Thursday, 10 January 2008

The HL7V3.0 Primer

What is HL7V3.0 ?

HL7V3.0 is a model based messaging specification which was proposed to resolve the weaknesses of HL7V2.x. The aim of a model based messaging standard is to create a standard set of structures for the sharing of healthcare information and support binding of standard clinical terminologies to the model. The creation of a standard clinical model with standard terminology establishes a healthcare domain specific common language which would allow any healthcare application model to map onto the reference model. This mapping would allow interoperability with other applications apart from allowing re-use of healthcare data in the application which has mapped its model onto the model.

Methodology Overview

Figure below shows the methodology used in HL7V3.0

The different processes in the HL7V3.0 methodology to obtain the message specifications are described below

Use Case Modeling:

The Use case Model consists of the following definitions

Ø Scope statement: A high level use case for the entire project

Ø Use case: Describes specific situations in which communication between healthcare entities is needed

Ø Actor: An entity which initiates or participates in the use case. Discovered in the process of developing use cases

Ø Use Case Diagram: Provides a graphical form to develop the use case model from the business process analysis and makes it easy to show the relationship between use cases

Interaction Modeling:

The Interaction Model

Ø Describes roles to which systems may claim conformance

Ø Fine-grained abstraction; every system will claim several roles

Ø Not standardizing system or application functions, only messaging roles

Ø Basis for contractual agreement

Ø Potential basis for conformance testing

Ø Captured in a MODEL, a TABLE, and a DIAGRAM

Each Interaction consists of:

Ø Trigger event :Event dependency usually expressed as the state of one or more classes

Ø Message ID : Each interaction sends one particular message

Ø Sender role

Ø Receiver role

Ø Receiver responsibility :A specific functional responsibility for the receiver to initiate another (consequential) interaction

Information Modeling:

The information model consists of the following

Ø A detailed and precise definition for the information from which the data content of HL7 messages are drawn.

Ø Follows object-oriented modeling and diagramming techniques, and is centered on a depiction of the classes that form the basis for the objects in HL7 messages.

Ø Provides a means for expressing and reconciling differences in data definition independent of message structure.

Ø Forms a shared view of the information domain used across all HL7 messages.

The Information Model defines

Ø classes, attributes, data types, and relationships

Ø vocabulary domains, code systems, and value sets

Ø states, trigger events, and transitions

Reference Information Model:

The model which forms the basis for the development of HL7V3.0 message specifications labeled RIM (Reference Information Model) is an object oriented model where the information is organized into classes that have attributes and that maintain associations with other classes. The information-related content of HL7 3.0 protocol specification standards is drawn from this model. RIM uses notation based on the Uniform Modeling Language (UML) and information modeling conventions outlined in the HL7 message development framework. RIM provides an explicit representation of the semantic and lexical connections that exist between the information carried in the fields of HL7 messages.

RIM is conveniently fashioned into three areas – semantic control, structured documents and message control.

The backbone of RIM consists of four primary classes and two linking classes. The four primary classes in the semantic content partition of the RIM are

Act: An Act is defined as intentional healthcare related action in the business domain of HL7. An instance is a record of an act. Acts definitions (master files), orders, plans, and performance records (events) are all represented by an instance of Act. An act is represented by a brick colored block in the RIM as following example

Act has the following sub-classes
Ø Account
Ø ControlAct
Ø DeviceTask
Ø DiagnosticImage
Ø Diet
Ø FinancialContract
Ø FinancialTransaction
Ø InvoiceElement
Ø Observation
Ø Participation
Ø PatientEncounter
Ø Procedure
Ø PublicHealthCase
Ø SubstanceAdministration
Ø Supply
Ø WorkingList

Entity: Entity is a physical thing or organization and grouping of physical things. A physical thing is anything that has extent in space, mass. Excludes information structures, electronic medical records, messages, data structures, etc. An Entity is represented by a green colored block in the RIM as following example

Entity has the following sub-classes

Ø Container
Ø Device
Ø LanguageCommunication
Ø LivingSubject
Ø ManufacturedMaterial
Ø Material
Ø NonPersonLivingSubject
Ø Organization
Ø Person
Ø Place

Role : For people, role is usually positions, jobs, or ‘hats’ and for places and things are what these places or things are normally used for. Each role is “played by” one Entity and is usually “scoped” by another. Thus the role of Patient is played by (usually) a person and scoped by the provider from whom the patient will receive services. Similarly, an Employee role is scoped by the employer. A role is represented by a yellow colored block in the RIM as follows

Role has the following sub-classes:

Ø Access
Ø Employee
Ø LicensedEntity
Ø Patient

Participation: Role exists only within the scope of one act. Acts have multiple participants, each of which is an entity in a role. Role signifies competence while participation signifies performance. Participation is represented in the RIM by the following block

Participation has ManagedParticipation sub-class.

