LDL-C 2.48 mmol/L,
低密度脂蛋白胆固醇 96 mg/dL, LDL Cholesterol 101 mg/dL. No chart can put
them on one line until something decides they are the same measurement and
how their numbers compare. Standardization is that decision. It runs offline
on your machine, needs no model API key, and when it cannot decide, it leaves
the reading uncoded and says why instead of guessing.
The steps one reading goes through
Answers then read the series, not the raw names: a trend across the three
reports is one line, and every number in the answer links back to the row and
the file it came from.
Sources of readings
Complaints and diagnoses go to a separate vocabulary, ICPC-3:
头疼 is
NS01 and 高血压 is KD73. A complaint is never coded as a lab test.
Readings that are not coded
An uncoded reading is still stored and still answers questions. It stays in its own series under the name it was printed with, so it is never mixed into another test.
If a common name is not recognized, report it: that is how new languages and
hospital spellings get added (see Improve the vocabulary).
Try it
Names only, from a terminal, offline after the first download:Limits
- It is lexical. A name the vocabulary does not hold returns
unresolved, never the nearest match. Recall grows by adding rows, not by guessing. - It codes measurements, not vendor scores. Body Battery or a stress score is one company’s formula and has no LOINC code.
- A code says what was measured, not whether it is normal. Highs and lows come from the range and flag the report printed.
- The vocabulary is versioned.
mirobody.BUNDLE_VERSIONnames the LOINC release and cut. After an upgrade,mirobody recoderecodes stored readings under the new vocabulary, and each change is kept in the coding history.
Improve the vocabulary
Runmirobody resolve "<name as printed>". If the answer is wrong or empty,
report the term
or add one row to mirobody/res/loinc/resolver_overrides.tsv with a test case.
Russian, Estonian and Turkish report spellings arrived this way.
The vocabulary’s scope, licence, measurements and the code behind each step are
in the standardization reference.