something i’d add is that Hanson refers to medical error as being “plausibly our third largest cause of death.” as i understand it that’s pretty exaggerated/ not true (see: https://www.mcgill.ca/oss/article/critical-thinking-health/medical-error-not-third-leading-cause-death), and as such it’s pretty irresponsible to suggest that. i don’t think it “signals” anything particularly good about hanson as an interlocutor on this topic.
Even then, "medical error kills people" is not at all the same as "an awful lot of modern medicine is just snake oil".
For instance, there's an egregious case covered recently in Ireland. "Died because she was not given antibiotics in time" is not the same as "Pfft, antibiotics are way over-prescribed and the best thing we could do is cut down on the unnecessary spending on them".
I’m glad this ended up somewhere more interesting than “maybe our entire health care system is useless.”
IMO (I may be biased) the strongest indictment of our current system is comments like Jake Seliger above. Federal bureaucracy strangles attempts to make progress on life threatening conditions.
The justification for this is safety, but you have to consider how much a person with a condition with 100% mortality is worried about side effects of a potentially life saving drug.
"but a lot of the stuff I work on in Australia is downstream of US pharma companies getting FDA approval"
From what I've read (apologies, I don't have a source for this off the top of my head) the U.S. is by far the biggest and most lucrative healthcare market in the world, so anyone developing a drug essentially has to be able to sell in the U.S. to make any money. Consequently, the FDA is the rate-limiting factor for much in pharma.
The problem with reciprocity is that the defining and formative experience of the modern FDA is the Thalodomide incident. Where European regulators broadly approved the wonderful new morning-sickness drug, and the stubborn FDA said "no, you haven't *proven* that it is safe by our standards; American women will just have to put up with morning sickness". Boo FDA!
And then ten thousand European mothers gave birth to children with flippers where their arms and legs should be(*). But not American mothers. Yay FDA!
Do you want another ten thousand flipper-babies? Because reciprocity is how you get flipper-babies. Or at least, that's what all the FDA's PR is going to be focusing on, along with the bit where you're an uncaring monster who wants flipper-babies so long as the mad scientists can play with their potions and Big Pharma can rake in the profits.
They've been playing that game for eighty years, and they're quite good at it. Big Pharma has figured out how to make boringly reliable profits playing the FDA's game, and nobody else is in a position to play at that level.
"Robin’s argument is strongest against prevention, least strong against treatment."
Agreed. I get that it feels intuitive to say that an ounce of prevention is better than a pound of cure, but for that to work, you need to know:
1. That you're likely to get the condition you want to prevent,
2. That the action will prevent or minimize the condition, and
3. That the preventative action won't cause iatrogenic harms as bad as or worse than the thing you're looking to prevent.
When you look at population-level interventions, they're often poorly targeted and benefit small populations. So you can end up distributing 10,000 pounds of prevention for every ounce of cure.
This isn't always true, of course, but who has an incentive to reverse course? Not pharma companies who have strong incentives to broaden their target populations. Patients don't have the expertise to look at large populations. Eventually research hospitals get around to vetting old treatments, but there's not as much glory as in looking for new ones.
I think it's also worth keeping in mind that there are plenty of efficacious preventative measures people can take against major health problems, the most effective ones just generally fall under the category of "lifestyle interventions" which people don't think of as "medicine." But it's because of medical research that we actually know that they're effective.
I think there are plenty of medical interventions that are obviously medical interventions (most vaccines) that qualify as sufficiently efficacious to merit broad application. It's just really difficult to parse which ones work and which don't without good evidence that takes 'too long' to implement, because people are dying today and 'we need something!'
I used to be much more convinced by your characterization of 'lifestyle interventions' as something that "it's because of medical research that we actually know they're effective." It seems to me that people used to be a lot more healthy and have much better nutrition before a massive glut of health and nutrition advice.
I guess it's possible that all the advice was spurred on by people behaving worse and worse, but if that's the case I'm not sure what the argument in favor of the advice is. Either it's iatrogenic harm or it's ineffective. I don't believe that the counterfactual of a world without the last 50 years of nutrition advice would result in people being more obese/unfit than today.
It's not like only nutrition advice rooted in the last 50 years is rooted in medical research. There's plenty we've learned before the 1970s that we still didn't know for most of human history. In the time of Galen, would people have known anything about the link between diet or exercise and diabetes?
That's kind of my point, though. A ton of people followed the advice to switch to low-fat foods, eliminate cholesterol, and to transition their lipid consumption away from animal fats to polyunsaturated vegetable oils. They mostly exchanged pork and cow for chicken. That doesn't seem to have had any positive health effects at a population level.
I knew a guy from my grandpa's generation who grew up on a farm. He used to talk about the meat they ate, preserved in animal fat before they had reliable refrigerators. (Space in the ice box was limited, I guess?) "I'd get home from school, dig out a slab of pork from the jar of lard, slap it between two slices of bread and slather some extra lard onto the sandwich. Looking back, I guess it was terrible for me, but I loved it."
To be fair, in some ways, our nutritional environment has probably gotten a lot worse than it used to be.
One thing I've speculated on occasionally is how much harm it might be doing to human health that we've dropped down to roughly 2000 calorie/day diets. Historically, most people have probably eaten significantly more than that, but also been active enough to have much higher maintenance levels. Getting all your micronutrients on 2000 calories a day is a very different proposition from getting them all on 4000 calories a day and burning off the excess. It's not like a slab of pork marinated in lard is actually bad for you, it's just that most people in the modern day aren't active enough to healthily maintain high calorie loads.
Maybe. I've heard a lot of bioplausibility arguments for why people are unhealthy and why intervention X - no Y - no Z! (etc.) will 'fix' this problem. Then people I know implement those solutions and don't get healthier, and large RCTs demonstrate short-term improvements, but then fail at long-term efficacy. I'm no longer as convinced by bioplausibility as I once was.
It's certainly true that US nutrition and activity have changed over the past century. This seems to have caused dramatically adverse health outcomes for a huge percentage of the country. I'll grant it's possible nutrition advice had no effect on that trend. Though my current priors point me more in the direction of advice being on net bad for health outcomes.
For anyone wanting to defend the proposition that nutrition advice has actually promoted better health, I'd question the counterfactual: would health outcomes be WORSE in the US in the absence of nutrition advice? What would that even look like, compared to the massive expansion of metabolic syndrome and nutrition-related disease?
I can see benefit in certain narrow cases, such as removing lead from children's toys. But stuff like the food guide pyramid or the new My Plate guide, or the thousands of admonitions about what to/not to eat seem like too much engineering biology that's doomed to fail in potentially deleterious ways.
Consider the thousands of dieting fads out there. "Diet and exercise" has been the perennial advice I grew up with for how to lose weight. Everyone KNEW this was the silver bullet, and that the only reason for people not getting there was because they were lazy, poorly disciplined, unwilling, gluttonous, etc. Yet RCT after RCT after meta-analysis has demonstrated that short-term weight losses are overshadowed by long-term net weight gain for dieters, and that this weight gain is in excess of what could have been expected from a non-dieter. So choose your diet, HFLC, keto, low fat, low calorie, IF, etc. So long as you choose any kind of dieting, you're choosing to gain weight over time. Yet this is the most common nutrition advice for people who are any kind of overweight, despite it being clearly ineffective for most people who in practice gain massive amounts of weight from serial dieting (including close friends and relatives of mine).
Why? Because it has been the most aggressively promoted nutrition advice, not because it has solid evidence to support its real-world efficacy writ large.
"The first detailed information concerning the aetiology, symptoms and treatment of diabetes derive from Rufus of Ephesus (98–117 AD), Aretaeus (120–200 AD) and Galen (128–200 AD). Caelius Aurelianus, a fifth century physician who translated Greek medical texts into Latin to transmit Greek medical views to Rome, mentions that Demetrius of Apamea, a Syrian–Greek colony in Phrygia, was the first to use the term diabetes, in the second century BC, but the writings of Demetrius have unfortunately been lost.
Rufus of Ephesus was a physician famous for his work on the variations of the pulse. He describes the symptoms of diabetes as ‘incessant thirst’ and immediate urination after drinking, which he calls ‘urinary diarrhoea’. He advocates induced vomiting after drinking cold water and mixtures consisting of several substances, such as ‘cyceon’ and ‘ptisane’ and cooked vegetables. He states that substances that promote urination should be avoided. Steam baths are also advised to cool the head whilst heating the body. The treatment is similar to that advised for fever, detailed below:
‘We then, apply to the hypochondrium a poultice made of dust from alphitonia mixed with vinegar and oil from roses or with grinded tender leaves of vine or with cotyledon, with pellitory (parietaria Judaica), with purslane (pulli pes) or with any other similar substance. Before lunch, we frequently administer juice of bistort to drink, of helenion mixed in dark wine, or a maceration of dates, of the fruit of mirth or of pears. At the beginning, we also make a venesection to the elbow. Sometimes we have to use stupefying medicaments.’
Aretaeus describes diabetes as a dreadful, chronic disease that develops over a long period of time and is not common. Furthermore, according to his views, the aetiology of diabetes is based on humidity and coldness in the organism and on chronic diseases of the small intestine and the kidneys with ‘hidden’ malignancy. Its symptoms are described as follows:
‘The flow is incessant... the patient is short-lived... for the melting of the flesh is rapid, the death speedy. Life is disgusting and painful... Thirst, unquenchable; excessive drinking, which, however, is disproportionate to the large quantity of urine, for more urine is passed; one cannot stop them either from drinking or making water. Or if for a time they abstain from drinking, their mouth becomes parched and their body dry; the viscera seems as if scorched up; they are affected with nausea, restlessness and a burning thirst; and at no distant term they expire. Thirst, as if scorched up with fire... The abdomen shriveled, veins protuberant, general emaciation, when the quantity of urine and the thirst have already increased... the disease appears to me to have the name diabetes as if from the Greek word (which signifies a siphon), because the fluid does not remain in the body...’
Additionally, patients may suffer from decay or sudden death, or have a disgusting and painful life .
The treatment for diabetes is similar to that for dropsy. Remedies are proposed to ‘treat’ the thirst originating from the stomach, such as purging the body with ‘sacred’ medicine (a mixture of many medicaments), compresses (made of spikenard, mastic, dates, raw quinces and rose oil), the sprinkling over the body of a mixture consisting of water boiled with fruits, styptic wine, shittah tree and hypocist, plus special food (milk with cereals, starch and wheat).
Galen of Pergamum, the most celebrated Greek physician of the Graeco-Roman era (128–200 AD) refers to diabetes in several of his writings. He believed that diabetes was a rare disease in which the kidneys are affected. Regarding the naming of the disease, the frequency of its appearance and its symptoms, Galen writes:
‘Some people refer to chamber-pot dropsy, others again to diabetes or violent thirst... For my own part I have until now only twice seen a disease in which patients suffered from an inextinguishable thirst, which forced them to drink enormous quantities; the fluid was urinated swiftly with a urine resembling water ... diabetes is a genuine kidney disease analogous to voracious appetite.’
He noted some further symptoms that accompanied diabetes, such as vomiting after drinking, diarrhoea, retention of the content of the stomach, continuous desire for urination or dysuria.
The aetiology of diabetes is attributed to malfunction of the ‘retentive’ faculties, as manifested by the continuous need for fluid and the inability of the kidneys to retain their content. Probably as a result of limited experience with diabetes, Galen does not provide details of treatment. The writer of the pseudo-Galenic treatise De remediis parabilibus does, however, advise the patient to take very thin membranes from the abdomen of young roosters, dry them under the sun and eat them. If this cannot be done, he suggests that the patient should grind together mountain copper, dry acorn, flower of the wild pomegranate and oak gall, add to honey of roses and cold water, and drink the resulting mixture."
I suspect from the sounds of it that this was type 1 diabetes, which in the time of Galen was invariably fatal. Type 2 diabetes, where diet and exercise has an important role, is largely a modern disease. Probably it was present to some extent in ancient times - gout, which has a similar pathogenesis of excess food and wine, certainly was ("the king of disease and the disease of kings.") But its effects - of foot ulcers, heart disease, kidney disease and blindness - would, I imagine, be much more apparent to ancient physicians than a high blood sugar reading. This link was discovered only recently.
Just be aware that lifestyle + most other primary preventive care measures probably do not do anything to influence health care at an aggregate level. Which is worth to keep in mind if you try to measure the effect of health care by looking at crude aggregate statistics on prevalence of diseases over time or across countries (none in this forum would do that, but you sometimes see analyses like that in less-serious newspapers).
Reduced health care costs on a national level only happens if the time period & treatment needed from people get seriously ill till they die gets shorter if people live longer before they get seriously ill (The compressed illness-hypothesis).
What might just as well happen if people increasingly avoid dying from non-communicable diseases in mid-life due to a healthier lifestyle & other successful preventive measures ("60 is the new 50" and all that), is that they are more likely to die with & from delayed degenerative diseases when they get older. (The extended illness-hypothesis.) These diseases are usually much more costly to deal with for insurance companies and/or national health care systems.
You have to die from something. If you and your significant others succeed in combatting the typical ilyfestyle illnesses early and in mid-life, you do not cheat death. You earn yourself an extension. That's all.
It's a question of chronic versus about disease. Solving metabolic syndrome and sedentary lifestyle would alone be expected to dramatically reduce health spending for the many associated conditions downstream. Sure everyone will die, but not everyone will be medicalized for chronic conditions for decades before they go. It's not clear the current medical system is optimized to do anything but promote chronic treatment bloat.
Hmmm...I don't have any data at hand, but I would stick to my general point that the longer people live before they die, the larger will be the percentage who get delayed degenerative diseases before they die. And these diseases are often extremely expensive - more so than treating sedetary lifestyles among people who can at least sit still.
Alzheimer patients for example often need 24/7 attendance by carers because they get up and walk around and do unpredictable things, and even unskilled carers are expensive when you need them all the time.
Then one can hope that we can find a cure for Alzheimer some day - there is promising research going on. Which is great of course, but just kicks the can further down the road, since this means that people will get something else instead before they die a bit further on in life. I am non sure which degenerative diseases that can be equally or even more expensive than Alzheimer - other types of dementia probably - but rest assured that something it will be. Because something it will have to be.
It would be good to have solid data to support either your view or my own, but I'm not sure how to quantify it exactly. All I have is personal experience of people who endure significant morbidity for year/decades, but who are also overweight and unfit, compared to people who are fit and active. I've watched both grow old and die, and the difference is often one of sudden versus chronic deterioration. Therein lies the difference in health resource utilization.
I think if you had a magic pill that cured heart disease, some 113 year old who was going to die of that tomorrow will die of a stroke within a week instead. Or something else once you cure that. This doesn't require them to incur hundreds of thousands of dollars of health system costs, though. It's true that the last year or so of your life is the most expensive, but this is often driven by chronic disease, and not everyone dies of chronic diseases.
Contrast this with a hypothetical 30 year old with non-alcoholic fatty liver disease. If they lose 80 pounds over the next year, the fat deposits in their liver will clear up and they'll avoid cirrhosis, a liver transplant, and maybe hepatocellular carcinoma. They might also avoid diabetes with attendant dialysis, peripheral arterial disease, and a kidney transplant. Instead of dying at 63 after millions of dollars of cumulative health resource utilization over a 3-decade ordeal, they die at age 83 after a 3-year fight with cancer that still cost a few hundred thousand dollars.
What you describe is the compressed morbidity hypothesis, and what I describe is the extended morbidity hypothesis. There is also a third possibility: the parallel-displacement-of-the-length-of-the-morbidity-period hypothesis. Which of them that will turn out to be correct is difficult to discover through research. Ideally, you should follow consequtive waves of birth cohorts till all of them are dead, and check if the average number of months/years needing expensive care over the life cycle goes up, down or stays constant for later birth cohorts.
I know of one study which did that in a local community, starting at age 75 (if I remember correctly, quoting from memory here) and followed two birth cohorts till all were dead. It supported the parallel displacement of the length of the morbidity period hypothesis. But apart from selection effects and other problems in the study, even when we get higher-quality studies (there is ongoing research on this) and find out which hypothesis is right today, we have no guarantee that the findings will stay valid 10-20-30 years from now, when the effects of population ageing on health and nursing care really starts to hit high-income countries. The Departments of Finance in the world certainly hope that you are right!
I am less sanguine, but keep an open (empirical) mind about which hypothesis that will turn out to be correct - since even better research today will not be able to give a sure answer about the future. At the end of the day, we just have to go on living and in due time we'll find out which hypothesis is correct:-)
As an Australian, I feel like this is one of the major strengths of a socialised single payer system. The government pays and also has power to influence prevention.
Some of the greatest triumphs in non-medical interventions that cut our healthcare spending - sunscreen/SunSmart (mostly administered through schools, no hat no play, getting children into the habit of applying sunscreen), free swimming lessons (also mostly administered through schools), and anti-smoking regulations.
People often say that Australia is kind of a nanny-state, and it kind of is, but IMO it's more efficient when the government can notice that they can cut ICU admissions by mandating bike helmets and then do that right away.
(Most recent example is the ban on engineered stone, due to silicosis).
wheres the take "most treatments are worse then useless, some are actually good(antibiotics after getting a bullet removed and you swell up like a balloon) the majority of patients are incapable of navigating the space"?
“Jake seems very dedicated to surviving as long as possible”
For me, living is mostly about keeping my wife, Bess, company. If not for her I doubt I’d have the fortitude or stamina to continue, and I’d have likely taken the opioid road by summer 2023. But she’s going to be very lonely when I die and so I’m trying to delay that moment as long as possible.
I think humans need projects and purpose to thrive, and part of mine has become 1. Explaining how clinical trials actually work—which I wish someone else had already done, but to my knowledge no one has. And also 2. The FDA is too slow, and so I’m hoping to somehow generate change, cause what I’ve gone through sucks, and it could’ve sucked much less if the FDA were keener on cost-benefit analyses.
On the testing question, I don't think it's just a question of inconvenience from the test. There are potential harms that come from excessive testing: https://youtu.be/yr_4RoPhtu4?si=RrA6KIFUfnkQzbit
Now, the existence of those harms doesn't negate the possible benefits of treatment. But it should change the nature of the discussion away from, "well, it's always better to test, because then you know." Because there's a difference between a positive test and real medical knowledge. At what point does testing become beneficial? We know it's somewhere between never test and test everything, but for many testing schema there's poor understanding of where potential benefit begins to outweigh the potential for iatrogenic harm.
> You could design a maximally cheap health care system from the top down where patients can't choose their physician, can't see a specialist without fulfilling specific guidelines, don't get access to any examinations that aren't evidence based, where there are long waiting periods for everything that isn't urgent, -- and this would save a lot of money probably without statistical detriment to outcomes. Patients would hate it though.
You've just described the UK's health system, the NHS. Historical, this has been extremely popular (as in, the NHS regularly polling as the UK's best gesture). Recently less popular, mostly because of increasing waiting lists to a point where they're currently probably longer than a cost-effectiveness analysis would suggest is optimal.
The Irish equivalent is the HSE, and it's long been mired in controversy over waiting lists, lack of access to services, constantly massively over-running budgets, etc.
It is a good system because it does provide medical care for all, and the private practice consultants often shuffle off patients onto the public system, but the problem is this little bit in the proposal above "long waiting periods for everything that isn't urgent" - there are long waiting lists even for things that *are* urgent, or become urgent while you're waiting to be seen and treated. Having to wait 5+ years for a scan because you have a twinge in your tummy may be a good thing; having to wait 5+ years for a scan while your heart condition quietly gets worse isn't.
