Hacker News new | past | comments | ask | show | jobs | submit login
Rare and strange ICD-10 codes (johndcook.com)
68 points by zdw 50 days ago | hide | past | web | favorite | 50 comments

These are too funny: https://www.empr.com/home/features/the-strangest-and-most-ob...

Choice selection:

V91.07X - Burn Due to Water Skis on Fire(?!)

W61.42XA. Struck By Turkey, initial encounter. If a duck is involved, there's a code for that too. (W61.62X)

R46.1. Bizarre Personal Appearance.

R46.1 is more sensible than it might seem. Personal appearance that's far outside the norm may indicate a neurological deficit or psychiatric condition.

Either that or they work in art, fashion or entertainment.

"Bizarre" probably refers to something both more extreme than Marilyn Manson or face tattoos and far less interesting/well-coordinated.

At the clinic mom used to work at, they used "FLK" instead of R46.1, which stood for "Funny Looking Kid". It was shorthand for "Something doesn't look right with this patient, and we need to find out why"

Yup. My grandma was a nurse and she used this. Many genetic disorders have characteristic facial features, such as the Pierre Robinson sequence.

My favorite: T63.012D: Toxic effect of rattlesnake venom, intentional self-harm, subsequent encounter

So basically you have a case where someone not only intentionally wanted to harm themselves with rattlesnake venom once, but at least twice!

> So basically you have a case where someone not only intentionally wanted to harm themselves with rattlesnake venom once, but at least twice!

No, you misunderstand the terminology. "Subsequent encounter" means with the doctor not with the rattlesnake. AKA followup care during or after recovery.

> No, you misunderstand the terminology. "Subsequent encounter" means with the doctor not with the rattlesnake

You can reference ICD codes with the schema.org/code property of schema.org/MedicalEntity and subclasses. https://schema.org/docs/meddocs.html

"Subsequent encounter" is poorly defined. IMHO, there should be a code for this.

> "Subsequent encounter" is poorly defined.

"Poorly defined" is poorly defined. Explanations of when to use the D make perfect sense to me.

"The 7th character for “subsequent encounter” is to be used for all encounters after the patient has received active treatment of the condition and is receiving routine care for the condition during the healing or recovery phase. Examples of subsequent encounters include cast change or removal, x-ray to check healing status of a fracture, removal of external or internal fixation device, medication adjustments, and other aftercare and follow-up visits following active treatment of the injury or condition. Encounters for rehabilitation, such as physical and occupational therapy, are another example of the use of the “subsequent encounter” 7th character. For aftercare following an injury, the acute injury code should be assigned with the 7th character for subsequent encounter."

Some people (actually, just heard of one) inject themselves with snake poison for various benefits.

This guy has been doing it for 30 years now: https://www.youtube.com/watch?v=AcbqB0pFRPA

And also a longer but older documentary-short about the same person: https://www.youtube.com/watch?v=8q_m-rDUNw0

Edit: Apparently, the same guy (Steve Ludwin) actually has his own snake show on VICE nowadays. You can see all videos about him (and ton of snakes) here: https://video.vice.com/en_uk/topic/steve-ludwin

Like cutting near the bite and trying to suck out blood?

Why would I suck blood out of a venomous snake? :3

At least it isn't poisonous.

Y92.253 - Opera House as the Place of Occurrence of the External Cause

Operas are so painful to watch there's even a medical code for it.

No discussion of strange ICD-10 codes can omit this amazing book: https://www.icd10illustrated.com/

Looking at secstate's post reminds me of one of the neater things (at least to me) in FHIR models, that a patient can also represent an animal. I have only worked with HIEs and EHRs for humans so I don't know what veterinarians' EHRs may be using. Do they use race and ethnicity fields for animal type and breed?

Anyway, my only other IDC related story is that of going from ICD-9 to ICD-10, when decimals came into play. We had several projects where we had to go back and do global replaces on ICD-9 codes to play nicely with the ICD-10 codes.

>Anyway, my only other IDC related story is that of going from ICD-9 to ICD-10, when decimals came into play. We had several projects where we had to go back and do global replaces on ICD-9 codes to play nicely with the ICD-10 codes.

I like how ICD 9 and 10 don't use mutually disjoint sets of codes so there's a tiny overlap where the same code is present in both but means very different things. But only a tiny overlap so a QA cycle probably won't catch the issue.

