"They thoughtlessly order tests and thoughtlessly obey the results."
Hacker News, come on. You're better than this.
Taking it from the top: The obvious take is that the new tools this is referring to is EMR and things like Watson. Will return to this in a moment.
Subjective and objective data both a play a role in medicine. The eye of an experienced person can often see in a blink what would be missed by someone looking only at numbers in a chart. Gestalt, or the fast system of Kahneman, is invaluable when time is a serious concern. But noone starts out that way. The slower, methodical plod of consciously using bayesian thinking is how the art is learned. Hear hoofbeats, think horses, not zebras... trying to weigh all available data and attempting to chart a course that gives patients the best outcomes at the most reasonable costs. Nowadays additional hoops must be jumped through: laws constrain, institutions have policies that must be followed, and most of all care is dictated by what is allowed by the insurance company. Rather than an invisible hand, this is an invisible supervisor robbing much autonomy and initiative from a sense of worthwhile work. Furthermore the ever-present fear of litigation pushes towards a course with more testing than might be suggested by treatment and diagnosis alone: how would this course be defended if things go wrong, as they will for a certain number? All of these things individually stood to reason, but we as a society must keep in mind the cumulative weight of it all. Emergent phenomena isn't just a thing of programs and physics, it's a thing of human systems like healthcare.
Back to the article. A happy picture is painted of modern CT scans, yet it neglects the downsides. In 1980 the average per capita dose of radiation was 3.0 mSv, with 0.5 coming from medical imaging. It is now 5.5 mSv and rising, with medical imaging alone exceeding 3.0 mSv. Medical imaging is now a larger source of ionizing radiation than all other sources combined, with particularly high risks for those in utero or pediatrics. Like any other test or treatment, there is a risk/reward ratio. As technology improves, it is more likely to be adopted, not because earlier physicians were anti-technology Luddites, but because the improved technology changed that risk/reward ratio. We are more likely to use imaging with less exposure, or better yet use a modality without that risk.
Back to the Bayesian part of thinking... testing isn't perfect. I'd love to see a test that is 100% sensitive and 100% specific. But there are inevitably false positives and false negatives. Tools and tests need to be used in an appropriate situation. For example: I have a test that is 99% sensitive. Great! It'll catch someone with the disease, 99% of the time. So I can thoughtlessly order tests and thoughtlessly obey the results, right? Wrong. What happens if you use it to test for a rare disease that only 0.1% of the population will have? It depends on how specific the test is. How many false positives does it let in? If I test it on 1,000 folks indiscriminately, I'll end up with a basket of folks, only one of which actually has the disease. How many false positives got treated (and possibly harmed by that treatment)? Mammograms work this way (which have fallen a little out of favor in younger demographics without risk factors like the BRCAs), necessitating imaging and invasive biopsies that, upon further collection of data and review, seem not worthwhile for those under 40 and of questionable value under 50.
Tools are great! They need to be used appropriately though. Things have a cost, not just financial but physical and temporal. Indiscriminate use of tests and tools is the last thing anyone should want.
Nothing to add in a world of advancing technology? Bah. Most would love for its promises to come to fruition. EMR for example. We were promised time savings, with cross-talk between systems for better availability of data and improved patient safety. Mostly what has happened is administrators now have data used to push docs to see more and more patients (and spend less and less time with any one of them), all the while the paperwork stacks up. Somehow the paperwork never quite seemed to go away.
Maybe doctors don't reject tools that make their jobs easier. The article is full of tools that were eventually adopted, after all. I can point to many in development that have their ardent advocates, like point-of-care ultrasound among many others. Maybe they don't like tools that were sold as making their jobs easier but mostly don't, and instead benefit insurance companies and conglomerate administrators.
> Mostly what has happened is administrators now have data used to push docs to see more and more patients (and spend fewer and fewer with any one of them), all the while the paperwork stacks up. Somehow they never quite seemed to go away.
Under capitalism, old companies (like hospitals) don't really tend to adapt in response to market forces by actually changing anything as drastic as the shape/relative scale of their internal bureaucracy.