We do have private health insurance companies operating, but it's not entirely great. We had new entrants come in (particularly during the Celtic Tiger era) and then leave again, because the Irish market is so small it's just not profitable enough. That leaves about three or four main insurance companies, which can then pretty much charge what they want. It's not as yet universal that all employers offer a company health plan (my place of work offers an employee pension scheme, for instance, but not health coverage while my brother does have it because he works for a pharma plant that is ultimately American multinational-owned) so it's individuals who purchase private health insurance themselves, if they're not in a scheme for work.
So we don't have the same level of 'high tech medicine paid for by your company health insurance' as the USA and the HSE is not at the same level of efficiency as the NHS (the HSE website will often copy word-for-word content from the NHS and just put a little note at the bottom saying they did this, which always makes me laugh). Falling between two stools, as it were!
There are two types of private health insurer in Ireland:
Open membership insurers provide insurance to everybody who wants to buy it from them. Currently, there are four open membership insurers operating in Ireland: Irish Life Health, Laya Healthcare, Vhi Healthcare, and HSF Health Plan. However, HSF Health Plan offers only cash plans. Cash plans don’t include any inpatient cover.
Restricted membership insurers provide insurance to people who are members of a particular group. This is normally a vocational group or employees of a particular organisation and their dependants. For example, there are restricted membership schemes for members of An Garda Síochána."
I came here to say exactly the same thing, and confirm that despite all the complaining that is present (mostly about long waits for non acute surgery and many assesments, with long waits at busy city A&E being perhaps more of a serious issue) NHS is generally thought of as a Very Good Thing, and while some people do use private options, many people who COULD afford them in theory don't, and accept the waits.
I also feel that UK primary care doctors deal with a lot of stuff managed by specialists in other countries. For anecdatal example I lived the vast majority of my reproductive age in the UK and while I never had any even moderately serious issues, I used contraception including coil, I have two children and I am menopausal now -- and in those 25+ years I encountered gynecologists on exactly two occasions, both while in labour in hospital. The idea of having "my obgyn" is completely alien to the vast majority of women here.
I think the value of choice is overstated naturally by the fact that people who write opinion pieces are more competent at and interested in making such choices, while the "free at the point of care" (and for large parts of the population, including 'free prescription medication") is a huge feature.
How many British people have had a chance to try US healthcare though? And not just in an ER room or something but the full experience with Blue Shield health insurance in a major city?
Most Americans think that public transit sucks but they also haven’t had a chance to try European public transit properly. Americans also used to be quite happy with their restaurant ecosystem 50 years ago even though it’s atrocious by modern standards.
Most Brits I’ve talked to are woefully ignorant of how the American system actually works and become quite jealous when I tell them of my experience.
I'm sure that much of the British horror at the US system is based on ignorance, or rather on being only exposed via media reports to its failure mode, and not the success mode.
On the other hand, I've heard people from other European countries (both Eastern and Western EU) grumble a lot about NHS and people who have experienced eg French system wax lyrical about that. So yes, it's very contingent.
I was mostly saying that the system OP described as a hypothetical one and one patients would absolutely hate actually functions in a fairly developed country on the other side of the pond and is NOT hated in principle.
I don't use healthcare much but basically had I had to pay anything I'd not use it at all, because I couldn't afford it, so if I was looking at comparisons, I'd need to look at free healthcare for the poor, which (very quick look) I probably wouldn't qualify for. I don't know how that compares to "blue shield full coverage" in quality or comprehensiveness.
As complete anecdata, I did a bit of medical phone interpreting at some point and while it was mostly UK cases, occasionally I'd be connected to a US call. The way the medical personnel, but especially doctors, treated the people I was interpreting for (often elderly, sometimes confused, sometimes after injury or even DV) was beyond belief in brusque, patronizing pseudo efficiency. I also sometimes did calls related to health insurance admin and the whole thing seemed entirely Kafkaesque. These are obviously not proper data, and I'm sure much care is excellent and delivered in a compassionate way.
I've been known to suggest that the NHS needs to be making regular payments to the US, for the valuable service of giving them "But at least we're not like the damn bloody Yanks! Don't you dare say a bad word about the NHS, or you might get *American* health care instead!", every time some Brit decides to say some very deservedly bad words about the NHS.
The European countries that seem to be able to make this work better than the NHS, mostly use hybrid systems with a more substantial role for the private sector. But they mostly don't speak English, so Americans default to just using the NHS (and its Canadian equivalent) as the model case for "Socialized Medicine".
It always amuses me when Scott says "let me propose as a thought experiment the following absurdly utopian and/or dystopian healthcare system", and the Brits all rock up in the comments to say "you've just reinvented the NHS".
b) They literally wrote "This is what public health care is like in some countries" right afterwards. They *know* about the NHS and were deliberately describing it.
To a lesser extent, it also describes most private insurance plans in the US if you're not willing to pay out of pocket. You have to pick an in-network physician, need a referral to see a specialist, need every intervention to have prior authorization from the insurance company (based on their judgement of whether it's necessary), and while I hear wait times are better, they're certainly not great (I personally had to wait several months to get an infected mass surgically removed from my sinuses).
It's a question of magnitude, but it's not like the US is winning any awards for quality-of-care-per-dollar, so it's not a public-versus-private issue.
Hanson is writing like a Holocaust denier, only he's being a medicine denier instead. Holocaust deniers often don't say "the Holocaust didn't happen", at least not immediately. They're just "asking questions", but the questions are skewed.
So Hanson doesn't say "medicine doesn't ever work". But he says things that amount to "how do we know medicine works? Here's a study that shows it doesn't seem to work. And we know there's a lot of bad replication in the field of medicine." He compares the belief that medicine works to the belief that you can make money in casinos. If questioned on that point, he can reply "I never said outright that medicine doesn't work", but the obvious intent of that is to say that medicine doesn't work.
You're not *misrepresenting* him by summarizing his position as "medicine doesn't work" any more than you're misrepresenting the Holocaust denier by calling him a Holocaust denier for asking "How do we know that the Holocaust really happened? Evidence isn't very good" without saying anything outright.
I think if a suspected Holocaust denier said outright "I admit the Holocaust definitely happened", I would stop thinking of them as a Holocaust denier, no matter what other weird positions they held. I feel like Robin has explicitly said some medicine works. I think his position about marginal units of medicine is relatively common and plausibly correct.
"Some medicine works" isn't like a Holocaust denier saying "the holocaust definitely happened". It's more like a Holocaust denier saying "of course the Nazis killed some Jews". Which they usually do say.
Hanson is making very strong statements whose implication is that he doesn't think medicine works. The reason that you are frustrated with summaries like “Health Care Is About Signaling” and “Most Drugs Are Bad For You” is that even when the summaries aren't created by Hanson himself, people read what he wrote, understand what he's implying, and write the summaries.
You are granting excessive charity to him by noticing that his exact words don't lterally say "medicine doesn't work". It's like the news media lying. It's not literally making statements, but it's implying them.
Have you never encountered the position of "Sure the Holocaust happened, but it killed maybe a couple hundred thousand people"? If someone said the Holocaust happened but killed literally twelve Jews, surely you agree that would make them a denier, so how many marginal Jews does someone have to deny the deaths of before they become a denier of the Holocaust overall?
I think once someone expresses a more or less consensus position on a subject, it's irritating and useless to compare that person to Holocaust deniers, and Robin's position is pretty common among health economists. This is more like saying that 5.8 million Jews died when the consensus range is between 5.5 million and 6.5 million or something.
I think we disagree on what Robin's position is. I don't think it's an accident that he keeps saying things which make people use titles like "Most Drugs Are Bad For You".
I guess I'm going to have to listen to this podcast and make a transcript. Will report back.
> All the government has to do is offer to pay for free STI treatment, and then if you have sex too much it’s an “externality” and you’re “robbing the government” so the government should be allowed to step in and stop you.
I don't recall seeing this in relation to STIs, but quick searches for "obesity cost to NHS" and "alcohol cost to NHS" bring up a *lot* of hand-wringing articles on how much fat people and drinkers cost the NHS in additional treatment. This is in turn used to argue for sin taxes on drink and unhealthy food, restrictions on advertising, etc.
Then let me blow your mind: corn syrup basically isn't a thing here, to the extent that UK-based cooks have to use alternatives when cooking American recipes: https://www.nigella.com/ask/uk-version-of-corn-syrup
I know! Apparently corn syrup goes back to the 19th century (so Wikipedia tells me) and I suppose I can see that if sugar was still a luxury good so this was an acceptable substitute, and it was popularised in the USA due to a big marketing campaign in 1910:
"Corn syrup was available at grocery stores in the 19th century, as a generic product sold from a barrel. In 1902, the Corn Products Refining Company introduced clear, bottled corn syrup under the brand name of Karo Syrup. In 1910, the company launched one of the largest advertising campaigns ever seen. This included full-page advertisements in women's magazines and free cookbooks full of recipes that called for Karo brand corn syrup. In the 1930s, they promoted a new pecan pie recipe that featured corn syrup, followed by a similar, nut-free chess pie recipe, in a bid to drive sales. Later, they promoted it as an alternative to maple syrup for waffles. As cooking in the home declined in the 21st century, so that fewer people made candies or pies at home, commercial sales of Karo tended to dominate over the retail sales."
Watching US cooking shows/channels where they use vegetable oil also puzzles me because I'm going "but why not use shortening like margarine or butter or lard or suet or something instead?" I keep thinking of vegetable oil as "what you deep fry your chips in" and not "what you use to make cake or pastry". This site may be setting itself up as fancy French baking (with a Californian twist) but I'm still going "what the hell?" about putting rapeseed - I'm sorry, I mean canola oil for delicate American sensibilities - in cake batter, and I'm pretty sure genuine French chefs would be ready to storm the Bastille over this:
"Coconut oil, one of the best oils for baking, goes well with bananas, chocolate, and tropical flavors. It has bold, aromatic flavors and can even contribute to a DIY body care routine, moisturizing your skin and hair. As a solid at room temperature, you can use coconut oil as a butter substitute in cookies, cakes and more. With a lower smoke point than some other oils, coconut oil is an excellent option for low-heat recipes like Virgin Coconut Oil Lemon Curd."
A floor wax *and* a dessert topping, hmm? Substitute for butter - the flip I will! I remain horrified by the history as presented on Wikipedia; truly, American zing and innovation in convenience foods has a lot to answer for:
"The process of the hydrogenation of organic substances in gas form was developed by Paul Sabatier in the late 19th century. Building on James Boyce's 1890s work in the successful development of a consumable solid lard substitute, Cottolene, in the U.S. the liquid form of hydrogenation was perfected and patented by Wilhelm Normann in 1903.
Joseph Crosfield and Sons acquired Normann's patent ostensibly for use in the production of soap. Their chief chemist, Edwin C. Kayser, was hired by Procter & Gamble's business manager, John Burchenal, and they patented two processes to hydrogenate cottonseed oil. Although their initial intent was to completely harden oils for use as raw material for making soap, these processes ensured that the fat would remain solid at normal storage temperatures and could find use in the food industry.
After rejecting the names "Krispo" and "Cryst" (the latter for its obvious religious connotations), Procter & Gamble called the product Crisco, a modification of the phrase "crystallized cottonseed oil". They used advertising techniques that encouraged consumers not to be concerned about ingredients but to trust in a reliable brand. Further success came from the marketing technique of giving away free cookbooks in which every recipe called for Crisco. Crisco vegetable oil was introduced in 1960. In 1976, Procter & Gamble introduced sunflower oil under the trade name Puritan Oil, which was marketed as a lower-cholesterol alternative. In 1988, Puritan Oil was switched to 100% canola oil."
'So, we *were* gonna make soap, but then we decided nah, let's use it for cooking instead!'
"They used advertising techniques that encouraged consumers not to be concerned about ingredients but to trust in a reliable brand."
Oh, Columbia! This is why I find it very funny when Dylan loses it over lard but is happy to measure out cups of vegetable oil:
Corn syrup, the kind found in grocery stores and used in home baking, is nearly all glucose with just enough fructose to inhibit crystallization. While it might be theoretically cheaper than cane sugar, it costs several times as much in retail packaging. It is only about half as sweet as sugar. While it is used as a major ingredient in things like pecan pie, its main use is as an additive to inhibit the crystallization of sugar. While you can accomplish the same result by partially inverting sugar syrup with an acid such as cream of tartar heated high enough and long enough to break apart enough of the sucrose molecules, adding a dollop of corn syrup is easy and foolproof.
As for cakes with vegetable oil, you owe it to yourself to try chiffon cake. It may have been invented by a guy from California, but it is a brilliantly good recipe.
Yes, here in the Rest Of The World we have healthy corn syrup alternatives like golden syrup and good ol' white sugar, which can happily be eaten by the cup with no ill effects.
"Compared with regular sugar, it’s cheaper and sweeter, and is more quickly absorbed into your body. But eating too much high fructose corn syrup can lead to insulin resistance, obesity, Type 2 diabetes and high blood pressure.
...Fructose was initially thought to be a better choice for people with diabetes due to its low glycemic index. But only your liver cells can process fructose, and that’s where the problems begin.
"Fructose goes straight to your liver and starts a fat production factory,” Dr. Hyman says. “It triggers the production of triglycerides and cholesterol.” He explains that it’s actually the sugar — not the fat — that causes the most trouble for your cholesterol.
What’s even worse, Dr. Hyman notes, is high doses of fructose “punch little holes in your intestinal lining, causing what we call a leaky gut.” He explains that this allows foreign food proteins and bacterial proteins to enter into your bloodstream, which triggers inflammation, makes you gain weight and causes Type 2 diabetes.
Studies show that high fructose corn syrup increases your appetite and promotes obesity more than regular sugar."
Yeah. It should be noted that "high" fructose corn syrup has...about the same fraction of fructose (to total sugars) as sucrose (ie table sugar) does. Which is about (for HFCS) 45-55%, vs the almost exactly 50% that sucrose does. Because sucrose is 1:1 glucose:fructose in a molecule that your mouth mostly breaks into its constituents, while HFCS is already broken apart (by enzymatically converting glucose to fructose). By the time it enters your bloodstream, all the sucrose is already broken down into glucose and fructose.
So 'High' in HFCS is only relative to *regular* corn syrup, which is basically 0% (only a tiny amount) fructose, not relative to any absolute standard.
So all those studies? Yeah, they're garbage. If you want high doses of fructose, *eat an apple*. Or drink "unsweetened" fruit juice. From what I find online, apples are ~2:1 (ie 67%) fructose:glucose, and are basically pure sugars.
Indeed. The only reason that corn syrup is cheaper than cane sugar in the US is because of tariffs and import quotas - if we didn't have trade restrictions intended to "protect" US farmers, our sodas would be sweetened by real sugar just like the soda in Mexico and most of the rest of the world.
With all due respect to Nigella, she is wrong. US supermarket corn syrup is not high fructose. The closest thing in a UK supermarket to corn syrup is called "liquid glucose." Corn syrup has a slightly higher water content, but it is basically the same thing. Dark corn syrup would require mixing in a little molasses or black treacle.
"I see some of my patients more often than I think is medically indicated because they feel more comfortable if they see a doctor more often (I don’t know why)"
Some people feel that they need to be "seen", i.e. they feel better if they get to explain their problems to someone, who indicates they they are listening.
As a parent, if the antibiotic gets my kids back to school/daycare a day earlier, so I don’t have to take off work it’s worth so much. I would imagine this effect alone is driving much of the ear infection antibiotic use. There’s even sort of a natural experiment you could do looking at leave policies and antibiotic use for ear infections.
I wouldn't have guessed that doctors are prescribing something that's supposed to o be medically efficacious to a child because the parent says "I gotta go to work tomorrow."
I mean, the guidelines leave it up to doctor discretion, but they usually say something like, "you don't strictly need to go on amoxcillin as it will likely clear up on its own in 2-5 days, but I can give you this prescription and they will likely be better by tomorrow."
We actually skipped the antibiotic once to test it out, and they were home from school for 3 days since they can't go back with a fever, even though they were completely fine otherwise. Seemed like a waste of everyone's time, so we always took the antibiotic after that.
Also, this is a terrible example of waste, because amoxicillin is so cheap. The doctor's visit is what's expensive regardless if a prescription is made. But again, if a kid has a fever they can't go back to daycare until its completely gone, or you have a doctor's note saying it's just an ear infection and they're safe to return. So you have to sort of know your kid and be like, "is this fever and ear pain driven by a cold or an ear infection?" when you decide to take them in or not.
It seems odd if “medically efficacious” is defined in a way that says that getting over pain a day earlier so that the carer can get to work a day earlier is not “medically efficacious”.
That's not just for treatment, though, I've had conjunctivitis and the gritty, dry-eye feeling was enough to get me to purchase over-the-counter eye drops for relief. So maybe the doctor shouldn't prescribe something but just tell the patient/parent to go buy Optrex at the chemist?
Though it seems that there can be cases where a prescription is warranted - oh look, and here's ear infections again!
"Bacterial Conjunctivitis
More commonly associated with discharge (pus), which can lead to eyelids sticking together
Sometimes occurs with an ear infection"
And the HSE advises that you should go to the doctor if a baby or child is the one suffering:
I meant specifically eye drops with antibiotics for treatment, not for immediate relief. And yes parents should go to the doctor when children exhibit those signs. It's just apparent that even if the cause is most likely viral (which can be inferred from symptoms), they'll give antibiotics anyway.
If a treatment is highly effective on 1/10 of cases and has no significant negative effects on the other 9/10 of cases, it doesn’t sound wrong to just give the treatment right away before figuring out which type of case it is.
The point of contention with antibiotic overprescription is that it may harden pathogens in the environment. There can be temporary deleterious effects on microbiota but I believe this is far less of a concern.
As an adult I'll happily take antibiotics if it will cure my reasonably-severe pain one day earlier, and I don't see any reason to make medical decisions for my kids that are crueller than the decisions I'd make for myself.
Antibiotics often make it easier to get past daycare gatekeepers. "The kid will be fine to return to school after 24 hours" is a more legible rule for the daycare than, "Wait until the symptoms go down. It'll probably be a day or two, maybe a bit longer."
I'm glad he clarified, as that is both a more reasonable position and likely correct. From a mindset of "how do we improve healthcare spending?" it's quite likely that his stronger language is necessary to bend the conversation away from current trends of spending money on anything that might possible increase health outcomes, even if many cases are wasteful or even make health worse. I doubt anyone in this conversation has decided to turn down all medical care, but maybe a few people will spend less on pointless or low utility services now.
For myself, I've actually been turning down healthcare when offered by doctors/hospitals for several years. If I need some kind of test or treatment, I take the minimum and often say no to additional tests or follow-ups. There's almost no benefit to going back to a doctor for that doctor to tell you that you're healing okay for instance. When you get your original treatment they should provide information about healing timelines and care. You can often tell for yourself if you're healing, and unless you're having significant complications or things aren't healing, there's not much they can or will do differently.
I'm going to ask a dumb possibly inflammatory question. It is based on ignorance. I vaguely remember hearing often in media that black women have high maternal death rates. And I think black women have a lot more abortions per capita. Would that suggest their maternal death rate is in part explained by the U S. method of including in maternal death rates anyone who has had an abortion recently? Or, to clear up another basic question. Let's say Hassidic women have 5 kids on average, and south Korean women have 0.5% kids on average. Would we expect Hassidic women to have 10x that maternal death rate, or the same? Is it counted per woman lifetime or per category justifying occurrence?
