FHIR R4 integration

Send a prescription as a FHIR R4 Bundle and get the risk analysis back as a FHIR Bundle of RiskAssessment + DetectedIssue. Ideal for hospital systems (EHR/HIS) that already speak FHIR.

Copia um prompt pronto com fluxo de integração, payload e exemplos.

Endpoint

POST /api/v1/fhir/$check (API key auth, scope interactions:check).

Query	Default	Description
format	fhir	fhir returns a FHIR Bundle; internal returns the native SauBit response (prescription_risk_level, clinical_alerts, …).

Resources accepted in the Bundle

FHIR resource	Mapped to
MedicationRequest / MedicationStatement	Medication (name, dose, frequency, route). Status stopped/cancelled/entered-in-error/draft is ignored.
Condition	Patient clinical condition.
AllergyIntolerance	Allergy.
Observation	Lab/context by LOINC code or text: potassium (2823-3), INR (6301-6), QTc, clearance/eGFR (33914-3), weight (29463-7), pregnancy (82810-3).
Patient	Age (from birthDate) and sex (gender).
Encounter	Care unit/setting (class/type) → per-unit governance (ICU/pediatrics/oncology).
DiagnosticReport	Contained Observations → labs; conclusionCode → condition.
Practitioner	Accepted for completeness; no effect on the deterministic analysis.

• The drug name comes from medicationCodeableConcept.text, coding[].display/code or medicationReference.display.
• Dose: dosageInstruction[0].doseAndRate[0].doseQuantity (g converted to mg).
• Frequency: dosageInstruction[0].timing.repeat (e.g. frequency:1, period:8, periodUnit:"h" becomes 8/8h).

Example

curl -X POST "https://api.saubit.com.br/api/v1/fhir/\$check" \
  -H "X-API-Key: sm_live_..." \
  -H "Content-Type: application/fhir+json" \
  -d '{
    "resourceType": "Bundle",
    "type": "collection",
    "entry": [
      { "resource": { "resourceType": "MedicationRequest", "status": "active",
        "medicationCodeableConcept": { "text": "Ibuprofeno" } } },
      { "resource": { "resourceType": "MedicationRequest", "status": "active",
        "medicationCodeableConcept": { "text": "Naproxeno" } } },
      { "resource": { "resourceType": "Condition", "code": { "text": "Insuficiência renal" } } }
    ]
  }'

Response (FHIR)

{
  "resourceType": "Bundle",
  "type": "collection",
  "total": 3,
  "entry": [
    {
      "resource": {
        "resourceType": "RiskAssessment",
        "status": "final",
        "method": { "text": "SauBit clinical engine" },
        "prediction": [
          { "outcome": { "text": "Adverse drug event / prescription risk" },
            "qualitativeRisk": { "text": "high" }, "probabilityDecimal": 0.82 }
        ]
      }
    },
    {
      "resource": {
        "resourceType": "DetectedIssue",
        "status": "preliminary",
        "severity": "high",
        "code": { "text": "drug_disease: NSAID may worsen renal function" },
        "detail": "Avoid in this condition…",
        "implicated": [ { "display": "Ibuprofeno" } ]
      }
    }
  ]
}

Severity mapping → DetectedIssue.severity: CONTRAINDICADO/GRAVE → high, MODERADO → moderate, LEVE → low. The prescription risk becomes a RiskAssessment (qualitativeRisk + probabilityDecimal = score/100).

Errors

Invalid bundles return a FHIR OperationOutcome with issue[].severity = "error":

{ "resourceType": "OperationOutcome",
  "issue": [ { "severity": "error", "code": "invalid",
    "diagnostics": "No MedicationRequest/MedicationStatement with a valid medication in the Bundle." } ] }

The analysis behind it is exactly the same as /interactions/check — including clinical_alerts, governance and the fail-safe policy.