Responses to “Why are doctors so unwilling to run tests?”

The original post was surprisingly controversial!

A week ago, after hiccupping for ~70 hours for still-unknown reasons, I published an essay criticizing doctors specifically and the medical establishment more generally for not being willing to prescribe more diagnostic tests.

First, thank you everyone who shared kind words and sent in home remedies for hiccups. None of them worked, but I appreciate you nonetheless. The hiccups DID eventually stop, still no idea what caused them or why, and my primary care provider and I are in wait-and-see mode. The H. Pylori test and the blood tests all turned up negative.

The essay about testing got shared around a bit, and sparked a lot of interesting conversation among friends. Notably, basically everyone who works in medicine (doctors, nurses, insurers, etc) had a strong reaction against the piece. Here are some responses, and my thoughts on those responses.

The medical system is overburdened.

This was the most common complaint to come up in conversation.

The basic idea is that we should discourage doctors from prescribing tests that are ‘unnecessary’ (and implicitly, we should have strong conceptions of what makes a test ‘necessary’) because there are not enough resources in the broader medical system. We want to ensure that the medical system has capacity for folks with more life threatening diseases — cancer, stroke, Alzheimer’s — and overprescribing tests for people who are approximately fine will end up reducing that capacity. Sometimes, this is framed in terms of class differences. Specifically, hospitals and practices in poorer areas do not have enough testing capacity, and overprescribing in richer areas will result in fewer available tests that end up being more expensive.

I understand the impulse here. During COVID, we really did need to ration tests because there was a limited supply of them. And we wanted to ensure that regular testing capacity was available for those who were most at risk. But I think many of the doctors who push back with this line of reasoning are missing some second order economic incentives.

The high demand for COVID testing led to higher supply of COVID test providers until, eventually, getting COVID tests was dirt cheap. More generally, there is not a finite pool of tests. We can create more reagents and produce more syringes and set up more clinics if there is demand for these things. If every doctor on the planet decided to make stool tests a part of their annual physical, the price of any given stool test should go down in the long term as suppliers flood in to meet demand. The same is true of more expensive tests like MRIs. They are expensive now because we do not mass produce these things. But we do not mass produce these things because there isn’t enough demand to justify figuring out how to do it. And if the reason there isn’t enough demand is because doctors are worried that MRIs are expensive, well, there’s a bit of circular logic embedded in the premise.

Interestingly, many of the folks who pushed back against the idea of prescribing more tests for non-critical cases were, simultaneously, very gung-ho about the idea of building more luxury housing as a method to alleviate the housing crisis. It seemed intuitive that building more luxury housing would end up easing the burden on lower income housing because supply is supply is supply. The same intuition did not transfer to tests. Just make more, and everything will become cheaper!

I think you could argue that supply is inelastic because healthcare is highly regulated or because certain tests are hard to mass produce, but I kind of don’t believe either of these things. Unless there is literal anti-competitive cartel behavior on behalf of the government, increasing demand should result in more shops trying to jump through healthcare hoops and more human ingenuity going towards producing cheaper tests. As an example of the latter, an MRI machine will run a hospital somewhere between $1-3m, because there are only like 5 companies that make them and they haven’t innovated on the core product in decades. Meanwhile, this startup has figured out how to make a decent MRI for 50k. If more doctors prescribed more MRIs, more hospitals and medical practices would start looking to procure an MRI machine, and then would probably pick up the $50k one!

Patient burden

Many folks felt that over testing would burden the patient.

Jacob

Surprised nobody mentioned what is probably the central issue with overtesting, which is iatrogenesis, either mild (making people constantly worried and running to the doctor for tests for reassurance) or significant (unnecessary biopsies, etc). Like it’s not that it’s wasteful to test everyone for everything, it is reasonably likely to cause net harm

From the DMs

telling millions of people “hey, you might have cancer, we see some lumps in your scans, but don’t worry about it, it’s probably nothing, don’t give it a second thought until your next scan next year” holds up well theoretically but causes a lot of harm/distress…the “lot of harm/distress” quantification -- there’s plenty of evidence that medical anxiety is a real thing, contributes to real physical harm and distress, and that’s ignoring all the other factors (cost on patients, cost on healthcare system, invasiveness, etc)

This was the second most common push back. We don’t want to give a lot of tests that will mostly result in a lot of false positives, because the patients falsely diagnosed will 1) suffer from medical anxiety and 2) potentially undergo treatment that they didn’t need, side effects and all.

I found this totally uncompelling.