US maternal mortality rates are based on "if you die within a year of pregnancy, that gets counted". So there is some humming and hawing over "if you die of a cardiac event ten months after having a baby, was that due to the pregnancy or not?" Certainly, if you are in bad health, then pregnancy and delivery will put a further strain on the system, but in that case it's not the pregnancy as such that caused it.
It seems that black Americans have a lot worse outcomes for several diseases:
"The death rate for Blacks/African Americans is generally higher than whites for COVID-19, heart disease, stroke, cancer, asthma, influenza and pneumonia, diabetes, HIV/AIDS, and homicide."
(Yes, obviously "homicide" is not a disease, but the department of health includes it, so what can I do?)
"One possible contributing factor: The Centers for Disease Control and Prevention (CDC) says African Americans are more likely to die at early ages for all causes, as young African Americans are living with diseases that are typically more common at older ages for other races. For example:
High blood pressure is common in 12% vs. 10% of blacks vs. whites aged 18-34 years, respectively. It is common in 33% vs. 22% of those aged 35-49 years, respectively.
Diabetes is common in 10% of blacks aged 35-49 compared to 6% of whites.
Stroke is present in 0.7% of blacks aged 18-34 compared to 0.4% of whites the same age. Stroke is common in 2% of African Americans compared to 1% of whites aged 35-49 and 7% vs. 4%, respectively, in those aged 50-64.
The CDC said that social factors compared to others in the U.S., specifically whites, affect African Americans at younger ages: unemployment, living in poverty, not owning a home, cost-prohibitive effects of trying to see an MD, smoking, inactive lifestyle, or obesity.
A white paper from Cigna went further, acknowledging mental health disparities between African Americans and white patients. They noted blacks are 20% more likely to report psychological distress and 50% less likely to receive counseling or mental health treatment due to the aforementioned underlying socioeconomic factors.
Another area of health care there is a disparity is among renal disease. Blacks and African Americans can suffer from kidney failure at as much as 3 times the rate of Caucasians, according to the National Kidney Foundation. Black patients represent as much of a third of all patients in the U.S. receiving dialysis for kidney failure, though they don’t represent anywhere near that proportion of the U.S. population, they added. Individuals who are black alone, the Office of Minority Health says, make up 12.7% of the U.S. population."
Kidney disease seems to be down to genetic factors:
"African Americans develop kidney failure at more than four times the rate of white Americans. In part, this increased risk can be attributed to two genetic APOL1 coding variants commonly found in African Americans. The NIH-supported APOLLO study will explore how APOL1 gene variants affect kidney transplantation outcomes, with the goal of improving the outcomes of both kidney donors and recipients."
Oh, and "it's never lupus"? Well, unless you're a black female:
"Systemic lupus erythematosus is two to three times more common among African American women than among white women. It also is more common in Hispanic/Latina, Asian, and Native American women."
EDIT: I got interested in digging out maternal mortality rates and comparing them due to the, ahem, expansive definition of the CDC leading the pro-choice lot to claim that pregnancy was horrendously dangerous and abortion was in fact way safer and that without abortion so many more women would die due to being pregnant. I was interested to compare "mortality rates in countries without abortion or before abortion was introduced, or with abortion but compared to the USA". Clearly, if the CDC is claiming that being pregnant is killing off black women, then the call for "reproductive justice" is justified, right?
Regarding Lindley's "Paradox"... It isn't really a paradox. The Bayesian answer is simply correct, and the frequentist answer, if it differs, is just wrong.
BUT! This is so ONLY if the context of the question is as assumed. The Bayesian analysis in Lindley's paradox assumes a prior in which there is some some substantial probability (say, 1/2) that the true effect is EXACTLY zero, and otherwise (say with probability 1/2) the true effect is some non-zero quantity that we have no reason to expect to be close to but not exactly zero (perhaps our prior is uniform over (-10,+10)). Then, if you gather a lot of data, and find that your 95% confidence interval is (0.00005,0.00045) and your p-value is 0.006, you should not reject the null hypothesis, because the low prior probability that a non-zero effect just happened to be that close to zero, when the uniform prior on (-10,+10) provides no reason to expect that, more than out-weighs the low probability of getting an observed effect of this size if the true effect is zero.
In most real problems, including the one under discussion, there is no reason to think that non-zero effect sizes that are quite close to zero are unlikely. One can imagine lots of reasons that medicine might be a little bit effective, but not highly effective. Also, as noted in the comments quoted, the sample size is not actually all that big considering that mortality over a year or so is only about 1% (so presumably there were on the order of 10000 deaths, making effects of a few percent on the death rate hard to detect).
In particular, the effectiveness of medicine is a sum of many small effects. Everyone agrees that some medical interventions are helpful on net and some are harmful. The total measured effect of the study should be a sum of positive and negative terms. The idea that there is a substantial probability that the effect adds up to exactly zero is absurd. A reasonable prior distribution is more like a Gaussian or simply a uniform distribution.
Yes, the assumption that there is any substantial probability of an effect that is exactly zero is also wrong. However, the prior for the effect size should probably not be Gaussian, but instead something with heavier tails (equivalently, a higher central peak), because the individual positive and negative effects of medicine are not necessarily SMALL, as needed to conclude that the combined effect has a Gaussian prior.
It's funny we made the same point but described in opposite ways—I described it as the frequentist version being correct and the Bayesian one being wrong.
The frequentist solution here gives the correct answer to the question you *should* have asked (how sure are we about the sign of the effect?). The Bayesian solution gives the correct answer to the question the question you *actually* asked (is there a nonzero effect?).
I was going to bitch about Hanson based on his presumed background, which I now realise would be an ad hominem attack and not support the point I was trying to make in any way about "eff you for proposing making medical care more expensive to cut down on 'waste' because people died from not having sixpence for the dispensary doctor", but I got derailed looking up "am I sure he was born in relative privilege to my family background, maybe he *did* grow up with no running water and an outdoor privy" and found this plum from a 2003 article:
" Robin Hanson is a married, 44-year-old father of two who teaches economics at George Mason University, a commuter school with aspirations that's plunked amid the affluent sprawl of northern Virginia."
"A commuter school with aspirations". I'm laughing so hard at that, in an admittedly mean-spirited way, that I've loosened my grip on my wrath and let it fall away, and am now enjoying somebody else being snarky about the GMU economists 😁
I don't quite understand the ear infection discussion. Is there a reasonable case to be made that antibiotics for childhood ear infections are ineffective? Or is the case just "meh, they'll probably clear up eventually, let 'em suffer".
I say this as someone who almost never sees the inside of a doctor's office except when one of my kids has a (real or possible) ear infection, which seems to happen every goddamn six months.
Yeah, it's very easy to say "oh, that doesn't need antibiotics, it'll clear up in a couple of days" but that's because it's not aware of screaming in pain small kids who can't sleep or eat and the frazzled parents who have no idea if this is something mild or something really serious, and just want the screaming to stop. Sure, you can dose your kids up on Calpol, but that evokes uncomfortable associations of Victorian Soothing Syrups which drugged infants to death:
"Significant levels of morphine and alcohol were known ingredients in this product. The medicine claimed to treat teething for children. The product was popularly used among mothers. Yet the copious amounts of alcohol and morphine had the potential to induce coma, addiction, and death for infants. The product originally contained 65 milligrams of morphine before legislation and litigation reduced the amount over the next few decades. The American Medical Association labeled the product as a “baby killer.”
The product was eventually discontinued and removed from the market in the 1930s."
Drat those meddlesome FDA-adjacent types who interfered with a parent's right to hop their infant kids up on morphine!
"The syrup contained morphine 65 mg per ounce, as well as alcohol. One teaspoonful had the morphine content equivalent to 20 drops of laudanum (opium tincture); and it was recommended that babies 6 months old receive no more than 2-3 drops of laudanum.
One teaspoonful contained enough morphine to kill the average child. Many babies went to sleep after taking the medicine and never woke up again, leading to the syrup's nickname: the baby killer.
...The Pure Food and Drug Act instituted in the United States in 1906 forced companies to disclose the active ingredients on drug packaging. Companies also had to ensure that the purity level of the drugs was not below the levels established by the US Pharmacopeia or National Formulary. A similar law, the Food and Drugs Act, was passed in Canada in 1920 in a similar attempt, and also to ensure that drugs were properly marketed."
Also, earache may not just be an ear infection simpliciter but a sign of something else, depending if there are associated symptoms:
"Earache and pain can be caused by many things, but sometimes it's not known by what.
Here are some of the most common causes:
Possible causes of earache symptoms.
Symptoms Possible condition
Ear pain with toothache Children teething, dental abscess
Ear pain with change in hearing Glue ear, earwax build-up, an object stuck in the ear (do not try to remove it yourself – see a GP), perforated eardrum (particularly after a loud noise or accident)
Ear pain with pain when swallowing Sore throat, tonsillitis, quinsy (a complication of tonsillitis)
This seemed to be in a section that made the broader claim that benefits that aren’t “medical benefits” don’t get counted. Having a more comfortable stay, or enjoying your meals, or being with friends, or having less pain, doesn’t count unless it’s been defined as part of a “medical” outcome.
My understanding of the literature on this was that antibiotics reduce the length of the infection by a relatively small amount. We tried to avoid them for moderate-seeming ear infections because I have read that frequent use of antibiotics can cause autoimmune disorders in adulthood, but we were also lucky enough that our kids were just uncomfortable and not inconsolable. Certainly if they’d had a high fever we would have given them some.
Thank you, WindUponWaves, for the discussion of maternal mortality stats in the USA versus other nations. I had heard for a long time that the comparison wasn't apples-to-apples, but hadn't seen a clear description of what was off until your comment.
I've heard there are similar issues with infant mortality as well. E.g. in France, if a baby dies within 24 hours of birth, it's counted as a stillbirth rather than an infant death, whereas in the US, it's counted if they came out of the womb alive even if they die immediately afterwards.
Glad to be of service! I'm just chuffed that I got a mention in a highlights post at all, it's not often that you get a honor that really makes you proud to contribute. That notetaking system I developed where I write down *EVERYTHING* that catches my interest, and file it away for later reference, is really paying off!
>And I remember that during residency I worked with a doctor whose answer to all painful-but-not-otherwise-dangerous conditions was “nobody ever died of pain”. This guy probably had the lowest medical spending in the hospital, and maybe the lowest side effect rate in the hospital, and probably many other valuable records, but I would not have wanted to be his patient.
As The American Taxpayer, I would much prefer every doctor be a clone of this guy. If people want white glove treatment they should pay out of pocket, keeping people alive at the lowest cost is good enough for public spending on health.
The doctor is straight up wrong, though, about pain not killing people. If nothing else, chronic pain has been repeatly shown to increase suicide risk.
I have no sympathy for people who hold themselves hostage with suicide threats if they aren't provided with elective/not health necessary medical procedures, or for using increased suicide prevalence as an argument for preemptively giving in to demands.. In my opinion it's the same as a four year old saying "if you don't give me candy I'm gonna hold my breath until I die!" and should discouraged, not rewarded.
If people are actually going to kill themselves if not provided with medically unnecessary medical care they should just be institutionalized on suicide watch until they calm down, for their own safety.
To clarify, if people are suicidal due to a rational self-reflection of their own chronic health issues and prospective quality of life, it's a free country and I respect their decision.
I object to allowing people to self-diagnose as suicidal and self-prescribe whatever they want people to give them, while holding their own lives hostage as justification.
>That’s a very odd statement on its own, as people who take their own life for any reason obviously have mental health issues - which need treatment.
I would give a snarky [citation needed], but instead I'll just note that your "obviously" is nothing of the sort. Many cultures, including basically all Western cultures today, support suicide in cases where no mental illness is present.
Euthanasia is one such case in western cultures. We don't assume that people requesting euthanasia are mentally ill. In other cultures, honor and grief have both been viewed as valid reasons for suicide, without any presumption of mental illness. You're attaching an "obviously" to something that is clearly not obvious at all, which would be viewed by many cultures as insultingly wrong, and which is probably a minority opinion in western cultures today.
Okay, let me jump in with "you are fucking stupid". We're not talking about "mild discomfort", we're talking about *pain*. Pain bad enough that people would be driven to kill themselves without relief.
It's not "take two aspirin and suck it up, buttercup" levels. I had to listen to my mother crying and moaning in pain when the morphine patches for her lung cancer weren't enough. So to Dr. Pain and to you, all I can say is I hope you bastards get to experience those levels, one day, and see if your opinion changes.
There's a midpoint between "overprescription of painkillers" and "people can't handle mild or even severe but temporary levels of pain" and "no pain relief at all, if you can't pay high prices for relief then just lie there and suffer".
You're trying to cover this under "medically unnecessary" but there are all too many doctors who ignore pain on the grounds that "nobody ever died of pain" and have no idea of the actual levels of pain involved, to the point of being dangerous to patients whose pain is ignored and the underlying condition not treated.
They can just deal with it. My understanding is that overperscription and overemphasis of painkillers effectively caused the opioid epidemic, I'd rather avoid adding to it.
This is a somewhat extreme position which thankfully is not shared by many... how would you feel if you or a family member had metastatic cancer to bones, which is excruciatingly painful? "No one ever died of pain" is not much help
Overprescription has little to do with the previous heroin or the current fentanyl epidemics. When the price of Heroin dropped at the beginning of this century, it became a cheaper high, and more people became addicted because usage went up.
Fentanyl is cheaper than heroin (because it's easy to manufacture) so it's now become the most popular illegal opioid. So the illegal opioid epidemic has little to do with the overprescription of regulated painkillers.
In the case of Oxycodone, it's a little more nuanced than the simple overprescription as being the problem. Evidence came out in discovery in the lawsuit against Purdue Pharma, that the Sackler family made an effort to get higher dosages prescribed as the standard dosages for milder pain. Thus addicting people who otherwise wouldn't have been addicted. From the Wikipedia article on Oxycodone...
> In 2019, The New York Times ran a piece confirming that Richard Sackler, the son of Raymond Sackler, told company officials in 2008 to "measure our performance by Rx's by strength, giving higher measures to higher strengths".[124] This was verified with documents tied to a lawsuit – which was filed by the Massachusetts attorney general, Maura Healey – claiming that Purdue Pharma and members of the Sackler family knew that high doses of OxyContin over long periods would increase the risk of serious side effects, including addiction.[125] Despite Purdue Pharma's proposal for a US$12 billion settlement of the lawsuit, the attorneys general of 23 states, including Massachusetts, rejected the settlement offer in September 2019.[126]
And this is where Hanson is at least a little right that health care is about showing that we care. If I would pay for an anti-pain treatment for myself or my kid, I would find it outrageous that insurance not pay for it. But we are agreeing that it really is about values, not life expectancy at this point! (Notwithstanding the point made elsewhere in the comments that pain increases risk of suicide)
I feel like this whole debate lacks a dimension because it seems to be assuming health-care is somehow only about keeping people alive. Surely health care is also about increasing life quality. Even from a society level cost-benefit perspective, a government should probably pay for treatment that keeps people healthy enough to do their job, rather than living as invalids!
A commenter on one of the previous posts said something like, "If [some intervention / health insurance / medical spending] is only improving my life expectancy by 3 years, I could take it or leave it." (That's a gross paraphrase, sorry.) But what I thought is, if some intervention only makes you live 3 extra years, you might not want it, but on the other hand, if it has a 20% chance of making you live 15 extra years, does that change your calculations? I know that's the same expected value, but it doesn't feel the same to me, and it's my LIFE after all.
Three years seems like kind of a big deal to me, tbh. Not if it's three years of lying in a nursing home unaware of my surroundings or my own name, but yes if it's three years I can interact with people I love and maybe do some things I enjoy.
> And I remember that during residency I worked with a doctor whose answer to all painful-but-not-otherwise-dangerous conditions was “nobody ever died of pain”.
This doctor is straight up wrong; at a minimum, chronic pain is known to be a risk factor for suicide.
If I had a probably-terminal cancer, my preferred treatment would be to be submerged in an ice-bath and have some giant needles thrust into my arteries and veins, and have all my blood flushed out, replaced by a mixture of saline, anti-coagulant, and para-formaldehyde. This is how I clean the blood out of rodent brains for preservation. That's how I'd want to preserve my brain. Works best if you do it with the heart still beating. I'd wash down some caffeine pills, asprin, and NSAIDs with hard alcohol first, to help keep my heart rate up and thin my blood a bit further and make sure I didn't mind the discomfort of the procedure. Then my head could be removed and properly preserved. Once uploaded, I'm sure digital-me will be grateful for the improved quality of the brain scan which results from such a cleanly preserved brain.
Well, the procedure would be lethal. I wouldn't want someone else to be charged with murder. So unless the procedure becomes legalized in time for my death, I'll have to perform it on myself.
Kenny is right, this is generally step 1 of cryonics (although they typically wait until after your heart has stopped, which makes the process suboptimal).
But there's an important reason I don't simply say, "I'd opt in for euthanasia and cryonics, if terminally ill."
The reason is: I don't want my brain stored at sub-zero temperatures for a long time. I want it to be preserved via CLARITY (brain clearing) so that it will be stable at room temperature, more mechanically cohesive, and able to be optically imaged in thick slices. Then I want my brain to be imaged. Not 100 years from now. Like, right after I die. And then I want someone to try to turn my digitized neurons into a Whole Brain Emulation. Why wait, we're nearly ready! If I live another 10 years, we'll surely be ready for that. I hope to live another 40 years or so, so there should be plenty of time. No more need for cryogenics, just straight to uploading.
Do you expect your consciousness to be preserved? I don't see how that could do so. It would create an imperfect copy that has some traits in common with you. But it wouldn't be you. From your perspective it would be the same as dying
Consciousness is lost every time you go under general anesthesia. So by that definition, I've died at least twice already. I'm not particularly worried about that aspect. But yes, since it would kill me, I wouldn't do it unless I was already close to death. Thus, the discussion of incurable cancer.
It would be more like... having a digital child, but one that was more similar to you than a child would normally be (and had some of your memories). So a digital twin. Not exactly me, not fully different. Something I would expect to have enough in common with me that I would feel pleased about it existing in the world, as I would feel if I had the option to wish a twin into existence or raise children. But being digital would mean it would have super powers. Be able to be immortal, and clone itself, and travel at the speed of light through fiberoptic networks, and many other advantages. Maybe someday travel the galaxy by leaping from laser-relay to laser-relay.
The fidelity of an early scan can't be expected to be perfect. But the scanning technology is non-destructive of the preserved brain tissue, so later on the scan can be repeated with improved scanning tech for higher fidelity.
> All the government has to do is offer to pay for free STI treatment, and then if you have sex too much it’s an “externality” and you’re “robbing the government” so the government should be allowed to step in and stop you.
Unironically, yeah? Like this just seems straightforwardly true? If tax dollars are taken to pay for STI treatment (and don't strictly limit under what circumstances it's provided), then either the government has the authority to regulate activity which results in STIs, or you will be in the situation where people who behaved more carefully are being forced to subsidized other peoples' irresponsible behavior. (If this sounds similar to the situation with student loans, that's not a coincidence).
If you want to make some sort of argument that this is net positive utility or whatever then you can do that, but you have to actually show that; it's not a reductio ad absurdum.
This brings up a kind of meta question about what kinds of powers a government should have, categorically. You’re saying there is a solid logic to the “pay for STI treatment —> interfere in sex life” position. But this kind of thing will be applied (1) incorrectly and (2) selectively.
