ahrq-picme-assessment

AHRQ PiCMe Assessment

Use the AHRQ PiCMe framework to systematically assess a research gap across 6 dimensions.

HARD-GATE

<HARD-GATE>

• Input must be a GapRecord with status: "complete"
• All 6 dimensions (P/I/C/M/E + overall verdict) must be completed; none may be skipped
• Each dimension must have an independent score (1-5) and a textual rationale
• overall_verdict must be one of "strong" | "moderate" | "weak"

</HARD-GATE>

Pipeline

1. Precondition check: verify completeness of the input GapRecord; confirm the domain field is valid
2. Population (P): identify the target population/system/dataset the gap concerns; assess clarity of definition (1-5)
3. Intervention (I): identify the proposed intervention/method/solution; assess operationalizability (1-5)
4. Comparator (C): identify the comparison baseline (existing SOTA, no intervention, alternative approach); assess baseline reasonableness (1-5)
5. Metrics (M): identify the evaluation metrics; assess their measurability and relevance (1-5)
6. Evidence (E): assess the strength of existing evidence supporting the existence of the gap (1-5)
7. Overall verdict: judge overall quality from the mean of the 5 dimensions (strong ≥ 3.5 / moderate 2.5-3.4 / weak < 2.5); generate a research question draft
8. Output: return the PiCMeAssessment object

Output Format

{
  "gap_id": "gap_001",
  "dimensions": {
    "population": { "score": 4, "description": "Target population description", "rationale": "..." },
    "intervention": { "score": 3, "description": "Intervention/method description", "rationale": "..." },
    "comparator": { "score": 3, "description": "Comparison baseline description", "rationale": "..." },
    "metrics": { "score": 4, "description": "Evaluation metric description", "rationale": "..." },
    "evidence": { "score": 4, "description": "Evidence strength description", "rationale": "..." }
  },
  "mean_score": 3.6,
  "overall_verdict": "strong",
  "research_question_draft": "Research question draft (1 sentence)",
  "improvement_suggestions": ["Suggestion 1", "Suggestion 2"]
}

</output>