NHS North West Genomics
2.2.5 - ci-build United Kingdom flag

NHS North West Genomics - Local Development build (v2.2.5) built by the FHIR (HL7® FHIR® Standard) Build Tools. See the Directory of published versions

Report Types

Laboratory reports come in two broad kinds. Which kind a report is decides how easily it fits a FHIR Document. A FHIR Document requires a Composition with human-readable narrative, so the kind that fits worst is the discrete result (for example a list of reportable variants) - the opposite of what is often assumed.

Terms

Term used in this guide Also called Meaning
Narrative (interpretive) report unstructured / semi-structured report, opinion-based or consultative report, document-based report The result is the opinion of a clinical scientist or pathologist, written as text - often under headings, sometimes as a PDF
Discrete (numeric) result structured / coded result, analytical result, observation-based result, atomic data The result is a set of values produced by an analyser or pipeline - a code, a value, units, a reference range and a flag for each test
Synoptic report structured report, dataset report A narrative report whose key items are recorded as discrete data from a defined dataset (for example the RCPath cancer datasets, or the variants in a genomic report)

Where a single word is wanted on each side, this guide uses report for the narrative kind and result for the discrete kind.

In messaging the difference shows as:

  Narrative report Discrete result
HL7 v2 OBX of type FT / TX (often many lines with \.br\), NTE, or ED (an embedded PDF) OBX of type NM, CWE / CE, SN, with OBX-6 units, OBX-7 range, OBX-8 flag
FHIR DiagnosticReport.conclusion / presentedForm, Observation.valueString, Composition.section.text Observation.valueQuantity / valueCodeableConcept, referenceRange, interpretation

Which Reports Are Usually Narrative

Usually narrative (text with headings, or PDF):

  • Genomic and molecular pathology - Genomic Laboratory Hub reports, NGS panels, whole genome sequencing and variant interpretation are today mostly sent as a PDF. The reportable variants can be sent as discrete data alongside it (for example the ctDNA examples), but the PDF remains the authorised report
  • Cytogenetics - karyotype (ISCN string) and FISH, with interpretation
  • Haematology interpretation - blood film comments, bone marrow aspirate and trephine, flow cytometry immunophenotyping
  • Histopathology / surgical pathology - clinical details, macroscopic, microscopic, conclusion. Cancer resections follow the RCPath datasets (synoptic), but are often still sent as text with a table inside it
  • Specialist and referral tests - results from reference laboratories, often PDFs attached by the local LIMS rather than re-keyed (see Inter Laboratory Workflow)
  • Interpretive comments on discrete results

Usually discrete:

  • quantitative molecular monitoring (for example BCR-ABL1)
  • routine clinical biochemistry and haematology
  • most immunoassays and serology with numeric or qualitative results

Report Patterns

Each diagram shows the FHIR resources, the HL7 v2 segment each one comes from, and the element that links them.

Discrete Results

For FHIR Messages and Resource Access. A report is one DiagnosticReport (the OBRs sharing OBR-3). Each DiagnosticReport.result is either a panel - a test group, one OBR with several OBX, as an organizer Observation Panel whose hasMember are the results - or a single result (an OBR with one OBX).

Panel - for example a chromosome analysis (karyotype and FISH):

flowchart LR
    DR["DiagnosticReport<br/>ORU_R01 report<br/>OBR-3, OBR-22, OBR-25"]
    P["Observation - panel<br/>test group (organizer)<br/>OBR-4 e.g. Chromosome analysis"]
    R1["Observation - result<br/>OBX e.g. ISCN karyotype"]
    R2["Observation - result<br/>OBX e.g. FISH"]
    DR -- result --> P
    P -- hasMember --> R1
    P -- hasMember --> R2

Single results - for example the reportable variants of a ctDNA panel, each linked directly from the report:

flowchart LR
    DR["DiagnosticReport<br/>ORU_R01 report<br/>OBR-3, OBR-22, OBR-25"]
    R1["Observation - result<br/>OBX e.g. Variant"]
    R2["Observation - result<br/>OBX e.g. Variant"]
    DR -- result --> R1
    DR -- result --> R2

Both patterns can be mixed in one report - for example a panel (one OBR) and a single result (a second OBR) from one request:

flowchart LR
    DR["DiagnosticReport<br/>OBRs sharing OBR-3"]
    P["Observation - panel<br/>OBR 1"]
    R1["Observation - result<br/>OBX"]
    R2["Observation - result<br/>OBX"]
    S["Observation - result<br/>OBR 2 / OBX"]
    DR -- result --> P
    P -- hasMember --> R1
    P -- hasMember --> R2
    DR -- result --> S

Examples: Shire cytogenetics (panel, karyotype), Shire cytogenetics with FISH (panel with two results) and OMICS DSS ctDNA test results (single results) - all FHIR Messages.