Incorrectly: we get some behavior/outcome thing wrong and end up oppressing people without even getting a monetary benefit.
Selectively: different political factions will be motivated to intensely focus on some behaviors in this way (being slutty) and not others (owning a gun). (I am a liberal so stuff conservatives might be wrong about is more available to me, sorry; I acknowledge there will be examples in the other direction.)
And what if it is just not possible to reliably identify “effects of behavior you can control” vs “treatment you need for reasons everyone would agree you are blameless for”? Or if it is so cumbersome that x% of our health care budget goes to blame-shifting rather than care?
Based on these objections, I don’t think we should go down this road very much as a way of saving money on health care. (I also just think it’s wrong, but will not try to directly convince you that my feelings are correct.)
> But this kind of thing will be applied (1) incorrectly and (2) selectively...
Of course it will. The government will do this for almost everything it implements.
> Based on these objections, I don’t think we should go down this road very much as a way of saving money on health care.
I don't know why the obvious solution of "don't have the government do so much" isn't the first idea to come to mind. You think that the issues you just pointed out aren't going to apply to government run programs, even if they aren't explicitly trying to deal with moral hazard?
> I don't know why the obvious solution of "don't have the government do so much" isn't the first idea to come to mind. You think that the issues you just pointed out aren't going to apply to government run programs, even if they aren't explicitly trying to deal with moral hazard?
Perhaps you are right and these objections aren’t good enough, as they prove too much (as you said, much of what the government does could be applied I’m misguided ways), and I’m mostly just saying I think it’s wrong.
But the reason I think it’s wrong is that it’s *not* like other government actions in some important ways. There is just a limit to how much the government should interfere in private lives. Most people agree that there is a good case for interference when a person’s actions harm others (breaking traffic laws, spying for an enemy country), but not when it doesn’t. It just doesn’t seem right to have society ban something as fundamental as romantic relationships because our social insurance system might incur costs. Some people are just really bothered by the idea that they have to partially bear the costs of actions by others, while others are more bothered by the idea that someone who needs help is being allowed to suffer. Most systems that help people have some errors in both directions. Which ever one bothers you more determines your politics to a large degree.
> There is just a limit to how much the government should interfere in private lives.
I agree, but I would consider "take a bunch of everyone's money" and "pay for people's <medical treatments or whatever else>" to be interference. The question, "if I'm being forced to pay for something, shouldn't I have some say over that thing and be allowed to take measures to reduce that spending?" is, in my mind, intended to point to the conclusion I described above, rather than to complete totalitarianism, although I suppose some people might conclude otherwise.
> Some people are just really bothered by the idea that they have to partially bear the costs of actions by others, while others are more bothered by the idea that someone who needs help is being allowed to suffer.
I think most people are on board with reducing suffering. But where does the idea come from that just because someone is suffering, we can't ask what the cause was, or put a limit on what relief is actually offered? Some people are probably suffering because of high taxes; can I just say that I need a lower tax rate and it's a violation of my privacy for anyone to ask why?
Also, this is not just a question of "does it seem right to force someone to pay for someone else's costs"? I think subsidizing demand has not only failed to make things like medicine and college affordable, it's actively made the problems worse, including for many of the people it's supposed to try to help. Biden's recently announced housing subsidy will do the same; without being able to increase supply, the entire subsidy is likely to be absorbed by existing owners. And in general, many social welfare programs seem to be poorly run and overall inefficient. So there's both philosophical and practical considerations to keep in mind.
Can the respective positions of Hanson and Scott be boiled down into a prediction they can make a bet on, or on the minimal concrete policy proposal which one would say 'yes' to and the other 'no'? I feel like the conversation as it's currently proceeding is an interminable debate about what someone does or doesn't mean when speaking fairly broadly and imprecisely.
When reviewing "The Origins of Woke", Scott notes how difficult it is to fairly review a book with deliberate omissions/logical gaps for the sake of making a better persuasive case. Back-and-forths like this show that Scott should be doing it for EVERY politics-adjacent argument he encounters, not just those where the author is honest enough to admit it!
Just in the last month, the Hanson Health Care posts, Hanania Woke posts, Coffeepocalypse post and (if you stretch) Emotional Support Animal post all had this problem. First, Scott interprets a rule or argument way too literally, instead of the vibes-based "let's have less of this sort of thing" statement that was clearly intended. Then, he wastes an entire post providing evidence in favor of the thing, explanation of the incentives that lead to the thing, or simply wringing his hands in confusion. Don't bother offering evidence when it wasn't asked for!
“nobody ever died of pain”, said Dr. Mengele while vivisecting twin children.
I mean, how is this doctor practicing medicine? It's not even factually correct, excruciating pain can cause cardiac arrest, people do die of extreme pain.
On a different note, kudos to MrP for engaging in a very civil and productive debate on this.
Statistical power does not only depend on the sample size. If I have 1 million samples and I'm trying to estimate whether an event happens 1-in-a-million or 2-in-a-million, my power is very low.
If I have N samples and I'm estimating an event with true probability p, then assuming p is small, my standard deviation on the estimate of p will be around sqrt(p/N), so my relative error will be ~sqrt(1/pN). For example, if I have N=1 million samples but mortality happens with p=0.1% probability, my relative error on the estimate of this probability will be sqrt(0.001)=3%. If I want 2-sigma confidence (to get a 95% confidence interval), I have to multiply this by 2. This means that I can only estimate mortality rate up to a 6% relative error.
So if an intervention decreases mortality rate by 5%, I don't even have the power to detect this with a sample size of 1,000,000 people, assuming people in my sample only die at a rate of 0.1%. Cremieux's intuition is dead wrong.
And this is particularly true for this study, where (1) takeup was low (they say the letters increased insurance by 1.3 percentage points, so corresponding to roughly 26,000 additional insured individuals) (2) the main specification is an IV specification, which tend to have much larger standard errors than ordinary OLS comparisons.
The only thing id add here is that we keep on jumping between population and individual effects and duties and responsibilities. Most medicines don't work most of the time for most people, except insulin for diabetes and some antibiotiscs for some infections. However, getting more people more access to more medications has greatly increased statistical life expectancy, even accounting for economic growth, less pollution, safer work etc. However, no one person has a duty to take a medication that will most likely not increase their survival just because a doctor has been told that giving out more drugs to more people will increase the survival of the average of the population. And a doctor's duty is to the patient in front of them. Howeverrrr, because of the great successes of broad guidelines and standardisation of care, many people (not just clinicians) have started to think of all health through the broad based indiscriminate lens. This is statistically naive, and often evidentially poorly founded. And..... It's annoying. It can be paternalistic and feel dismissive. But it's hard coded into our guidelines and insurance systems. The way I think about it is - screening. There are very strict criteria for setting up a screening programme and the evidence is reviewed regularly. However, nowadays any tom dick or harry can do a metaanalysis of a bunch of RCTs and say something like "this magic statin reduces risk of cv death by 5%" and it'll be all over the news and in the guidelines without the public health or the staticians getting involved and death rates never budge and we'll wonder why
Hol' up. Not a single reference in these comments regarding the Emotional Support Animal Racket post which maybe, just maaaaaaaybe, provides an obvious example of ineffective medical spending?
Gorillas really can play all the b-ball they want unobserved.
The Emotional Support Animal *Racket*, involves very little medical spending. The animals in question are ones whose owners/supportees already purchased and will maintain as pets, whether they're called ESAs or not; the only involvment of the medical community is a doctor signing a form letter, which probably isn't going to involve more than fifteen minutes of their billable time.
Non-racket Emotional Support Animals might involve a more substantial expense, for training and/or for purchasing and maintaining and animal that you otherwise wouldn't. This may be a waste, but that isn't established by the existence of a parallel racket.
This is incorrect. The medical spending in question isn't the money spent on the animals or the effort on the psychiatrist's part, it's the money spent on the psychiatrist's time. This is generally billed per session, not per minute of effort. I have no clue how much exactly this nets out to - certainly some people are already seeing a psychiatrist and include their request during an appointment that would have otherwise been purchased*, but these need to be balanced by those willing to purchase multiple sessions as mentioned in Scott's post.
* But if the 15 minutes aren't being spent on anything more valuable - or, more precisely, that there is no more valuable substitute on offer - then how much of the non-ESA therapy time is of zero or negative value?
Seems like the maternal death rate discrepancy could be easily compensated by subtracting the base mortality rate in that age range. What's the death rate of X year old women who were pregnant in the past year? What's the death rate of X year old women who were not pregnant in the past year? What's the difference? There you go. I suppose you'd also have to adjust for other demographic con-founders, but even doing a sloppy job of that is going to be better than not doing this math at all.
> Okay, I was pretty on board with maybe I was just strawmanning Robin and he thinks most medicine works and it’s just that we overspend at the margin — but the podcast is called “Most Drugs Are Bad For You”! Someone who listens to podcasts - is this just mistitled?
To answer the narrow question: there were two guests on that podcast and "most drugs are bad for you" is a direct quote from the other guest.
That being said, I don't think "he thinks most medicine works and it’s just that we overspend at the margin" is an accurate summary of his position. He wants to cut spending on medicine in half. That doesn't imply that it "mostly works".
It seems like you think cutting back by say 10% and cutting back by 50% are basically the same position?
I think Hanson's position is that we don't have the data to know whether 10% or 50% of medical spending is useless or even harmful, but he suspects it's closer to 50.
(If you like, you can phrase the issue as Scott: The US isn't getting much for its extra medical spending relative to Britain, and Robin: Britain isn't getting much for its extra medical spending relative to Cuba.)
I recommend this book: Medical Nihilism - by professor Jacob Stegenga -
"A highly controversial argument for a sceptical view about the effectiveness of modern medicine
Draws on interdisciplinary research and grounds the arguments in medical examples
Accessible to readers from any academic background" -
"This book argues that if we consider the ubiquity of small effect sizes in medicine, the extent of misleading evidence in medical research, the thin theoretical basis of many interventions, and the malleability of empirical methods, and if we employ our best inductive framework, then our confidence in medical interventions ought to be low." - Published by Oxford University Press.
Regarding the issue about p-values, I think it is better thought of as a claim relating the size of the effect to the p-values observed. In general the argument that, if this were a true effect and not totally trivial in size then it would have a much lower p-value is totally valid (I'm not completely convinced that 1.4 million is beyond the range we should expect to pick up true but very tiny effects [1]).
However, it's misleading to think of 1.4 million as being the study size. In reality it's the number of respondents who adopted healthcare because of the intervention but wouldn't have otherwise who represent the effective population size. So yes, it's a very good argument that the practical impact of sending nagging letters on health is vanishingly small but I don't believe that translates to showing that the effect of actually getting healthcare is trivial. And I think Goldin's claim that if you select the study sized based on power considerations based on prior estimates is a good response.
However, the bit about reshuffling the groups doesn't seem to be a particularly strong argument. I mean, imagine you got the results by p-hacking then you'd still expect the same thing to be true because you choose the outcome examined based on exactly the fact that it was a statistical outlier. But here they seem to be reporting the obvious variables of interest not some odd cross tab.
--
1: Yes, given a finite number of variables you need to have some fall off in number of true correlations with the true effect size below some value simply because every true correlation must occur at some finite value. However, we don't really have a finite number of variables -- or at least have a very large number -- once we start studying composed effects (eg effect on health outcomes of a nagging letter).
Also, regarding the social benefit of medicine, I think this is a particularly difficult issue because what is of benefit to society may diverge quite sharply from what is of benefit to the individual.
After all, given a finite lifespan it's not at all clear extending that lifespan is of any benefit to overall social utility. Everyone is going to have to mourn their parents and friends at some point (some benefit to less mourning per life year) and if the choice is between giving resources to the elderly or replacing them with young healthy people it seems like net utility may favor the latter.
This leads to the odd result that the mere existence of medical treatments might be net harmful because people want to feel cared for and not coldly abandoned to death and if you've done all you can they will feel that way.
I also think it's worth defending the psychological benefit of feeling you are being cared for and treated. After all, look at how much people are willing to pay for homeopaths or Reiki healers who mostly just offer a personal interaction and placebo.
I think there is a strong argument for decoupling the expert work of diagnosis and care from the human relationship work. Maybe a new kind of nurse practitioner whose job is mostly just listening and interacting with the patient and just futzing with relatively easy shit to make them feel listened to and like someone is concerned about them and to handle the check-in requirements for dispensing controlled or risky substance. Then maybe AI analysis of the recorded interaction to flag warning signs.
Regarding negative externalities: one cost-negative approach to negative externalities would be for the government to stop subsidizing things that are bad for us, e.g. grains and tobacco.
If 95 out of 100 people die from a risky surgery that looks, in isolation, bad; if 100 of them would die without the surgery it looks less bad. In either case, though, probably not a surgery that should be covered with insurance.
I like to joke that Lindley's paradox is paradoxical, because the frequentist solution is so egregiously wrong that it loops back around into being more Bayesian and more correct than Lindley's idea of a "Bayesian hypothesis test".
First, let's back up, and ask what the *actually correct* solution to Lindley's original problem of birth rates. Lindley tested a dataset of tens of thousands of births, found about 52% of them were male, he found this gave an exceedingly large Bayes factor in support of the idea that exactly 50% of human births are male. Do human births have a sex ratio of exactly 50%? No! We have overwhelming evidence that sex ratio at birth is slightly skewed towards men.
So why did Lindley find a large probability of a balanced sex ratio? It's because his prior was @#$%!ing stupid. His prior assigned an equal probability to two hypotheses:
1. Exactly 50% of babies are male.
2. Any sex ratio between 0% and 100% is equally likely.
So, his probability density is 0.5 everywhere between 0 and .5, and everywhere from 0.5 to 1. But at *exactly* 0.5, the probability density is @#$%ing *infinity*. If you decide your prior density for *anything* is infinity, your prior is bad and you should feel bad. You will get ridiculous answers like this.
Natura non facit saltus. The idea that exactly, precisely, 50% of births are male is ridiculous, because 50% is a *single number*, and there are infinitely many real numbers. The prior should assign a large probability that the sex ratio is *close* to 50%, but not that it will be *exactly* equal to 50% and not 50.00000000001%.
This is not how any sensible Bayesian statistician would approach this question. Bayesian hypothesis testing approaches like Lindley's are what happens when you have a frequentist LARPing Bayesian statistics, and trying to construct a hypothesis-testing framework that sounds like what they're familiar with.
But why does the chi-square test give a sensible result? Well, it turns out all the common frequentist procedures (T-tests, chi-square tests, etc.) have a dual, Bayesian interpretation for large samples: You can read the p-value as the probability of a *positive* (rather than negative) effect size. In this case, Lindley's p-value of 1.2% is the correct, Bayesian answer to the question "What's the probability that the sex ratio at birth is actually skewed towards girls?"
But this is a complete disaster of an answer if you interpret it like most people do, as an answer to the question "What's the probability of a perfectly balanced sex ratio?" (0%! The answer is 0%! Nothing is perfectly balanced like that!)
Going back to the original topic: the correct interpretation of Goldin's results is that he's conclusively proven that healthcare will probably not kill you. On the one hand, that's reassuring. On the other hand, it doesn't really answer the question we actually care about: "is healthcare spending cost-effective"?
>Suppose you were to take the individuals in our treatment and control groups and randomly re-shuffle them into (fake) treatment and control groups, and compare the difference in the mortality rates between the fake groups. You wouldn't expect to find an effect, but there might some differences just due to random noise. In Appendix Figure A.VII (below) we do this 1000 times, and compare the difference between the real treatment and control groups (our estimated effect from the study) to the distribution of the differences between these fake-groups. This tells us whether the difference between the treatment and control groups that we observe in the study (shown by the red line) is likely due to chance -- the figure below suggests that the answer is no, because it is more extreme than almost all of the fake comparisons.
I don't think this argument works. When I simulate a biased or p-hacked result, the estimated effect is also in the tail of the placebo distribution. So as a test to distinguish between real and spurious results, it doesn't work. See:
Actually, I misunderstood. Goldin is just restating the definition of a p-value (defined from a permutation test instead of the classical way). The permutation distribution is calculated under the null hypothesis (when the treatment has no effect, we can reshuffle the treatment labels and not affect the outcome), so the number of estimates larger than the actual estimate is the p-value (the probability of getting a more extreme result assuming H0).
something i’d add is that Hanson refers to medical error as being “plausibly our third largest cause of death.” as i understand it that’s pretty exaggerated/ not true (see: https://www.mcgill.ca/oss/article/critical-thinking-health/medical-error-not-third-leading-cause-death), and as such it’s pretty irresponsible to suggest that. i don’t think it “signals” anything particularly good about hanson as an interlocutor on this topic.
I mentioned that elsewhere, but by the time it made it into my post I think someone told him and he'd edited it out of his.
There was an ultimate source, in the BMJ, but the source article turned out to have faulty reasoning.
Even then, "medical error kills people" is not at all the same as "an awful lot of modern medicine is just snake oil".
For instance, there's an egregious case covered recently in Ireland. "Died because she was not given antibiotics in time" is not the same as "Pfft, antibiotics are way over-prescribed and the best thing we could do is cut down on the unnecessary spending on them".
https://www.breakingnews.ie/ireland/aoife-johnston-inquest-coroner-records-verdict-of-medical-misadventure-in-uhl-1617852.html
yeah that’s my take too
The article that bjw shared also names “treatment delay” as potentially lethal medical error.
Certainly, this would increase as medical spending decreases.
It is therefore possible that deaths from medical error go down as medical spending goes up.
So regardless what the proportion of deaths “by medical error” really is, it’s not even clear in what direction that should affect medical spending.
Johns Hopkins study, I think... https://hub.jhu.edu/2016/05/03/medical-errors-third-leading-cause-of-death/
Ah ok fair enough, thanks for following up
I’m glad this ended up somewhere more interesting than “maybe our entire health care system is useless.”
IMO (I may be biased) the strongest indictment of our current system is comments like Jake Seliger above. Federal bureaucracy strangles attempts to make progress on life threatening conditions.
The justification for this is safety, but you have to consider how much a person with a condition with 100% mortality is worried about side effects of a potentially life saving drug.
This seems interesting - https://en.m.wikipedia.org/wiki/Right-to-try_law
(I don’t live in the US but a lot of the stuff I work on in Australia is downstream of US pharma companies getting FDA approval)
"but a lot of the stuff I work on in Australia is downstream of US pharma companies getting FDA approval"
From what I've read (apologies, I don't have a source for this off the top of my head) the U.S. is by far the biggest and most lucrative healthcare market in the world, so anyone developing a drug essentially has to be able to sell in the U.S. to make any money. Consequently, the FDA is the rate-limiting factor for much in pharma.
Isn't the EU just as big as the US? But I guess nationalized medical insurance means the payouts are smaller?
Couldn't the latter problem be solved with a FDA rule that "it's acceptable for a doctor to prescribe anything which is approved by the EU"?
es regarding size, but almost all European countries’ public systems impose severe restrictions on what they’ll pay.
The second idea is called reciprocity and it is also a good one. Japan, for example, has approved a Covid drug far more effective than Paxlovid: https://cbuck.substack.com/p/life-liberty-and-the-pursuit-of-health and it should be available in the US.
The problem with reciprocity is that the defining and formative experience of the modern FDA is the Thalodomide incident. Where European regulators broadly approved the wonderful new morning-sickness drug, and the stubborn FDA said "no, you haven't *proven* that it is safe by our standards; American women will just have to put up with morning sickness". Boo FDA!
And then ten thousand European mothers gave birth to children with flippers where their arms and legs should be(*). But not American mothers. Yay FDA!
