MII IG Modul Patho
2027.0.0-ballot - Germany

MII IG Modul Patho - Local Development build (v2027.0.0-ballot) built by the FHIR (HL7® FHIR® Standard) Build Tools. See the Directory of published versions

Terminologies

For coding the entire pathology report as well as the grouper observations, LOINC terms from the DOC.* class or the Doc scale are used.

The generic individual observations themselves should be coded as precisely as possible in the way they were observed, i.e. on which specimen type, with which method, etc.

The numerous possibilities of describing observations with code-value pairs are limited, for non-numeric observation results, to two (three) basic options:

  1. The Observation.code represents the type of the observation and the Observation.value the non-numeric observation value.
  • Observation.code: extent of extraprostatic tumor spread
  • Observation.valueCodebleConcept: focal
  1. The Observation.code is expressed in a way that does not code the action of the observation but represents a statement about a finding. In this way the Observation.value is a qualifier that codes the presence or absence of the finding.
  • Observation.code: perineural sheath invasion
  • Observation.valueCodebleConcept: yes/no indicator or Observation.valueBoolean
  1. The Observation.code is represented by a statement about a finding as in 2., the Observation.value is omitted.
  • Observation.code: perineural sheath invasion
  • Observation.valueCodebleConcept:
  • Observation.dataAbsentReason: not-applicable

For the first option, all LOINC terms of the types "Lab" and "Clinical" are available. If no suitable LOINC code is available, a SNOMED CT code is used that comes from the hierarchy axis 363787002 (Observable entity) or 386053000 (Evaluation procedure). The Observation.values are preferably SNOMED CT codes or corresponding LOINC answer codes.

For the second option, the Observation.code is preferably a SNOMED CT code from the hierarchy axes 404684003 (Clinical finding), 413350009 (Finding with explicit context) or 272379006 (Event). The Observation.value is preferably represented by the SNOMED CT axis 362981000 (Qualifier value). Postcoordinated SNOMED CT expressions should be avoided.

For the third option, a SNOMED CT code from the above-mentioned hierarchy axes should preferably be used. An Observation.dataAbsentReason.value for the missing Observation.value should be used to indicate why this value is missing.

If no suitable LOINC terms or SNOMED CT codes can be found, further standardized code systems (e.g. HL7 V2.x or V3, ICD-10, ICD-11, ICD-O-3, ADT/GEKID) are to be preferred over codes from local code systems.

For synoptic reports on tumors, the SNOMED CT Clinical Implementation Guide for Cancer Synoptic Reporting (https://confluence.ihtsdotools.org/display/DOCCANSIG/SNOMED+CT+Clinical+Implementation+Guide+for+Cancer+Synoptic+Reporting) should be taken into account.

The units of measurement for measured values must be given in UCUM units so that the results can be converted into one another. Validators can be used to ensure that the units used are permissible.

For the coding of specimens and procedures, SNOMED CT is used.


Stainings

For staining procedures the following approach is recommended:

Where no SNOMED CT concept exists: SNOMED CT carries no code for many antibodies — p63, for example, is not contained in the International Edition. In these cases a local code in Substance.code is permissible; it should live in a namespace of the institution and be referenced via Specimen.processing.additive. For a structured statement of the antibody type, 1236876007 Type of antibody used in immunohistochemistry technique is additionally available.

In addition, a postcoordinated staining process can be coded:

  • staining process(procedure):usingSubstance=dye or =antibody
  • staining process(procedure):directSubstance=target antigen
  • or a combination of these codings.

With postcoordinated codes it should be noted, however, that they are hardly evaluable without a suitable terminology server. For this reason our recommendation would be to first transfer the relationships between staining processes and their respective staining substances to the FHIR information model and to represent them by means of .processing.procedure and .processing.additive.

The processing procedures used in this module are collected in the ValueSet MII VS Patho Processing Procedure [SNOMED CT]. A dedicated ValueSet for staining substances is in preparation.


Module ValueSets

In addition to the terminologies mentioned above and further international terminologies (ICD-O-3 and UICC TNM), the module Pathology Report defines its own ValueSets. Please note that none of the ValueSets contains an expansion. This has to be performed with a terminology server before use.

In addition, several data elements in the FHIR resources are coded by HL7 V2.x codes.

The following table is generated automatically from the FHIR package and lists one row per ValueSet with the profiles and elements that bind it:

TitleURLDescriptionUsage (profile: element, strength)
MII VS Patho All [LOINC]https://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-all-loincValueSet for general LOINC VS bindingMII PR Patho Base Observation: Observation.code.coding (preferred),MII PR Patho Composition: Composition.section.code.coding (required)
MII VS Patho Collection Method [SNOMED CT]https://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-collection-method-snomed-ctValueSet for the clinical collection methodMII PR Patho Specimen: Specimen.collection.method (extensible)
MII VS Patho Composition Type KDLhttps://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-composition-type-kdlValueSet for the categorization of Pathology Report documentsMII PR Patho Composition: Composition.type.coding (extensible)
MII VS Patho Composition Type [Snomed CT]https://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-composition-type-snomed-ctValueSet for the types a pathology Composition can haveMII PR Patho Composition: Composition.type.coding (extensible),MII PR Patho Composition: Composition.category.coding (extensible)
MII VS Patho Container Type [SNOMED CT]https://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-container-type-snomed-ctValueSet for container type based on IHE PaLM's ContainerEntityClassType ValueSetMII PR Patho Specimen: Specimen.container.type (required)
MII VS Patho Media Modality [SNOMED CT]https://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-media-modality-snomed-ctValueSet with some of the main Media.modalities needed in the pathological scopeMII PR Patho Attached Image: Media.modality (extensible)
MII VS Patho Problem List [SNOMED CT]https://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-problem-list-snomed-ctValueSet containing the recommended vocabulary for describing problems, according to the IHE Patient Care Coordination Technical Framework (Vol. 2)MII PR Patho Problem List Item: Condition.code (extensible)
MII VS Patho Processing Procedure [SNOMED CT]https://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-processing-procedure-snomed-ctSpecimen processing procedureMII PR Patho Specimen: Specimen.processing.procedure (extensible)
MII VS Patho Report Category HL7https://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-report-category-hl7ValueSet for the categorization of Pathology Report documentsMII PR Patho Report: DiagnosticReport.category (extensible)
MII VS Patho Report Code [LOINC]https://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-report-code-loincValueSet for the report type codes of pathology DiagnosticReportsMII PR Patho Report: DiagnosticReport.code.coding (extensible)
MII VS Patho Section Types [LOINC]https://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-section-types-loincValueSet for IHE PaLM APSR's section typesMII PR Patho Finding: Observation.category (required)
MII VS Patho Service Request Codehttps://www.medizininformatik-initiative.de/fhir/ext/modul-patho/ValueSet/mii-vs-patho-service-request-codeValueSet for a preferred set of ServiceRequest codesMII PR Patho Service Request: ServiceRequest.code (extensible)