Hacker News Reader

Comment by m0zg | Hacker News Reader

I think eventually we'll have to sidestep most of this fraud entirely by enabling self-serve early diagnostics and cranking up the doctor-replacing technologies to the max. Machines are good at recognizing patterns in non-adversarial environments not starved for compute. There's every possibility that they could do diagnostics better than a human could. Treatment will be up to human doctors for the foreseeable future, though.

Too bad Theranos salted the earth for much of this innovation for the next decade. But it's still happening, starting with EKG and ultrasound.

View on HN

>But it's still happening, starting with EKG and ultrasound.

Whole-body ultrasound? How would an AI understand how to interpret images with low signal:noise? Who will be performing the actual ultrasound image capture? Who will be ordering the imaging tests? How will an AI understand a patient that can't meaningfully interact with an AI?

What's the utility of "EKG" [sic] for diagnosing non-cardiac problems? Or, do you come from the perspective that all medical problems can be diagnosed with an ECG, urine analysis, ultrasound (of what?), blood test, and CT scan? Do you realize that an ECG is more complex than a single vector ECG via Kardia mobile?

I thought we were trying to reduce healthcare costs, but you're recommending over-utilization of expensive diagnostics?

Why are you limiting treatment to doctors only, why not have AI do everything?

Everything you've written demonstrates an extremely limited understanding of the practice of medicine, a characteristic far too common shared by those who are trying to "disrupt" medicine.

Sure, I realize there's much that the current tech does not cover. But you gotta start somewhere, and through extremely wide deployment tens of thousands of lives will be saved, and important feedback will be collected to improve products. $400 watch that does a dozen other things and that the user would buy even if it didn't have ECG is not "expensive diagnostics".

This is where things are at, whether you like it or not, especially for the working poor who can't afford to fork over a few hundred dollars every time something feels "off". The current crop of devices is not affordable to them, but 5-10 years from now they'll be bundled in cereal boxes. Hook up sensors to wearables, catch diseases early and at a massive scale, cure or treat them before they get extremely expensive. To me it looks like it's the only way out of the ripoff "healthcare" scheme we're in right now. I'm pretty convinced the solution to the US healthcare problem will be technological in nature.

And don't tell me universal healthcare will fix things. It won't, because US doctors (unlike their counterparts elsewhere) expect to be earning a pretty penny after they go through the trouble (and expense) of educating themselves for a decade, and that's not going to change. There simply needs to be less demand for doctors, and the ones that remain should be able to use their time efficiently.

Smart/Self-serving diagnostic tools occasionally fail for legal feasibility reasons before they are adopted by the market. Consider a smart heart monitoring system, maybe connected to a canary that will alert a healthcare professional or first responders if something is amiss. Absolutely nobody wants to be liable for misdiagnosing acute symptoms, which may lead to death or false positives.

It doesn't even have to be acute symptoms, either. A large healthcare company a friend worked at was working on a smart heart monitor that for some reason they wanted to connect to an app. In their head, the app would be used by both the patient and their doctor who would check on the patient. If they have a lot of patients, that's an overwhelming amount of data streaming in 24/7. The time commitment to really evaluate those data for anomalies was too high, as were the stakes. Doctors were scared that if they missed something they might be held liable for unrelated medical emergencies which is terrifying for them. At the time, algorithms were really not good enough to catch the really subtle markers of an impending medical problem. At least according to the guy that told me this, he was a biomed/hardware engineer not a programmer or statistician by any means.

Even if you had some legalese associated with the product to make this a non-issue, doctors still wouldn't trust the idea. So they pivoted the technology to something else that was more traditional. It is much harder to pursue justice for an algorithm than a human. I guess what I want to get at here is: sometimes the software and hardware are working as intended, but subtleties can be lost on the technical side (programmers, biomed engineers) can kill a product. Stuff like market research, liability and legality, and public perception of the tech.

Sure, there will be cases where your device will say "I'm at a loss here, go see a doctor". The point is, right now the majority of (very expensive!) doctor visits are a total waste of everyone's time, and your money.

There's no rational reason why I need to drive to a lab for bloodwork or simple diagnostics, and have a human do it: a robot at a local convenience store should be able to do the job. There's no rational reason why I need a doctor to misdiagnose my elbow pain as something requiring surgical intervention (I just stopped using my laptop in bed and it went away). Another doctor ordered a ($900) MRI to diagnose pain in my left shoulder and then told me I have a SLAP tear there based on extremely blurry results (again $$$$, surgical intervention, and 6 months of recovery). Bullshit, bud, I don't - it's a non-dominant arm, and I'm someone who doesn't do a lot of throwing or work above shoulder level. It was that dang laptop in bed again. Someone with weaker critical thinking skills and google-fu would be thousands of dollars poorer thanks to taking that doctor's "advice". In fact they could end up with a permanently fucked up shoulder and elbow, too.

Did you know that medical errors are the third leading cause of death in the US, right after heart attack and cancer? Seems to me that "human" doctors are also a dangerous proposition.