Narrative Reports

Rendered report (PDF) - an OBX of type ED (Base64 PDF) or RP (a link) becomes a DocumentReference and a Binary. The DiagnosticReport stays the parent: presentedForm.url gives the PDF directly, and the DocumentReference points to the same PDF and back to the report:

flowchart LR
    DR["DiagnosticReport<br/>ORU_R01 report"]
    DOC["DocumentReference<br/>OBX type ED / RP<br/>OBX-3 document type"]
    BIN["Binary<br/>the PDF<br/>OBX-5 ED-5 Base64 data"]
    DR -- presentedForm.url --> BIN
    DOC -- content.attachment.url --> BIN
    DOC -- context.related --> DR

A rendered report is normally exchanged in one of two ways (see the general rule above):

Text report in a FHIR Message or Resource Access - the text is in OBX of type ST / FT (or TX lines), and becomes a result whose value is the text - for example the ISCN karyotype and FISH results in the Shire cytogenetics example:

flowchart LR
    DR["DiagnosticReport<br/>ORU_R01 report"]
    R["Observation - result<br/>valueString<br/>OBX type ST / FT<br/>(line breaks become new lines)"]
    DR -- result --> R

Text report in a FHIR Document (Document Exchange) - the link is simply Composition to section to narrative. The OBX of type ST / FT is used to create the section: OBX-3 gives section.code (and title), and OBX-5 (with \.br\ as line breaks) gives section.text:

flowchart LR
    C["Composition<br/>FHIR Document header<br/>Composition.text"]
    S["section<br/>section.code from OBX-3"]
    N["Narrative<br/>section.text<br/>from OBX-5 (ST / FT)"]
    C -- section --> S
    S -- text --> N

Where the narrative is only available as a PDF, the section entry is the DocumentReference (and Binary) - for example the Presented Format section of the ctDNA FHIR Document, and the Genomics Report Targeted Variant document.

Discrete Results in a FHIR Document (Document Exchange)

A Composition can also carry discrete results. The Composition is the parent: the DiagnosticReport is an entry in a section, and links on to its panels and results as in the FHIR Message. The Observations are also entries in the sections: a section entry can be the panel Observation or an individual result directly (for example each reportable variant in a Findings section). The section's narrative is generated from those results - see Why Discrete Results Are Harder to Send as a FHIR Document:

flowchart LR
    C["Composition<br/>FHIR Document header<br/>Composition.text"]
    S["section<br/>section.text = generated<br/>table of results"]
    DR["DiagnosticReport<br/>ORU_R01 report"]
    P["Observation - panel<br/>OBR-4"]
    R1["Observation - result<br/>OBX"]
    R2["Observation - result<br/>OBX"]
    C -- section --> S
    S -- entry --> DR
    S -- entry --> P
    S -- entry --> R2
    DR -- result --> P
    P -- hasMember --> R1
    P -- hasMember --> R2

Examples: ctDNA UGR Phase 2 - EU Laboratory Report, ctDNA and Lynch syndrome (FHIR Documents).

Why Discrete Results Are Harder to Send as a FHIR Document

A FHIR Document is a Bundle of type document whose first entry is a Composition. HL7 Europe Composition-eu-lab (the parent of the Composition Genomic Report) requires at least one section and makes section.text mandatory (1..1). The narrative is the attested, human-readable content of the document.

For a narrative report this fits naturally: the laboratory already wrote the narrative. Its headings (findings, implications, recommended actions) become sections, the authorised text becomes section.text, and what is coded is added alongside it. The document is a wrapper for content that already exists.

For a discrete result there is usually nothing to put there:

  • No narrative exists. A set of variants or a quantitative result is a set of values, not text; the ORU^R01 has no text form of it beyond a few NTE comments. The sender SHALL make one to meet section.text 1..1
  • The narrative is generated, not authored. The sending system creates a table from the coded data. It SHALL agree with the coded data - a value rounded differently, or a lost flag or comment, is a clinical safety issue - so it has to be generated from the resources, and tested as clinical content
  • Who attests it is unclear. The laboratory authorised the values (or the PDF), not the rendered table. The document's attested content is something no one in the laboratory has seen (Open - how this is covered in the clinical safety case, DCB0129 / DCB0160)
  • The sections have to be invented. A narrative report has headings; discrete results only have test groups (OBR), so each test group becomes a section, with the test group code as section.code
  • It is extra work for no new information. The receiver files and displays the coded Observations; the narrative adds a second copy that must be kept in step. The ctDNA FHIR Document example has placeholder section text ("see PDF") rather than a rendering of the variants, which shows the difficulty
  • It does not suit other routes. HL7 v2 ORU^R01 and FHIR Messages carry discrete results without a Composition. HL7 Europe DiagnosticReport-eu-lab requires the DiagnosticReport.composition extension, which is why this guide's Diagnostic Report is derived from EU Core instead

In short, the FHIR Document model suits narrative reports, where a document already exists. For discrete results it adds an artificial document that has to be generated.

References