A Problem-Oriented Medical Record Ontology

This is an attempt to demonstrate how a few best practices in ontology engineering can be applied to address the problem of the lack of a unified vocabulary for Computer-based Patient Records (CPR) [20]. This developed from an initial effort to develop a Patient Record ontology (in the absence of any), an initial presentation - during the 2006 F2F - on how a major Coronary Artery Bypass Graft (CABG) Procedure Guideline could be implemented using Notation 3 rules and OWL, a recent presentation [1] to the HCLSIG in ISWC on CPRs and how Semantic Web representation standards can be applied to CPR requirements, and the development of two use cases: 1) demonstrating how GRDDL can be applicable as a technology for enabling Clinical Content Management systems and 2) modeling an angiogram finding:

[ a cpr:clinical-description;
   dc:date "2006-12-06"^^xsd:date;
   foaf:maker [ 
                foaf:name "Chimezie Ogbuji";
                a cpr:person  ];
   rdfs:label 'Cardiac Catheterization @ 2006-12-06';
   cpr:description-of
    [ a ptrec:CardiacCatheterization, cpr:diagnostic-act;
      inf:realizes
       [ a inf:diagrammatic-object;
         edns:interpreted-by [ a cpr:clinician ];
         foaf:depicts 
             [ a galen:Stricture, cabg:ImportantStenosis,cpr:pathological-structure, cpr:medical-problem; 
               galen:hasSeverity "50";
               galen:hasSpecificLocation galen:CoronaryArtery;           
       ]
    ]
]

For the terms described the prefix cpr is associated with the namespace URL http://purl.org/cpr/1.0/problem-oriented-medical-record.owl#


The goal is to define a minimal set of terms that connect representations from well defined healthcare information & process models (such as HL7 RIM) with more expressive foundational ontologies [3] through the use of the criteria outlined in the traditional Problem-oriented Medical Record (POMR) structure. This separation is necessary [4] in order to provide sound ontological commitment for heavily adopted models of health care information. The approach taken for building a vocabulary around the POMR structure takes after the recommendations outlined in Ontologies for Knowledge Representation in a Computer-Based Patient Record [6].

The Problem-Oriented Medical Record

From A Problem Oriented Approach to the Computerized Patient Record:

The traditional structure / organization of the POMR is:

Foundational / Upper Ontologies

Wikipedia has a (mildly) useful description of upper / foundational ontologies - along with a list of a few. It probably can be said that foundational ontologies (such as BFO and DOLCE) are typically more rigorously oriented around philosophy and linguistics.

Grounding HL7 RIM Ontologically

The cpr:clinical-description class is an attempt to reconcile the ontological inconsistencies with the archetype of a recording of an act and the thing the recording describes (which could be either a continuant / endurant or an occurrant / perdurant).


The Classes

cpr:patient-record

A patient record is a universal [17].

cpr:person

The person whom the patient record is about. An anchor for associating (non-lingitudinal) demographic and genetic data with the patient. Also a bridge for other vocabularies which describe people and their related data (vCard and FOAF for instance). Galen has the notion of a role which is missing from FOAF but essential for medical record ontologies.

A person is a universal.

cpr:patient

A patient is a defined class [17].

cpr:clinician

Restricts the person class to only those that play clinical roles (from galen).

A clinician is a defined class.

cpr:clinical-description

Corresponds (at least syntactically) with the HL7 RIM Act Class, insofar as it is the class of clinical recordings (a REPRESENTATIONAL ARTIFACT [17]) of a natural phenomenon (cpr:medical-problems, cpr:clinical-acts, and anything else of relevance) by an individual (foaf:maker and dc:creator are used for this relation as a bridge to agent and provenance vocabularies - respectively). This recording manifests itself as the patient record. Note that it is crucial that the recording be associated with a point on a time scale in order to later facilitate temporal reasoning. Expressive, interoperable, and freely available knowledge representations [18] can be used or the natural unicode ordering of ISO 8601 can be leveraged intuitively [19] with rules. Towards this end, the instances of the cpr:clinical-description class can be associated with a TemporalEntity from the OWL-time vocabulary.

A clinical-description is a defined class.

This class is a partial attempt to resolve the ontological inconsistencies [4] of the HL7 RIM design intent (especially with regards to the act of recording or documenting).

Clinical descriptions can be directly associated with provenance data (via Dublin Core vocabulary terms such as dc:source, dc:date):

cpr:anatomical-structure

An anatomical-structure a universal.

cpr:immaterial-anatomical-entity

An immaterial-anatomical-entity a universal.

cpr:anatomical-space

cpr:anatomical-surface

cpr:anatomical-point

cpr:physiological-state

A physiological-state is a defined class.

cpr:physiological-role

A physiological-role is a defined class.

cpr:physiological-process

A physiological-process is a universal.

From OBR:

cpr:medical-problem

A medical-problem is a defined class.

Whether or not a medical problem is active is an important distinction within the POMR model [2].

A medical problem is typically associated with a measureable severity (especially an objective problem - cpr:medical-sign), an onset modifier, and a time interval.

How subtle is the distinction between symptoms, findings, and diseases? See: http://en.wikipedia.org/wiki/Medical_sign#Signs_versus_symptoms

cpr:medical-sign

Wikipedia entry:

See: http://en.wikipedia.org/wiki/Medical_sign#Signs_versus_symptoms

cpr:symptom

From Wordnet:

So a symptom is related to a cpr:disease by some relation. INTERPRETANT-OF is a relationship between a SIGN-OR-SYMPTOM and a cpr:disease. Typically interpreted by the cpr:patient.

cpr:disease

From Wordnet:

Corresponds directly with pathological universals [17] from OBR [3]: pathological entity, continuant, or process. Typically interpreted by a cpr:physician. This constraint corresponds well with OBJECTIVE-MEDICAL-SIGN (Clinical Act Theory [13]) which uses the relationship INTERPRETED-BY for this purpose.

cpr:immaterial-pathological-entity

cpr:pathological-space

cpr:pathological-surface

cpr:pathological-structure

From OBR:

cpr:pathological-process

From OBR:

" ..processes that transform a physiological into a pathological state, or one pathological state into another, are instances of pathological process."

cpr:pathological-process = galen:NAMEDPathologicalProcess or galen:PathologicalBodyProcess

cpr:pathological-state

Processes and states (conditions) (from DOLCE):

cpr:clinical-act

The root class for clinical acts described in a patient record. This hierarchy corresponds with Elisabeth Bayegan's CareActType class [6].

cpr:diagnostic-act

cpr:screening

Equivalent (syntactically) with rim:Observation

Corresponds well with a DIAGNOSTIC-PROCEDURE (Medical-procedures Theory [13]):

cpr:history-screening

cpr:family-history-screening

cpr:social-circumstance-screening

cpr:clinical-examination

cpr:laboratory-test

cpr:therapeutic-act

cpr:physical-therapy

cpr:psychological-therapy

cpr:medical-therapy

cpr:substance-administration

From HL7 RIM:

cpr:surgical-procedure

Corresponds well with GALEN's notion of SurgicalDeed and HL7 RIM's notion of a Procedure (defined below):

Finally, from the notion of a THERAPEUTIC-OR-PREVENTIVE-PROCEDURE:

cpr:medical-device

cpr:medication

Corresponds well with PHARMACOLOGIC-SUBSTANCE:

Or from Wordnet:

Related HL7 RIM ActCodes:

The Roles/Properties

cpr:interpretant-of

Relates a medical sign with associated phenomena.

cpr:description-of

Relates a clinical description with the phenomena it describes

Functions v.s. Processes

A GRDDL Transform

An XSLT transformation can be written to extract RDF statements (which adhere to this ontology) from HL7 CDA documents as a means to seperate the structural nature of healthcare information from it's unambigious interpretation (to the extent that this is possible with a well-founded target ontology). See Health Care: Querying an XML-based clinical data using an standard ontology

For the details, see: POMROntology and GRDDL

Ontology files

You can load the ontology in SWOOP - using the nightly java webstart build.


Papers and References

  1. A Comprehensive Representation Methodology for CPRs

  2. A Problem Oriented Approach to the Computerized Patient Record

  3. A Strategy for Improving and Integrating Biomedical Ontologies

  4. HL7 RIM: An Incoherent Standard

  5. DOLCE: Descriptive Ontology of Linguistics and Cognitive Engineering

  6. Ontologies for Knowledge Representation in a Computer-Based Patient Record

  7. Open Galen OWL Ontology

  8. DOLCE OWL Ontology - full

  9. HL7 RIM OWL Ontology

  10. Manchester OWL Syntax

  11. bio-zen-plus.owl

  12. openEHR Archetypes

  13. Laboratory for Applied Ontology (LOA) - Medical Theories

  14. HL7 Reference Information Model

  15. Health Level Seven Releases Updated Clinical Document Architecture (CDA) Specification.

  16. Ontology: Philosophical and Computational

  17. Towards a Reference Terminology for Ontology Research and Development in the Biomedical Domain

  18. Time Ontology in OWL

  19. Practical Temporal Reasoning with Notation 3

  20. National Institutute of Medicine, The Computer-Based Patient Record: An Essential Technology for Health Care - Revised Edition., 1998, ISBN: 0309055326.

HCLS/POMROntology (last edited 2007-01-31 20:08:11 by vtmail)