>The point is, right now the majority of (very expensive!) doctor visits are a total waste of everyone's time, and your money.

Honestly, what are you talking about? A doctor's visit costs between $0 and $150, depending on insurance/government, and it is a focused encounter to determine the source of a medical complaint and a most likely effective treatment for that problem.

You want to stick a one-lead ECG on someone and expect that it's supposed to spit out a diagnosis, or heaven forbid, a recommendation to see a doctor who can actually figure out how to diagnose and treat the patient.

Your statement in fact points to the uselessness of this proposed medical device if it can't actually figure out what's going on and instead will send you to an actual expert (physician). How reliable would it be for everything else it "diagnoses"?

>> A doctor's visit costs between $0 and $150

And who do you think covers the remaining $150-infinity?

>> You want to stick a one-lead ECG

You're not arguing in good faith. Goodbye.

>And who do you think covers the remaining $150-infinity?

Primary care office visits of low complexity are typically less than $150. Suggesting otherwise is a sign of profound ignorance.

>You're not arguing in good faith. Goodbye.

You haven't come up with a single situation where a non-cardiac medical problem can be diagnosed solely with a handheld single vector ECG.

You don't mention anything about false positives or false negatives, sensitivity, specificity, pathophysiology, anatomy, pharmacology, diagnostics, or medical decision making.

You have no understanding of how your magical tech device would bridge every aspect of the practice of medicine, but you make an audacious claim with a bit of hand waving, saying "this will work better than physicians, and cost far less".

>Did you know that medical errors are the third leading cause of death in the US, right after heart attack and cancer? Seems to me that "human" doctors are also a dangerous proposition.

Where's your citation?

Here's what the CDC says about leading causes of death:

https://www.cdc.gov/nchs/hus/index.htm

Heart disease 257.6 (2000) 168.5 (2015) 165.5 (2016) Cancer 199.6 (2000) 158.5 (2015) 155.8 (2016) Chronic lower respiratory diseases 44.2 (2000) 41.6 (2015) 40.6 (2016) Unintentional injuries 34.9 (2000) 43.2 (2015) 47.4 (2016) Stroke 60.9 (2000) 37.6 (2015) 37.3 (2016) Alzheimer's disease 18.1 (2000) 29.4 (2015) 30.3 (2016) Diabetes 25.0 (2000) 21.3 (2015) 21.0 (2016) Influenza and pneumonia 23.7 (2000) 15.2 (2015) 13.5 (2016) Nephritis, nephrotic syndrome and nephrosis 13.5 (2000) 13.4 (2015) 13.1 (2016) Suicide 10.4 (2000) 13.3 (2015) 13.5 (2016)

Notice that medical errors are conspicuously absent from your list. Super convenient: if you don't count 'em, it looks like nobody died.

https://www.npr.org/sections/health-shots/2016/05/03/4766361...

Attributing death to a medical error instead of the underlying condition that would have been the cause of death anyway is missing the forest through the trees.

If you go into the hospital because you have stage IV colon cancer, and you die because a palliative surgery resulted in infection, you died because of the colon cancer, not the infection.

>This is where things are at, whether you like it or not, especially for the working poor who can't afford to fork over a few hundred dollars every time something feels "off".

But you think they'll fork over hundreds of dollars for misguided tech toys that won't even address their actual medical problem?

>Hook up sensors to wearables, catch diseases early and at a massive scale, cure or treat them before they get extremely expensive.

Which diseases are you talking about? Hypertension? Diabetes? Obesity? Smoking? You don't need fancy medical toys to diagnose these problems, yet they're responsible for most morbidity.

Do you think a medical toy will somehow detect cancer? How exactly do you think this would work? Someone feels "off" one day so they plug in a "device"... ECG? Ultrasound? Urine analysis?

>To me it looks like it's the only way out of the ripoff "healthcare" scheme we're in right now.

The ripoff isn't from physicians practicing medicine, it's from outsiders who interfere with the patient-physician relationship to get their slice of the pie, ie insurance companies, healthcare MBAs, device manufacturers, pharmaceutical companies. You think tech will replace physicians, but it's obvious you're completely misguided regarding the actual source of problems in healthcare.

You don't have to be rude...

> What's the utility of "EKG" [sic] …

What’s the Difference Between ECG and EKG?

The fact of the matter is that an ECG and an EKG are the exact same thing. That’s right, the most surprising difference between an ECG and an EKG is that there is no difference at all. Both ECG and EKG stand for electrocardiogram.

So, if an ECG is the same thing as an EKG, then why are there two different abbreviations? It’s actually quite simple—when the word electrocardiogram is translated into the German language, it is spelled Elektro-kardiographie. EKG is just the way some people choose to say ECG based on this translation.

From http://neurosky.com/2015/05/ecg-vs-ekg-whats-the-difference/

EKG also avoids confusing patients. ECG and EEG are a little to close to each other. And don't get me started on explaining what EGG is when you already have ECG and EEG to worry about.

I think IBM Watson salted the earth more than Theranos.

Reply on news.ycombinator.com