The most common refrain was ‘false positives for cancer,’ so we can roll with that. Let’s say a patient gets a false positive for a cancer diagnosis. Obviously, up front, we know it’s a false positive, but the patient does not. Using some

. ݁₊ ⊹ . ݁˖ .Bayesian analysis . ݁₊ ⊹ . ݁˖ .

we can determine the likelihood the patient has cancer given the positive test. Because cancer incidence base rate is pretty low, the likelihood is probably like 5% even if the test is 99% accurate. That is, most people will still get negative results, and even the people who get positive results are pretty unlikely to have cancer.¹

I was sent an article by a physician that was mixed about full body screening. In order to make the case against full body screening, this doctor writes:

Let’s do the math: For every 100 people scanned, 30 undergo follow-up investigations. Of those 30, roughly 1 has cancer. The other 29 endured anxiety, additional procedures, and potential complications for nothing.

This is the fundamental problem: whole-body MRI is so sensitive it finds almost everything. The challenge is distinguishing signal from noise—and that distinction often requires invasive testing to know for sure.

If I told you that there was a 3% absolute risk of you having cancer, you would rightfully want to get that shit checked out and take precautions! Things that can result in a 3% absolute risk of cancer include:

• Heavy alcohol use for over a decade

• High dose radiation exposure

• Long term immunosuppression

All of these people should regularly get tested. If you get a positive test, you’re about as likely to have cancer as someone who got a glancing blow from Chernobyl. At the very least, you would want to get another test done, which is basically what we already encourage patients to do when they get a hit by a potentially life changing diagnosis. Or, put another way, I think it is totally fine for 29 people to have some additional medical anxiety if it means catching a malignant cancer.

Moreover, it is incredibly paternalistic to assume that patients shouldn’t get checked out regularly because some of them may naturally follow up on the results of those tests. And the cost of that paternalism is measured in lives — people who could have spotted an actual cancer diagnosis early if they just tested more.

I know that the medical system needs to think about tradeoffs at the population level, but if you assume cost is not an issue, in my mind there would have to be an absolutely massive amount of ‘patient anxiety’ to justify even a single preventable death. If we fiated that MRIs were essentially free, would we really prevent people from taking them because of patient anxiety? If not, what are we even talking about?

Note too that there are certain tests that we do just do constantly. As I wrote in the original piece:

Every trip to the ER almost always results in a suite of tests, as does every physical. Hell, they’ll take your blood pressure and temp and oxygen levels before you even see a doctor. They’re thrilled, they’re practically jumping for joy to take your blood pressure / temp / O2. I don’t think these tests are all that different from a blood panel or a stool test. And I just refuse to believe that we’ve coincidentally settled on the exact optimal amount of testing for every person, which just so happens to be ‘once a year-ish.’

What makes blood pressure testing or O2 testing different from, like, a stool test, at least when it comes to patient anxiety? In my mind, the only differences are cultural acceptance among the medical establishment and cost. Whenever a test becomes cheap enough we magically stop caring about patient anxiety entirely and instead just order more tests.

I don’t mean to be dismissive of the folks who argue for patient anxiety. But it seems to me like this is a case of gigabraining — mistaking second order effects for first order effects in order to have a convenient just-so explanation for an unintuitive and contrarian take. I think the amount that we should build our medical system around minimizing patient anxiety is close to a rounding error, and when weighed against actual lives I think most people would agree.

The test results won’t matter anyway

Scott Alexander writes:

There’s a saying - never do a test unless some possible test outcome will change your plan.

Your doctor could order CRP - a nonspecific test that’s elevated whenever anything is wrong - but if they don’t have some plan for what they would do when CRP is high, what’s the point?

Level 50 Lapras adds:

The analogy in the software case would be where you have no ability to change the system anyway, so knowing the exact cause is pointless.

I’m a bit confused. Why are there tests that only show ‘general malaise’? I agree that that seems like a bad test.

Maybe I’m just too much of a software engineer, but to me, gathering data is upstream of making a diagnosis. In the same way that you may pepper a codebase with logs to find a bug, you do the testing so you can figure out what the diagnosis ought to be. In many situations you can take a reasonable first stab at a diagnosis based on how the patient describes their symptoms (which is just first line data gathering).

But if you can’t make a confident diagnosis because the symptoms are non-specific or because the previous diagnosis can be obviously ruled out, you should gather more data, which will then change your plan.

