… is confusing.
I meant to write about the use of generative AI (misleading name for Large Language Models or LLMs), only to stumble across research into using them to generate billing codes, which go into a patient’s permanent records, and OMG, they think getting a third of them correct is acceptable. That’s two-thirds of them wrong. And in other research, it’s been shown that billing-code errors tend to occur most often in complicated cases, which, presumably includes whole-body disorders like XLH. Sigh.
Anyway, I’ll be talking later about the garbage-in/garbage-out aspects of generative AI in the medical context over the course of a few essays after my summer break (it’s a bigger topic than I’d first realized), but for today, I’d like to talk about an issue that will likely add to poor medical care due to non-human review of medical records: the proliferation of names for medical conditions.
The chronic hypophosphatemia community has been dealing with the naming issue since 1980 or thereabouts. That’s when it became fairly widely understood that what was previously called Vitamin D Resistant Rickets had very little to do with Vitamin D, and instead was a form of hypophosphatemia, ergo the new name X-linked Hypophosphatemia. Somewhere along the line, there was an attempt to include the autosomal population, ultra-rare as it is, with the name “familial hypophosphatemia.” I’ve also seen “FGF23-mediated hypophosphatemia,” which EXcludes the genetic hypophosphatemia that doesn’t feature excessive FGF23, and INcludes tumor-induced osteomalacia, so as to share only patients who share the same root cause of the phosphate wasting. Some of the names also still have “rickets” attached to the end, but I’m going to pretend they don’t, so my blood pressure doesn’t spike.
So how do we get a consensus on the most accurate and most useful name, so we can make medical records uniform and ensure that the readers (human or otherwise) all have the same baseline understanding of the condition the name describes? The problem is, in part, that there’s no real equivalent of a single medical dictionary that can act as an arbiter of the “correct” meaning of medical terms, like Merriam-Webster does for (American) English or the Chicago Manual of Style does for general non-fiction writing.
I suppose the Online Mendelian Inheritance in Man website (OMIM) comes close to a definitive authority for genetic disorders, but as best I can tell, the sources for choosing names are extremely dated (the citation for XLH is from 2009, before it was widely known outside the patient community that it’s not a purely pediatric disorder), and there’s no public/transparent process for correcting the names. XLH is still called XLHRD for “hypophosphatemic rickets, x-linked dominant” even though it’s widely known now (not in 2009) that the condition is far more than a pediatric disorder involving the growth plates. Without some sort of process for updating the entries, OMIM will continue to become more and more outdated, especially as specific genes are discovered for conditions that were known to be genetic, but the exact gene location hadn’t been identified. Case in point: Generalized Arterial Calcification in Infancy is still called GACI in this database, even though experts in the condition use a diagnosis of ENPP1 deficiency.
What we need, essentially, is a cross between the OMIM and the ICD billing code system, which covers all medical conditions (not just genetic ones), and, most importantly, is kept current, with input from scientists and patients with the relevant expertise in specific conditions (not just the general field). A sort of peer review process, you could say.
I know people get tired of hearing me say that words matter, so getting the names of our diagnoses correct matters. XLH is NOT a form of rickets; XLH is a genetic form of hypophosphatemia. GACI is not idiopathic (unknown cause) any longer, as described in OMIM; it’s usually an ENPP1 deficiency with an identified genetic cause.
Getting the name correct is important to patients insofar as it’s necessary for access to the best possible treatment. You can’t (absent special procedures) get burosumab unless you’re diagnosed with XLH pr TIO (the only conditions it’s approved for), so if your records instead say you’ve got vitamin D resistant rickets (or some odd name for TIO), you’re outta luck. (Plus, you really need a new clinician whose knowledge of XLH is more recent than the 1970s.)
But beyond that, it’s going to be increasingly important for the reliability of medical records, especially if, despite all efforts to stop it, health care is handed over to AI to decide what treatment patients should receive and can access.
Useful analysis of data requires that the underlying data be reliable, and generally that requires standardization. Billing codes are an attempt at standardization, which would make the names of conditions somewhat obsolete. But billing codes are often chosen incorrectly for patients with rare disorders, which means that researchers reviewing that data will miss a lot of patients whose diagnosis is mis-coded. If, for example, researchers are looking for patients using the correct code of E83.31 (familial hypophosphatemia), they’re going to miss all the patients who were coded E55 (nutritional rickets) instead. That’s not an unlikely event either. For most of my life, including when I was in my fifties, that’s the only billing code related to XLH that could be found anywhere in my records!
If billing codes are reliable, then the alternative is for researchers to do a search on diagnosis names. But there’s no reliable uniformity there either. A search for XLH requires also looking for XLH, XLHR, XLHRD, hypophosphatemic rickets, vitamin d resistant rickets, familial hypophosphatemia, familial hypophosphatemic rickets, and I’m not sure what-all else.
You may think I’m exaggerating, and over time there’s a building consensus, so if researchers are looking at data from just, say, the last twenty years, surely they don’t need to include VDRR in the search. Yeah, no. It’s still being used. I just read an article published last month in a supposedly peer-reviewed medical journal, by authors who are implicitly claiming expertise about the chronic hypophosphatemia community, that states “X-linked hypophosphatemic rickets (XLH) is the most common form of hereditary vitamin-resistant rickets….”) (emphasis added).
