Infermedica Symptom Assessment REST API

Known Gotchas

• ⚠ INTERVIEW IS STATEFUL AND ADAPTIVE: Infermedica's symptom interview is stateful; automated stateless assumption creates incoherent_interview for assessments not maintaining the full evidence array across question-answer cycles; automated must accumulate evidence across all interview turns
• ⚠ EVIDENCE ARRAY GROWS EACH TURN: Each new symptom/answer must be appended to the evidence array, not replace it; automated replace assumption creates evidence_lost for interview turns that overwrite previous evidence; automated must append new evidence to existing evidence array
• ⚠ CONCEPT IDs ARE MEDICAL CODES: Infermedica uses its own medical concept IDs (e.g., s_21 for headache); automated ICD-code assumption creates concept_not_found for symptoms specified with ICD codes instead of Infermedica IDs; automated must use Infermedica concept IDs from the concept library or NLP parser
• ⚠ INITIAL EVIDENCE SEEDS THE INTERVIEW: The first API call should include chief complaint as initial evidence; automated empty-start assumption creates generic_questions for interviews without initial symptom to focus assessment; automated must provide initial symptom evidence to begin targeted interview
• ⚠ TRIAGE SHOULD ALWAYS BE LAST: Triage API call (care level recommendation) should be called after sufficient evidence collection; automated early-triage assumption creates low_confidence_triage for triage calls without sufficient symptom evidence; automated should collect adequate symptoms before requesting triage recommendation