In HL7 v2, information travels as messages composed of reusable segments. Segments can be grouped into repeating or optional collections called groups. A message is sent when something happens, such as a patient being admitted or a result being finalized, so v2 is a push model: the source system tells downstream systems about events as they occur.
Why it is still everywhere
HL7 credits v2 with strict processes for forward and backward compatibility, so an interface built against an older version keeps working as the standard moves on. The same page lists the costs. Many elements are used only in narrow circumstances. The data elements supported, and even where a given element sits in a message, can vary between implementations. Local extensions called Z-segments are opaque without a manual explanation from the sender, and v2 messages usually carry no human-readable version of their content.
For a buyer, this means every v2 interface needs its own specification and testing against the actual feed from the actual site. Two sites can send differently shaped messages for the same event.
Example and the path to FHIR
A lab system sends a result message to the hospital EHR the moment a test is verified, and the EHR files it to the chart. If a newer app needs the same data as FHIR, HL7 publishes a V2-to-FHIR implementation guide (STU 1) that maps v2 messages, segments, datatypes and vocabulary to FHIR R4 Bundles.
Our article FHIR vs HL7v2 covers when to use each in an integration project.