Metadata-based Assessment of SNOMED CT and LOINC Adoption Across 800 000 Swiss Patients
This study presents the first nationwide empirical baseline of SNOMED CT and LOINC adoption across 800,000 Swiss patients, revealing that while these standards are widely used, significant institutional heterogeneity and limited cross-institutional overlap in code usage currently hinder full semantic harmonization.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the world of healthcare as a massive, bustling library where every patient's story is written down. For decades, these stories were kept in thousands of different languages and dialects. One doctor might write "heart attack," while another writes "myocardial infarction," and a third scribbles a local shorthand that only they understand. This is the problem of interoperability: if you want to read all the stories together to find patterns or cure diseases, you first need everyone to speak the same language. To solve this, the medical world created two giant dictionaries. The first, called SNOMED CT, is like a massive encyclopedia for everything a doctor might see, feel, or do—from a broken bone to a specific type of allergy. The second, LOINC, is a specialized dictionary just for lab tests, ensuring that a blood test for "sugar" in one hospital is recognized as the exact same thing in another. The big question for researchers is: Are hospitals actually using these dictionaries, or are they still writing in their own secret codes?
This paper takes a giant, nationwide peek into the Swiss healthcare system to answer that question. Instead of reading millions of individual patient files—which would be a privacy nightmare—the researchers looked at the "metadata," which is like checking the library's card catalog without opening the books. They analyzed the digital footprints of 817,856 patients across six major Swiss university hospitals. Their goal was to see if these hospitals were really using the SNOMED CT and LOINC dictionaries to label their data, or if they were just pretending to while secretly sticking to their old, local ways.
Here is what they found: The hospitals are definitely using the dictionaries. In fact, SNOMED CT and LOINC are the stars of the show, accounting for 87% of all the coded data instances (specifically 60% for SNOMED CT and 32% for LOINC). That's a lot of data—roughly 717 million individual pieces of information! However, the story gets a bit messy when you look closer. While everyone is using the same dictionaries, they aren't necessarily using the same words from them.
Think of it like a group of friends trying to build a giant Lego castle together. They all have the same box of bricks (the dictionaries), but Hospital A is building a tower using only red bricks, while Hospital B is using blue bricks for the same tower. The researchers found that while there were 6,557 different SNOMED CT codes and 3,693 LOINC codes in total, most of them were unique to just one or two hospitals. For example, in the "Substance" category (like drugs or chemicals), there were thousands of distinct codes, but almost none of them were shared by all six hospitals. It's as if every hospital invented its own specific shade of "red" for the same Lego piece.
The only time everyone agreed on the exact same code was for the most basic, routine things. The top ten most used SNOMED CT codes were things like "Female," "Male," "Intravenous route," and "Accepted" (meaning the patient agreed to the study). Similarly, the top ten LOINC codes were all standard blood tests like "Platelets" or "Potassium." It seems that for the boring, everyday stuff, everyone speaks the same language. But as soon as things get specific—like describing a complex injury or a rare organism—everyone starts speaking their own dialect again.
The researchers also looked at how "deep" the hospitals went into the dictionaries. Imagine a dictionary where you can look up "Animal," then "Mammal," then "Dog," then "Golden Retriever." Some hospitals stopped at "Dog," while others went all the way to "Golden Retriever." The study found that this "depth" varied wildly. Some hospitals were very specific, while others stayed general. This isn't necessarily a mistake; sometimes a hospital might have a very detailed code for a specific body part, while another hospital describes the same part by combining a general code with a "left" or "right" tag. Both methods work, but they look different to a computer trying to compare them.
So, what's the verdict? The paper suggests that while Switzerland has made huge progress in adopting these international standards, simply having the dictionaries isn't enough to make the data perfectly interchangeable. The "semantic harmonization"—the state where everyone agrees on the exact same word for the exact same thing—is still a work in progress. The authors argue that we can't just assume the data is ready to be mixed and matched; we need to keep working on getting everyone to use the same specific words, not just the same dictionary. It's a reminder that even when we all have the same map, we still need to agree on exactly which path to take.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.