Skip to main content

Preprocessed ePI (p(ePI))

p(ePI) (Preprocessed ePI) is the intermediate state of an ePI after semantic annotations have been embedded by Preprocessors, preparing it for Lens execution.

Creation Process​

Raw ePI → Preprocessor(s) → p(ePI)

Preprocessors transform ePI by:

  1. Analyzing narrative text
  2. Identifying medically relevant sections
  3. Linking to standard terminologies
  4. Embedding annotations in FHIR structure

Structure​

p(ePI) retains the ePI Bundle structure but adds:

In the Composition resource:

{
  "extension": [{
    "url": "http://hl7.eu/fhir/ig/gravitate-health/StructureDefinition/HtmlElementLink",
    "extension": [{
      "url": "elementClass",
      "valueString": "diabetes-warning"
    }, {
      "url": "concept",
      "valueCodeableConcept": {
        "coding": [{
          "system": "http://snomed.info/sct",
          "code": "73211009",
          "display": "Diabetes mellitus"
        }]
      }
    }]
  }]
}

HTML Class Attributes​

In the Composition narrative (text.div):

<p class="diabetes-warning">
  Use with caution in patients with diabetes.
</p>

Stacking​

p(ePI) may result from multiple Preprocessors:

  • First preprocessor: ePI → p(ePI)₁
  • Second preprocessor: p(ePI)₁ → p(ePI)₂
  • Final p(ePI) contains all annotations

Usage​

p(ePI) is consumed by:

Key Characteristics​

  • Original content preserved: Text unchanged from ePI
  • Semantically enriched: Linked to medical concepts
  • Machine-readable: Annotations enable automated processing
  • Lens-ready: Prepared for attention modification

Examples​

See Gravitate-Health IG processed ePI examples.