技能 数据科学 开放证据审查与文献合成

开放证据审查与文献合成

v20260423
openevidence-core-workflow-b
这是一个用于系统性医疗证据审查的综合工作流。它支持用户检索临床研究、构建精选引文集合、使用GRADE等专业框架评估证据等级,并自动生成符合学术规范格式(如AMA)的结构化报告。适用于撰写高质量的文献综述和临床研究报告。
获取技能
324 次下载
概览

OpenEvidence — Evidence Review & Citations

Overview

Search medical evidence, manage citations, and generate formatted evidence reports
through OpenEvidence. Use this workflow to find clinical studies for a specific
question, build citation collections for literature reviews, or produce structured
evidence summaries with graded recommendations. This is the secondary workflow —
for DeepConsult research synthesis, see openevidence-core-workflow-a.

Instructions

Step 1: Search the Evidence Database

const results = await client.evidence.search({
  query: 'SGLT2 inhibitors cardiovascular outcomes type 2 diabetes',
  filters: {
    study_type: ['rct', 'meta_analysis', 'systematic_review'],
    year_range: { min: 2020, max: 2026 },
    evidence_level: ['1a', '1b', '2a'],
  },
  limit: 25,
  sort: 'relevance',
});
console.log(`Found ${results.total} studies`);
results.items.forEach(s =>
  console.log(`  [${s.evidence_level}] ${s.title} (${s.journal}, ${s.year}) — ${s.citations} citations`)
);

Step 2: Build a Citation Collection

const collection = await client.citations.create({
  name: 'SGLT2i CV Outcomes Review — April 2026',
  study_ids: results.items.slice(0, 15).map(s => s.id),
  tags: ['cardiology', 'diabetes', 'sglt2i'],
});
console.log(`Collection ${collection.id}: ${collection.study_count} studies`);
await client.citations.addByDoi(collection.id, { doi: '10.1056/NEJMoa2034577' });

Step 3: Grade Evidence and Extract Key Findings

const graded = await client.evidence.grade(collection.id, {
  framework: 'GRADE',  // GRADE | Oxford | USPSTF
  outcome: 'major_adverse_cardiovascular_events',
});
graded.findings.forEach(f =>
  console.log(`${f.outcome}: ${f.grade} (${f.certainty}) — ${f.summary}`)
);
console.log(`Overall recommendation: ${graded.recommendation}`);

Step 4: Generate a Formatted Evidence Report

const report = await client.reports.generate({
  collection_id: collection.id,
  format: 'structured',
  sections: ['clinical_question', 'search_strategy', 'evidence_table', 'grade_summary', 'references'],
  citation_style: 'AMA',
});
console.log(`Report generated: ${report.page_count} pages`);
console.log(`Download: ${report.download_url}`);

HIPAA Notice

  • HIPAA-compliant and SOC 2 Type II certified — never include patient identifiers
  • Use de-identified clinical scenarios only; ensure BAA is in place before handling PHI

Error Handling

IssueCauseFix
401 UnauthorizedInvalid API key or expired sessionRegenerate key in OpenEvidence dashboard
404 Study not foundDOI not indexed or incorrect IDSearch by title or check DOI format
422 Invalid filterUnsupported evidencelevel or studytypeUse allowed values from client.schema.filters()
429 Rate limitedExceeded 60 queries/minuteAdd backoff; batch searches where possible
503 Grading unavailableGRADE engine under maintenanceRetry after 5 minutes or use Oxford framework

Output

A successful workflow returns ranked evidence results with evidence levels, a curated
citation collection, GRADE assessments with certainty ratings, and a downloadable
structured report in AMA citation format.

Resources

Next Steps

See openevidence-sdk-patterns for authentication and HIPAA-compliant configuration.

信息
分类 数据科学
名称 openevidence-core-workflow-b
版本 v20260423
大小 3.81KB
更新时间 2026-04-28
语言