metadata
license: mit
language:
- en
task_categories:
- text-generation
tags:
- clarus
- clinical
- epro
- patient-reported-outcomes
- quad-coupling
pretty_name: >-
Clinical Quad Coupling ePRO Compliance Device Update Notification Bias Site
Coaching Governance Interim v0.1
Clarus Clinical Quad Coupling ePRO Integrity v0.1
What this dataset is
This dataset tests whether a model can detect ePRO integrity risk driven by four interacting nodes.
Quad coupling nodes
- Compliance drop or patterned missingness
- Device or app update and reminder configuration
- Site influence or coaching and backfill behavior
- Governance pressure from interim reads or submission reliance
Input
- One vignette
Output
Return strict JSON only.
Required output JSON keys
- epro_integrity_risk
- risk_type
- driver_nodes
- recommended_action
- action_detail
- rationale
- confidence
Files
- data/train.csv
- data/test.csv
- scorer.py
Run scoring
Create JSONL predictions
{"id":"EPRO-T01","output":"{...your json...}"}
Run
python scorer.py --gold_csv data/test.csv --preds_jsonl preds.jsonl