Do you want another ten thousand flipper-babies? Because reciprocity is how you get flipper-babies. Or at least, that's what all the FDA's PR is going to be focusing on, along with the bit where you're an uncaring monster who wants flipper-babies so long as the mad scientists can play with their potions and Big Pharma can rake in the profits.
They've been playing that game for eighty years, and they're quite good at it. Big Pharma has figured out how to make boringly reliable profits playing the FDA's game, and nobody else is in a position to play at that level.
* Do not google this if you're at all squeamish
"Robin’s argument is strongest against prevention, least strong against treatment."
Agreed. I get that it feels intuitive to say that an ounce of prevention is better than a pound of cure, but for that to work, you need to know:
1. That you're likely to get the condition you want to prevent,
2. That the action will prevent or minimize the condition, and
3. That the preventative action won't cause iatrogenic harms as bad as or worse than the thing you're looking to prevent.
When you look at population-level interventions, they're often poorly targeted and benefit small populations. So you can end up distributing 10,000 pounds of prevention for every ounce of cure.
This isn't always true, of course, but who has an incentive to reverse course? Not pharma companies who have strong incentives to broaden their target populations. Patients don't have the expertise to look at large populations. Eventually research hospitals get around to vetting old treatments, but there's not as much glory as in looking for new ones.
I think it's also worth keeping in mind that there are plenty of efficacious preventative measures people can take against major health problems, the most effective ones just generally fall under the category of "lifestyle interventions" which people don't think of as "medicine." But it's because of medical research that we actually know that they're effective.
I think there are plenty of medical interventions that are obviously medical interventions (most vaccines) that qualify as sufficiently efficacious to merit broad application. It's just really difficult to parse which ones work and which don't without good evidence that takes 'too long' to implement, because people are dying today and 'we need something!'
I used to be much more convinced by your characterization of 'lifestyle interventions' as something that "it's because of medical research that we actually know they're effective." It seems to me that people used to be a lot more healthy and have much better nutrition before a massive glut of health and nutrition advice.
I guess it's possible that all the advice was spurred on by people behaving worse and worse, but if that's the case I'm not sure what the argument in favor of the advice is. Either it's iatrogenic harm or it's ineffective. I don't believe that the counterfactual of a world without the last 50 years of nutrition advice would result in people being more obese/unfit than today.
It's not like only nutrition advice rooted in the last 50 years is rooted in medical research. There's plenty we've learned before the 1970s that we still didn't know for most of human history. In the time of Galen, would people have known anything about the link between diet or exercise and diabetes?
That's kind of my point, though. A ton of people followed the advice to switch to low-fat foods, eliminate cholesterol, and to transition their lipid consumption away from animal fats to polyunsaturated vegetable oils. They mostly exchanged pork and cow for chicken. That doesn't seem to have had any positive health effects at a population level.
I knew a guy from my grandpa's generation who grew up on a farm. He used to talk about the meat they ate, preserved in animal fat before they had reliable refrigerators. (Space in the ice box was limited, I guess?) "I'd get home from school, dig out a slab of pork from the jar of lard, slap it between two slices of bread and slather some extra lard onto the sandwich. Looking back, I guess it was terrible for me, but I loved it."
To be fair, in some ways, our nutritional environment has probably gotten a lot worse than it used to be.
One thing I've speculated on occasionally is how much harm it might be doing to human health that we've dropped down to roughly 2000 calorie/day diets. Historically, most people have probably eaten significantly more than that, but also been active enough to have much higher maintenance levels. Getting all your micronutrients on 2000 calories a day is a very different proposition from getting them all on 4000 calories a day and burning off the excess. It's not like a slab of pork marinated in lard is actually bad for you, it's just that most people in the modern day aren't active enough to healthily maintain high calorie loads.
Maybe. I've heard a lot of bioplausibility arguments for why people are unhealthy and why intervention X - no Y - no Z! (etc.) will 'fix' this problem. Then people I know implement those solutions and don't get healthier, and large RCTs demonstrate short-term improvements, but then fail at long-term efficacy. I'm no longer as convinced by bioplausibility as I once was.
It's certainly true that US nutrition and activity have changed over the past century. This seems to have caused dramatically adverse health outcomes for a huge percentage of the country. I'll grant it's possible nutrition advice had no effect on that trend. Though my current priors point me more in the direction of advice being on net bad for health outcomes.
For anyone wanting to defend the proposition that nutrition advice has actually promoted better health, I'd question the counterfactual: would health outcomes be WORSE in the US in the absence of nutrition advice? What would that even look like, compared to the massive expansion of metabolic syndrome and nutrition-related disease?
I can see benefit in certain narrow cases, such as removing lead from children's toys. But stuff like the food guide pyramid or the new My Plate guide, or the thousands of admonitions about what to/not to eat seem like too much engineering biology that's doomed to fail in potentially deleterious ways.
Consider the thousands of dieting fads out there. "Diet and exercise" has been the perennial advice I grew up with for how to lose weight. Everyone KNEW this was the silver bullet, and that the only reason for people not getting there was because they were lazy, poorly disciplined, unwilling, gluttonous, etc. Yet RCT after RCT after meta-analysis has demonstrated that short-term weight losses are overshadowed by long-term net weight gain for dieters, and that this weight gain is in excess of what could have been expected from a non-dieter. So choose your diet, HFLC, keto, low fat, low calorie, IF, etc. So long as you choose any kind of dieting, you're choosing to gain weight over time. Yet this is the most common nutrition advice for people who are any kind of overweight, despite it being clearly ineffective for most people who in practice gain massive amounts of weight from serial dieting (including close friends and relatives of mine).
Why? Because it has been the most aggressively promoted nutrition advice, not because it has solid evidence to support its real-world efficacy writ large.
"In the time of Galen, would people have known anything about the link between diet or exercise and diabetes?"
Seemingly not, their view of diabetes was that it was a kidney disease because of the excessive urination:
https://link.springer.com/article/10.1007/s00125-008-0981-4
"The first detailed information concerning the aetiology, symptoms and treatment of diabetes derive from Rufus of Ephesus (98–117 AD), Aretaeus (120–200 AD) and Galen (128–200 AD). Caelius Aurelianus, a fifth century physician who translated Greek medical texts into Latin to transmit Greek medical views to Rome, mentions that Demetrius of Apamea, a Syrian–Greek colony in Phrygia, was the first to use the term diabetes, in the second century BC, but the writings of Demetrius have unfortunately been lost.
Rufus of Ephesus was a physician famous for his work on the variations of the pulse. He describes the symptoms of diabetes as ‘incessant thirst’ and immediate urination after drinking, which he calls ‘urinary diarrhoea’. He advocates induced vomiting after drinking cold water and mixtures consisting of several substances, such as ‘cyceon’ and ‘ptisane’ and cooked vegetables. He states that substances that promote urination should be avoided. Steam baths are also advised to cool the head whilst heating the body. The treatment is similar to that advised for fever, detailed below:
‘We then, apply to the hypochondrium a poultice made of dust from alphitonia mixed with vinegar and oil from roses or with grinded tender leaves of vine or with cotyledon, with pellitory (parietaria Judaica), with purslane (pulli pes) or with any other similar substance. Before lunch, we frequently administer juice of bistort to drink, of helenion mixed in dark wine, or a maceration of dates, of the fruit of mirth or of pears. At the beginning, we also make a venesection to the elbow. Sometimes we have to use stupefying medicaments.’
Aretaeus describes diabetes as a dreadful, chronic disease that develops over a long period of time and is not common. Furthermore, according to his views, the aetiology of diabetes is based on humidity and coldness in the organism and on chronic diseases of the small intestine and the kidneys with ‘hidden’ malignancy. Its symptoms are described as follows:
‘The flow is incessant... the patient is short-lived... for the melting of the flesh is rapid, the death speedy. Life is disgusting and painful... Thirst, unquenchable; excessive drinking, which, however, is disproportionate to the large quantity of urine, for more urine is passed; one cannot stop them either from drinking or making water. Or if for a time they abstain from drinking, their mouth becomes parched and their body dry; the viscera seems as if scorched up; they are affected with nausea, restlessness and a burning thirst; and at no distant term they expire. Thirst, as if scorched up with fire... The abdomen shriveled, veins protuberant, general emaciation, when the quantity of urine and the thirst have already increased... the disease appears to me to have the name diabetes as if from the Greek word (which signifies a siphon), because the fluid does not remain in the body...’
Additionally, patients may suffer from decay or sudden death, or have a disgusting and painful life .
The treatment for diabetes is similar to that for dropsy. Remedies are proposed to ‘treat’ the thirst originating from the stomach, such as purging the body with ‘sacred’ medicine (a mixture of many medicaments), compresses (made of spikenard, mastic, dates, raw quinces and rose oil), the sprinkling over the body of a mixture consisting of water boiled with fruits, styptic wine, shittah tree and hypocist, plus special food (milk with cereals, starch and wheat).
Galen of Pergamum, the most celebrated Greek physician of the Graeco-Roman era (128–200 AD) refers to diabetes in several of his writings. He believed that diabetes was a rare disease in which the kidneys are affected. Regarding the naming of the disease, the frequency of its appearance and its symptoms, Galen writes:
‘Some people refer to chamber-pot dropsy, others again to diabetes or violent thirst... For my own part I have until now only twice seen a disease in which patients suffered from an inextinguishable thirst, which forced them to drink enormous quantities; the fluid was urinated swiftly with a urine resembling water ... diabetes is a genuine kidney disease analogous to voracious appetite.’
He noted some further symptoms that accompanied diabetes, such as vomiting after drinking, diarrhoea, retention of the content of the stomach, continuous desire for urination or dysuria.
The aetiology of diabetes is attributed to malfunction of the ‘retentive’ faculties, as manifested by the continuous need for fluid and the inability of the kidneys to retain their content. Probably as a result of limited experience with diabetes, Galen does not provide details of treatment. The writer of the pseudo-Galenic treatise De remediis parabilibus does, however, advise the patient to take very thin membranes from the abdomen of young roosters, dry them under the sun and eat them. If this cannot be done, he suggests that the patient should grind together mountain copper, dry acorn, flower of the wild pomegranate and oak gall, add to honey of roses and cold water, and drink the resulting mixture."
I suspect from the sounds of it that this was type 1 diabetes, which in the time of Galen was invariably fatal. Type 2 diabetes, where diet and exercise has an important role, is largely a modern disease. Probably it was present to some extent in ancient times - gout, which has a similar pathogenesis of excess food and wine, certainly was ("the king of disease and the disease of kings.") But its effects - of foot ulcers, heart disease, kidney disease and blindness - would, I imagine, be much more apparent to ancient physicians than a high blood sugar reading. This link was discovered only recently.
Gout was related to wine because there was lead in wine, and is only related to high-protein foods.
Sure, from the perspective of an individual.
Just be aware that lifestyle + most other primary preventive care measures probably do not do anything to influence health care at an aggregate level. Which is worth to keep in mind if you try to measure the effect of health care by looking at crude aggregate statistics on prevalence of diseases over time or across countries (none in this forum would do that, but you sometimes see analyses like that in less-serious newspapers).
Reduced health care costs on a national level only happens if the time period & treatment needed from people get seriously ill till they die gets shorter if people live longer before they get seriously ill (The compressed illness-hypothesis).
What might just as well happen if people increasingly avoid dying from non-communicable diseases in mid-life due to a healthier lifestyle & other successful preventive measures ("60 is the new 50" and all that), is that they are more likely to die with & from delayed degenerative diseases when they get older. (The extended illness-hypothesis.) These diseases are usually much more costly to deal with for insurance companies and/or national health care systems.
You have to die from something. If you and your significant others succeed in combatting the typical ilyfestyle illnesses early and in mid-life, you do not cheat death. You earn yourself an extension. That's all.
It's a question of chronic versus about disease. Solving metabolic syndrome and sedentary lifestyle would alone be expected to dramatically reduce health spending for the many associated conditions downstream. Sure everyone will die, but not everyone will be medicalized for chronic conditions for decades before they go. It's not clear the current medical system is optimized to do anything but promote chronic treatment bloat.
Hmmm...I don't have any data at hand, but I would stick to my general point that the longer people live before they die, the larger will be the percentage who get delayed degenerative diseases before they die. And these diseases are often extremely expensive - more so than treating sedetary lifestyles among people who can at least sit still.
Alzheimer patients for example often need 24/7 attendance by carers because they get up and walk around and do unpredictable things, and even unskilled carers are expensive when you need them all the time.
Then one can hope that we can find a cure for Alzheimer some day - there is promising research going on. Which is great of course, but just kicks the can further down the road, since this means that people will get something else instead before they die a bit further on in life. I am non sure which degenerative diseases that can be equally or even more expensive than Alzheimer - other types of dementia probably - but rest assured that something it will be. Because something it will have to be.
It would be good to have solid data to support either your view or my own, but I'm not sure how to quantify it exactly. All I have is personal experience of people who endure significant morbidity for year/decades, but who are also overweight and unfit, compared to people who are fit and active. I've watched both grow old and die, and the difference is often one of sudden versus chronic deterioration. Therein lies the difference in health resource utilization.
I think if you had a magic pill that cured heart disease, some 113 year old who was going to die of that tomorrow will die of a stroke within a week instead. Or something else once you cure that. This doesn't require them to incur hundreds of thousands of dollars of health system costs, though. It's true that the last year or so of your life is the most expensive, but this is often driven by chronic disease, and not everyone dies of chronic diseases.
Contrast this with a hypothetical 30 year old with non-alcoholic fatty liver disease. If they lose 80 pounds over the next year, the fat deposits in their liver will clear up and they'll avoid cirrhosis, a liver transplant, and maybe hepatocellular carcinoma. They might also avoid diabetes with attendant dialysis, peripheral arterial disease, and a kidney transplant. Instead of dying at 63 after millions of dollars of cumulative health resource utilization over a 3-decade ordeal, they die at age 83 after a 3-year fight with cancer that still cost a few hundred thousand dollars.
What you describe is the compressed morbidity hypothesis, and what I describe is the extended morbidity hypothesis. There is also a third possibility: the parallel-displacement-of-the-length-of-the-morbidity-period hypothesis. Which of them that will turn out to be correct is difficult to discover through research. Ideally, you should follow consequtive waves of birth cohorts till all of them are dead, and check if the average number of months/years needing expensive care over the life cycle goes up, down or stays constant for later birth cohorts.
I know of one study which did that in a local community, starting at age 75 (if I remember correctly, quoting from memory here) and followed two birth cohorts till all were dead. It supported the parallel displacement of the length of the morbidity period hypothesis. But apart from selection effects and other problems in the study, even when we get higher-quality studies (there is ongoing research on this) and find out which hypothesis is right today, we have no guarantee that the findings will stay valid 10-20-30 years from now, when the effects of population ageing on health and nursing care really starts to hit high-income countries. The Departments of Finance in the world certainly hope that you are right!
I am less sanguine, but keep an open (empirical) mind about which hypothesis that will turn out to be correct - since even better research today will not be able to give a sure answer about the future. At the end of the day, we just have to go on living and in due time we'll find out which hypothesis is correct:-)
As an Australian, I feel like this is one of the major strengths of a socialised single payer system. The government pays and also has power to influence prevention.
Some of the greatest triumphs in non-medical interventions that cut our healthcare spending - sunscreen/SunSmart (mostly administered through schools, no hat no play, getting children into the habit of applying sunscreen), free swimming lessons (also mostly administered through schools), and anti-smoking regulations.
People often say that Australia is kind of a nanny-state, and it kind of is, but IMO it's more efficient when the government can notice that they can cut ICU admissions by mandating bike helmets and then do that right away.
(Most recent example is the ban on engineered stone, due to silicosis).
wheres the take "most treatments are worse then useless, some are actually good(antibiotics after getting a bullet removed and you swell up like a balloon) the majority of patients are incapable of navigating the space"?
“Jake seems very dedicated to surviving as long as possible”
For me, living is mostly about keeping my wife, Bess, company. If not for her I doubt I’d have the fortitude or stamina to continue, and I’d have likely taken the opioid road by summer 2023. But she’s going to be very lonely when I die and so I’m trying to delay that moment as long as possible.
I think humans need projects and purpose to thrive, and part of mine has become 1. Explaining how clinical trials actually work—which I wish someone else had already done, but to my knowledge no one has. And also 2. The FDA is too slow, and so I’m hoping to somehow generate change, cause what I’ve gone through sucks, and it could’ve sucked much less if the FDA were keener on cost-benefit analyses.
On the testing question, I don't think it's just a question of inconvenience from the test. There are potential harms that come from excessive testing: https://youtu.be/yr_4RoPhtu4?si=RrA6KIFUfnkQzbit
Now, the existence of those harms doesn't negate the possible benefits of treatment. But it should change the nature of the discussion away from, "well, it's always better to test, because then you know." Because there's a difference between a positive test and real medical knowledge. At what point does testing become beneficial? We know it's somewhere between never test and test everything, but for many testing schema there's poor understanding of where potential benefit begins to outweigh the potential for iatrogenic harm.
Occam’s razor suggests that Robin is employing a Motte-and-Bailey throughout.
Perhaps you could establish that by quoting him.
This started with attention-grabbing sensationalist claims and ended with banality, so I agree.
A skillful deployment, but yes; Scott is being very charitable.
> You could design a maximally cheap health care system from the top down where patients can't choose their physician, can't see a specialist without fulfilling specific guidelines, don't get access to any examinations that aren't evidence based, where there are long waiting periods for everything that isn't urgent, -- and this would save a lot of money probably without statistical detriment to outcomes. Patients would hate it though.
You've just described the UK's health system, the NHS. Historical, this has been extremely popular (as in, the NHS regularly polling as the UK's best gesture). Recently less popular, mostly because of increasing waiting lists to a point where they're currently probably longer than a cost-effectiveness analysis would suggest is optimal.
The Irish equivalent is the HSE, and it's long been mired in controversy over waiting lists, lack of access to services, constantly massively over-running budgets, etc.
It is a good system because it does provide medical care for all, and the private practice consultants often shuffle off patients onto the public system, but the problem is this little bit in the proposal above "long waiting periods for everything that isn't urgent" - there are long waiting lists even for things that *are* urgent, or become urgent while you're waiting to be seen and treated. Having to wait 5+ years for a scan because you have a twinge in your tummy may be a good thing; having to wait 5+ years for a scan while your heart condition quietly gets worse isn't.
We do have private health insurance companies operating, but it's not entirely great. We had new entrants come in (particularly during the Celtic Tiger era) and then leave again, because the Irish market is so small it's just not profitable enough. That leaves about three or four main insurance companies, which can then pretty much charge what they want. It's not as yet universal that all employers offer a company health plan (my place of work offers an employee pension scheme, for instance, but not health coverage while my brother does have it because he works for a pharma plant that is ultimately American multinational-owned) so it's individuals who purchase private health insurance themselves, if they're not in a scheme for work.
So we don't have the same level of 'high tech medicine paid for by your company health insurance' as the USA and the HSE is not at the same level of efficiency as the NHS (the HSE website will often copy word-for-word content from the NHS and just put a little note at the bottom saying they did this, which always makes me laugh). Falling between two stools, as it were!
https://www.hia.ie/consumer-information/faq
"Where can I buy health insurance?
There are two types of private health insurer in Ireland:
Open membership insurers provide insurance to everybody who wants to buy it from them. Currently, there are four open membership insurers operating in Ireland: Irish Life Health, Laya Healthcare, Vhi Healthcare, and HSF Health Plan. However, HSF Health Plan offers only cash plans. Cash plans don’t include any inpatient cover.
