生成并核对系统综述筛选流程
将文献筛选数据映射到PRISMA流程框,核对各阶段数量,并生成流程图代码与结果段落。
完整提示词
<context>
The PRISMA 2020 flow diagram is the first thing reviewers check in a systematic review, and its most common faults are numbers that do not add up, records and reports treated as the same thing, full-text exclusions without reasons, and other sources (citation searching, websites, contacting authors) folded into database counts. PRISMA 2020 has templates for new reviews and for updates, each with an optional second column for "identification of studies via other methods". It counts records at screening, reports at full text, and studies at inclusion, because one study can have several reports.
</context>
<task>
Build the PRISMA flow report from these numbers:
<screening_numbers>
{{screening_numbers}}
</screening_numbers>
1. Map every number to its PRISMA 2020 box: identification (records from each database and register; other methods by source), records removed before screening (duplicates, marked ineligible by automation tools, other reasons), records screened, records excluded, reports sought for retrieval, reports not retrieved, reports assessed for eligibility, reports excluded with reasons, new studies included and reports of included studies. For an update, add the previous review's studies and reports and the totals.
2. Check the arithmetic at each step, separately for each column. Databases and registers: identified minus removed before screening equals screened; screened minus excluded equals sought; sought minus not retrieved equals assessed; assessed minus excluded equals included reports. Other methods have no record-screening stage: reports identified by each source are sought for retrieval, then sought minus not retrieved equals assessed, and assessed minus excluded equals included reports. Check that exclusion reasons sum to each column's full-text exclusion total, and that included reports from both columns add up to the total reports of included studies. Show each sum.
3. If anything does not add up or a box is missing, do not adjust or invent numbers. Show the gap, give the likely causes (for example a study with several reports, records found in both columns, an unrecorded exclusion), and leave the box as [MISSING] or [CHECK: expected N, given M].
4. Write the diagram content box by box, and drawable code for it.
5. Write the study selection paragraph for the results section, in past tense, reporting the counts, the main full-text exclusion reasons in descending order, and the number of studies and reports included. Mention any studies that seemed to meet the criteria but were excluded, if the author named them.
</task>
<constraints>
- Never change, round or fill in a number. Every number in the output appears in the input or is a sum you showed.
- Keep records, reports and studies distinct; flag wording in the input that blurs them.
- Full-text exclusion reasons must be specific ("wrong comparator", "conference abstract only"), not "irrelevant"; flag vague ones for the author to recode.
- Use PRISMA 2020 terminology, not PRISMA 2009 box names.
</constraints>
<output_format>
## Arithmetic check
A table: step | calculation | result | OK or mismatch.
## Flow diagram content
The boxes in order, grouped into Identification, Screening and Included, with the database and other-methods columns side by side where both exist.
## Diagram code
A Mermaid flowchart (top-down) in a code block that renders the boxes and arrows, with exclusion boxes to the side. Put every node label in double quotes, for example `A["Records identified from databases (n = 812)"]`, because parentheses and commas in unquoted labels break rendering; use `<br>` for line breaks inside a box.
## Study selection paragraph
One paragraph ready to paste, with "(Figure 1)" where the diagram is cited.
## Missing items
Each [MISSING] or [CHECK] item and the question to answer.
</output_format>可替换内容
- {{screening_numbers}}
- [MISSING]
- [CHECK: expected N, given M]
- [CHECK]
分类与标签
学习教育、系统综述、文献筛选、PRISMA流程图