and I interview users (doctors). Nowdays, doctors really just want algorithms to follow. They want to know what the fastest/cheapest way to diagnose problems is.

Frankly, they aren't particularly GOOD at diagnosing problems outside of these kinds of algorithms - for example, a likely cause of many Morgellons' disease cases is actually tropical rat mite (see Nick Mann http://www.telegraph.co.uk/health/8626767/Morgellons-Disease...). Doctors tend to diagnose delusional parasitosis just because it doesn't fit the algorithms.

If doctors are just going to follow a basic set of diagnostic procedures, we can easily automate that using the data that are put into modern electronic health record (EHR) systems. We have your vitals, we know the tests that have been done and the outcomes, and so on. As one (very blunt) doctor expressed to me, doctors just want to follow the diagnosis algorithm, he paid $100k to be a data entry technician, and I think that was pretty much on the money. That's not the only time I've heard that, either.

If doctors are going to be replaced, there's not much that CAN'T. It's just a question of designing a program/robot for the particular task, which is becoming increasingly cheap.

A funny fact. I am working on automation of various finance operations in different organisations , and in workshops often the business users will ask for complete automation; "I want to press the button and like that, reports produced." People fail to realise that the easiest thing to automate is pressing a button.

I am sure myself I am assuming the convenience of the technology that will take my place rather than realising the threat.

I have a hard time believing doctors will be replaced in large numbers. The nature of their jobs may shift away from hypothesis generation but ultimately a human will be making the final diagnosis, choosing from several options that have been vetted algorithmically. Legally & practically a human needs to arbitrate that process, people would feel uncomfortable otherwise.

Even if human doctors become demonstrably worse at it than the machines?

If? It's already happened: http://en.wikipedia.org/wiki/Mycin#Results (and this was in the 70s).

Yeah, in a very limited set of circumstances, and still with a pretty significant miss rate.

But the day will come when the machines are near-perfect.

And people already trust computers more than they did in the 70s. Perhaps not if you ask them directly, but the reality is that they do, with many areas of their lives, and without a second thought.

Then why isn't there a website to get medical diagnosis on a self-serve basis? (I'm not asking rhetorically, genuinely curious)

The FDA, mostly. They have a multi billion dollar industry to keep in business.

Patients are humans.

Why? what people do not understand about automation is that it doesn't need to be perfect - just good enough. Doctors make mistakes all the time, if say you replace doctors diagnostic duties with say IBM Watson MD then as long as he's just as bad as your average doctor it won't matter. The big open question in automation whether it's in transportation or medicine is liability, but that something that insurance companies can solve easily between them selves :)

Diagnosis can be automated to an extent, but we're far from being able to automate the clinical observation tasks necessary to gather data to feed the diagnostic algorithm. Automating clinical observations will require huge advances in NLP, image recognition, and robotic manipulators that are many decades away at best.

Shameless plug:

We are working towards a solution that takes in data to improve health outcomes: https://www.wellframe.com

It's a super exciting space and time to be in, but it's also super challenging, so we are hiring! Let me know if you have any questions!

Doctors don't just diagnose and move on. Maybe you've heard of surgeons?

http://en.m.wikipedia.org/wiki/Robot-assisted_surgery

>a likely cause of many Morgellons' disease cases is actually tropical rat mite

Many? You made that conclusion from a single self-reported case of finding that bug? Seems rather far-fetched to me.

Do I think that 100% of Morgellons' sufferers have a medically explainable cause? Nope, there's hypochondriacs everywhere.

Do I think that doctors are way too eager to dismiss diagnoses that don't fit their criteria? Yup. I have a sister who has an absolutely zebra diagnosis (astroblastoma tumor, <0.5% of cases, with typical characteristics), given by top doctors. It was a case punctuated with all kinds of medical accusations regarding the diagnosis.

I write Clinical Decision Support tools for a living, for a high-level nonprofit. Our content providers suck ASS. Do I worry that I might be contributing to inappropriate diagnoses/non-diagnoses? Yes, it's a thing I bear in mind every day, every meeting, and I have gone to bat to improve.

I'll take it as a "yes" - you have jumped from a single self-reported case to "a likely cause of many Morgellons' disease cases is actually tropical rat mite".

No - as stated, I have jumped to "there are many cases that aren't covered by generic diagnostic algorithms", and the doctors I'm interviewing who are espousing "I want to be a data entry technician" type philosophies.

Independent of the examples, that's a worrying philosophy. I write the tools for this - in theory, I assume the liability when Shit Doesn't Work - not that we don't do our best to push this off to Other Organizations With Medical Expertise.

Reply on news.ycombinator.com