[derefr]

Short-term point sampling can do you more harm than good—this is as true in medicine as it is in investing.

There's nothing wrong with getting an MRI every year if you aren't actually looking at it on its own, but rather diffing it in a time-series, and extracting chartable statistics from it, to see if anything in your body is getting progressively [adjective]er each year.

This approach won't actually tell you if you've got six months to live, but it's not supposed to. It's to show you that your liver has been getting 10% more scarred over with each visit, so—with evidence!—you'd better lay off the booze already. Or to find something like arteriosclerosis before it presents clinically.

[jacquesm]

You are tying up an MRI machine that might be put to better use.

[dionidium]

You're implying that MRI machines are 100% utilized and that any usage prevents some other usage. Surely that's not actually the case.

[jacquesm]

You don't need them to be 100% utilized, it's not the machines that are the problem but the technicians operating them and the doctors interpreting the results and so on.

There is a whole team clustered around an MRI machine and tying up the machine is effectively tying up the team.

If everybody got an MRI every year on the off chance they might have something then the machines would not be available to those that actually need them (because they have symptoms that require an MRI to narrow in on a cause).

And that's besides the whole fall-out further down the chain in case something does show up that does not generate symptoms and even further down the chain where probably un-necessary operations will be performed on symptom-less patients.

Doing the right amount of screening is a really hard problem. Getting an MRI done every year just because you can afford it is probably not even near the right amount.

[derefr]

The idea, here, would be to make MRIs both cheaper, safer, and less trusted for their predictive power.

Imagine if you could get an MRI the way you currently get a blood pressure test or ultrasound or x-ray. Imagine a simplified MRI machine in every GP's office 40 years from now. Imagine the machine having a networked software layer integrated that doesn't even allow the raw static-image data out to the doctors, but instead just reports out weighted-average feature-change velocity heatmaps.

Presumably, the only people getting an MRI done every year now are the ridiculously rich, who could buy an MRI machine of their own, and a hospital wing to put it in. But the theory behind getting one isn't bad. It's just scarcity, and the effect scarcity has on the level of trust we put in the few samples we can beg, borrow or steal.

[jacquesm]

An MRI is not some magical device, it's more like an advanced way to do tomography and interpreting the results and preparing you for the scan is non-trivial and likely will remain so even if the cost of the MRI machine drops to '0'.

So you're going to need the crew of attendants and interpreters with associated costs regardless of the price of the machine.

What cheap MRI machines will do is marginally reduce the cost of a scan, which would be a good thing but over the life of the machine it is the people that make up the bigger chunk of the cost.

Mobile MRI machines are currently truck sized, something that could be moved around in a suitcase and set up on the spot would be a game changer in the third world (as would be cheaper MRI machines to begin with), but for the first world it wouldn't be so much of a change.

What did change things dramatically was the reduction in the cost of ultra-sound but the complexity of an ultra-sound scanner compared to an MRI machine is orders of magnitude less, interpreting the results is still not the job of the tech administering the scan (though in the case of say a look at an infant in utero even a lay person can make out what is what and that's already useful).

Healthcare costs are an interesting subject of study, you could easily make a career out of that.