Animal was removed from the base Patient resource spec in r4.

I did see that after I posted last night. It looks like they are now an extension.


"V97.33XA: Sucked into jet engine, initial encounter"

There's a fairly well known incident where a sailor was sucked into the intake of an A6 Intruder, and lived: https://youtu.be/AF55oyAJDBk

That happened to a F-16 crewchief at my base. He went to duck under the intake instead of walking around like he was supposed to, and the airflow grabbed him by the field jacket and pulled him in. He did not live.

Not everything that has happened once can have its own code. There must be things that happen more frequently that don't have their own code.

I wasn't seriously suggesting they should. Just noting it is possible to survive an "initial encounter". I wouldn't have guessed so had I not seen that prior.

"Initial encounter" means the first time seeing the medical professional and not the first time the accident happened.

Ahh, thanks...that makes sense. Still holds though. No coding an initial encounter if your first stop is a coroner.

"Is a swimming injury in a prison pool medically different than a swimming injury in a YMCA pool?"

They are medically but they will be covered differently by insurance. A private insurance company would probably not cover a swimming injury in a prison pool as they would pass that off to the government

Every time I nark on ICD-9, -10, -11, someone claims there's some hidden benefit as yet unrevealed to us mortals. Just clap louder and Tinkerbell will fly.

Is this true? Even in finely contrived circumstances using the biggest suspensions of disbelief?

I've looked a few times, have found no studies showing ICD-* improve patient outcomes, reduce costs, or any other healthcare benefit. (Consultants and insurers do okay, though.)

When will we learn? This is just another rehash of ontology vs folksonomy, symbolic vs machine learning AI, data dictionaries vs screen scrapping, etc.

Source: Implemented 5 health exchanges, was active during transition from ICD-9 to ICD-10.

The benefits seem pretty obvious to me. I work for a hospital and a few years ago we discovered that we had a serious problem with sepsis. Every hospital struggles with sepsis, but our sepsis mortality rates were higher than our peers. As a result, we invested significant resources into improving how we handle sepsis, and our mortality rates have gone down.

None of that would have happened if ICD codes didn't exist.

Frankly, ICD codes aren't even there to improve outcomes or reduce costs. They exist primarily for billing purposes. Anything else we get out of them is a bonus.

They exist primarily for billing purposes. Anything else we get out of them is a bonus.

Yes. If you work in insurance and have the right code, you know you can pay it without pending the claim and sending a letter requesting additional information. If you have the medical records but no code, you may have to ask for additional info to determine if it is covered.

I'm still struggling with the notion that an insurer, and not the health care providers, now determine medical necessity.

You can get a doctor to write a letter of medical necessity. Sometimes that will get it covered.

But, yeah, the US healthcare system has issues.

Medical necessity isn't a binary condition. Two providers presented with the same patient chart will frequently come to different conclusions. Insurers (both private and government) do employ doctors to perform medical necessity reviews.

Very interesting. Thanks.

Is that because sepsis diagnosis weren't being captured in the transcribed notes?

Is this kinda related to the checklist manifestos, a la Atul Gawanda?

The part that required ICD codes was when we compared our mortality rates to peer hospitals. You could theoretically do this without codes, but a standardized way to classify patients makes it orders of magnitude easier. There are companies that ingest data from hospitals and then give it back with comparisons to other hospitals. That's how we were able to identify the problem.

Hospitals aren't flush with a lot of spare cash, so if doing this required a bunch of complicated text processing then it probably wouldn't have happened, and trying to compare our notes to notes from other institutions just multiplies the nightmare. With ICD codes we all speak the same language.

When I was working on a project that involved ICD codes, one thing that kept cropping up in my mind was even how consistently the codes would be used.

I other fields of endeavor, when I see such voluminous, nuanced classification codes, the real result is that it becomes confusing as to 1) what code any given subject should be classified as, 2) the people doing the classification cannot know all the codes available and therefore will classify to something close that they know... even if a closer code exists, 3) there can be general disagreement on what any given code means and where dividing lines sit between similar codes, and 4) in medical offices, where several codes may be available and could reasonably apply, I've seen choices made based on the financial differences between the codes where they existed (not to mention that all the previous problems in coding a condition for the practitioner is also a problem for those deciding how codes may/may not be covered or paid for).