Restricted membership insurers provide insurance to people who are members of a particular group. This is normally a vocational group or employees of a particular organisation and their dependants. For example, there are restricted membership schemes for members of An Garda Síochána."
I came here to say exactly the same thing, and confirm that despite all the complaining that is present (mostly about long waits for non acute surgery and many assesments, with long waits at busy city A&E being perhaps more of a serious issue) NHS is generally thought of as a Very Good Thing, and while some people do use private options, many people who COULD afford them in theory don't, and accept the waits.
I also feel that UK primary care doctors deal with a lot of stuff managed by specialists in other countries. For anecdatal example I lived the vast majority of my reproductive age in the UK and while I never had any even moderately serious issues, I used contraception including coil, I have two children and I am menopausal now -- and in those 25+ years I encountered gynecologists on exactly two occasions, both while in labour in hospital. The idea of having "my obgyn" is completely alien to the vast majority of women here.
I think the value of choice is overstated naturally by the fact that people who write opinion pieces are more competent at and interested in making such choices, while the "free at the point of care" (and for large parts of the population, including 'free prescription medication") is a huge feature.
How many British people have had a chance to try US healthcare though? And not just in an ER room or something but the full experience with Blue Shield health insurance in a major city?
Most Americans think that public transit sucks but they also haven’t had a chance to try European public transit properly. Americans also used to be quite happy with their restaurant ecosystem 50 years ago even though it’s atrocious by modern standards.
Most Brits I’ve talked to are woefully ignorant of how the American system actually works and become quite jealous when I tell them of my experience.
I'm sure that much of the British horror at the US system is based on ignorance, or rather on being only exposed via media reports to its failure mode, and not the success mode.
On the other hand, I've heard people from other European countries (both Eastern and Western EU) grumble a lot about NHS and people who have experienced eg French system wax lyrical about that. So yes, it's very contingent.
I was mostly saying that the system OP described as a hypothetical one and one patients would absolutely hate actually functions in a fairly developed country on the other side of the pond and is NOT hated in principle.
I don't use healthcare much but basically had I had to pay anything I'd not use it at all, because I couldn't afford it, so if I was looking at comparisons, I'd need to look at free healthcare for the poor, which (very quick look) I probably wouldn't qualify for. I don't know how that compares to "blue shield full coverage" in quality or comprehensiveness.
As complete anecdata, I did a bit of medical phone interpreting at some point and while it was mostly UK cases, occasionally I'd be connected to a US call. The way the medical personnel, but especially doctors, treated the people I was interpreting for (often elderly, sometimes confused, sometimes after injury or even DV) was beyond belief in brusque, patronizing pseudo efficiency. I also sometimes did calls related to health insurance admin and the whole thing seemed entirely Kafkaesque. These are obviously not proper data, and I'm sure much care is excellent and delivered in a compassionate way.
I've been known to suggest that the NHS needs to be making regular payments to the US, for the valuable service of giving them "But at least we're not like the damn bloody Yanks! Don't you dare say a bad word about the NHS, or you might get *American* health care instead!", every time some Brit decides to say some very deservedly bad words about the NHS.
The European countries that seem to be able to make this work better than the NHS, mostly use hybrid systems with a more substantial role for the private sector. But they mostly don't speak English, so Americans default to just using the NHS (and its Canadian equivalent) as the model case for "Socialized Medicine".
It's also the system in the Nordics.
It always amuses me when Scott says "let me propose as a thought experiment the following absurdly utopian and/or dystopian healthcare system", and the Brits all rock up in the comments to say "you've just reinvented the NHS".
a) It was Kristian who wrote that not, Scott
b) They literally wrote "This is what public health care is like in some countries" right afterwards. They *know* about the NHS and were deliberately describing it.
To a lesser extent, it also describes most private insurance plans in the US if you're not willing to pay out of pocket. You have to pick an in-network physician, need a referral to see a specialist, need every intervention to have prior authorization from the insurance company (based on their judgement of whether it's necessary), and while I hear wait times are better, they're certainly not great (I personally had to wait several months to get an infected mass surgically removed from my sinuses).
It's a question of magnitude, but it's not like the US is winning any awards for quality-of-care-per-dollar, so it's not a public-versus-private issue.
And the NHS spends about half per-capita of what the US spends if you include all public and private spending!
Of course, waiting lists aren't a designed-in feature, they emerge from resource constraints.
Typo pointing out:
> a co-author of al study
Meanwhile, ironically, "type-o" is neither "typo" nor a "typo" for "typo".
Scott's being a quokka here.
Hanson is writing like a Holocaust denier, only he's being a medicine denier instead. Holocaust deniers often don't say "the Holocaust didn't happen", at least not immediately. They're just "asking questions", but the questions are skewed.
So Hanson doesn't say "medicine doesn't ever work". But he says things that amount to "how do we know medicine works? Here's a study that shows it doesn't seem to work. And we know there's a lot of bad replication in the field of medicine." He compares the belief that medicine works to the belief that you can make money in casinos. If questioned on that point, he can reply "I never said outright that medicine doesn't work", but the obvious intent of that is to say that medicine doesn't work.
You're not *misrepresenting* him by summarizing his position as "medicine doesn't work" any more than you're misrepresenting the Holocaust denier by calling him a Holocaust denier for asking "How do we know that the Holocaust really happened? Evidence isn't very good" without saying anything outright.
I think if a suspected Holocaust denier said outright "I admit the Holocaust definitely happened", I would stop thinking of them as a Holocaust denier, no matter what other weird positions they held. I feel like Robin has explicitly said some medicine works. I think his position about marginal units of medicine is relatively common and plausibly correct.
"Some medicine works" isn't like a Holocaust denier saying "the holocaust definitely happened". It's more like a Holocaust denier saying "of course the Nazis killed some Jews". Which they usually do say.
Hanson is making very strong statements whose implication is that he doesn't think medicine works. The reason that you are frustrated with summaries like “Health Care Is About Signaling” and “Most Drugs Are Bad For You” is that even when the summaries aren't created by Hanson himself, people read what he wrote, understand what he's implying, and write the summaries.
You are granting excessive charity to him by noticing that his exact words don't lterally say "medicine doesn't work". It's like the news media lying. It's not literally making statements, but it's implying them.
Have you never encountered the position of "Sure the Holocaust happened, but it killed maybe a couple hundred thousand people"? If someone said the Holocaust happened but killed literally twelve Jews, surely you agree that would make them a denier, so how many marginal Jews does someone have to deny the deaths of before they become a denier of the Holocaust overall?
I think once someone expresses a more or less consensus position on a subject, it's irritating and useless to compare that person to Holocaust deniers, and Robin's position is pretty common among health economists. This is more like saying that 5.8 million Jews died when the consensus range is between 5.5 million and 6.5 million or something.
I think we disagree on what Robin's position is. I don't think it's an accident that he keeps saying things which make people use titles like "Most Drugs Are Bad For You".
I guess I'm going to have to listen to this podcast and make a transcript. Will report back.
> All the government has to do is offer to pay for free STI treatment, and then if you have sex too much it’s an “externality” and you’re “robbing the government” so the government should be allowed to step in and stop you.
I don't recall seeing this in relation to STIs, but quick searches for "obesity cost to NHS" and "alcohol cost to NHS" bring up a *lot* of hand-wringing articles on how much fat people and drinkers cost the NHS in additional treatment. This is in turn used to argue for sin taxes on drink and unhealthy food, restrictions on advertising, etc.
I am all in favor of this .. quantify externalities and tax them in a cost effective way. Corn syrup should be taxed not subsidized
Then let me blow your mind: corn syrup basically isn't a thing here, to the extent that UK-based cooks have to use alternatives when cooking American recipes: https://www.nigella.com/ask/uk-version-of-corn-syrup
I know! Apparently corn syrup goes back to the 19th century (so Wikipedia tells me) and I suppose I can see that if sugar was still a luxury good so this was an acceptable substitute, and it was popularised in the USA due to a big marketing campaign in 1910:
https://en.wikipedia.org/wiki/Corn_syrup
"Corn syrup was available at grocery stores in the 19th century, as a generic product sold from a barrel. In 1902, the Corn Products Refining Company introduced clear, bottled corn syrup under the brand name of Karo Syrup. In 1910, the company launched one of the largest advertising campaigns ever seen. This included full-page advertisements in women's magazines and free cookbooks full of recipes that called for Karo brand corn syrup. In the 1930s, they promoted a new pecan pie recipe that featured corn syrup, followed by a similar, nut-free chess pie recipe, in a bid to drive sales. Later, they promoted it as an alternative to maple syrup for waffles. As cooking in the home declined in the 21st century, so that fewer people made candies or pies at home, commercial sales of Karo tended to dominate over the retail sales."
Watching US cooking shows/channels where they use vegetable oil also puzzles me because I'm going "but why not use shortening like margarine or butter or lard or suet or something instead?" I keep thinking of vegetable oil as "what you deep fry your chips in" and not "what you use to make cake or pastry". This site may be setting itself up as fancy French baking (with a Californian twist) but I'm still going "what the hell?" about putting rapeseed - I'm sorry, I mean canola oil for delicate American sensibilities - in cake batter, and I'm pretty sure genuine French chefs would be ready to storm the Bastille over this:
https://latourangelle.com/blogs/general/what-oils-are-best-for-baking
"Coconut oil, one of the best oils for baking, goes well with bananas, chocolate, and tropical flavors. It has bold, aromatic flavors and can even contribute to a DIY body care routine, moisturizing your skin and hair. As a solid at room temperature, you can use coconut oil as a butter substitute in cookies, cakes and more. With a lower smoke point than some other oils, coconut oil is an excellent option for low-heat recipes like Virgin Coconut Oil Lemon Curd."
A floor wax *and* a dessert topping, hmm? Substitute for butter - the flip I will! I remain horrified by the history as presented on Wikipedia; truly, American zing and innovation in convenience foods has a lot to answer for:
https://en.wikipedia.org/wiki/Crisco
"The process of the hydrogenation of organic substances in gas form was developed by Paul Sabatier in the late 19th century. Building on James Boyce's 1890s work in the successful development of a consumable solid lard substitute, Cottolene, in the U.S. the liquid form of hydrogenation was perfected and patented by Wilhelm Normann in 1903.
Joseph Crosfield and Sons acquired Normann's patent ostensibly for use in the production of soap. Their chief chemist, Edwin C. Kayser, was hired by Procter & Gamble's business manager, John Burchenal, and they patented two processes to hydrogenate cottonseed oil. Although their initial intent was to completely harden oils for use as raw material for making soap, these processes ensured that the fat would remain solid at normal storage temperatures and could find use in the food industry.
After rejecting the names "Krispo" and "Cryst" (the latter for its obvious religious connotations), Procter & Gamble called the product Crisco, a modification of the phrase "crystallized cottonseed oil". They used advertising techniques that encouraged consumers not to be concerned about ingredients but to trust in a reliable brand. Further success came from the marketing technique of giving away free cookbooks in which every recipe called for Crisco. Crisco vegetable oil was introduced in 1960. In 1976, Procter & Gamble introduced sunflower oil under the trade name Puritan Oil, which was marketed as a lower-cholesterol alternative. In 1988, Puritan Oil was switched to 100% canola oil."
'So, we *were* gonna make soap, but then we decided nah, let's use it for cooking instead!'
"They used advertising techniques that encouraged consumers not to be concerned about ingredients but to trust in a reliable brand."
Oh, Columbia! This is why I find it very funny when Dylan loses it over lard but is happy to measure out cups of vegetable oil:
https://www.youtube.com/shorts/hxXoMegR2nY
Corn syrup, the kind found in grocery stores and used in home baking, is nearly all glucose with just enough fructose to inhibit crystallization. While it might be theoretically cheaper than cane sugar, it costs several times as much in retail packaging. It is only about half as sweet as sugar. While it is used as a major ingredient in things like pecan pie, its main use is as an additive to inhibit the crystallization of sugar. While you can accomplish the same result by partially inverting sugar syrup with an acid such as cream of tartar heated high enough and long enough to break apart enough of the sucrose molecules, adding a dollop of corn syrup is easy and foolproof.
As for cakes with vegetable oil, you owe it to yourself to try chiffon cake. It may have been invented by a guy from California, but it is a brilliantly good recipe.
Yes, here in the Rest Of The World we have healthy corn syrup alternatives like golden syrup and good ol' white sugar, which can happily be eaten by the cup with no ill effects.
" good ol' white sugar, which can happily be eaten by the cup with no ill effects"
Yes, too much sugar is bad for you. But corn syrup is high fructose, which has added risk factors:
https://health.clevelandclinic.org/avoid-the-hidden-dangers-of-high-fructose-corn-syrup-video
"Compared with regular sugar, it’s cheaper and sweeter, and is more quickly absorbed into your body. But eating too much high fructose corn syrup can lead to insulin resistance, obesity, Type 2 diabetes and high blood pressure.
...Fructose was initially thought to be a better choice for people with diabetes due to its low glycemic index. But only your liver cells can process fructose, and that’s where the problems begin.
"Fructose goes straight to your liver and starts a fat production factory,” Dr. Hyman says. “It triggers the production of triglycerides and cholesterol.” He explains that it’s actually the sugar — not the fat — that causes the most trouble for your cholesterol.
What’s even worse, Dr. Hyman notes, is high doses of fructose “punch little holes in your intestinal lining, causing what we call a leaky gut.” He explains that this allows foreign food proteins and bacterial proteins to enter into your bloodstream, which triggers inflammation, makes you gain weight and causes Type 2 diabetes.
Studies show that high fructose corn syrup increases your appetite and promotes obesity more than regular sugar."
Corn syrup and high fructose corn syrup are different things.
Yeah. It should be noted that "high" fructose corn syrup has...about the same fraction of fructose (to total sugars) as sucrose (ie table sugar) does. Which is about (for HFCS) 45-55%, vs the almost exactly 50% that sucrose does. Because sucrose is 1:1 glucose:fructose in a molecule that your mouth mostly breaks into its constituents, while HFCS is already broken apart (by enzymatically converting glucose to fructose). By the time it enters your bloodstream, all the sucrose is already broken down into glucose and fructose.
So 'High' in HFCS is only relative to *regular* corn syrup, which is basically 0% (only a tiny amount) fructose, not relative to any absolute standard.
So all those studies? Yeah, they're garbage. If you want high doses of fructose, *eat an apple*. Or drink "unsweetened" fruit juice. From what I find online, apples are ~2:1 (ie 67%) fructose:glucose, and are basically pure sugars.
Indeed. The only reason that corn syrup is cheaper than cane sugar in the US is because of tariffs and import quotas - if we didn't have trade restrictions intended to "protect" US farmers, our sodas would be sweetened by real sugar just like the soda in Mexico and most of the rest of the world.
With all due respect to Nigella, she is wrong. US supermarket corn syrup is not high fructose. The closest thing in a UK supermarket to corn syrup is called "liquid glucose." Corn syrup has a slightly higher water content, but it is basically the same thing. Dark corn syrup would require mixing in a little molasses or black treacle.
Heh, interestingly enough smokers and drinkers are actually a net benefit to the government budget thanks to dying before the age of retirement.
And I did see people argue that those who have casual sex without a condom are horrible people and should be stopped.
"I see some of my patients more often than I think is medically indicated because they feel more comfortable if they see a doctor more often (I don’t know why)"
Some people feel that they need to be "seen", i.e. they feel better if they get to explain their problems to someone, who indicates they they are listening.
As a parent, if the antibiotic gets my kids back to school/daycare a day earlier, so I don’t have to take off work it’s worth so much. I would imagine this effect alone is driving much of the ear infection antibiotic use. There’s even sort of a natural experiment you could do looking at leave policies and antibiotic use for ear infections.
I wouldn't have guessed that doctors are prescribing something that's supposed to o be medically efficacious to a child because the parent says "I gotta go to work tomorrow."
I mean, the guidelines leave it up to doctor discretion, but they usually say something like, "you don't strictly need to go on amoxcillin as it will likely clear up on its own in 2-5 days, but I can give you this prescription and they will likely be better by tomorrow."
We actually skipped the antibiotic once to test it out, and they were home from school for 3 days since they can't go back with a fever, even though they were completely fine otherwise. Seemed like a waste of everyone's time, so we always took the antibiotic after that.
Also, this is a terrible example of waste, because amoxicillin is so cheap. The doctor's visit is what's expensive regardless if a prescription is made. But again, if a kid has a fever they can't go back to daycare until its completely gone, or you have a doctor's note saying it's just an ear infection and they're safe to return. So you have to sort of know your kid and be like, "is this fever and ear pain driven by a cold or an ear infection?" when you decide to take them in or not.
It seems odd if “medically efficacious” is defined in a way that says that getting over pain a day earlier so that the carer can get to work a day earlier is not “medically efficacious”.
Also eye drops for pink-eye. This is usually viral and not bacterial, but doctors will immediately prescribe anyway.
That's not just for treatment, though, I've had conjunctivitis and the gritty, dry-eye feeling was enough to get me to purchase over-the-counter eye drops for relief. So maybe the doctor shouldn't prescribe something but just tell the patient/parent to go buy Optrex at the chemist?
Though it seems that there can be cases where a prescription is warranted - oh look, and here's ear infections again!
"Bacterial Conjunctivitis
More commonly associated with discharge (pus), which can lead to eyelids sticking together
Sometimes occurs with an ear infection"
And the HSE advises that you should go to the doctor if a baby or child is the one suffering:
https://www2.hse.ie/conditions/conjunctivitis/
I meant specifically eye drops with antibiotics for treatment, not for immediate relief. And yes parents should go to the doctor when children exhibit those signs. It's just apparent that even if the cause is most likely viral (which can be inferred from symptoms), they'll give antibiotics anyway.
If a treatment is highly effective on 1/10 of cases and has no significant negative effects on the other 9/10 of cases, it doesn’t sound wrong to just give the treatment right away before figuring out which type of case it is.
> no significant negative effects
The point of contention with antibiotic overprescription is that it may harden pathogens in the environment. There can be temporary deleterious effects on microbiota but I believe this is far less of a concern.
As an adult I'll happily take antibiotics if it will cure my reasonably-severe pain one day earlier, and I don't see any reason to make medical decisions for my kids that are crueller than the decisions I'd make for myself.
You send your kids to school, don't you? I bet you wouldn't put yourself through that again. Talk about cruel.
I go to work, that seems no better than school.
Antibiotics often make it easier to get past daycare gatekeepers. "The kid will be fine to return to school after 24 hours" is a more legible rule for the daycare than, "Wait until the symptoms go down. It'll probably be a day or two, maybe a bit longer."
I'm glad he clarified, as that is both a more reasonable position and likely correct. From a mindset of "how do we improve healthcare spending?" it's quite likely that his stronger language is necessary to bend the conversation away from current trends of spending money on anything that might possible increase health outcomes, even if many cases are wasteful or even make health worse. I doubt anyone in this conversation has decided to turn down all medical care, but maybe a few people will spend less on pointless or low utility services now.
For myself, I've actually been turning down healthcare when offered by doctors/hospitals for several years. If I need some kind of test or treatment, I take the minimum and often say no to additional tests or follow-ups. There's almost no benefit to going back to a doctor for that doctor to tell you that you're healing okay for instance. When you get your original treatment they should provide information about healing timelines and care. You can often tell for yourself if you're healing, and unless you're having significant complications or things aren't healing, there's not much they can or will do differently.
