5 ms·

> make things like health care inexpensive enough... At least half of health care is "my magic book says that if you cough a lot and have a bit of temperature,

> make things like health care inexpensive enough... At least half of health care is "my magic book says that if you cough a lot and have a bit of temperature, you need to take these pills three times a day and call me again in a week". This can be automated. At least half of the remaining part is "if you run these tests and get these results, you need to take these other pills and call me again in a week". This can also be automated. I was discussing "expert systems" capable of doing the first part in the '80s. The reason we don't yet have them is political. I am still waiting for the Uber / AirBnB of healthcare (as in, "ignore the law, fix it after getting 100 million users").

grandalf 10y ago
> At least half of health care is "my magic book says that if you cough a lot and have a bit of temperature, you need to take these pills three times a day and call me again in a week". This can be automated. You're referring to non-serious concerns expressed to a primary care physician that are handled by a simple differential diagnosis. These are common but I think you'd find that they make up a very small percentage of total healthcare cost. The issue is this: Suppose the magic AI tells you that if you take the pills and the problem goes away, that everything is fine. Then two years later you develop a more serious condition that was missed by the AI. People want to be able to hold someone accountable for healthcare mistakes, which is one of the roles doctors serve. Most of the costs of healthcare are due to surgical procedures or hospital stays. The magic book AI isn't going to do much to prevent those. Surgeons are scarce because they are required (by culture) to be overqualified and make very few mistakes. If we were OK with a more normal error rate for surgeries, then more than 1% of med students would be accepted into surgical residencies. Because of scarcity, prices are high, because of expectations of extreme quality, lawsuits are costly and insurance is costly. There are already expert systems in use to help docs manage the early stages of a differential diagnosis in many healthcare systems. This is driven by cost because when a doc follows a non-fruitful branch of the decision tree too far it wastes money. So I think there is room for a lot more technology in healthcare, but rather than political the problem has to do with the social role occupied by doctors and the human need for authority and certainty when dealing with things that are full of uncertainty and risk.
- saosebastiao 10y ago
  For a very long time ATMs were suspicious to users. What if I asked for $40 and it only gives me $20? Why can't I get $23.45? I need to speak to a human because surely my (entirely commonplace) problem is unique enough that a machine can't handle it. What if I ask for a receipt and the machine has a bug that transfers my entire asset portfolio to the Netherlands?! There are plenty of social important and authority-dominated areas of our lives that have been taken over by technology. Health care can easily be one of them if we let it. It will go through the typical technology adoption curve, but going through it will happen nonetheless. And yes, people still prefer to interact with human professionals. When they know the costs of doing so, most choose a competent alternative, especially so for things the alternative is good enough at doing. It's the entire reason legalzoom and webmd exist. The real problem is that we are hell bent on obscuring the cost of health care and demanding way beyond our means.
  - grandalf 10y ago
    I think those are good points, but very few health conditions lend themselves to a smart chatbot replacing a doctor. Things like preventative care would be much improved by AI, but that is not protected by licensing. So why hasn't it become more common? Largely because people don't really care about it. Health is viewed as being binary, and our system is geared to address severe trauma (via the ER) and routine surgeries (hip replacements, coronary bypass operations, etc.). Primary care is improving, but in our current landscape it's a gateway to the more scarce resources of specialists. Most people who see a primary care doctor regularly are far from healthy and the doctor deals with managing multiple complex disease processes and specialists. Yes, you could replace the primary care doc for a patient with ongoing heart disease, kidney disease, and mobility issues with an algorithm, but that's not the hard part.

grandalf 10y ago

> At least half of health care is "my magic book says that if you cough a lot and have a bit of temperature, you need to take these pills three times a day and call me again in a week". This can be automated. You're referring to non-serious concerns expressed to a primary care physician that are handled by a simple differential diagnosis. These are common but I think you'd find that they make up a very small percentage of total healthcare cost. The issue is this: Suppose the magic AI tells you that if you take the pills and the problem goes away, that everything is fine. Then two years later you develop a more serious condition that was missed by the AI. People want to be able to hold someone accountable for healthcare mistakes, which is one of the roles doctors serve. Most of the costs of healthcare are due to surgical procedures or hospital stays. The magic book AI isn't going to do much to prevent those. Surgeons are scarce because they are required (by culture) to be overqualified and make very few mistakes. If we were OK with a more normal error rate for surgeries, then more than 1% of med students would be accepted into surgical residencies. Because of scarcity, prices are high, because of expectations of extreme quality, lawsuits are costly and insurance is costly. There are already expert systems in use to help docs manage the early stages of a differential diagnosis in many healthcare systems. This is driven by cost because when a doc follows a non-fruitful branch of the decision tree too far it wastes money. So I think there is room for a lot more technology in healthcare, but rather than political the problem has to do with the social role occupied by doctors and the human need for authority and certainty when dealing with things that are full of uncertainty and risk.

saosebastiao 10y ago
For a very long time ATMs were suspicious to users. What if I asked for $40 and it only gives me $20? Why can't I get $23.45? I need to speak to a human because surely my (entirely commonplace) problem is unique enough that a machine can't handle it. What if I ask for a receipt and the machine has a bug that transfers my entire asset portfolio to the Netherlands?! There are plenty of social important and authority-dominated areas of our lives that have been taken over by technology. Health care can easily be one of them if we let it. It will go through the typical technology adoption curve, but going through it will happen nonetheless. And yes, people still prefer to interact with human professionals. When they know the costs of doing so, most choose a competent alternative, especially so for things the alternative is good enough at doing. It's the entire reason legalzoom and webmd exist. The real problem is that we are hell bent on obscuring the cost of health care and demanding way beyond our means.
- grandalf 10y ago
  I think those are good points, but very few health conditions lend themselves to a smart chatbot replacing a doctor. Things like preventative care would be much improved by AI, but that is not protected by licensing. So why hasn't it become more common? Largely because people don't really care about it. Health is viewed as being binary, and our system is geared to address severe trauma (via the ER) and routine surgeries (hip replacements, coronary bypass operations, etc.). Primary care is improving, but in our current landscape it's a gateway to the more scarce resources of specialists. Most people who see a primary care doctor regularly are far from healthy and the doctor deals with managing multiple complex disease processes and specialists. Yes, you could replace the primary care doc for a patient with ongoing heart disease, kidney disease, and mobility issues with an algorithm, but that's not the hard part.