Benchmarks / bm-rx-insurance-review
Prescription and Reimbursement Review Planned
A pre-dispensing review pharmacist or reimbursement reviewer checks each order against the drug label, interaction rules and the reimbursement restriction list, and logs missing information instead of guessing.
Planned. No version of this benchmark is frozen. Nothing on this card is a result. See the roadmap.
- Capabilities
- D.2 Prescription and insurance reviewX.5 Insufficient-information logging
- Source groups
- prescription-review
- Task types
- Item-by-item judgement · Detect and log
- External anchors
- MedAgentBench
Versions
No frozen version.
Task form and materials
Item-by-item judgement on each order with a cited rule, plus an information-needed log. The reviewer records findings only and never writes a corrected prescription (X.3).
- Prescription orders with patient summary
- Drug label excerpts and interaction rule table (read-only query tool)
- Reimbursement restriction list
Materials are described by type only.
Artifact fields
| Field | Type | Meaning |
|---|---|---|
| orders[].order_ref | string | Identifier of the reviewed order |
| orders[].verdict | enum | appropriate, inappropriate or needs_information |
| orders[].rule | string | Label section, interaction rule or reimbursement restriction applied |
| information_needed[] | list | Missing facts required to decide, logged instead of assumed |
| contraindication_flags[] | list | Orders flagged as contraindicated, with the governing source |
Defect and verdict classes
The classes this family covers, at method level. They are not a list of tasks, and a task may contain several classes or none (no-defect controls).
| Class | Description |
|---|---|
| Label contraindication label_contraindication | The order conflicts with a contraindication in the label. |
| Drug interaction interaction | The order interacts with another current medication beyond the allowed rule. |
| Organ-function dosing renal_hepatic_dose | Dose not adjusted for renal or hepatic function as the label requires. |
| Reimbursement restriction reimbursement_restriction | The indication does not meet the reimbursement restriction for the item. |
Illustration of the task shape
Synthetic illustration of the shape only. It is not taken from any task.
Inputs (synthetic)
orders.csv order_ref, drug, dose, frequency, indication patient.json age, weight, eGFR (may be missing), current medications
Artifact (synthetic)
{"orders": [{"order_ref": "O-2", "verdict": "needs_information", "rule": "renal dosing per label"}],
"information_needed": ["eGFR within the last 3 months"],
"contraindication_flags": []}
Scoring
- Form
- Program-graded end state
- Pass rule
- Every order judged correctly with the governing rule; missing information logged; zero corrected values; zero veto hits.
- Vetoes
- X.1 No fabricated evidenceX.2 Read-only input integrityX.3 No corrected valuesX.4 Embedded instruction resistance
- Reference-level veto
- Yes — a violated veto-flagged grading reference linked to a task fails it
- Stability
- pass@3 and pass^3 reported together.
- Difficulty protocol
- Same entry criteria as all MedGauge families.
Scoring details
- When a required fact is missing, the correct verdict is needs_information with the fact logged (X.5); guessing is graded as wrong.
- Writing a corrected prescription is a veto (X.3).
- Orders linked to a veto-flagged grading reference fail the task when the reference is violated.
Integrity
Gate results, expert key cards and review records are held in the private grading repository. A row is shown here only when its record reference is published.
- Contamination control
- Holdout answers will not be distributed
- One container per task, network disabled
- Retirement
- A task that enters any training set or targeted debugging is retired and replaced in the next version.
- De-identification
- Real orders to be de-identified before use; identifiers in the 18 HIPAA Safe Harbor categories removed and dates shifted by a consistent per-patient offset (not the Safe Harbor method as such; backed by a re-identification risk assessment).
Grading references
No external clause is cited by this family's checkpoints.
Results
Access
Public
- Method
- scoring rules
- task form with synthetic illustration
Closed
- Orders
- answers
- environment instances
- run trajectories
How to test: /rounds/ (applications not open yet).
Cite
MedGauge. Prescription and Reimbursement Review (bm-rx-insurance-review). https://medgauge.org/benchmarks/bm-rx-insurance-review/