I'm going to ask a dumb possibly inflammatory question. It is based on ignorance. I vaguely remember hearing often in media that black women have high maternal death rates. And I think black women have a lot more abortions per capita. Would that suggest their maternal death rate is in part explained by the U S. method of including in maternal death rates anyone who has had an abortion recently? Or, to clear up another basic question. Let's say Hassidic women have 5 kids on average, and south Korean women have 0.5% kids on average. Would we expect Hassidic women to have 10x that maternal death rate, or the same? Is it counted per woman lifetime or per category justifying occurrence?
Probably not, if the current method of measurement wasn't introduced until 2003; black women already had much higher rates of mortality back then https://www.scientificamerican.com/article/why-maternal-mortality-rates-are-getting-worse-across-the-u-s/
US maternal mortality rates are based on "if you die within a year of pregnancy, that gets counted". So there is some humming and hawing over "if you die of a cardiac event ten months after having a baby, was that due to the pregnancy or not?" Certainly, if you are in bad health, then pregnancy and delivery will put a further strain on the system, but in that case it's not the pregnancy as such that caused it.
It seems that black Americans have a lot worse outcomes for several diseases:
https://minorityhealth.hhs.gov/blackafrican-american-health
"The death rate for Blacks/African Americans is generally higher than whites for COVID-19, heart disease, stroke, cancer, asthma, influenza and pneumonia, diabetes, HIV/AIDS, and homicide."
(Yes, obviously "homicide" is not a disease, but the department of health includes it, so what can I do?)
https://www.pfizer.com/news/articles/health_disparities_among_african_americans
"One possible contributing factor: The Centers for Disease Control and Prevention (CDC) says African Americans are more likely to die at early ages for all causes, as young African Americans are living with diseases that are typically more common at older ages for other races. For example:
High blood pressure is common in 12% vs. 10% of blacks vs. whites aged 18-34 years, respectively. It is common in 33% vs. 22% of those aged 35-49 years, respectively.
Diabetes is common in 10% of blacks aged 35-49 compared to 6% of whites.
Stroke is present in 0.7% of blacks aged 18-34 compared to 0.4% of whites the same age. Stroke is common in 2% of African Americans compared to 1% of whites aged 35-49 and 7% vs. 4%, respectively, in those aged 50-64.
The CDC said that social factors compared to others in the U.S., specifically whites, affect African Americans at younger ages: unemployment, living in poverty, not owning a home, cost-prohibitive effects of trying to see an MD, smoking, inactive lifestyle, or obesity.
A white paper from Cigna went further, acknowledging mental health disparities between African Americans and white patients. They noted blacks are 20% more likely to report psychological distress and 50% less likely to receive counseling or mental health treatment due to the aforementioned underlying socioeconomic factors.
Another area of health care there is a disparity is among renal disease. Blacks and African Americans can suffer from kidney failure at as much as 3 times the rate of Caucasians, according to the National Kidney Foundation. Black patients represent as much of a third of all patients in the U.S. receiving dialysis for kidney failure, though they don’t represent anywhere near that proportion of the U.S. population, they added. Individuals who are black alone, the Office of Minority Health says, make up 12.7% of the U.S. population."
Kidney disease seems to be down to genetic factors:
https://www.niaid.nih.gov/research/diseases-disproportionately-affecting-minorities
"African Americans develop kidney failure at more than four times the rate of white Americans. In part, this increased risk can be attributed to two genetic APOL1 coding variants commonly found in African Americans. The NIH-supported APOLLO study will explore how APOL1 gene variants affect kidney transplantation outcomes, with the goal of improving the outcomes of both kidney donors and recipients."
Oh, and "it's never lupus"? Well, unless you're a black female:
"Systemic lupus erythematosus is two to three times more common among African American women than among white women. It also is more common in Hispanic/Latina, Asian, and Native American women."
EDIT: I got interested in digging out maternal mortality rates and comparing them due to the, ahem, expansive definition of the CDC leading the pro-choice lot to claim that pregnancy was horrendously dangerous and abortion was in fact way safer and that without abortion so many more women would die due to being pregnant. I was interested to compare "mortality rates in countries without abortion or before abortion was introduced, or with abortion but compared to the USA". Clearly, if the CDC is claiming that being pregnant is killing off black women, then the call for "reproductive justice" is justified, right?
Regarding Lindley's "Paradox"... It isn't really a paradox. The Bayesian answer is simply correct, and the frequentist answer, if it differs, is just wrong.
BUT! This is so ONLY if the context of the question is as assumed. The Bayesian analysis in Lindley's paradox assumes a prior in which there is some some substantial probability (say, 1/2) that the true effect is EXACTLY zero, and otherwise (say with probability 1/2) the true effect is some non-zero quantity that we have no reason to expect to be close to but not exactly zero (perhaps our prior is uniform over (-10,+10)). Then, if you gather a lot of data, and find that your 95% confidence interval is (0.00005,0.00045) and your p-value is 0.006, you should not reject the null hypothesis, because the low prior probability that a non-zero effect just happened to be that close to zero, when the uniform prior on (-10,+10) provides no reason to expect that, more than out-weighs the low probability of getting an observed effect of this size if the true effect is zero.
In most real problems, including the one under discussion, there is no reason to think that non-zero effect sizes that are quite close to zero are unlikely. One can imagine lots of reasons that medicine might be a little bit effective, but not highly effective. Also, as noted in the comments quoted, the sample size is not actually all that big considering that mortality over a year or so is only about 1% (so presumably there were on the order of 10000 deaths, making effects of a few percent on the death rate hard to detect).
In particular, the effectiveness of medicine is a sum of many small effects. Everyone agrees that some medical interventions are helpful on net and some are harmful. The total measured effect of the study should be a sum of positive and negative terms. The idea that there is a substantial probability that the effect adds up to exactly zero is absurd. A reasonable prior distribution is more like a Gaussian or simply a uniform distribution.
Yes, the assumption that there is any substantial probability of an effect that is exactly zero is also wrong. However, the prior for the effect size should probably not be Gaussian, but instead something with heavier tails (equivalently, a higher central peak), because the individual positive and negative effects of medicine are not necessarily SMALL, as needed to conclude that the combined effect has a Gaussian prior.
It's funny we made the same point but described in opposite ways—I described it as the frequentist version being correct and the Bayesian one being wrong.
The frequentist solution here gives the correct answer to the question you *should* have asked (how sure are we about the sign of the effect?). The Bayesian solution gives the correct answer to the question the question you *actually* asked (is there a nonzero effect?).
I was going to bitch about Hanson based on his presumed background, which I now realise would be an ad hominem attack and not support the point I was trying to make in any way about "eff you for proposing making medical care more expensive to cut down on 'waste' because people died from not having sixpence for the dispensary doctor", but I got derailed looking up "am I sure he was born in relative privilege to my family background, maybe he *did* grow up with no running water and an outdoor privy" and found this plum from a 2003 article:
" Robin Hanson is a married, 44-year-old father of two who teaches economics at George Mason University, a commuter school with aspirations that's plunked amid the affluent sprawl of northern Virginia."
"A commuter school with aspirations". I'm laughing so hard at that, in an admittedly mean-spirited way, that I've loosened my grip on my wrath and let it fall away, and am now enjoying somebody else being snarky about the GMU economists 😁
I always find it funny that it’s the Third best university in the DC Area named after someone named George.
"I get nervous about this argument, because it implies that if the government helps you in any way then they ought to have power over you"
Thank you for saying this.
I don't quite understand the ear infection discussion. Is there a reasonable case to be made that antibiotics for childhood ear infections are ineffective? Or is the case just "meh, they'll probably clear up eventually, let 'em suffer".
I say this as someone who almost never sees the inside of a doctor's office except when one of my kids has a (real or possible) ear infection, which seems to happen every goddamn six months.
Yeah, it's very easy to say "oh, that doesn't need antibiotics, it'll clear up in a couple of days" but that's because it's not aware of screaming in pain small kids who can't sleep or eat and the frazzled parents who have no idea if this is something mild or something really serious, and just want the screaming to stop. Sure, you can dose your kids up on Calpol, but that evokes uncomfortable associations of Victorian Soothing Syrups which drugged infants to death:
https://library.usa.edu/mrs-winslows-soothing-syrup
"Significant levels of morphine and alcohol were known ingredients in this product. The medicine claimed to treat teething for children. The product was popularly used among mothers. Yet the copious amounts of alcohol and morphine had the potential to induce coma, addiction, and death for infants. The product originally contained 65 milligrams of morphine before legislation and litigation reduced the amount over the next few decades. The American Medical Association labeled the product as a “baby killer.”
The product was eventually discontinued and removed from the market in the 1930s."
Drat those meddlesome FDA-adjacent types who interfered with a parent's right to hop their infant kids up on morphine!
https://www.pharmacytimes.com/view/pharmacys-past-the-soothing-syrup-known-for-causing-death-in-thousands-of-babies-
"The syrup contained morphine 65 mg per ounce, as well as alcohol. One teaspoonful had the morphine content equivalent to 20 drops of laudanum (opium tincture); and it was recommended that babies 6 months old receive no more than 2-3 drops of laudanum.
One teaspoonful contained enough morphine to kill the average child. Many babies went to sleep after taking the medicine and never woke up again, leading to the syrup's nickname: the baby killer.
...The Pure Food and Drug Act instituted in the United States in 1906 forced companies to disclose the active ingredients on drug packaging. Companies also had to ensure that the purity level of the drugs was not below the levels established by the US Pharmacopeia or National Formulary. A similar law, the Food and Drugs Act, was passed in Canada in 1920 in a similar attempt, and also to ensure that drugs were properly marketed."
Also, earache may not just be an ear infection simpliciter but a sign of something else, depending if there are associated symptoms:
https://www.nhs.uk/conditions/earache/
"Earache and pain can be caused by many things, but sometimes it's not known by what.
Here are some of the most common causes:
Possible causes of earache symptoms.
Symptoms Possible condition
Ear pain with toothache Children teething, dental abscess
Ear pain with change in hearing Glue ear, earwax build-up, an object stuck in the ear (do not try to remove it yourself – see a GP), perforated eardrum (particularly after a loud noise or accident)
Ear pain with pain when swallowing Sore throat, tonsillitis, quinsy (a complication of tonsillitis)
Ear pain with a fever Ear infection, flu, cold"
MrP changed his view to a more-nuanced one after he and I had a back-and-forth on this. It was a rather nice civil discussion.
This seemed to be in a section that made the broader claim that benefits that aren’t “medical benefits” don’t get counted. Having a more comfortable stay, or enjoying your meals, or being with friends, or having less pain, doesn’t count unless it’s been defined as part of a “medical” outcome.
My understanding of the literature on this was that antibiotics reduce the length of the infection by a relatively small amount. We tried to avoid them for moderate-seeming ear infections because I have read that frequent use of antibiotics can cause autoimmune disorders in adulthood, but we were also lucky enough that our kids were just uncomfortable and not inconsolable. Certainly if they’d had a high fever we would have given them some.
"Comments From The Rest Of You Yokels"
Ah, God bless and kape yer honour, sor, safe an' well, shure aren't you afther gettin' the right of us? Indade and begob, yokels we are! 😁
I feel so Seen and Represented!
I am deliberately delaying my shower slightly by posting this, so I can officially be one of the unwashed masses.
Lol
Thank you, WindUponWaves, for the discussion of maternal mortality stats in the USA versus other nations. I had heard for a long time that the comparison wasn't apples-to-apples, but hadn't seen a clear description of what was off until your comment.
I've heard there are similar issues with infant mortality as well. E.g. in France, if a baby dies within 24 hours of birth, it's counted as a stillbirth rather than an infant death, whereas in the US, it's counted if they came out of the womb alive even if they die immediately afterwards.
Glad to be of service! I'm just chuffed that I got a mention in a highlights post at all, it's not often that you get a honor that really makes you proud to contribute. That notetaking system I developed where I write down *EVERYTHING* that catches my interest, and file it away for later reference, is really paying off!
>And I remember that during residency I worked with a doctor whose answer to all painful-but-not-otherwise-dangerous conditions was “nobody ever died of pain”. This guy probably had the lowest medical spending in the hospital, and maybe the lowest side effect rate in the hospital, and probably many other valuable records, but I would not have wanted to be his patient.
As The American Taxpayer, I would much prefer every doctor be a clone of this guy. If people want white glove treatment they should pay out of pocket, keeping people alive at the lowest cost is good enough for public spending on health.
The doctor is straight up wrong, though, about pain not killing people. If nothing else, chronic pain has been repeatly shown to increase suicide risk.
I have no sympathy for people who hold themselves hostage with suicide threats if they aren't provided with elective/not health necessary medical procedures, or for using increased suicide prevalence as an argument for preemptively giving in to demands.. In my opinion it's the same as a four year old saying "if you don't give me candy I'm gonna hold my breath until I die!" and should discouraged, not rewarded.
If people are actually going to kill themselves if not provided with medically unnecessary medical care they should just be institutionalized on suicide watch until they calm down, for their own safety.
To clarify, if people are suicidal due to a rational self-reflection of their own chronic health issues and prospective quality of life, it's a free country and I respect their decision.
I object to allowing people to self-diagnose as suicidal and self-prescribe whatever they want people to give them, while holding their own lives hostage as justification.
>That’s a very odd statement on its own, as people who take their own life for any reason obviously have mental health issues - which need treatment.
I would give a snarky [citation needed], but instead I'll just note that your "obviously" is nothing of the sort. Many cultures, including basically all Western cultures today, support suicide in cases where no mental illness is present.
Euthanasia is one such case in western cultures. We don't assume that people requesting euthanasia are mentally ill. In other cultures, honor and grief have both been viewed as valid reasons for suicide, without any presumption of mental illness. You're attaching an "obviously" to something that is clearly not obvious at all, which would be viewed by many cultures as insultingly wrong, and which is probably a minority opinion in western cultures today.
Okay, let me jump in with "you are fucking stupid". We're not talking about "mild discomfort", we're talking about *pain*. Pain bad enough that people would be driven to kill themselves without relief.
It's not "take two aspirin and suck it up, buttercup" levels. I had to listen to my mother crying and moaning in pain when the morphine patches for her lung cancer weren't enough. So to Dr. Pain and to you, all I can say is I hope you bastards get to experience those levels, one day, and see if your opinion changes.
There's a midpoint between "overprescription of painkillers" and "people can't handle mild or even severe but temporary levels of pain" and "no pain relief at all, if you can't pay high prices for relief then just lie there and suffer".
You're trying to cover this under "medically unnecessary" but there are all too many doctors who ignore pain on the grounds that "nobody ever died of pain" and have no idea of the actual levels of pain involved, to the point of being dangerous to patients whose pain is ignored and the underlying condition not treated.
Why should we care about health enough to pay for it but not concerned enough about pain to pay for it?
They can just deal with it. My understanding is that overperscription and overemphasis of painkillers effectively caused the opioid epidemic, I'd rather avoid adding to it.
This is a somewhat extreme position which thankfully is not shared by many... how would you feel if you or a family member had metastatic cancer to bones, which is excruciatingly painful? "No one ever died of pain" is not much help
Overprescription has little to do with the previous heroin or the current fentanyl epidemics. When the price of Heroin dropped at the beginning of this century, it became a cheaper high, and more people became addicted because usage went up.
Fentanyl is cheaper than heroin (because it's easy to manufacture) so it's now become the most popular illegal opioid. So the illegal opioid epidemic has little to do with the overprescription of regulated painkillers.
In the case of Oxycodone, it's a little more nuanced than the simple overprescription as being the problem. Evidence came out in discovery in the lawsuit against Purdue Pharma, that the Sackler family made an effort to get higher dosages prescribed as the standard dosages for milder pain. Thus addicting people who otherwise wouldn't have been addicted. From the Wikipedia article on Oxycodone...
> In 2019, The New York Times ran a piece confirming that Richard Sackler, the son of Raymond Sackler, told company officials in 2008 to "measure our performance by Rx's by strength, giving higher measures to higher strengths".[124] This was verified with documents tied to a lawsuit – which was filed by the Massachusetts attorney general, Maura Healey – claiming that Purdue Pharma and members of the Sackler family knew that high doses of OxyContin over long periods would increase the risk of serious side effects, including addiction.[125] Despite Purdue Pharma's proposal for a US$12 billion settlement of the lawsuit, the attorneys general of 23 states, including Massachusetts, rejected the settlement offer in September 2019.[126]
But why stop there? What requires keeping people alive?
The Overton window
And this is where Hanson is at least a little right that health care is about showing that we care. If I would pay for an anti-pain treatment for myself or my kid, I would find it outrageous that insurance not pay for it. But we are agreeing that it really is about values, not life expectancy at this point! (Notwithstanding the point made elsewhere in the comments that pain increases risk of suicide)
I feel like this whole debate lacks a dimension because it seems to be assuming health-care is somehow only about keeping people alive. Surely health care is also about increasing life quality. Even from a society level cost-benefit perspective, a government should probably pay for treatment that keeps people healthy enough to do their job, rather than living as invalids!
Which is one of many reasons why people other than the patient paying for the doctor is a terrible idea.
A commenter on one of the previous posts said something like, "If [some intervention / health insurance / medical spending] is only improving my life expectancy by 3 years, I could take it or leave it." (That's a gross paraphrase, sorry.) But what I thought is, if some intervention only makes you live 3 extra years, you might not want it, but on the other hand, if it has a 20% chance of making you live 15 extra years, does that change your calculations? I know that's the same expected value, but it doesn't feel the same to me, and it's my LIFE after all.
Three years seems like kind of a big deal to me, tbh. Not if it's three years of lying in a nursing home unaware of my surroundings or my own name, but yes if it's three years I can interact with people I love and maybe do some things I enjoy.
I agree, especially if I think of them as the next three years, and not some three years far off in the less imaginable distance.
> And I remember that during residency I worked with a doctor whose answer to all painful-but-not-otherwise-dangerous conditions was “nobody ever died of pain”.
This doctor is straight up wrong; at a minimum, chronic pain is known to be a risk factor for suicide.
If I had a probably-terminal cancer, my preferred treatment would be to be submerged in an ice-bath and have some giant needles thrust into my arteries and veins, and have all my blood flushed out, replaced by a mixture of saline, anti-coagulant, and para-formaldehyde. This is how I clean the blood out of rodent brains for preservation. That's how I'd want to preserve my brain. Works best if you do it with the heart still beating. I'd wash down some caffeine pills, asprin, and NSAIDs with hard alcohol first, to help keep my heart rate up and thin my blood a bit further and make sure I didn't mind the discomfort of the procedure. Then my head could be removed and properly preserved. Once uploaded, I'm sure digital-me will be grateful for the improved quality of the brain scan which results from such a cleanly preserved brain.
Why not under general anesthesia? Your heart will still be beating, but the discomfort will be gone.
Well, the procedure would be lethal. I wouldn't want someone else to be charged with murder. So unless the procedure becomes legalized in time for my death, I'll have to perform it on myself.
Ok this makes sense. You really thought this through!
Why not do cryonics instead?
I think that’s roughly how cryonics works. You need some sort of antifreeze/anticoagulant to prevent the freeze from destroying things.
Kenny is right, this is generally step 1 of cryonics (although they typically wait until after your heart has stopped, which makes the process suboptimal).
But there's an important reason I don't simply say, "I'd opt in for euthanasia and cryonics, if terminally ill."
