Consensus-Anchored Multi-Corpus Clinical Chatbot for Long COVID
A new clinician-facing chatbot integrates four evidence sources—expert consensus guidance, PubMed literature, registered trials, and living systematic reviews—to support Long COVID clinical decisions. The system always anchors responses in consensus guidance while allowing parallel retrieval of other sources. In an automated evaluation on 50 questions, it achieved comparable mean ratings to OpenEvidence with higher scores and lower variability in an LLM-judged comparison.
Key facts
- Chatbot organizes four sources: expert-curated consensus guidance, PubMed literature, registered interventional trials, living systematic reviews.
- Consensus guidance is always included to frame responses.
- Remaining sources are retrieved in parallel when selected by the user.
- Exploratory automated evaluation on 50 clinician-facing questions.
- Chatbot showed comparable mean ratings to OpenEvidence.
- Numerically higher scores and lower score variability in LLM-judged comparison.
- Long COVID poses challenge for clinical decision support due to distributed evidence.
- Sources have different update cycles, evidentiary roles, and levels of clinical maturity.
Entities
Institutions
- arXiv
- OpenEvidence
- Semantic Scholar
- arXivLabs