Just to be really specific about this, in the original essay, I mentioned that the doctor really didn’t want me to take an H. Pylori test. This is a test that, if it turned up positive, would immediately have given a clear diagnosis with a clear change to the treatment. There wasn’t a single person who could mount a good case for why I shouldn’t have had the H. Pylori test.

More generally, I got this fatalistic attitude from a lot of doctors and medical professionals. “I shouldn’t order the test because it’s not going to matter.” I don’t get it, what are they teaching in the med schools? Order the tests, it may matter!

Certificate of Need Laws

My cofounder, Cliff, previously used to work in healthcare, and he came out as a very strong supporter of the overall essay. His overall point was that our government tries way too hard to cap market forces when it comes to healthcare, often leading to really stupid and preventable shortages, and spent a while talking about Certificate of Need Laws and how stupid these things are.

I hadn’t heard of this before, but having read about it, I agree — they are stupid!

Certificate of Need Laws (literally CON laws, you can’t make this up) make it so that you cannot build more medical capacity in a region unless a federal or state agency affirms that that region needs more medical capacity. That is, a hospital needs to go to the state and say, “hey, we want to get a new MRI machine, we can afford to buy it and we have a willing seller” and then the state can just be like “no.” From wiki:

CONs have been criticized for granting monopoly privileges to existing hospitals and healthcare facilities, thereby driving down the number of hospitals and hospital beds in a community. One study found that CON laws resulted in 50% fewer hospitals per 100,000 persons, and 12% fewer beds at the typical hospital.

Believe it or not, it gets worse. A lot of state regulators aren’t experts in public health, obviously. So when they get a request to increase medical capacity at a hospital, they will seek out experts to help advise them on whether or not a particular proposal is reasonable. Who do they ask? Other hospitals. This is a massive conflict of interest! “Well, currently all the people who live in the boonies drive 4 hours to our hospital to get their MRI scans…so they don’t really need an MRI machine out there, in that other hospital branch with the other insurance network. They are perfectly well served by our hospital in our insurance network 😁😁😁”

CONs are sometimes sold in bankruptcy as an asset,[9] and the CON requirement is sometimes used by competitors to block the reopening of existing hospitals.[10] In June 2023, a Tennessee administrative law judge blocked the opening of a new hospital in Rutherford County by Vanderbilt University. The state had initially approved the hospital and granted it a certificate of need. But three existing providers intervened, claiming that there was not a need for another facility in the area.

What the fuck!

In 2018, the Department of Health and Human Services recommended that states repeal their CON laws because they are a significant reason for increasing medical costs, and they reduce patients’ healthcare choices

There are still 35 states that have these on the books, by the way. I’ve become markedly less libertarian over the last ~2 years or so, but wow does this make me want to reboot the entirety of our legislative code.

I guess maybe it does make sense to reduce testing when you have regulatory capture that prevents more supply from meeting demand.

Anecdotes

Basically every doctor I spoke to had some justification for the existing system. But there were a lot of people who weren’t doctors who commented, and their stories were all some variant of mine.

Fedaiken:

I feel your frustration here. I had a medical issue years ago where doctor after doctor just dismissed me and I came to the conclusion that you have to own your medical outcomes and direct the doctors just as you would any other service employee. Much harder as the domain is way more complex and the consequences very real. However, my experience demonstrated to me, that only I cared enough about my health to really guide the process. And that means demanding tests and not taking no for an answer. IMO

KSaucy:

This hits close to home for me because i once went to the ER for stomach pain. The ambulance ($3k bill) was like “probably the taco bell” and the triage team was... nonresponsive? To be fair it was 3am but i was screaming, begging for help, explaining that i was in the worst pain of my life and if they couldnt treat me immediately i at least needed painkillers. I was told to go outside and get some at a gas station- which i tried to do and failed

Anyway it was a burst appendix. I was literally 24 hours to death, in a hospital, and still could have died that day

The takeaway from all of this is that you CANNOT trust in the medical system on your own. You NEED an advocate, even as an adult. The system is overworked and being financially preyed upon by people outside of it.

Luke:

I’ll never forget the time I went to a gastroenterologist with inexplicable stomach pains and he was like yeah we don’t know what it is, probably IBS (which means we don’t know what it is, but your stomach hurts, I guess), though it could be IBD (stomach hurts and we DO know why and can help), but it’s probably not IBD based on random things you’ve said or something. And I asked if there was a test he was like yes, but I don’t think it’s IBD, so it would be a waste. And then he said, I swear to God, that I didn’t seem to be the type of person to waste the time and effort required to get a test done.