The reason is: I don't want my brain stored at sub-zero temperatures for a long time. I want it to be preserved via CLARITY (brain clearing) so that it will be stable at room temperature, more mechanically cohesive, and able to be optically imaged in thick slices. Then I want my brain to be imaged. Not 100 years from now. Like, right after I die. And then I want someone to try to turn my digitized neurons into a Whole Brain Emulation. Why wait, we're nearly ready! If I live another 10 years, we'll surely be ready for that. I hope to live another 40 years or so, so there should be plenty of time. No more need for cryogenics, just straight to uploading.
Do you expect your consciousness to be preserved? I don't see how that could do so. It would create an imperfect copy that has some traits in common with you. But it wouldn't be you. From your perspective it would be the same as dying
Consciousness is lost every time you go under general anesthesia. So by that definition, I've died at least twice already. I'm not particularly worried about that aspect. But yes, since it would kill me, I wouldn't do it unless I was already close to death. Thus, the discussion of incurable cancer.
It would be more like... having a digital child, but one that was more similar to you than a child would normally be (and had some of your memories). So a digital twin. Not exactly me, not fully different. Something I would expect to have enough in common with me that I would feel pleased about it existing in the world, as I would feel if I had the option to wish a twin into existence or raise children. But being digital would mean it would have super powers. Be able to be immortal, and clone itself, and travel at the speed of light through fiberoptic networks, and many other advantages. Maybe someday travel the galaxy by leaping from laser-relay to laser-relay.
The fidelity of an early scan can't be expected to be perfect. But the scanning technology is non-destructive of the preserved brain tissue, so later on the scan can be repeated with improved scanning tech for higher fidelity.
In the same vein, is you 10 years ago also dead? Surely you are no longer the same as the consciousness in your body 10 years ago?
> All the government has to do is offer to pay for free STI treatment, and then if you have sex too much it’s an “externality” and you’re “robbing the government” so the government should be allowed to step in and stop you.
Unironically, yeah? Like this just seems straightforwardly true? If tax dollars are taken to pay for STI treatment (and don't strictly limit under what circumstances it's provided), then either the government has the authority to regulate activity which results in STIs, or you will be in the situation where people who behaved more carefully are being forced to subsidized other peoples' irresponsible behavior. (If this sounds similar to the situation with student loans, that's not a coincidence).
If you want to make some sort of argument that this is net positive utility or whatever then you can do that, but you have to actually show that; it's not a reductio ad absurdum.
This brings up a kind of meta question about what kinds of powers a government should have, categorically. You’re saying there is a solid logic to the “pay for STI treatment —> interfere in sex life” position. But this kind of thing will be applied (1) incorrectly and (2) selectively.
Incorrectly: we get some behavior/outcome thing wrong and end up oppressing people without even getting a monetary benefit.
Selectively: different political factions will be motivated to intensely focus on some behaviors in this way (being slutty) and not others (owning a gun). (I am a liberal so stuff conservatives might be wrong about is more available to me, sorry; I acknowledge there will be examples in the other direction.)
And what if it is just not possible to reliably identify “effects of behavior you can control” vs “treatment you need for reasons everyone would agree you are blameless for”? Or if it is so cumbersome that x% of our health care budget goes to blame-shifting rather than care?
Based on these objections, I don’t think we should go down this road very much as a way of saving money on health care. (I also just think it’s wrong, but will not try to directly convince you that my feelings are correct.)
> But this kind of thing will be applied (1) incorrectly and (2) selectively...
Of course it will. The government will do this for almost everything it implements.
> Based on these objections, I don’t think we should go down this road very much as a way of saving money on health care.
I don't know why the obvious solution of "don't have the government do so much" isn't the first idea to come to mind. You think that the issues you just pointed out aren't going to apply to government run programs, even if they aren't explicitly trying to deal with moral hazard?
> I don't know why the obvious solution of "don't have the government do so much" isn't the first idea to come to mind. You think that the issues you just pointed out aren't going to apply to government run programs, even if they aren't explicitly trying to deal with moral hazard?
Perhaps you are right and these objections aren’t good enough, as they prove too much (as you said, much of what the government does could be applied I’m misguided ways), and I’m mostly just saying I think it’s wrong.
But the reason I think it’s wrong is that it’s *not* like other government actions in some important ways. There is just a limit to how much the government should interfere in private lives. Most people agree that there is a good case for interference when a person’s actions harm others (breaking traffic laws, spying for an enemy country), but not when it doesn’t. It just doesn’t seem right to have society ban something as fundamental as romantic relationships because our social insurance system might incur costs. Some people are just really bothered by the idea that they have to partially bear the costs of actions by others, while others are more bothered by the idea that someone who needs help is being allowed to suffer. Most systems that help people have some errors in both directions. Which ever one bothers you more determines your politics to a large degree.
> There is just a limit to how much the government should interfere in private lives.
I agree, but I would consider "take a bunch of everyone's money" and "pay for people's <medical treatments or whatever else>" to be interference. The question, "if I'm being forced to pay for something, shouldn't I have some say over that thing and be allowed to take measures to reduce that spending?" is, in my mind, intended to point to the conclusion I described above, rather than to complete totalitarianism, although I suppose some people might conclude otherwise.
> Some people are just really bothered by the idea that they have to partially bear the costs of actions by others, while others are more bothered by the idea that someone who needs help is being allowed to suffer.
I think most people are on board with reducing suffering. But where does the idea come from that just because someone is suffering, we can't ask what the cause was, or put a limit on what relief is actually offered? Some people are probably suffering because of high taxes; can I just say that I need a lower tax rate and it's a violation of my privacy for anyone to ask why?
Also, this is not just a question of "does it seem right to force someone to pay for someone else's costs"? I think subsidizing demand has not only failed to make things like medicine and college affordable, it's actively made the problems worse, including for many of the people it's supposed to try to help. Biden's recently announced housing subsidy will do the same; without being able to increase supply, the entire subsidy is likely to be absorbed by existing owners. And in general, many social welfare programs seem to be poorly run and overall inefficient. So there's both philosophical and practical considerations to keep in mind.
Can the respective positions of Hanson and Scott be boiled down into a prediction they can make a bet on, or on the minimal concrete policy proposal which one would say 'yes' to and the other 'no'? I feel like the conversation as it's currently proceeding is an interminable debate about what someone does or doesn't mean when speaking fairly broadly and imprecisely.
When reviewing "The Origins of Woke", Scott notes how difficult it is to fairly review a book with deliberate omissions/logical gaps for the sake of making a better persuasive case. Back-and-forths like this show that Scott should be doing it for EVERY politics-adjacent argument he encounters, not just those where the author is honest enough to admit it!
Just in the last month, the Hanson Health Care posts, Hanania Woke posts, Coffeepocalypse post and (if you stretch) Emotional Support Animal post all had this problem. First, Scott interprets a rule or argument way too literally, instead of the vibes-based "let's have less of this sort of thing" statement that was clearly intended. Then, he wastes an entire post providing evidence in favor of the thing, explanation of the incentives that lead to the thing, or simply wringing his hands in confusion. Don't bother offering evidence when it wasn't asked for!
“nobody ever died of pain”, said Dr. Mengele while vivisecting twin children.
I mean, how is this doctor practicing medicine? It's not even factually correct, excruciating pain can cause cardiac arrest, people do die of extreme pain.
On a different note, kudos to MrP for engaging in a very civil and productive debate on this.
Cremieux misunderstands statistics, as usual.
Statistical power does not only depend on the sample size. If I have 1 million samples and I'm trying to estimate whether an event happens 1-in-a-million or 2-in-a-million, my power is very low.
If I have N samples and I'm estimating an event with true probability p, then assuming p is small, my standard deviation on the estimate of p will be around sqrt(p/N), so my relative error will be ~sqrt(1/pN). For example, if I have N=1 million samples but mortality happens with p=0.1% probability, my relative error on the estimate of this probability will be sqrt(0.001)=3%. If I want 2-sigma confidence (to get a 95% confidence interval), I have to multiply this by 2. This means that I can only estimate mortality rate up to a 6% relative error.
So if an intervention decreases mortality rate by 5%, I don't even have the power to detect this with a sample size of 1,000,000 people, assuming people in my sample only die at a rate of 0.1%. Cremieux's intuition is dead wrong.
Thank you, I knew something like this must be the right response but couldn’t be bothered with the calculations.
And this is particularly true for this study, where (1) takeup was low (they say the letters increased insurance by 1.3 percentage points, so corresponding to roughly 26,000 additional insured individuals) (2) the main specification is an IV specification, which tend to have much larger standard errors than ordinary OLS comparisons.
The only thing id add here is that we keep on jumping between population and individual effects and duties and responsibilities. Most medicines don't work most of the time for most people, except insulin for diabetes and some antibiotiscs for some infections. However, getting more people more access to more medications has greatly increased statistical life expectancy, even accounting for economic growth, less pollution, safer work etc. However, no one person has a duty to take a medication that will most likely not increase their survival just because a doctor has been told that giving out more drugs to more people will increase the survival of the average of the population. And a doctor's duty is to the patient in front of them. Howeverrrr, because of the great successes of broad guidelines and standardisation of care, many people (not just clinicians) have started to think of all health through the broad based indiscriminate lens. This is statistically naive, and often evidentially poorly founded. And..... It's annoying. It can be paternalistic and feel dismissive. But it's hard coded into our guidelines and insurance systems. The way I think about it is - screening. There are very strict criteria for setting up a screening programme and the evidence is reviewed regularly. However, nowadays any tom dick or harry can do a metaanalysis of a bunch of RCTs and say something like "this magic statin reduces risk of cv death by 5%" and it'll be all over the news and in the guidelines without the public health or the staticians getting involved and death rates never budge and we'll wonder why
Hol' up. Not a single reference in these comments regarding the Emotional Support Animal Racket post which maybe, just maaaaaaaybe, provides an obvious example of ineffective medical spending?
Gorillas really can play all the b-ball they want unobserved.
The Emotional Support Animal *Racket*, involves very little medical spending. The animals in question are ones whose owners/supportees already purchased and will maintain as pets, whether they're called ESAs or not; the only involvment of the medical community is a doctor signing a form letter, which probably isn't going to involve more than fifteen minutes of their billable time.
Non-racket Emotional Support Animals might involve a more substantial expense, for training and/or for purchasing and maintaining and animal that you otherwise wouldn't. This may be a waste, but that isn't established by the existence of a parallel racket.
This is incorrect. The medical spending in question isn't the money spent on the animals or the effort on the psychiatrist's part, it's the money spent on the psychiatrist's time. This is generally billed per session, not per minute of effort. I have no clue how much exactly this nets out to - certainly some people are already seeing a psychiatrist and include their request during an appointment that would have otherwise been purchased*, but these need to be balanced by those willing to purchase multiple sessions as mentioned in Scott's post.
* But if the 15 minutes aren't being spent on anything more valuable - or, more precisely, that there is no more valuable substitute on offer - then how much of the non-ESA therapy time is of zero or negative value?
Seems like the maternal death rate discrepancy could be easily compensated by subtracting the base mortality rate in that age range. What's the death rate of X year old women who were pregnant in the past year? What's the death rate of X year old women who were not pregnant in the past year? What's the difference? There you go. I suppose you'd also have to adjust for other demographic con-founders, but even doing a sloppy job of that is going to be better than not doing this math at all.
The point of numbers is not to get an accurate idea of the size of the problem, the point of numbers is to get more funding for the problem.
> Okay, I was pretty on board with maybe I was just strawmanning Robin and he thinks most medicine works and it’s just that we overspend at the margin — but the podcast is called “Most Drugs Are Bad For You”! Someone who listens to podcasts - is this just mistitled?
To answer the narrow question: there were two guests on that podcast and "most drugs are bad for you" is a direct quote from the other guest.
That being said, I don't think "he thinks most medicine works and it’s just that we overspend at the margin" is an accurate summary of his position. He wants to cut spending on medicine in half. That doesn't imply that it "mostly works".
It seems like you think cutting back by say 10% and cutting back by 50% are basically the same position?
I think Hanson's position is that we don't have the data to know whether 10% or 50% of medical spending is useless or even harmful, but he suspects it's closer to 50.
(If you like, you can phrase the issue as Scott: The US isn't getting much for its extra medical spending relative to Britain, and Robin: Britain isn't getting much for its extra medical spending relative to Cuba.)
I recommend this book: Medical Nihilism - by professor Jacob Stegenga -
"A highly controversial argument for a sceptical view about the effectiveness of modern medicine
Draws on interdisciplinary research and grounds the arguments in medical examples
Accessible to readers from any academic background" -
"This book argues that if we consider the ubiquity of small effect sizes in medicine, the extent of misleading evidence in medical research, the thin theoretical basis of many interventions, and the malleability of empirical methods, and if we employ our best inductive framework, then our confidence in medical interventions ought to be low." - Published by Oxford University Press.
Regarding the issue about p-values, I think it is better thought of as a claim relating the size of the effect to the p-values observed. In general the argument that, if this were a true effect and not totally trivial in size then it would have a much lower p-value is totally valid (I'm not completely convinced that 1.4 million is beyond the range we should expect to pick up true but very tiny effects [1]).
However, it's misleading to think of 1.4 million as being the study size. In reality it's the number of respondents who adopted healthcare because of the intervention but wouldn't have otherwise who represent the effective population size. So yes, it's a very good argument that the practical impact of sending nagging letters on health is vanishingly small but I don't believe that translates to showing that the effect of actually getting healthcare is trivial. And I think Goldin's claim that if you select the study sized based on power considerations based on prior estimates is a good response.
However, the bit about reshuffling the groups doesn't seem to be a particularly strong argument. I mean, imagine you got the results by p-hacking then you'd still expect the same thing to be true because you choose the outcome examined based on exactly the fact that it was a statistical outlier. But here they seem to be reporting the obvious variables of interest not some odd cross tab.
--
1: Yes, given a finite number of variables you need to have some fall off in number of true correlations with the true effect size below some value simply because every true correlation must occur at some finite value. However, we don't really have a finite number of variables -- or at least have a very large number -- once we start studying composed effects (eg effect on health outcomes of a nagging letter).
Also, regarding the social benefit of medicine, I think this is a particularly difficult issue because what is of benefit to society may diverge quite sharply from what is of benefit to the individual.
After all, given a finite lifespan it's not at all clear extending that lifespan is of any benefit to overall social utility. Everyone is going to have to mourn their parents and friends at some point (some benefit to less mourning per life year) and if the choice is between giving resources to the elderly or replacing them with young healthy people it seems like net utility may favor the latter.
This leads to the odd result that the mere existence of medical treatments might be net harmful because people want to feel cared for and not coldly abandoned to death and if you've done all you can they will feel that way.
I also think it's worth defending the psychological benefit of feeling you are being cared for and treated. After all, look at how much people are willing to pay for homeopaths or Reiki healers who mostly just offer a personal interaction and placebo.
I think there is a strong argument for decoupling the expert work of diagnosis and care from the human relationship work. Maybe a new kind of nurse practitioner whose job is mostly just listening and interacting with the patient and just futzing with relatively easy shit to make them feel listened to and like someone is concerned about them and to handle the check-in requirements for dispensing controlled or risky substance. Then maybe AI analysis of the recorded interaction to flag warning signs.
Regarding negative externalities: one cost-negative approach to negative externalities would be for the government to stop subsidizing things that are bad for us, e.g. grains and tobacco.
If 95 out of 100 people die from a risky surgery that looks, in isolation, bad; if 100 of them would die without the surgery it looks less bad. In either case, though, probably not a surgery that should be covered with insurance.
I like to joke that Lindley's paradox is paradoxical, because the frequentist solution is so egregiously wrong that it loops back around into being more Bayesian and more correct than Lindley's idea of a "Bayesian hypothesis test".
First, let's back up, and ask what the *actually correct* solution to Lindley's original problem of birth rates. Lindley tested a dataset of tens of thousands of births, found about 52% of them were male, he found this gave an exceedingly large Bayes factor in support of the idea that exactly 50% of human births are male. Do human births have a sex ratio of exactly 50%? No! We have overwhelming evidence that sex ratio at birth is slightly skewed towards men.
So why did Lindley find a large probability of a balanced sex ratio? It's because his prior was @#$%!ing stupid. His prior assigned an equal probability to two hypotheses:
1. Exactly 50% of babies are male.
2. Any sex ratio between 0% and 100% is equally likely.
So, his probability density is 0.5 everywhere between 0 and .5, and everywhere from 0.5 to 1. But at *exactly* 0.5, the probability density is @#$%ing *infinity*. If you decide your prior density for *anything* is infinity, your prior is bad and you should feel bad. You will get ridiculous answers like this.
Natura non facit saltus. The idea that exactly, precisely, 50% of births are male is ridiculous, because 50% is a *single number*, and there are infinitely many real numbers. The prior should assign a large probability that the sex ratio is *close* to 50%, but not that it will be *exactly* equal to 50% and not 50.00000000001%.
This is not how any sensible Bayesian statistician would approach this question. Bayesian hypothesis testing approaches like Lindley's are what happens when you have a frequentist LARPing Bayesian statistics, and trying to construct a hypothesis-testing framework that sounds like what they're familiar with.
But why does the chi-square test give a sensible result? Well, it turns out all the common frequentist procedures (T-tests, chi-square tests, etc.) have a dual, Bayesian interpretation for large samples: You can read the p-value as the probability of a *positive* (rather than negative) effect size. In this case, Lindley's p-value of 1.2% is the correct, Bayesian answer to the question "What's the probability that the sex ratio at birth is actually skewed towards girls?"
But this is a complete disaster of an answer if you interpret it like most people do, as an answer to the question "What's the probability of a perfectly balanced sex ratio?" (0%! The answer is 0%! Nothing is perfectly balanced like that!)
Going back to the original topic: the correct interpretation of Goldin's results is that he's conclusively proven that healthcare will probably not kill you. On the one hand, that's reassuring. On the other hand, it doesn't really answer the question we actually care about: "is healthcare spending cost-effective"?
*that* would require a cost-benefit analysis.
I thoroughly enjoyed this series.
I would love to at some point see Scott engage one of Corey Doctorow's pieces, on any topic (eg. corporate looting: https://doctorow.medium.com/https-pluralistic-net-2024-05-23-spineless-invertebrates-ac092ac1de39)
RE“ but the podcast is called “Most Drugs Are Bad For You”! Someone who listens to podcasts - is this just mistitled?”
The title is quoting Sebastian Brunemeier, CEO of ImmuneAge.
>Suppose you were to take the individuals in our treatment and control groups and randomly re-shuffle them into (fake) treatment and control groups, and compare the difference in the mortality rates between the fake groups. You wouldn't expect to find an effect, but there might some differences just due to random noise. In Appendix Figure A.VII (below) we do this 1000 times, and compare the difference between the real treatment and control groups (our estimated effect from the study) to the distribution of the differences between these fake-groups. This tells us whether the difference between the treatment and control groups that we observe in the study (shown by the red line) is likely due to chance -- the figure below suggests that the answer is no, because it is more extreme than almost all of the fake comparisons.
I don't think this argument works. When I simulate a biased or p-hacked result, the estimated effect is also in the tail of the placebo distribution. So as a test to distinguish between real and spurious results, it doesn't work. See:
https://michaelwiebe.com/blog/2021/01/randinf
Actually, I misunderstood. Goldin is just restating the definition of a p-value (defined from a permutation test instead of the classical way). The permutation distribution is calculated under the null hypothesis (when the treatment has no effect, we can reshuffle the treatment labels and not affect the outcome), so the number of estimates larger than the actual estimate is the p-value (the probability of getting a more extreme result assuming H0).