返回提示词目录第 1382 页

编制研究证据分级与结果汇总表

帮你梳理各研究结局的效应与证据确定性,生成分级汇总表,供评审团队确认。

完整提示词

<context>
A GRADE summary of findings (SoF) table tells decision-makers, per important outcome, what the effect is and how sure we can be. Evidence from randomised trials starts at high certainty and from observational studies at low; it is rated down for risk of bias, inconsistency, indirectness, imprecision and publication bias (one or two levels each), and observational evidence may be rated up for a large effect, a dose-response gradient or when plausible confounding would reduce the effect. Experts rate imprecision against a threshold (minimal important difference or a decision threshold), not only on whether the CI crosses no effect, and avoid double-penalising one problem in two domains. They give absolute effects per 1,000 alongside relative effects, because a relative risk without a baseline risk misleads. They write plain-language statements with standard GRADE wording ("probably reduces" for moderate, "may reduce" for low, "very uncertain" for very low). Typical mistakes: rating whole reviews instead of outcomes, leaving downgrades unexplained, and inventing baseline risks.
</context>

<task>
Build a summary of findings for:
<question>
{{question}}
</question>
<results>
{{results}}
</results>
{{#baseline_risk}}
<baseline_risk>
{{baseline_risk}}
</baseline_risk>
{{/baseline_risk}}

1. List the outcomes (at most seven, critical first) and, for each, the threshold used to judge imprecision and importance. If no threshold is given, propose one, label it as an assumption for the team to confirm, and say what would change if it were different.
2. For each outcome, assess the five downgrading domains and, for observational evidence, the three upgrading domains. Give the judgement (not serious, serious, very serious) and a one-line reason that cites the supplied data.
3. Assign the overall certainty: high, moderate, low or very low.
4. Calculate absolute effects for dichotomous outcomes from the supplied baseline risk: risk with comparator per 1,000, risk with intervention per 1,000 and the difference with its CI. Show the arithmetic once. If no baseline risk is supplied, leave the cells as "baseline risk needed" and say where to get one.
5. Build the SoF table with footnote letters explaining every downgrade or upgrade.
6. Write one plain-language statement per outcome using GRADE's standard wording for its certainty level.
7. List decisions the team must confirm and data that are missing.
</task>

<constraints>
- Use only the supplied numbers. Never invent effect estimates, participant counts, baseline risks or risk-of-bias judgements.
- Do not rate the same limitation in two domains; if you considered it, note where it was counted.
- Keep relative and absolute effects consistent with each other; round absolute effects to whole numbers per 1,000 and show their CI.
- If results are missing for an outcome, keep the row and state "no studies" or "not reported" rather than dropping it.
- Ratings are provisional judgements for the review team, not final; say so once.
</constraints>

<output_format>
## Outcomes and thresholds
A table: outcome | importance | threshold | source of threshold.
## Certainty assessment
A table per outcome: domain | judgement | reason.
## Summary of findings table
Columns: outcome | relative effect (95% CI) | risk with comparator | risk with intervention | difference (95% CI) | participants (studies) | certainty | comments. Footnotes below.
## Plain-language statements
One sentence per outcome.
## Open questions
Bullets for the team.
</output_format>

可替换内容

  • {{question}}
  • {{results}}
  • {{#baseline_risk}}
  • {{baseline_risk}}
  • {{/baseline_risk}}

分类与标签

学习教育、系统综述、证据分级、研究结果汇总、效应量

前往 AI 图片创作