Data Types — from ST and NM to XPN, XAD, CX, and CWE

Fields are typed

Earlier lessons treated a field as “the text between two | characters.” That is true on the wire, but every field is also declared to have a data type that says how to interpret it ¹. A data type can be primitive (a single value) or composite (several components with defined meanings). Knowing a field’s type is what lets a parser turn 19800101 into a date and DOE^JANE^Q^^DR^MD into a structured name.

Primitive types you will see constantly

Type	Name	Example	Notes
ST	String	Final report	Short free text.
TX	Text	Patient reports...	Longer text; may contain formatting.
NM	Numeric	7.4	A number, no units attached.
DT	Date	20240601	YYYYMMDD.
DTM/TS	Date-Time	202406011230	Date with time, optional precision.
ID / IS	Coded value	F, AA	A code drawn from an HL7 or local table.

The single most important habit: a value’s meaning comes from its type plus its position, never from the text alone. F in a sex field means “female”; F in a result-status field means “final.” Both are the ID type pointing at different tables.

Composite types: a field with internal structure

Composite types use the component delimiter (^) to pack related pieces into one field. The figure below shows XPN, the extended person name:

A few composites carry most of the clinical weight:

• XPN — extended person name: family, given, middle, suffix, prefix, degree.
• XAD — extended address: street, other designation, city, state, zip, country.
• XCN — extended composite name & number for persons: an id and a name, used for providers (e.g., 004777^SMITH^JOHN).
• CX — extended composite id with check digit: an identifier plus its assigning authority and identifier type — how PID-3 carries a medical record number (100711^^^HOSP^MR).
• CWE — coded with exceptions: the workhorse for coded clinical data — a code, its display text, and the coding system it came from. The next course on terminologies returns to CWE in depth.

Why this matters

Most interface defects are type defects: a date in the wrong format, a coded field populated with free text, an identifier missing its assigning authority so two systems can’t agree whose “100711” it is ². Reading a segment well means reading each field as its declared type — the structure is there even when the delimiters make it look like plain text.