trial_id stringclasses 10
values | protocol_deviation stringclasses 3
values | staffing_drift stringclasses 2
values | adjudication_variance stringclasses 2
values | missingness_bias stringclasses 2
values | label stringclasses 3
values | signal stringclasses 10
values |
|---|---|---|---|---|---|---|
T501 | high | yes | high | high | collapse_risk | All nodes high with staffing drift. Process control collapses. |
T502 | low | no | low | low | coherent | All nodes low with stable staffing. Clean operations. |
T503 | high | no | high | low | tradeoff | High deviation and adjudication variance but missingness contained and staffing stable. |
T504 | medium | yes | high | high | tradeoff | Staffing drift plus high adjudication and missingness creates strain without full max collapse. |
T505 | high | yes | low | high | tradeoff | Deviation and missingness high with staffing drift but adjudication stable. |
T506 | high | yes | high | low | tradeoff | Deviation high with staffing drift and adjudication variance but missingness contained. |
T507 | low | yes | low | high | tradeoff | Staffing drift plus missingness bias creates governance strain. |
T508 | medium | no | high | high | tradeoff | High adjudication and missingness with no staffing drift. |
T509 | low | no | high | low | tradeoff | Adjudication variance high while other nodes stable. |
T510 | medium | no | low | low | coherent | Mostly stable with one medium deviation and no staffing drift. |
Clinical Quad Protocol Deviation Staffing Drift Adjudication Variance Missingness Bias v0.2
What this dataset does
It tests whether a model can detect operational collapse risk in trial conduct.
The quad nodes
- protocol_deviation
- staffing_drift
- adjudication_variance
- missingness_bias
Labels
coherent
- stable staffing
- low drift and low bias
- operations remain controlled
tradeoff
- mixed strain
- issues exist but do not meet collapse pattern
collapse_risk
- all level nodes high and staffing drift present
- process control fails across the trial pipeline
What changed in v0.2
- Fixed repo name typo from linical to clinical
- Version bumped so scorer updates are visible
- New scorer with validation, confusion, and error sampling
- Added risk_score and rule_pred diagnostics
Files
data/train.csv
data/test.csv
scorer.py
Run scoring
python scorer.py --preds_csv predictions.csv --gold_csv data/test.csv
- Downloads last month
- 7