Act Relationship: ActRelationship is association between a source Act and a target Act. A point from a later instance to a earlier instance OR point from collector instance to component instance.Act Relationship is represented in RIM by the following block

ActRelationship has no sub-classes

Role Link: Role Link is a relationship between two entity roles. For example Clinicians relationship with a Hospital and a Patients relationship with a Hospital to express the doctor and patient relationship. Role Link is represented in RIM by the following block

RIM Classes Example

Figure below shows the relationships between the different classes in the RIM with an example

In the above example the Act Examination is performed by Entity Person A in his role as GP and the Subject of the Examination is the Entity Person B in his role as Patient.


Class attributes are the core components of the information model. The attributes are the source for all the information content of HL7. There are three special kinds of attributes in the information model: identifier attributes, classifier attributes and state attributes.

Ø Identifier attributes: Identifier attributes can be used to identify an instance of a class. Values of identifier attributes never change. Examples of identifier attributes from the RIM include and, which uniquely identify a particular Entity or Act respectively

Ø Classifier Attributes: The classifier attributes are a critical aspect of the classes forming the backbone of the RIM. The classifier attributes are named "classCode".

Ø Structural Attributes: Structural attributes are those attributes whose coded values are needed to fully interpret the classes that they classify. ClassCode together with moodCode, typeCode and determinerCode are the structural attributes.

Ø State Attributes: A state attribute is used in subject classes (classes that a Technical Committee designates as the central focus of a collection of messages). It contains a value that concisely indicates the current state (named condition) of the class.

Message Design:

The Message Design in HL7V3 from RIM is obtained as follows Message Design Overview

Domain Message Information Model:


Ø Full description (model) of the information classes within a domain and the way in which they are being used

Ø Classes are ‘cloned’ (copied and adapted) and renamed (alias) according to the way they are used in the domain context

Ø Data Types and vocabulary for attributes are specified

Ø Every RIM class category has its own colour; this makes the model description more visual

To obtain D-MIM from RIM

Ø Select RIM classes to be included in D-MIM Select the subset of RIM classes relevant to the problem domain.

Ø Clone subset classes of the RIM If the same RIM class is selected to meet different information needs then create as many clones of the RIM base class as needed. Ø Select a subset of class relationships Add relationships between the D-MIM clones consistent with existing class relationships in the RIM.

Ø Select a subset of class attributes For each D-MIM clone select the applicable attributes from its RIM base class.

Ø Select a subset of attribute data types and vocabulary domains Assign data types and domains to D-MIM attributes that are consistent with assigned data types and domains in the RIM

Refined Message Information Model:

R-MIM is

Ø Same type of representation as D-MIM

Ø Aimed for a specific category of interaction (e.g. all inpatient admission related messages) Ø Special form of an R-MIM are the Common Message Element Types (CMETs).These express common and reusable ‘information units’, that are derived from a D-MIM (like all R-MIMs) but can be used in other D-MIMs (and therefore R-MIMs) as building blocks (e.g. ‘patient’)

To obtain R-MIM from D-MIM

Ø Create clones of D-MIM classes and attributes Further specialize D-MIM classes by creating additional clones for each unique collection of attributes

Ø Assign alias class and relationship role names Assign domain specific names to class clones Ø Eliminate unnecessary class hierarchies Remove classes with no attributes and relationships from the class hierarchy

Ø Finalize class relationships and cardinality Alter relationship cardinality to reflect the appropriate minimum and maximum class occurrences

Ø Finalize attribute data types and domains Replace data type and domain specifications with the most appropriate constrained version of each

Hierarchical Message Definition

HMD is a further restriction of an R-MIM:sequentialized/serialized; represented in a different format. To create a HMD

Ø Select a root class for the message Chose a class to serve as the starting point for message content

Ø Arrange classes and attributes hierarchically Follow the relationships from class to class and add the attributes to the HMD.

Ø Declare inclusion and repetition constraints For each attribute specify if its presence in message content is optional, mandatory, or conditional and specify the number of times the attribute may repeat

Ø Declare data type and domain value constraints Declare additional constraints on the data types and domain values

Ø Assign message element names Assign a final meaningful business name to each attribute in the HMD

HL7 V3.0 Data Types:

HL7 data types define the structure and constrain the allowable values of attributes in the RIM. The HL7 data type specification declares the set of data types, identifies components of complex types, and defines relationships between data types. The HL7 data type implementation technology specification defines constraints and operations for data types and describes how data types are to be represented in XML. Data types are reusable atomic building blocks used to create all HL7 V3 information structures. Each RIM attribute is assigned one data type; each data type is assigned to zero, one, or more RIM attribute. Few of HL7V3.0 data types are given below

