shield_lock
金大哥
← 返回 Skill 列表
extension
分类: 效率与办公无需 API Key

sr-update

Systematic review Phase 10 — Living review maintenance, search updates, and new evidence integration for EM/critical care SRs. Use when the user wants to update their review with new evidence, set up automated search alerts, decide whether to publish an update, or manage a living systematic review. Trigger on "update my review", "living review", "search update", "new evidence", "maintain my review", "quarterly update", "surveillance", "alerts", "stale review". Use ONLY when maintaining, updating, or keeping a review current.

person作者: TashanworldhubOpenAPI
open_in_new仓库download下载资源包

SR Phase 10: Review Updates and Living Review Maintenance

Your Role

Guide the user through updating their systematic review, setting up surveillance for new evidence, and deciding when to publish an update.

Prerequisites

  • Original search strings (from Phase 3)
  • Original search date (from search log)
  • Original included studies list (from Phase 4)
  • Publication status of original review

Workflow

Step 1: Triage — Is an Update Needed?

ASK:

  1. "When was your original search run?"
  2. "Has the review been published? If so, what did the conclusions say?"
  3. "Have there been any major new trials or practice-changing developments in this area?"

Decision rules for update:

  • <6 months old: "Your search is still current. No update needed. Set up an alert for quarterly checks."
  • 6-12 months old: "Consider updating before submission/publication. Many journals expect searches within 12 months of submission."
  • 12-24 months old: "An update is recommended. New evidence likely exists."
  • >24 months old: "Your review is likely outdated. A full update is strongly recommended."

Step 2: Search Update

RE-RUN the original search with a date filter:

PubMed: Append to the original string:

AND ("2023/01/01"[Date - Publication] : "3000"[Date - Publication])

Replace the start date with the original search date + 1 day.

Run the search:

QUERY="(original string) AND (\"2023/01/01\"[Date - Publication] : \"3000\"[Date - Publication])"
python3 -c "
import urllib.parse, json, urllib.request
q = urllib.parse.quote('''$QUERY''')
url = f'https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=pubmed&term={q}&retmax=0&retmode=json'
resp = urllib.request.urlopen(url)
data = json.load(resp)
print(f'New records since last search: {data[\"esearchresult\"][\"count\"]}')
"

Repeat for other databases as feasible.

Step 3: Screening Update

PROTOCOL:

  1. Screen ONLY new records (those published after the original search date)
  2. Follow the same eligibility criteria and process as the original review
  3. Use the same screening form and kappa targets
  4. Dual screening still required

Step 4: Decide on Update Type

| Situation | Recommended Update Type | |-----------|----------------------| | 0-2 new studies, conclusions unchanged | Brief update: Add to existing analysis, note in discussion | | 3-10 new studies, conclusions unchanged | Minor update: Re-run meta-analysis, update manuscript | | Any new studies, conclusions change | Major update: Full re-analysis, may change recommendation | | New population or outcome evidence emerges | Scope expansion: Consider separate review or major update | | New methodology (e.g., new synthesis method) | Methodological update: Re-analyze existing data |

Step 5: Living Review Infrastructure

If the user wants to maintain a living review:

  1. Set up PubMed auto-alerts: Instructions for My NCBI:

    • Save the search string in PubMed
    • Create an RSS feed or email alert
    • Set to "weekly" or "monthly"

  2. Quarterly surveillance protocol:

    • Every 3 months: re-run date-limited search
    • Screen new records (1 reviewer, 2nd for conflicts)
    • If ≥1 new eligible study found → escalate to full update
    • If no new studies → log "No new evidence found, review remains current"

  3. When to publish an update:

    • ≥3 new studies that could change the pooled estimate
    • New large trial (>500 participants) that could change conclusions
    • Methodological advance that improves precision
    • Live within 2 years of original publication

Step 6: Update Protocol

For a formal update (publishable):

  1. Register update on PROSPERO (existing record can be updated)
  2. Follow Cochrane handbook update guidance (Chapter 4)
  3. Report per PRISMA 2020 update guidelines
  4. Distinguish between "updated review" and "new review" in title

Step 7: Reporting the Update

In the updated manuscript:

  • Title: "[Title]: An Updated Systematic Review and Meta-Analysis"
  • Methods: "We updated our search on [date] and identified [N] additional studies."
  • Results: Report original + new studies combined. Label new studies in forest plot.
  • Discussion: "Since our original review, [N] additional trials have been published, [strengthening/changing] the evidence."

Scripts

scripts/search_update.sh

#!/bin/bash
# Run a date-limited PubMed search update
# Usage: ./scripts/search_update.sh "original search string" "2023-01-01"
QUERY=$(python3 -c "import urllib.parse; print(urllib.parse.quote('($1) AND (\"$2\"[Date - Publication] : \"3000\"[Date - Publication])'))")
curl -s "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=pubmed&term=$QUERY&retmax=0&retmode=json" | python3 -c "import json,sys; d=json.load(sys.stdin); print(f'New records: {d[\"esearchresult\"][\"count\"]}')"

Outputs to Generate

  1. sr-update-protocol.md — Plan for the review update
  2. sr-new-evidence-log.md — Tracking new studies identified
  3. sr-search-alert-setup.md — Instructions for auto-alerts
  4. sr-living-review-protocol.md — If maintaining a living review

Guardrails

  1. Do NOT claim an update if you only re-ran the search — An update must include screening, extraction, and synthesis of new evidence.
  2. Do NOT combine new and old studies without checking comparability — New studies may use different populations, interventions, or outcomes.
  3. Date filter is essential — Without it, you'll search the entire database again.
  4. Living reviews require infrastructure — "A living review requires a dedicated team, regular surveillance, and a commitment to timely updates. Be realistic about resources."
  5. Do NOT change eligibility criteria between versions — Use the SAME criteria. If you change them, it's a different review.

Edge Cases

| Situation | Response | |-----------|----------| | No new studies found | "No new evidence since the original search. Your review remains current. Consider setting up alerts and re-checking in 6 months." | | New studies contradict original findings | Flag prominently. Meta-analysis may show different results. Discuss reasons for discrepancy (different population, larger sample, longer follow-up). | | User published the review in a journal that doesn't allow updates | "Consider publishing the update as a separate 'rapid update' or letter to the editor. Or deposit the update on a preprint server." | | Cochrane review update | "Cochrane has specific update policies. Contact your Cochrane Review Group for guidance." | | Original data lost | "Without original data, you'll need to re-extract. This is a strong argument for data archiving." |

Handoff

This is the final phase. Once complete:

  1. Summarize: "Review update complete. [N] new studies added. Conclusions [unchanged/changed]."
  2. Final outputs: Updated manuscript + search log + evidence tracking log
  3. If living review: "Your living review infrastructure is in place. Next quarterly check: [date]."