Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Table of Contents
maxLevel8
minLevel2

 

Using OpenCDS Concepts

 

Introduction

OpenCDS is a Clinical Decision Support system that is built around the notion of “clinical concepts.”  There are many medical terminology systems and medical information exchange systems that refer to “concepts,” “concept descriptors (CDs),” or “concept unique identifiers (CUIs).” 

When we refer to an OpenCDS Concept we are referring to specific implementation techniques within OpenCDS which have a strong dependence on most of these ideas, but in a concrete implementation.

The purpose of this document is to explain how we use all of these terms in OpenCDS, and how they relate to the general clinical understanding of “concepts.”

 

What is an “OpenCDS Concept?”

An OpenCDS Concept is an implementation technique in OpenCDS.  As a big picture idea, OpenCDS has a structure and methods that are designed to allow the clinical user to develop decision support rules that are written using clinical concepts.  We call those particular concepts “OpenCDS Concepts”, and they provide an interface to the detailed data that represents an instance of the clinical concept.

This means that the clinical rule writer can work using clinical terminology that clinicians understand.  OpenCDS is designed to support rules written for the open-source Drools inferencing engine, using a domain specific language (DSL).  The DSL makes it possible to write the rules so that they read just like the way a clinician would describe them, using terms (OpenCDS Concepts) that the clinician uses every day.

A medical informaticist, terminologist, or vocabulary expert then produces mappings of those concepts to values in a code system which is used in the actual clinical data.  In many cases this will involve the use of terminology management systems with large databases, such as Apelon, First DataBank, and UMLS which support potentially large and internationally supported terminologies.  However, it is possible to also create simple XML files that map proprietary or special-case codes to OpenCDS Concepts.

The clinical rules use OpenCDS Concepts in preference to references to the raw data, and the terminology mappings provide implementations of those concepts as lists of codes from one or more code systems.  This separates the logic of the rules from the details of the data which the rules work on.

Therefore, an OpenCDS Concept is the interface between the clinical ideas and the data details that represent instantiations of the clinical concepts. 

This section of this document will explain the following items, and discuss how they are used in OpenCDS:

  • Concept Types
  • OpenCDS Concepts
  • Concept Instances
  • Concept Mappings and Enumerations

The diagram below may make this relationship easier to visualize:

 

Following sections will discuss the techniques used to relate the patient data to the clinical concepts or ideas in the following software:

  • JBoss Drools Guvnor (our supported authoring environment      for KnowledgeModules, aka “rules” or generically as “knowledge”)
  • OpenCDS Decision Support Service (our software service to      do clinical decision support)

 

Concept Types

Definition:

Concept Types in OpenCDS is the term we use to refer to Java classes in OpenCDS that have been created for every “concept descriptor” (aka CD) and “template” found in the input data structured as a virtual medical record (vMR).  

 

A Concept Type represents an entire category of information that can be found at several places in a clinical statement.  Clinical statements are the basic building blocks of the vMR.  The vMR is more fully discussed in a separate document named “Notes on OpenCDS internal Data Structure.”  In addition, there are also a few additional Concept Types that deal with “templates” and some structural elements of the vMR.

In this document, I will always capitalize Concept Type when I am referring to those Java classes, and use lower-case when I am referring to the more common or generic notions of “concept.”

In some cases, we have also post-coordinated more than one CD in the input data (such as “body site” and “laterality”), because there may be a need to make decisions based on the combination of the two, or common coding systems for a procedure or problem may not provide adequate specificity without being associated with both values as a set.  For example, we need to be able to say that a rash was present both on the right arm and the left leg.

Examples of Concept Types are a “ProcedureConcept,” an “AdverseEventAgentConcept,” a “ProblemConcept”, or a “MedicationClassConcept.” 

Because they are based on specific data elements in the vMR, OpenCDS supports a fixed list of these Concept Types, including a few that are broad enough to fill most needs for an “other” category.  While we can add more Concept Types if the need arises, we feel that the current list is useable.  Supporting additional Concept Types is non-trivial, so it will not be done casually.

