Evidentia
Evidentia is a provenance-aware, longitudinal, clinical diagnosis hypothesis-generation and Bayesian ranking system with next-best clinical questioning.
Core trick
Extracts HPO phenotype terms from clinical notes, reverse-indexes them against disease ontologies, and Bayesian-reranks candidate diagnoses as each note arrives.
Description
Evidentia is a provenance-aware, longitudinal, clinical diagnosis hypothesis-generation and Bayesian ranking system with next-best clinical questioning.
Data from clinical unstructured notes are extracted and normalized into Human Phenotype Ontology (HPO) terms and are reverse indexed search via the Monarch Initiative (symptoms-> disease ontology map) to generate a list of disease conditions. Reactome pathway look up establishes potential linkages that strengthen the claim of these constellation of symptoms, creating a ranking system per episode of care. As the notes progress over time, the weight of the evidence shifts accordingly, self-adapting to the greater clinical picture.
README excerpt
# Evidentia **A longitudinal clinical-reasoning demo that replays a patient's chart as an evolving differential, shows exactly what moved it, and names the single most informative question to ask next.** > _Provenance-aware longitudinal hypothesis routing with next-best clinical questioning._ > ⚕ **Synthetic educational prototype. Includes labeled scenario augmentation. Relative > probabilities are confined to a retrieved candidate universe and are not clinically > calibrated. Not for diagnosis or patient care.** --- ## What it does A patient's diagnostic clues are scattered across years, specialties, and free text. Each encounter is documented as a snapshot — but the diagnosis lives in the **trajectory**. Evidentia is a **note-centric** workspace that reconstructs that trajectory and makes it auditable: - **Encounter trajectory** (top, horizontal) — a longitudinal timeline of encounters left→right, with the leading hypothesis after each and a trend line of the selected hypothesis climbing over time. Click a stop to open that encounter; **＋ New note** to start a consult. - **Clinical record** (center) — the encounter's note/transcript/FHIR, paginated, with the substrings the compiler lifted into typed observations **highlighted inline**. Hover a highlight for the HPO-normalized observation; the color is that finding's favor toward the **currently selected hypothesis** (green supports, red argues against, gray neutral), so the same text recolors as you swi