返回提示词目录第 1187 页

开展无荟萃分析的研究综合

基于所给研究分组汇总效应与偏倚,生成结果表、方向图和报告清单,并说明证据局限。

完整提示词

<context>
Many reviews cannot pool effects, yet their "narrative synthesis" often means study-by-study description and an unstated, informal vote count of significant results. The SWiM (Synthesis Without Meta-analysis) reporting guideline asks reviewers to state how studies were grouped and why; which standardised metric was used (for example a standardised mean difference, odds ratio, or simply direction of effect); which synthesis method was used and why (combining P values, summarising effect estimates with median and range, or vote counting based on direction of effect, never on statistical significance); how studies were prioritised (for example by risk of bias); how heterogeneity was explored; how certainty was assessed; how data are presented (tables, effect direction plots, harvest or albatross plots); and the limits of the method. Vote counting by significance is misleading because small underpowered studies produce "no effect" results that are absence of evidence.
</context>

<task>
Synthesise without meta-analysis:
<question>
{{question}}
</question>
<studies>
{{studies}}
</studies>
{{#why_not_pooled}}
Reason pooling was not done: {{why_not_pooled}}
{{/why_not_pooled}}

1. Check the reason pooling was not done. If the data look poolable (same outcome type, effect estimates with variance available), say so and suggest plan-meta-analysis, then continue with the structured synthesis.
2. Group studies by a stated, pre-specifiable logic (intervention type, outcome domain, population) and justify the groups.
3. Choose a standardised metric per group and convert reported results where the data allow, showing each conversion; where they do not, fall back to direction of effect and say so.
4. Choose and justify a synthesis method per group: summary of effect estimates (median, interquartile range, range), combining P values, or vote counting on direction of effect with a sign test and its binomial probability and a 95% CI for the proportion favouring the intervention.
5. Present a results table per group with each study's design, n, metric, direction, risk of bias, and an effect direction plot in text form (study | outcome | direction arrow | sample size symbol | risk of bias).
6. Describe heterogeneity informally (by setting, dose, risk of bias, measure) and report a sensitivity check that drops high-risk-of-bias studies.
7. State certainty per outcome with GRADE domains in brief, noting that imprecision is harder to judge without a pooled CI.
8. Fill a SWiM checklist with where each item is addressed.
</task>

<constraints>
- Never count studies by statistical significance; vote count only on direction of effect, and explain why.
- Use only supplied data; do not invent effect sizes, variances or study details. Show any conversion formula you use.
- Do not describe the synthesis as a meta-analysis or report a pooled effect.
- Keep study-level description short; the synthesis is organised by group, not by paper.
</constraints>

<output_format>
## Synthesis approach
Grouping logic, metric, method and why pooling was not done.
## Study grouping
A table: group | studies | reason.
## Results by group
Per group: a results table, the effect direction plot and two to four sentences of synthesis.
## Heterogeneity and certainty
Heterogeneity, sensitivity check and a certainty line per outcome.
## Reporting checklist
A table: SWiM item | where addressed | note.
</output_format>

可替换内容

  • {{question}}
  • {{studies}}
  • {{#why_not_pooled}}
  • {{why_not_pooled}}
  • {{/why_not_pooled}}

分类与标签

学习教育、文献综述、证据综合、效应方向图、偏倚评估

前往 AI 图片创作