sauce.ai / phenotype

Validated phenotyping algorithms, ranked

Name a condition. The service searches PubMed, Europe PMC and the citation graph for studies that validated a case-ascertainment algorithm in claims or EHR data, spells each algorithm out (codes, counting logic, OMOP SQL), links it to its validation references, and ranks the algorithms by pooled accuracy and strength of evidence for your use and data.

1. Search + one-hop citation snowball; model-suggested papers kept only if they resolve in PubMed.
2. Screen abstracts; read included papers (open-access full text when available).
3. Extract algorithms and accuracy figures, each backed by a verbatim quote.
4. Pool across validations, grade the evidence (QUADAS-2 / GRADE-style), rank.

Condition

Condition

Synonyms comma-separated, optional

Your study

Intended use decides which accuracy metrics drive the ranking

Your data

Coding era

Country optional

Expected prevalence optional, e.g. 0.003 — gives PPV/NPV at your prevalence

Email optional — report is emailed when done

Search depth

Records to screen

Papers to read in full

Citation snowballing Add model-suggested known papers (verified in PubMed)

Notes for the reviewer optional, e.g. "pediatric", "Medicare", "need incident cases"

Output is a machine-assisted evidence review, not a substitute for reading the papers. Figures that cannot be matched to a verbatim quote are flagged; code lists often live in supplements the service cannot read.

Find algorithms