November 9, 2026

How to Update an Old Literature Search Efficiently

An old literature search is not a document to reread; it is a set of decisions you can re-run. The efficient path is to recover the original eligibility criteria and search logic, then update only the parts that changed:

Written byWisPaper TeamAI Research Workflow Team
Editorial illustration for How to Update an Old Literature Search Efficiently

An old literature search is not a document to reread; it is a set of decisions you can re-run. The efficient path is to recover the original eligibility criteria and search logic, then update only the parts that changed: the date window, the terms, and the sources. This guide walks clinicians and medical researchers through that sequence with checkpoints, a worked example, and explicit completion criteria.

What does "updating" a literature search actually mean?

Updating means extending a prior search to capture records published or indexed since the last run, while keeping the original question and eligibility criteria stable enough that old and new results remain comparable. Cochrane's guidance frames searching as a planned activity derived from eligibility criteria, using multiple sources, combining free-text terms with subject headings, and documenting the process. An update inherits that structure and adds a time boundary.

Two things distinguish an update from a fresh search:

  • A fixed anchor. The original population, intervention or exposure, comparator, and outcomes stay in place unless you deliberately record a change.
  • A moving edge. The date window, newly coined terms, and any new sources form the delta you are adding.

If you cannot reconstruct the original criteria, you are not updating — you are rebuilding, and you should say so in your methods note.

Step 1: Recover the original search as an artifact

Before touching any database, assemble the original search record. Cochrane's chapter on searching and selecting studies emphasizes documenting the process, which is exactly what makes an update cheap. Your recovery checklist:

  • The review question and eligibility criteria (population, exposure or intervention, outcomes, study designs).
  • The databases and other sources originally searched.
  • The full search strings, including free-text terms and subject headings.
  • Any filters or limits applied.
  • The date of the last run and the date window it covered.
  • The screening decisions and how multiple reports of one study were linked.

If the original record is partial, reconstruct what you can and mark the gaps explicitly. A documented gap is a limitation; an undocumented one is a defect.

Checkpoint: You can state the last search date, the sources used, and the eligibility criteria in writing. If not, pause and reconstruct before searching.

Step 2: Decide what stays fixed and what changes

Not every element of an update should move. Separate your decisions into three buckets:

ElementKeep fixedChange deliberatelyRecord the reason
Review question and eligibility criteriaYes, by defaultOnly with a documented protocol amendmentWhy the question shifted
Date windowStart from last run dateExtend to todayLast run date and new end date
Free-text termsCore conceptsAdd newly coined terms and synonymsWhich terms were added
Subject headingsExisting headingsAdd or remap changed headingsVocabulary changes
SourcesOriginal setAdd a source only if justifiedWhy the source was added
Filters and limitsOriginal limitsRe-examine each limitWhether the limit still serves the question

This table is the artifact that keeps an update honest. It also becomes the methods paragraph you will need later.

Step 3: Rebuild the search string with a clear delta

Rebuild the string concept by concept rather than editing the old string in place. PubMed's search details show how terms were translated and combined, which helps you see what the original string actually did versus what you intended. Field tags, phrase searching, and filters are all documented PubMed behaviors you can use to make the delta precise.

A practical rebuild sequence:

  1. Re-enter each concept block from the original string.
  2. Add new free-text synonyms for each concept, keeping them inside the correct block.
  3. Confirm subject headings still map to the intended concepts.
  4. Apply the date limit as the final step so you can see the unfiltered yield first.
  5. Save the search so the next update starts from a known state.

PubMed supports saved searches, which turns the next update into a re-run rather than a reconstruction. Google Scholar's documented features — author and title searching, date limits and sorting, Cited by, Related articles, and email alerts — are useful for catching records a database string may miss and for monitoring forward citations of key papers.

Checkpoint: Run the rebuilt string without the date limit and confirm it retrieves a sample of known relevant papers from the original search. If it does not, the string has drifted.

Step 4: Run the delta across sources

Cochrane recommends using multiple sources, so an update should touch each source in the original set. For each source, record the run date, the string used, and the raw yield. This is where a scholarly search interface helps: you can run the concept blocks, capture the results, and move directly into screening.

If you are working inside WisPaper, the scholarly search interface is the natural place to run the rebuilt string and collect candidate papers, and the Scholar Search entry point explains the paper-finding workflow for readers new to the tool. Save the papers you intend to screen into My Library so the update has a durable working set rather than a set of open browser tabs.

Checkpoint: Every source in the original set has a dated run and a raw yield count.

