The RIM has ~40 basic classes, but the most important ones are color coded for easy recognition (inspired by Peter Coad):

• Act — things that people, machines, or organizations do. For example: a clinical observation, a procedure, a measurement.
• ActRelationship — the relationship between two Acts. For RDF users, this could be thought of as a predicate. For example, the severity of some condition, the fulfillment of a prescription, or, perhaps surprisingly, the assertion that some measurement is a reference range for another.
• Entity — some real-world thing, possibly tangible. For example, a person, organization, device or medication.
• Role — the reason or justification for some entity participating in an act.
• Participation — an instance of participating in some role in an act. For example, Dr. Leonid Rogozov participated in an emergency appendectomy as a surgeon, patient, and, if we was diligent, records custodian.

At the right is a typical RIM diagram using the color-coding described above.

Suppose we want to write down a fairly detailed clinical observation:

Grade 4 anaphylactic reaction to penicillin as evidenced by the combination of wheals (hives), acute respiratory distress (ARD) and systemic hypotension (LBP) following a penicillin injection.

As a set of observations and relationships, it might look like the image on the right.

That image may be somewhat informative if viewed directly by a physician, but it lacks all the structure and use of common coding systems necessary for machine interpretation.

We can add codes to the observation (recall that they are Acts). Codes perform several functions:

• Move term standardization to organizations with appropriate domain knowledge, e.g LOINC for laboratory tests, SNOMED for pathology (originally).
• Separate coded semantics from information model for portability between different information models or even databases.
• Modularize the vocabularies to minimize those necessary for any given task or domain.

This is a step towards machine-readability and interoperability, but so far, we are still counting on the information model have sufficient relationship terms to convey every detail relating two observations.

Adding codes to the ActRelationships permits us to use a coding system to provide the detailed relationships we want to convey. This reduces significantly the expressivity requirements on the information model, though in practice, the codes in this example still come from HL7, the same folks who define the information model.

There is still some consistency between RMIMs; the structure attributes like mood and typeCode remain the same and, if you know the mapping back to RIM, it's fairly intuitive to read data expressed in some RMIM as RIM objects. At the right is a small excerpt of the RMIM for C-CDA capturing the reference ranges for measured values. Below is what that looks like in XML:

Some XML elements represent a cyclic structure. At the right, the entryRelationship element connects one Observation to another. This captures the Acts and ActRelationships we saw in the rimITS example. The diagnosis and severity from the rimITS example can be expressed in C-CDA as:

  <Observation classCode="OBS" moodCode="EVN">
    <code code="ASSERTION" displayName="Assertion"
       codeSystem="2.16.840.1.113883.5.4" codeSystemName="HL7 ActCode"/>
    <value xsi:type="CD" code="241938005" displayName="penicillin-induced anaphylaxis (disorder)"
       codeSystem="2.16.840.1.113883.6.96" codeSystemName="SNOMED CT"/>

    <entityRelationship typeCode="SUBJ"
         xsi:type="Observation" classCode="OBS" moodCode="EVN">
        <code code="SEV" displayName="Severity Observation"
           codeSystem="2.16.840.1.113883.5.4" codeSystemName="HL7 ActCode"/>

        <value xsi:type="CD" code="423132009"
           codeSystem="2.16.840.1.113883.6.96" codeSystemName="SNOMED CT"/>

      </@@@source>
    </@@@inboundRelationship>