UCUM units

In one line

UCUM (Unified Code for Units of Measure) defines machine-readable codes for units — mg/dL, mmol/L, mm[Hg], kg/m2 — so software can validate, convert and compare quantities instead of guessing from free text.

Why a unit needs a code at all

A number without a trustworthy unit is not data — it's a hazard. "Glucose 100" is reassuring in mg/dL and a medical emergency in mmol/L (≈ 1800 mg/dL). Humans disambiguate from context; software cannot, unless the unit is carried in a form it can parse. Free-text units ("mg per deciliter", "mg/dl", "MG/DL") defeat that. UCUM gives each unit exactly one canonical, machine-parseable code, so a value travels with an unambiguous meaning.

A grammar, not a list

The clever part: UCUM doesn't enumerate every unit, it defines a grammar that composes them from a small set of base units. That means even rare or compound units are expressible and still computable:

• Algebraic composition — mg/dL, mL/min/{1.73_m2} (eGFR), ug/(kg.min) (infusion rate).
• Prefixes — k, m, u, n attach to base units (mmol = millimole).
• Bracketed special units — non-metric or customary units wear square brackets: mm[Hg] (blood pressure), [iU] (international unit), [degF].
• Annotations — descriptive, non-semantic text in braces: {breaths}/min, {cells}/uL.

Because each code is parseable, a system can check that two units are commensurable (both pressures, both concentrations) and convert between them automatically.

Case sensitivity — the trap

UCUM is case-sensitive by design, and this catches people: mg is milligram but Mg is megagram; m is metre but M is mega; s is second but S is siemens. Lowercasing or upper-casing a unit string "to normalise it" silently corrupts the data. Carry UCUM codes verbatim.

Where it shows up in digital health

• FHIR — Observation.valueQuantity.code (and Quantity everywhere) is bound to UCUM, with the unit's display in unit and the system fixed to http://unitsofmeasure.org.
• LOINC — every LOINC term's example units are expressed in UCUM, so LOINC and UCUM are used together: LOINC names the test, UCUM names the unit of its result.
• Devices — IEEE 11073 MDC device units map to UCUM, so a connected monitor's readings arrive in a known unit.
• Cross-border exchange — any lab or registry sharing quantities across systems relies on UCUM to make conversion safe.

Common pitfalls

• Normalising case — the single most dangerous "cleanup"; never do it.
• Storing the display string as the code — "mg/dL" the label is for humans; the UCUM code is what software must key on.
• Dropping the unit on import — a value with no unit, or a defaulted unit, is worse than no value, because it looks usable.
• Confusing mass and molar — mg/dL ↔ mmol/L conversion needs the molar mass; UCUM enables the check but the clinical factor is yours to get right.

Key takeaways

• UCUM makes units computable: validate, convert, compare — not guess.
• It's a grammar, so any compound unit is expressible and still machine-readable.
• Case matters; carry codes verbatim.
• The cautionary tale: glucose in mg/dL read as mmol/L is an ~18-fold error. Units are clinical data — treat them with the same rigour as the value itself.