osce-rubric-review

OSCE Rubric Review

Use For

• Broad first-pass audits of existing OSCE rubrics.
• Clarifying questions before rubric analysis when case context, station constraints, modality, or educator priorities could change the review.
• Structured Rubric Maker improvement suggestions for discovered issues.
• Finding issues that may later need content validation or transformation.

Do Not Use For

• Adjudicating a disputed concern, proposed fix, severity, or consensus recommendation; use content-validation.
• Direct adaptation, restyling, score expansion, or template application as the main task; use osce-rubric-transform.
• Mechanical import from source files; use rubric-import.
• New post-encounter-note rubric drafting; use post-encounter-note-rubric.
• Generating synthetic student notes or transcripts; use generate-student-artifact.
• Dry-run grading a sample artifact against a rubric or producing a trial grade sheet; use grading-dry-run.
• Analyzing a trial grade sheet to identify rubric improvements; use evaluate-dry-run.
• Live patient simulation case design; treat as out of scope for this rubric-focused plugin.

Sibling Sequence

Use osce-rubric-review to discover and frame rubric issues before dry-run testing. Use evaluate-dry-run when the user already has a case, artifact, and trial grade sheet and wants improvements based on how the rubric behaved in practice. Escalate selected findings to content-validation when their clinical validity, severity, fairness, safety impact, or best fix is contested or high-stakes. Use osce-rubric-transform after review, validation, or dry-run evaluation when the user wants the rubric content adapted or rewritten.

Workflow

1. Read the rubric and any case scenario, station instructions, learner level, modality, or evaluator constraints.
2. Audit with the six rubric quality pillars: alignment, safety, observability, objectivity, feasibility, and reliability.
3. Focus on clinical substance before wording. Flag mismatches, missing critical steps, vague anchors, double-barrel items, mode mismatches, and score levels that are not distinguishable.
4. If context is incomplete, generate 2-4 clarifying questions with concrete answer options.
5. When asked for revisions, return structured suggestions with location, current_value, suggested_value, reasoning, priority, row, field, and optional sub.

Output Guidance

• For a quick review, lead with high-priority findings and line/row references.
• For structured suggestions, use locations like 0:QuestionName or 2:ScoringLogic.Score3.
• Keep current_value exact when editing existing text.
• Use priority values: critical, high, medium, low.

References

• Load references/rubric-schema.md for the shared schema, mode semantics, and structured suggestion conventions.
• Load references/review-framework.md for the six-pillar audit and clarifying-question workflow.
• Load references/suggestion-contracts.md when producing or validating structured improvement suggestions.