The authoritative list of supported Concept Types can always be found in the source code as separate files in the module opencds-vmr-1_0-internal as separate Java classes in the package org.opencds.vmr.v1_0.internal.concepts.  These files have the name of the Concept Types as you would see them in a Drools Rule.  They are also listed for a different purpose (building enumerations) in a Java class named “org.opencds.common.terminology.OpenCDSConceptTypes.java”. 

The current list of Concept Types as of the time this document was written is shown on the next page:

 

 

Section
Column

AdverseEventAffectedBodySiteConcept

AdverseEventAffectedBodySiteLateralityConcept

AdverseEventAgentConcept

AdverseEventConcept

AdverseEventCriticalityConcept

AdverseEventSeverityConcept

AdverseEventStatusConcept

BrandedMedicationConcept

CDSInputTemplateConcept

CDSOutputTemplateConcept

ClinicalStatementEntityInRoleRelationshipConcept

ClinicalStatementRelationshipConcept

ClinicalStatementTemplateConcept

DataSourceTypeConcept

DoseTypeConcept

DosingSigConcept

EncounterCriticalityConcept

EncounterTypeConcept

EntityRelationshipConcept

EntityTemplateConcept

EntityTypeConcept

EthnicityConcept

EvaluatedPersonRelationshipConcept

GenderConcept

GenericMediccationConcept

GoalCodedValueConcept

GoalCriticalityConcept

GoalFocusConcept

GoalStatusConcept

GoalTargetBodySiteConcept

GoalTargetBodySiteLateralityConcept

ImmunizationConcept

InformationAttestationTypeConcept

InformationRecipientPreferredLanguageConcept

InformationRecipientTypeConcept

ManufacturerConcept

MedicationClassConcept

MedicationConcept

ObservationCodedValueConcept

ObservationCriticalityConcept

ObservationFocusConcept

ObservationInterpretationConcept

ObservationMethodConcept

ObservationTargetBodySiteConcept

ObservationTargetBodySiteLateralityConcept

ObservationUnconductedReasonConcept

Column

PreferredLangugageConcept

ProblemAffectedBodySiteConcept

ProblemAffectedBodySiteLatgeralityConcept

ProblemConcept

ProblemImportanceConcept

ProblemSeverityConcept

ProblemStatusConcept

ProcedureApproachBodySiteConcept

ProcedureApproachBodySiteLateralityConcept

ProcedureConcept

ProcedureCriticalityConcept

ProcedureMethodConcept

ProcedureTargetBodySiteConcept

ProcedureTargetBodySiteLateralityConcept

RaceConcept

ResourceTypeConcept

RoleConcept

SubstanceAdministrationApproachBodySiteConcept

SubstanceAdministrationApproachBodySiteLateralityConcept

SubstanceAdministrationCriticalityConcept

SubstanceAdministrationGeneralPurposeConcept

SubstanceAdministrationTargetBodySiteConcept

SubstanceAdministrationTargetBodySiteLateralityConcept

SubstanceDeliveryMethodConcept

SubstanceDeliveryRouteConcept

SubstanceFormConcept

SubstanceManufacturerConcept

SupplyConcept

SupplyCriticalityConcept

SupplyTargetBodySiteConcept

SupplyTargetBodySiteLateralityConcept

SystemUserPreferredLanguageConcept

SystemUserTaskConctextConcept

SystemUserTypeConcept

UndeliveredProcedureReasonConcept

UndeliveredSubstanceAdministrationReasonConcept

VmrOpenCdsConcept

VMRTemplateConcept

 

 

OpenCDS Concepts

Definition

An OpenCDS Concept is a specific instance of an OpenCDS Concept Type that identifies a clinical concept familiar or useful to clinicians in the context of decision support, and has been identified and assigned an ID in OpenCDS using Apelon DTS.

...

  • Go to Apelon.opencds.org
  • Login with the      following parameters (password is “welcome2opencds”).

...

Image Added

 

The master set of concepts can be generated for local use within the OpenCDS service from this central Apelon repository using a utility in the opencds-decision-support-service module named org.opencds.terminology.OpenCDSConceptsFileCreator.  