It gets worse, given how little most clinicians know about XLH. Researchers might even have to include vitamin D dependent rickets (VDDR, an actual, correct diagnosis for a completely different disorder) in their search, and then individually review the patient’s data to see if VDRR was meant, so the diagnosis is actually XLH, because, really, how many non-endocrinologist clinicians know the difference between the ultra-rare VDDR and XLH? (I met a patient with VDDR once who was convinced they had XLH, and would not consider the possibility that they were confusing the two conditions, so goodness only knows what they told their clinicians and what ended up in their records, especially their primary care provider’s notes.)
There isn’t anything much we can do about all the existing misdiagnoses in our records—as far as I can tell, you can ask that a note be added to your file that explains away any misdiagnoses, but the old record won’t be edited, which means that a computer search of the records will still pick up on the misdiagnosis and may not ever see the correcting note. (And if the researcher sees the note, what are the chances that they’ll believe the patient over the clinician who made the diagnosis?)
So what CAN we do? If we can’t fix past mistakes, we need to work toward correct coding and consistent naming going forward. And I think that will ultimately require, as I suggested above, some sort of reference resource that everyone agrees to use. Until then, patients are going to have to be more vocal about insisting that everyone in the healthcare setting use the correct names for our diagnoses.
I’ve been thinking about the naming issue a lot lately, not just in the context of AI, but also in light of the recent news about another medical condition getting a new and improved name. The relevant experts (endocrinologists, just like the folks who treat us, except a different subspecialty) have renamed Polycystic Ovary Syndrome (PCOS) to Polyendocrine Metabolic Ovarian Syndrome (PMOS). Patients cheered the name change, hoping that the new one would convince clinicians to take the patients’ symptoms more seriously, since the addition of “metabolic” and “endocrine” reinforces the biochemical source of the symptoms, making it harder to discount the condition as psychosomatic. But as I’m sure you know, simply announcing a name change doesn’t in fact change anything. The only clinicians who will take seriously the fact that the condition is a metabolic disorder are the same people who already know it’s not a psychosomatic one.
You’ve heard me talk about the seventeen-year statistic—the time it takes for a new health discovery to actually reach common use in doctors’ offices. There was even a news article recently about a new example, where a well-documented breakthrough in dental treatment (a medicated, prescription toothpaste that can obviate the need for filling a cavity if it’s caught early enough) will be recommended by recent dental-school graduates who were taught the toothpaste should be the default treatment, but anyone who graduated several years ago will instead reach for the drill. (Key quote: “A dentist that graduates in 2026 versus a dentist that graduates in 1999, they might have two completely different treatment plans.”) I’m sure this breakthrough is talked about in continuing medical/dental education, but the state-of-the-art treatment recommendation is apparently not, in fact, breaking through to day-to-day use in dental chairs by anyone who learned differently in dental school.
Similarly, in the PCOS/PMOS setting, and in the absence of the passage of time or a huge PR campaign, I expect that the thought process of the majority of clinicians won’t change, but will be frozen at whatever name and preconception they were taught in med school. Which means that, unless patients speak up loudly, a whole generation of doctors will have to retire before PMOS becomes the standard name. After all, we’ve known for close to fifty years that XLH has (almost) nothing to do with vitamin D, and the VDRR name persists, albeit with decreasing frequency. It’s not the patients’ fault that clinicians haven’t kept up with the science, but I think it’s clear that what we’re doing isn’t working, and we need to do more.
I’d like to tell the PCOS/PMOS patients (and patient-focused experts) that they’re in for an uphill battle, and they’ll need to be loud and emphatic in insisting on the use of the correct (new) name. Patient advocacy is still growing, learning from past mistakes, and I think a big mistake is that we (meaning patient advocates in general, but also including the XLH advocates) have been too polite, too deferential, too willing to be sidelined as cute mascots, nothing more. We’ve been asking to be heard as if we were beggars at the table instead of the people who have the greatest stakes of everyone involved in the conversation.
For the longest time, the standard advice for patient advocacy has been to meet scientists where they’re most comfortable, with facts and logic, not emotion. “Don’t get angry, don’t alienate the gods of the health care system,” we were told. But you know what? Following the rules, never getting emotional, hasn’t worked to change the general understanding of XLH in the health care community (outside the handful of patient-focused experts). Perhaps if we’d named and shamed from the beginning, when we’d first realized how harmful it was to refer to XLH with a term that excluded adult patients, we might have long since eradicated the myth that adults don’t need treatment.
Once again, we can’t undo the past, so we need to look forward to what we can do. And like the PMOS patients who should be correcting anyone in the health care system who refers to their diagnosis as PCOS, chronic hypophosphatemia patients need to correct anyone who refers to our diagnosis by an incorrect and/or outdated term. Not abusively, but firmly and repeatedly until it finally gets through. Before another seventeen (or fifty) years passes.
***
Please note that the author is a well-read patient, not a doctor, and is not offering medical or legal advice.
If you’d prefer to read this blog as a newsletter, sent straight to your inbox, please sign up here.