# OpenOnco evaluation and application-paper working outline

Status: protocol and methods outline, not a completed study or submission.
No clinical effectiveness, time savings or safety benefit has been measured.

## Proposed research question
Can reviewers reproduce deterministic source-cited drafts and locate
rule/source discrepancies using the public synthetic review packet?

## Methods available now
Versioned YAML KB, Python rules engine, browser Pyodide execution, source
identifiers, bilingual audit pages and a DLBCL first-line synthetic packet.
Freeze the Git commit and KB revision before evaluation. Record environment,
scenario IDs, expected engineering contracts and complete traces. Technical
contract passes are distinct from clinical correctness and sign-off.

## Evaluation to perform
1. Recruit qualified oncology reviewers; record credentials and conflicts.
2. Independently review synthetic cases in a familiar disease area.
3. Record missing input, unsafe output, source mismatch and severity.
4. Report inter-reviewer disagreement and adjudication; retain negative results.
5. Measure completion time only with a defined comparator and consent.
6. Report denominators, uncertainty, exclusions and KB review limitations.

## Manuscript outline
Background; governance and intended use; architecture; source licensing;
reproducible evaluation methods; results once collected; limitations;
data/code availability; contributions; conflicts; ethics statement as applicable.

## Publication gates
Accountable real authors and clinical co-authors; actual evaluation dataset;
institutional determination of applicable ethics requirements; journal scope
and current submission instructions; no fabricated affiliations or results.
Conference abstracts follow the same evidence gates. No journal/conference
submission has occurred. Patient data cannot be published without the
repository's consent, de-identification and ethics requirements.