 

Concept Mapping Specifications

Definition:

A concept mapping specification in OpenCDS is an OpenCDS Concept that has an associated Determination Method and may or may not specify an associated Code System. 

...

It should be noted that the name of a concept instance has no computable value, it is informative for human consumption.  While we try to give meaningful names to the files, all of the files in the directory are loaded by OpenCDS at startup, regardless of the file name.  All of the computable knowledge is specified by XML within the concept instance.

 

Concept Mapping Instances

Definition:

Concept mapping instances in OpenCDS relate a concept mapping specification to one or more codes from the one or more associated Code System(s) as specified by the associated Determination Method.

...

  • Apelon DTS with the built-in interface included in OpenCDS      to produce XML files,
  • any other terminology management system that can be      configured to produce XML files, or
  • by hand-crafted XML files.      

...

Concept mappings are intended to be maintained locally by multiple contributors in a distributed fashion, similar to knowledge modules.  This is in contrast to the OpenCDS Concepts, which are intended to be maintained centrally as described above.
 

Concepts in Guvnor

While it is possible to write rules for OpenCDS without using Guvnor, and without referring to OpenCDS Concepts at all, we support Guvnor and strongly recommend writing your rules in reference to OpenCDS Concepts.  The combination of Guvnor and OpenCDS Concepts produces rules that have visible and easily understood logic, and can be readily reviewed for accuracy by a clinician.

  • Guvnor is a tool for doing two things:
    • writing rules (which we sometimes refer to as       “knowledge” about “clinical concepts”), and
    • testing the logical validity of those rules

...

It is a best practice to write thorough tests in Guvnor for your rules, and to run them whenever there is any change to the rules in a KnowledgeModule (aka a “Package” in Guvnor).  This can be done in Guvnor with a single click to run all the tests for a single package.

 

Concept Types

The same Java classes that define Concept Types in OpenCDS are imported into Guvnor using the jar file from the OpenCDS module named “opencds-vmr-v1_0-internal.”  This then gives you exactly the same definitions of Concept Types that are used in the web service.  You will also need the OpenCDS Common jar from the OpenCDS module named “opencds-common.”:

Image Added

 

If you select the “Edit” tab instead of the “Assets” tab, you will be able to view (and edit) the contents of the model:

Image Added

 

If you click the “Advanced View” button, you will be able to edit the model as text, and you may need to do this:

Image Added

 

When you first create a new model in Guvnor by importing the two jars from OpenCDS, this list will be in a random order, and it will include some elements from the common jar that you don’t need.  We provide a sorted list of the import statements, which you can paste into this window after deleting the generated content.  This sorted list is maintained in the OpenCDS source code as a resource named “ImportStatementsForOpenCDSGuvnor.txt” within the opencds-vmr-1_0-internal module

...

Once you have the model defined, leave it alone!  Changing the model may break rules that are based on the model.

 

Guvnor Enumerations start with OpenCDS Concepts

There is a utility in the opencds-decision-support-service module named “org.opencds.terminology.GuvnorEnumerationCreator.java” which creates a list of enumerations that can be placed into Guvnor Package Enumeration assets by using copy and paste. 

...

A typical package in Guvnor (aka a KnowledgeModule in OpenCDS) should have the following Enumerations:

Image Added

 

If you click the “Open” button from one of the enumerations listed, you will see something that looks like this:

Image Added

 

Every enumeration listed will have a “determinationMethodCode” entry, which is generated by the tool.  Those which have an associated “openCdsConceptCode” will list those enumerations as well. 

Refer to the official Guvnor documentation, and to our document “OpenCDS Guvnor 5.3 DSL Best Practices” for more information.

 

Domain Specific Language (DSL)

Enumerations make it possible to write rules using DSLs.  These DSLs, when properly designed, may make reference to a Guvnor Enumeration, which then populates a drop-down list in the rule, listing all the OpenCDS Concepts that are included in the enumeration.  DSLs are essentially a macro language that allows the use of plain clinical statement language that is non-technical, and should use a similar language to that a clinician would use to describe what the rule is supposed to do.