• AD: Postal Address
• ANY: DataValue
• BAG: Bag
• BL: Boolean
• CD: Concept Descriptor
• CE: Coded With Equivalents
• CS: Coded Simple Value
• ED: Encapsulated Data
• EN: Entity Name
• GTS: General Timing Specification
• HIST: History
• II: Instance Identifier

• INT: Integer Number
• IVL: Interval
• LIST: Sequence
• MO: Monetary Amount
• ON: Organization Name
• PN: Person Name
• PPD: Parametric Probability Distribution
• PQ: Physical Quantity
• REAL: Real Number
• RTO: Ratio
• SC: Character String with Code
• SET: Set
• ST: Character String
• TEL: Telecommunication Address
• TN: Trivial Name
• TS: Point in Time
• UVP: Uncertain Value - Probabilistic

HL7V3.0 Vocabulary Specification:

The HL7 V3 Vocabulary Specification defines vocabulary domains used in RIM coded Attributes and coded data type properties. A vocabulary domain is an abstract collection of interrelated concepts. It describes the purpose or intent of a coded attribute. A value set is a collection of coded concepts drawn from a single code system. A vocabulary domain may be associated with many value sets. A context expression provides a means of designating which value sets within a given domain are applicable for a given usage context. A coded concept has a concept code assigned by the coding system and a concept designation which names the referenced concept. A coded concept may be a parent concept for a collection of additional concepts within the same code system.

HL7 XML Schema Generator:

The HMD’s constrained with HL7 Vocabulary Specification and HL7 Data Type specification to a HL7XML schema Generator gives the XSD the schema definition of an HL7V3.0 message. Figure below shows the set up for the generation of XSD.

Links, References, Acknowledgements and Copyright


  1. Thanks for this useful primer. I have a question though: what do and Entity.Id actually identify?

    Allow me to clarify my question.

    For "Act", you write: "An Act is defined as intentional healthcare related action in the business domain of HL7. An instance is a record of an act. Acts definitions (master files), orders, plans, and performance records (events) are all represented by an instance of Act.". The first sentence let me assume that "Act" refers to real actions that a clinician might do, such as examining a patient, ordering a lab test, etc. But in your second sentence, you write that Act instances are records of such acts, thus information structures about such acts.

    For "Entity", you write: "Entity is a physical thing or organization and grouping of physical things. A physical thing is anything that has extent in space, mass. Excludes information structures, electronic medical records, messages, data structures, etc. ". In this case, it seems thus that real people such as you and me are referred to. Strange however is that you don't mention "Entity instances" here.

    To be more concrete, take the following scenario: I have patient John Doe in front of me and I observe (by looking at his left hand) a scar on his left hand and I record this fact a bit later in the EHR, what are then the things involved ?
    (1) Are John Doe, his left hand and the scar on his hand "instances of Entity" (all three are extent in space and have mass),
    (2) Are the references in the record "instances of Entity" ? (Probably not because these are information structures)
    (3) Is my looking at John Doe an "instance of Act" ?
    (4) Is my registration of the fact that I see a scar on his hand an "instance of Act" ?

    It becomes even more confused when further down, you write :"Identifier attributes: Identifier attributes can be used to identify an instance of a class. Values of identifier attributes never change. Examples of identifier attributes from the RIM include and, which uniquely identify a particular Entity or Act respectively ".

    Now with "particular Entity", do you then mean things such as John Doe, his left hand, and the scar on his hand, and if so, does the value than contain an identifier for these things, or is the .id for the information structures about these things ? If the former, and suppose that my nurse also registers in her nursing record that John Doe has a scar on his left hand, the .id would be the same independent of whether I or she writes it down. If the latter, the .ids would be different.

    For "particular Act", it seems to be different: me and my nurse can't do the same "looking" (as if we were both looking through the same eyes), but it still is unclear whether the stands for the looking or for the "record" thereof. I can indeed look once, and register twice. If I register the same fact twice (imagine I forgot whether I wrote it down already, and don't bother to look it up, or, more often, different parts of an EHR ask the same information to be completed) are there then two different .ids ?

    Thanks for your help.

  2. HL7 v3 makes no distinction between the act itself or the record of that act. "If you didn't chart it, it didn't happen".

    The is an instance identifier. For acts, these are most likely system assigned meaningless identifiers used for something akin to a primary key. The advantage is that you can refer to them by id, rather than by having to replicate the whole thing in a model. Somewhat akin to a pointer v. passing values by copy. Nothing dramatic. Just a way to uniquely identifier a particular act as distinct from every other act in the universe.

  3. A bit pointless as your diagrams are such low quality they cannot be viewed by a human - have a look at your RIM classes example!. A pity because it would be a good tutorial if the material was visable.

  4. Just as a suggestion, readers of this blog may find “Unofficial Developer's Guide to HL7v3 Basics” eBook useful in context of this topic. The book can be downloaded at -