/v1/extract/contexts

A SNOMED code alone does not tell you whether the patient has the condition. "Nie toute fievre" (denies any fever) and "antecedent familial de cancer du sein" (family history of breast cancer) both mention a concept that is not an active patient problem. extract/contexts applies rule-based ConText analysis (EN + FR lexicons) to each span and returns five context axes, each with the exact trigger phrase that fired so the decision is auditable.

Pass the spans (and optional codes) you got from /v1/extract/concepts.

Request body

Response

{
  "contexts": [
    {
      "code": "386661006",
      "span": "fievre",
      "context": {
        "assertion": { "value": "negated", "confidence": 0.95, "trigger": "nie toute" },
        "subject": { "value": "patient", "confidence": 0.9, "trigger": null },
        "temporality": { "value": "current", "confidence": 0.8, "trigger": null },
        "certainty": { "value": "confirmed", "confidence": 0.8, "trigger": null },
        "role": { "value": "finding", "confidence": 0.8, "trigger": null }
      },
      "context_confidence": 0.9,
      "readable_note": "Fever - negated (patient denies)"
    }
  ]
}

Examples

curl -X POST https://api.obscyro.com/v1/extract/contexts \
  -H "Authorization: Bearer $OBS_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Il nie toute fievre. Sa mere est atteinte d un cancer du sein.",
    "language": "fr",
    "concepts": [
      { "span": "fievre", "code": "386661006" },
      { "span": "cancer du sein", "code": "254837009" }
    ]
  }'

Tips

• Every fired axis includes a trigger phrase, so you can show clinicians why a concept was marked negated or historical.
• A negated or family-subject assertion usually means the concept should not be added to the active problem list, even if extract/concepts resolved it confidently.
• Run /v1/extract/concepts first to get the spans, then pass them here in one call.

Errors