...

  • Patient has had 3 inpatient encounters for asthma in the      past 1 year
  • Patient has had 2 outpatient encounters for diabetes      mellitus in the past 6 months
  • Patient has had 1 emergency encounter for acute myocardial      infarction in the past 10 days
  • Patient has had 1 well-child encounter for EPSDT in the      past 1 year

 

More information about using DSLs is found in the OpenCDS document named “OpenCDS Guvnor 5.3 DSL Best Practices.” 

 

Concept Mappings in Guvnor

We don’t think it is a good idea to reference specific codes in Guvnor.  Your rules should be written against OpenCDS Concepts, and they should be tested against OpenCDS Concepts, without worrying about how the concepts will get populated at run-time while you are in Guvnor.  Run-time mappings can then be updated or modified without changing the rules at all.

For this reason, it is a best practice to do final integration testing of OpenCDS in the run-time environment rather than Guvnor by using the OpenCDS end-to-end testing tool.  You will need to build a library of test data instances.  This library can be enhanced and resubmitted at any time, and should be used as a final validation of both updated rules and updated concept mappings. 


Concepts in OpenCDS Decision Support Service

 

OpenCDS Decision Support Service is the run-time implementation of OpenCDS.  Everything about it is designed to support the separation of clinical concepts as used in rules from the values in the data which is to be evaluated.

It has been said before, and I’ll repeat it here:  Write your rules against concepts, and separate the mappings of the codes found in your data from the rules.  OpenCDS provides internal methods at run-time which link each concept instance reference in your rules to every data element which has a mapping to the same concept instance in the concept mappings.

 

Concept Types in OpenCDS Internal Data

Concept Types have already been pretty thoroughly discussed above.  The important things to remember are that

  • They are broad categories,
  • They correspond to all the CD data elements in the vMR,
  • They are relatively fixed (adding a new Concept Type      requires a number of modifications to the OpenCDS software), and
  • They are implemented as OpenCDS Concepts, maintained in      Apelon DTS, and implemented by user-specified concept mapping      specifications which are ultimately mapped to concept codes in terminology      systems.

 

OpenCDS Internal Data, Concept Mapping Specifications, and Concept Mapping Instances

OpenCDS DSS maps input data to what we call the internal vMR structure.  While the external structure (the vMR schema) provides options to submit data in either listed (relational) structure or nested (object) structure, the internal structure is strictly the listed (relational) structure.

...

Concepts versus   Data

ProblemCode   = 493.1

Intrinsic   Asthma

ProblemCode   = 493.2

Chronic Obstructive   Asthma

ProblemCode   = 510

Emphysema

Chronic Asthma

 

True

 

Asthma

True

True

 

Pulmonary Disease

True

True

True

 

 

Updating OpenCDS Concept Mapping Specifications and Mappings

OpenCDS concept mapping specifications and mappings can be maintained in Apelon DTS.  Here is a screen shot showing maintenance of the OpenCDS Concept for Asthma in Apelon DTS:

Image Added

 

OpenCDS provides a tool to download them from Apelon DTS, named “OpenCDSConceptsFileCreator.”  Instructions for the use of this tool are described earlier in the document, in Section A.2.

Concept mapping specifications and mapping instances can alternatively be maintained manually in XML files.  This may be adequate for users with a limited number of concepts that have relatively stable code values.

 

Change Log

Date

Author

Notes

12-17-2011

David Shields

Initial document

12-20-2011

Ken Kawamoto

Minor edits; added more details on use of Apelon.

12-20-2011

David Shields

Add supporting graphics and examples, separate “OpenCDS   Concepts” from “concept instances.”

12-21-2011

David Shields

More examples, improved graphics, explain both OpenCDS and   user maintained concepts better.

4-6-2012

David Shields

Reworded definitions, improved graphics, worked on better   linkage from ideas about OpenCDS Concepts to the actual setup files used for   OpenCDS.

 

 

 

 

 

 

...