Given all that, it seems to me any coding system like this, which apparently fails to account for the limits of human judgement and usage patterns, will ultimately fail. Naturally, I assume that the goal in constructing such a system is the collection of nuanced data, but expect the result to be a muddle of that data as the system is misused.

And based on the medical offices I have worked with over the years, there is nothing about the medical profession that will mitigate those issues.

You can always find occasional errors and discrepancies but the vast majority of coding is done correctly. Medical coders go through an extensive training and certification process.

I understand that there are training/certification processes for medical coders and I expect that to certainly help. But is there a basis to the claim that errors are occasional and that coding is done correctly? Or is that just an anecdotal assertion? My observations are just anecdotal, but do come from a fair amount of on the ground experiences.

I expect their to be limits on the effectiveness of that sort of thing. My original comment dealt with the complexity issue, but there's how well that training/certification ages over time... such that common codes are handled well due to daily reinforcement, but less common codes will tend to fare worse. Unless there are required maintenance trainings and re-certification requirements (I'm not aware of that either way), I expect on average for there to be issues in practice.

While there has been medical procedure coding of some sort for a very long time, this rather more complex coding that we're generally discussing with this story is still in rather early days. Nonetheless, I would be interested to see an actual study on how the complexity of coding systems like ICD fare at actually capturing the information they try to; I just tried searching the ACM Digital Library, but didn't find anything... though that could just be weakness of my search terms.

It's not just an anecdotal assertion. I know that public and private payers conduct audits and reviews on a subset of claims to verify accurate coding. Private insurers don't release their audit results but if you search around you can probably find something from CMS for Medicare claims.

Certified medical coders are required to complete several hours of additional training every year in order to maintain their status.

ICD-9 had been in use for many years. ICD-10 is only slightly more complex. Most provider organizations typically only use a small subset of the available codes so it's not too hard to keep track of the common ones.

Hidden benefits have nothing to do with it. ICD is primarily targeted toward billing (claims). As long as providers are paid by separate payers (regardless of whether the payers are private insurers or some sort of government single payer system) there has to be some sort of consistent coding system to describe patient conditions.

When you have a chronic, relatively rare disease and are moving countries.

Mmmm, love me some ICD code hilarity

Less funny is the human condition reduced to billing codes as only available option in the age of electronic health records moving towards precision medicine.

What if clinicians were able to describe what they observed at a granularity they found appropriate in terms they went to great (and expensive) lengths to learn?[1]

What if untrained people could describe patients in a way that could be equally meaningful (if not better informed)?[2]

What if instead of dead ending on an insurance ledger, the codes describing your medical condition were part and parcel of an extensible logical framework over which reasoning and inference engines are run (okay fine... call them AI & ML if you are feeling buzz word deficient)[3]

What if using these alternative codes afforded way of leveraging that science stuff grad students have been inflicting on little critters since time immemorial?[4]

Note: Most everything is available on GitHub and work is ongoing ...

1 [https://hpo.jax.org/app/] 2 [https://hpo.jax.org/app/help/layperson] 3 [https://en.wikipedia.org/wiki/Ontology_(information_science)] 4 [https://github.com/obophenotype]

ICD is primarily for billing and analytics, but it's not the only option. SNOMED CT is usually a better fit for clinical use cases.

You might also enjoy this: https://twitter.com/EPICEMRparody

There's a whole book of these: https://www.icd10illustrated.com/products/book

ICD-11 will also have codes for TCM, so no end to strange codes

You can already browse the draft here: https://icd.who.int/browse11/l-m/en

Some examples with interesting names:

SD24 Frequent protrusion of tongue disorder (TM1) https://icd.who.int/browse11/l-m/en#/http%3a%2f%2fid.who.int...

SD70 Qi goiter disorder (TM1) https://icd.who.int/browse11/l-m/en#/http%3a%2f%2fid.who.int...

SE90 Qi deficiency pattern (TM1) https://icd.who.int/browse11/l-m/en#/http%3a%2f%2fid.who.int...

SG29 Triple energizer meridian pattern (TM1) https://icd.who.int/browse11/l-m/en#/http%3a%2f%2fid.who.int...

Traditional Chinese Medicine? That would be interesting.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact