Comment by jhulla | Hacker News Reader

Comment by jhulla | Hacker News Reader

Coding errors abound in this space. For example, you can go to the CMS.gov to lookup ICD codes.

This is what you get when you lookup the word "finger"

https://www.cms.gov/medicare-coverage-database/staticpages/i...

It contains hundreds of entries.

Here are just a few selected at random:

S61.352S Open bite of right middle finger with damage to nail, sequela S61.353A Open bite of left middle finger with damage to nail, initial encounter S61.353D Open bite of left middle finger with damage to nail, subsequent encounter S61.353S Open bite of left middle finger with damage to nail, sequela S61.354A Open bite of right ring finger with damage to nail, initial encounter S61.354D Open bite of right ring finger with damage to nail, subsequent encounter S61.354S Open bite of right ring finger with damage to nail, sequela S61.355A Open bite of left ring finger with damage to nail, initial encounter S61.355D Open bite of left ring finger with damage to nail, subsequent encounter S61.355S Open bite of left ring finger with damage to nail, sequela S61.356A Open bite of right little finger with damage to nail, initial encounter S61.356D Open bite of right little finger with damage to nail, subsequent encounter S61.356S Open bite of right little finger with damage to nail, sequela S61.357A Open bite of left little finger with damage to nail, initial encounter S61.357D Open bite of left little finger with damage to nail, subsequent encounter S61.357S Open bite of left little finger with damage to nail, sequela S61.358A Open bite of other finger with damage to nail, initial encounter S61.358D Open bite of other finger with damage to nail, subsequent encounter S61.358S Open bite of other finger with damage to nail, sequela S61.359A Open bite of unspecified finger with damage to nail, initial encounter S61.359D Open bite of unspecified finger with damage to nail, subsequent encounter S61.359S Open bite of unspecified finger with damage to nail, sequela

Well, this is the notorious ICD-10 "upgrade" from ICD-9. Which has some "wonderful" codes, e.g. try https://www.google.com/search?q=icd+10+codes+funny

Everyone's unfair favorite is "V97.33XD: Sucked into jet engine, subsequent encounter."

Unfair because that doesn't mean you were stupid enough to do that twice, but had to see a doctor after the first visit.

"W55.41XA: Bitten by pig, initial encounter" is OK, it happens. But "W61.62XD: Struck by duck, subsequent encounter." WTF??? I'm from a hunting family, and I've never heard of smallish birds like ducks harming people in this way. Ditto "W61.12XA: Struck by macaw, initial encounter" and "W59.22XA: Struck By Turtle" (now, being bitten by a big snapping turtle is horrific, but being struck by any turtle???).

But this remains the most bizarre of all I've heard: "V91.07XD: Burn due to water-skis on fire, subsequent encounter."

OK, at least we can laugh about ICD-10. For now.

That is funny.

The wide range of codes is a great setup for coding errors: errors of omission, imprecision, using multiple codes to describe an event when a more accurate single code exists, etc.

In the best case scenario, every doctor/nurse/PA/etc would have a technician follow them around just to manage the coding on EMRs.

OK - I had to confirm the water-skis. You're not kidding:

https://www.cms.gov/medicare-coverage-database/staticpages/i...

Has Monty Python taken over the ICD?

I don't blame you, I should have tried that myself.

I'm still trying to figure out how that could be a legit code. Jet skis, of course. But ... ah, while trying to see if anyone else had a brainstorm, I found a better one that eliminates a lot of possibilities:

V90.27XA: Drowning and submersion due to falling or jumping from burning water-skis, initial encounter

Got that from http://skepticalscalpel.blogspot.com/2011/09/icd-10-codes-dr..., which suggests:

I have come up with an answer. Someone, possibly an actor from the Jackass series of movies, sets up his water skis on milk crates in the back yard. The skis are positioned over a kiddie pool. He mounts the water skis while an accomplice sprays them with gasoline. The skis are lit, and the man jumps or falls from the burning skis, submerging and drowning himself in the process. Luckily there is an ICD-10 code for that.

Maybe we're being trolled? You know, there are trolling motors, but they're (all?) slow, battery powered ones to avoid disturbing fish, not at all suitable for water skiing.

When I paid accident claims, I paid a surprising number of claims for 2nd degree burn on lower leg from motorcycle exhaust pipe. You make me wonder if there is a code specifically for that.

/weird but true

Looks like it's just the burn code: http://www.icd10data.com/ICD10CM/Codes/S00-T88/T20-T25/T24/T...

The funny thing is that we have ultra specific water ski codes, yet can't find a more specific one for "burn by motorcycle"

Pretty much every code added to a code set like the ICD is added because someone, somewhere needed it, and didn't have it in the previous version.

And ICD-10 codes are used for a variety of purposes -- statistical, billing, etc. -- so if its important for one of those uses, its going to get added. Especially since, with the degree of automation preferred now, if it a distinction could reasonably affect any decision in any of the roles that the code set is used for, it needs to be represented in the code set.

And ICD-10 codes are used for a variety of purposes -- statistical

And that's where it goes off the rails, this revision is trying to serve too many masters, in which actual health care is subservient to collection of data for the usual nosy suspects. Did anyone even contemplate the trade-offs?

And even then, look at all those codes for bites of fingers. Are you really insisting they will provide anything other than fancy looking statistics for self-important bureaucrats?

> And even then, look at all those codes for bites of fingers. Are you really insisting they will provide anything other than fancy looking statistics for self-important bureaucrats?

While I don't know the history of the particular codes, a lot of the hyperspecific codes are driven by needs of insurance payers (public and private) and their desire to incorporate elements that would otherwise be identified in (comparatively costly) non-automated review and provide the necessary resolution in coding that they can be identified and distinguished without that review and the associated cost.

Driving down that area of administrative costs is a different thing than providing fancy looking statistics for self-important bureaucrats.

non-automated review and provide the necessary resolution in coding that they can be identified and distinguished without that review and the associated cost.

What type of review - automated or manual - requires the categorical specificity of distinguishing between 359+ different types of marginally varying finger trauma from bites? This is just madness. That anyone would even suggest that that level of specificity was even useful, never mind required, is bureaucratic madness.

Step back from the tree and you can see the forest burning.

Believe it or not, these codes are used to define contracts and aid in payment all the way from patient to broker.

So, pretend I'm a doctor and tell me, will I get paid more for S61.354D Open bite of right ring finger with damage to nail, subsequent encounter

Or S61.355D Open bite of left ring finger with damage to nail, subsequent encounter ???

And with such specificity, why include this code:

S61.359D: Open bite of unspecified finger with damage to nail, subsequent encounter

I'm assuming "other finger" is the one between your little and middle ones, and we'll all sleep easier knowing that, while not found with jhulla's search, your left and right thumbs have not been neglected.

I actually used to pay insurance claims. Sometimes, the question being answeed is how many fingers we are paying for. To know that, we sometimes need specifics.

There were enough fraud claims where someone was wanting to be paid for a third leg amputation that we joked we needed LL (left) R (right) and M (middle) as our designations instead of just L and R.

Generally, you don't get paid an amount for a diagnosis, you get paid for the services you bill.

Whether a payer accepts that the specific services you billed for are medically appropriate and thus pays them depends, often, on the diagnosis they are treating. (And, often, there are three possibilities: the payer pays the claim directly, or the payer requires additional supporting documentation which is manually reviewed before making a decision, or the payer denies payment outright.)

The more detail diagnostic coding provides (and, for that matter, the more detail procedure coding provides), the more an automated system can move cases that would otherwise be in the manual review category (which adds costs for all parties, and delays) to an automated decision (which is quicker and cheaper for all parties.)

> What type of review - automated or manual - requires the categorical specificity of distinguishing between 359+ different types of marginally varying finger trauma from bites?

Quite likely, review for medical necessity of the particular treatments billed for those bites. Which probably have wildly varying costs, and different indication; precision of diagnostic coding reduces the degree of circumstances in which it is necessary to require additional supporting documentation to be sent and have a human review it; manual review adds time and cost to claim processing, and is something everyone -- payers, providers, and patients -- benefits from reducing.

> That anyone would even suggest that that level of specificity was even useful, never mind required, is bureaucratic madness.

Only to the extent that medical insurance is inherently "bureaucratic madness".

They also are used for insurance purposes, how much should a treatment cost given so so diagnosis. ICD-10 attempts to put things in a bit of a flattened hierarchy with prefixes and ranges.

Whether you are "struck by turtle" while playing Mario Kart, or smother yourself with a pillow [1], most codes are derived from recorded events. Then there's the algorithmically appended first, second, etc. encounters.

[1]: http://www.icd10data.com/ICD10CM/Codes/S00-T88/T66-T78/T71-

> Everyone's unfair favorite is "V97.33XD: Sucked into jet engine, subsequent encounter."

Not as unfair as it seems! Someone got sucked into an A-6 Intruder on a carrier with only minor injuries due to his helmet shredding the engine. https://www.youtube.com/watch?v=5jxcSY1AwrM

> Struck by duck

Could mean lovestruck? :)

> Struck By Turtle

Eagles have been known to drop turtles from flight, in order to crack them open. The ancient Greek playwright Aeschylus is supposed to have been killed by a tortoise in this manner.

Damn, I forgot that. So it's not completely insane.

Except in that, as I note elsewhere in this subthread, it has no more to do with providing better healthcare than the myriad insane variations of finger injuries that started this subthread.

While it seems overkill, it is unambiguous. Errors are easy to make in medicine.

When I was a neuro tech, I remember doing a nerve conduction study on a patient, on one arm only, and the patient had no abnormality, when went into my report. The doctor doing the final report for the record copy/pasted from his list of stanzas, and forgot to switch 'left' in his copypasta to 'right', which I caught. This particular doctor was perfectly competent; it's just that there's a thousand ways to make errors like this (and yes, copying from a list of stanzas isn't the best method, but neither is individually typing out the same diagnosis paragraphs again and again).

Being unambigous with which side of the body and which body part is very important in medicine.

OK, that's a good argument, although the system of coding each as a separate code without a "system" is crazy (for a bit I thought they were at least using Sinister/Dexter, but that didn't seem to be the case). But probably required due to legacy systems.

Although I'd say this is still an experiment in which will result in the fewest mistakes. Maybe someone should have paid Kaiser or the like to try this out first to find out some of the consequences?

The lack of a system to the coding is crazy. You find all sorts of nonsense next to each other. Unfortunately, there is a hell of a lot of politics in medicine - infrastructural change is extraordinarily difficult.

Moderate to large hospitals generally have full-time coders, whose job is to take the records generated during a patient's stay and figure out which codes apply. There's a lot of money in what's called 'up-coding', or using a similar code that pays more.

While lower population areas are losing their only hospitals due to "one size fits all" stuff like this (see e.g. http://www.usatoday.com/story/news/nation/2014/11/12/rural-h...), and in my "moderate" sized market, all the doctors I know of are now working for "the man", one of the two big hospital complexes, or otherwise outsourcing this sort of thing to them.

You're right, that's a huge problem. Critical access hospitals are really struggling, and many smaller hospitals and clinics are getting swallowed up by big hospitals somewhat nearby. A side effect of that is that the newly-acquired sites are often migrated to the records system of the new owner, which is creating silos of medical records. The problem is that the silos still aren't connected to one another, which means the vast majority of hospitals still have to transfer records through fax. I really wish I was joking about that.

Reply on news.ycombinator.com