MedGaugeCapabilities / D · Medical practice
D.19 Treatment option appraisal and contraindication detection Proposed
Given case material and candidate treatments, the agent fills a structured assessment form judging whether each option complies with guidelines and labels and flagging contraindications and items that need manual adjustment. Graded by checkpoints on each option; the agent does not write a corrected prescription (see X.3).
Proposed. This capability is under RFC and does not yet carry levels.
Evidence: planned. Only planned benchmark cards list this capability so far; no frozen benchmark version exists for it.
- Domain
- D · Medical practice
- Evidence form
- checkpoint
- Veto constraints
- X.9 Medical safety boundary
- Depends on
- A.6 Research evidence semanticsD.18 Diagnostic reasoning and differential
- Required by
- —
- Internal task families
- —
- External anchors
- PhysicianBench
- Version
- 0.1.0 · 0.1.0
Definition in detail
Treatment choices must fit guidelines, labels and the individual patient. Missed contraindications and unflagged dose adjustments are avoidable harms.
Inputs a task gives
- Case material with diagnoses, laboratory values and current medications
- A set of candidate treatment options
- Guideline excerpts
- Drug label excerpts
Artifact required
An assessment form: options[] (option, guideline_compliant, label_compliant, contraindications[] with evidence, needs_manual_adjustment with reason).
How it is graded
Checkpoints compare each option's fields with the reference; where a task cites grading references, there is one checkpoint per reference. The agent does not write a corrected prescription (X.3). X.9 is a one-shot veto.
In scope
- Judging each option against guideline and label
- Flagging contraindications from the patient's data
- Flagging options that need manual adjustment
Out of scope
- Writing a corrected prescription - prohibited under X.3
- Line-by-line prescription and payment review - graded under D.2
- Population-specific adjustments tested by paired cases - graded under X.10
Typical failures
Expected failure patterns; not measured rates.
- Approving an option without checking the laboratory value that contraindicates it
- Applying a label threshold from a different region or version than the one supplied
- Recommending a replacement regimen instead of flagging the issue
Standards: std-ada-soc-2026 · std-cds-cn-diabetes
Benchmarks
- Paired Population Cases Planned
Grading references
No grading reference is mapped to this capability.