Typically, you would perform as little work parsing each HL7 message as you need to:

• Only include the MSH segment and the segments that you need to extract data from.
• Use segment filtering to ignore unwanted segments.

For instance if you just wanted to extract patient demographics from an ADT feed then it would make sense to create a segment grammar like:

Getting Patient Demographics

This segment grammar would be optimal for extracting patient demographics out from an ADT feed since the PID segment will contain all the information that you require. It means that you do not have to spend any effort validating the rest of the segments in the message, or if your counterparty sends you an unexpected Z-segment it's not going to interfere with the data you really need.

If you are getting patient demographics from an ADT feed, read the ADT Feeds section.