It looks like that happens from a 10,000ft view, but what's really happening is that old companies are just dying, having been outcompeted by new small companies that "grew up in" the market environment where the changes were "the new normal." And then, eventually, the new, small companies acquire the big old dying companies for their brand value—so the resulting merged company has the appearance of the big old company having managed to turn over a new leaf.
When a company is only slightly relatively unfit (due to e.g. serving a market with inelastic demand, like medical care), it can take decades for their relative unfitness to deplete their resources to the point that they'd seek to be acquired. The current heavily-bureaucratic hospitals might be actively dying right now—it'll just take them another 50 years to become all-the-way dead.
Hacker News, come on. You're better than this.
Taking it from the top: The obvious take is that the new tools this is referring to is EMR and things like Watson. Will return to this in a moment.
Subjective and objective data both a play a role in medicine. The eye of an experienced person can often see in a blink what would be missed by someone looking only at numbers in a chart. Gestalt, or the fast system of Kahneman, is invaluable when time is a serious concern. But noone starts out that way. The slower, methodical plod of consciously using bayesian thinking is how the art is learned. Hear hoofbeats, think horses, not zebras... trying to weigh all available data and attempting to chart a course that gives patients the best outcomes at the most reasonable costs. Nowadays additional hoops must be jumped through: laws constrain, institutions have policies that must be followed, and most of all care is dictated by what is allowed by the insurance company. Rather than an invisible hand, this is an invisible supervisor robbing much autonomy and initiative from a sense of worthwhile work. Furthermore the ever-present fear of litigation pushes towards a course with more testing than might be suggested by treatment and diagnosis alone: how would this course be defended if things go wrong, as they will for a certain number? All of these things individually stood to reason, but we as a society must keep in mind the cumulative weight of it all. Emergent phenomena isn't just a thing of programs and physics, it's a thing of human systems like healthcare.
Back to the article. A happy picture is painted of modern CT scans, yet it neglects the downsides. In 1980 the average per capita dose of radiation was 3.0 mSv, with 0.5 coming from medical imaging. It is now 5.5 mSv and rising, with medical imaging alone exceeding 3.0 mSv. Medical imaging is now a larger source of ionizing radiation than all other sources combined, with particularly high risks for those in utero or pediatrics. Like any other test or treatment, there is a risk/reward ratio. As technology improves, it is more likely to be adopted, not because earlier physicians were anti-technology Luddites, but because the improved technology changed that risk/reward ratio. We are more likely to use imaging with less exposure, or better yet use a modality without that risk.
Back to the Bayesian part of thinking... testing isn't perfect. I'd love to see a test that is 100% sensitive and 100% specific. But there are inevitably false positives and false negatives. Tools and tests need to be used in an appropriate situation. For example: I have a test that is 99% sensitive. Great! It'll catch someone with the disease, 99% of the time. So I can thoughtlessly order tests and thoughtlessly obey the results, right? Wrong. What happens if you use it to test for a rare disease that only 0.1% of the population will have? It depends on how specific the test is. How many false positives does it let in? If I test it on 1,000 folks indiscriminately, I'll end up with a basket of folks, only one of which actually has the disease. How many false positives got treated (and possibly harmed by that treatment)? Mammograms work this way (which have fallen a little out of favor in younger demographics without risk factors like the BRCAs), necessitating imaging and invasive biopsies that, upon further collection of data and review, seem not worthwhile for those under 40 and of questionable value under 50.
Tools are great! They need to be used appropriately though. Things have a cost, not just financial but physical and temporal. Indiscriminate use of tests and tools is the last thing anyone should want.
Nothing to add in a world of advancing technology? Bah. Most would love for its promises to come to fruition. EMR for example. We were promised time savings, with cross-talk between systems for better availability of data and improved patient safety. Mostly what has happened is administrators now have data used to push docs to see more and more patients (and spend less and less time with any one of them), all the while the paperwork stacks up. Somehow the paperwork never quite seemed to go away.
Maybe doctors don't reject tools that make their jobs easier. The article is full of tools that were eventually adopted, after all. I can point to many in development that have their ardent advocates, like point-of-care ultrasound among many others. Maybe they don't like tools that were sold as making their jobs easier but mostly don't, and instead benefit insurance companies and conglomerate administrators.