Genuinely, I think some doctors are just... Lazy? They know if you get tests done, they have more work to do.

But then I can’t really explain why no doctors ever bothered to test my iron levels despite my complaining about fatigue for years. Men don’t get iron deficiency is the consensus, I guess? Well, I paid out of pocket and found out I do.

Daniel:

I had a similar experience recently with some stomach pain. The doctor wasn’t really interested in why I had the stomach pain and just prescribed some anti reflux meds. Felt like it was just a guess a check approach.

The anti reflux meds didn’t work, but my stomach pain did eventually go away, so it does lend credibility to just do the “first line treatment”/do nothing approach, but definitely frustrating from the patient perspective.

And many more. I know that the plural of anecdote is not data, but it’s not not data.

Hiccups

After all of my complaining about tests and how my special snowflake hiccups needed to be analyzed more deeply by the best doctors in the world, the most embarrassing thing happened: the hiccups stopped. It’s been about a week since they stopped, so fingers crossed they are gone for good.

It seems to me that the most likely proximal reason they stopped was the baclofen. I took like two doses of the stuff, it knocked me out both times, and then on the third day I woke up and no hiccups.

It gets worse: the tests that I demanded both turned up negative. So not only did the hiccups go away, they went away without us ever figuring out why they started in the first place. And while I’m still curious about what exactly happened, a) it’s basically impossible to justify getting more testing done at this point, b) it’s not clear that anything would even show up, and c) I have very limited interest in spending a lot of time doing tests for things that aren’t actively causing problems.

It’s worth noting that I very purposely did not follow the doctor’s prescription — 10mg of baclofen every 8 hours is enough to tranquilize a horse, and they wanted me to take this stuff for a week. According to Claude and multiple patient studies, it has withdrawal effects, which is something that the doctor flatly denied was true. The doctors had also prescribed a bunch of gerd meds for like 4 weeks, and I also stopped those way early since they were not working and giving me diarrhea. And after the fact many other doctors were very confused by the baclofen prescription, arguing that I should’ve been prescribed something else.

But whining about dosages or whatever is cope. In the original post, my friend David said:

In your case, that discomfort requirement is clearly met. But that doesn’t mean jumping straight to a test is best for society as a whole -- the symptoms may resolve on their own / with medication before you’d even get the results back, the test may/may not find anything, etc.

“Do the thing cheapest/least resource intensive thing that’s likely to work first, then move on to the alternative.”

David was right, the hiccups went away with third line treatment. I could’ve gotten more tests and waited weeks to figure out exactly the cause before taking any meds; but that would’ve been worse for me and more expensive than just taking the baclofen that the doctor prescribed.

Does that mean that getting more tests is a waste?

No, I think it is still worthwhile to test. I don’t think any of the justifications for more testing depend on whether or not symptoms generally resolve on their own.

My ideal world is one where we shift medical care to being proactive instead of reactive. Testing more is a necessary part of being in a proactive regime. Everyone in the medical system is aware of and values preventative care, whether that’s an annual teeth cleaning or a regular physical. Expanding the depth of that care is a benefit measured in terms of current and future lives. Under this framing, the fact that symptoms may resolve has little to do with our ability to catch diseases before they become more serious, or our ability to leverage more data to figure out root causes for more diseases.

When a patient does show symptoms, I think following first line and second line treatments are generally fine. But the bout of hiccups underscored for me just how hesitant doctors are to order tests even in cases where they have literally nothing else to go off because first and second line treatments have failed.

In a previous life, I trained diagnostic neural networks for rare diseases. One common failure mode of such models is that they will simply say that every patient is healthy, regardless of the actual data, because 99.9% of the time that is the correct answer. If a disease is rare, it’s easier for the model to just cheat than to try and learn some complicated pattern matching algorithm that lets it properly handle the diagnoses that actually matter.

I think doctors are stuck in the same sort of failure mode.

It’s an engrained cultural norm to do nothing. If you start from the premise that the symptoms will always just go away eventually, you end up doing nothing. The Hippocratic Oath biases everyone to doing nothing. I think this actively harms our ability to understand disease and treat patients. We are living through an era where patient satisfaction with the medical system is at all time lows, and where more and more people have chronic diseases that don’t just go away. A medical establishment that lacks curiosity as part of its standard operating procedure is going to fail in this environment every time.

I still think doctors should feel comfortable ordering more tests. Data gathering is good and necessary.

1. if this doesn’t intuitively make sense, drop this article and read this one instead.