Step 5: Deduplicate and screen in stages

Cochrane's guidance covers deduplicating records, screening in stages, and linking multiple reports of one study. Apply the same order to the delta:

  1. Deduplicate across sources and against the original included set.
  2. Screen titles and abstracts against the eligibility criteria.
  3. Screen full texts for the remaining records.
  4. Link multiple reports of the same study so a trial and its follow-up paper count once.

Screening in stages keeps the full-text workload proportional to the delta rather than the whole corpus. When you need to interrogate the papers you have collected, Library QA lets you ask questions based on papers uploaded to My Library or added from search results, which is useful for checking whether a candidate actually reports the outcome your question requires.

Checkpoint: You can report the number of records at each stage — identified, deduplicated, title/abstract screened, full-text screened, included.

Worked example: updating a search on a focused population and outcome

Suppose the original search asked whether a specific care-delivery exposure is associated with a defined clinical outcome in a defined adult inpatient population, and the last run covered literature through a date two years ago. Here is how the update proceeds.

Recover. The team locates the original eligibility criteria, the three databases searched, the full strings, and the last run date. One gap: the original subject headings were not recorded. They mark this as a limitation.

Decide. The question, population, exposure, and outcome stay fixed. The date window extends from the last run date to today. Two newly coined free-text synonyms are added to the exposure block. No new sources are added.

Rebuild. Each concept block is re-entered. The new synonyms sit inside the exposure block. The date limit is applied last. The rebuilt string retrieves the team's five known landmark papers from the original search, confirming no drift.

Run. Each of the three databases is run on the same day, with run dates and raw yields recorded. Candidate papers are saved to a working library.

Screen. Records are deduplicated, then screened in stages. Two candidate papers turn out to be reports from the same study and are linked. A third is excluded at full text because it reports a different outcome.

Assess. For each included study, the team records the design, the directness of the population and outcome to the review question, and the study's own limitations. They keep evidence level, applicability, and indirectness as separate columns rather than collapsing them into a single judgment.

Report. The methods note states the last search date, the new search date, the sources, the terms added, and the screening counts.

This example shows the core economy of an update: most of the work is recovery and decision-making, and the searching itself is a small, well-bounded step.

How do you judge evidence level, applicability, and indirectness separately?

These are four distinct judgments, and conflating them is the most common error in an update brief.

  • Evidence level describes the design and strength of the study's own inference — what kind of study it is and how it was conducted.
  • Applicability asks whether the study's setting, population, and care context resemble the context you care about.
  • Indirectness asks whether the study measured the population, exposure, comparator, or outcome your question specifies, or something adjacent to it.
  • Limitations are the study's internal weaknesses — the things that could bias its own results.

A study can be a strong design (high evidence level) yet indirect for your question because it measured a different outcome. Another can be directly on target but carry serious limitations. Record each in its own column so a reader can see which judgment drove your conclusion.

Step 6: Organize the update so the next one is cheaper

An update produces artifacts, not just a list of papers. Keep them together so the next run starts from a known state. If your team runs recurring updates, a Research Projects area gives the update a home where the search record, screening decisions, and included studies stay associated with the question.

For longer synthesis tasks — assembling the narrative around the included studies, or exploring what the updated evidence implies for the research question — Scholar Agent supports literature review and related research workflows. Use it after screening, when the included set is settled, so the synthesis is built on your decisions rather than replacing them.

Checkpoint: A colleague can pick up your artifacts and re-run the search without asking you questions.

Step 7: Verify references before circulating the brief

An update brief usually carries a reference list, and reference errors are easy to introduce when papers are added quickly. TrueCite checks BibTeX against real academic databases to flag hallucinated references, which is a concrete verification step before the brief goes to your team. Run it on the final reference list, not on a draft.

Checkpoint: Every reference in the brief resolves to a real record.

Reusable checklist for the next update

  • Original question and eligibility criteria recovered in writing
  • Original sources, strings, filters, and last run date recorded
  • Gaps in the original record marked as limitations
  • Fixed-versus-changed decisions tabulated with reasons
  • Search string rebuilt concept by concept
  • New terms added inside the correct concept blocks
  • Date limit applied last
  • Rebuilt string retrieves known relevant papers
  • Each source run on a recorded date with a raw yield
  • Records deduplicated across sources and against the original set
  • Screening completed in stages with counts at each stage
  • Multiple reports of one study linked
  • Evidence level, applicability, indirectness, and limitations recorded separately
  • Methods note states last search date, new search date, sources, and terms added
  • References verified before circulation

When is the updated search ready to share?

The update is ready when all of the following are true:

  1. The last search date and the new search date are both stated.
  2. Every source in the original set has a dated run and a raw yield.
  3. The terms added and the reason for each addition are recorded.
  4. Screening counts are reported at each stage.
  5. Multiple reports of the same study are linked.
  6. Each included study has separate entries for evidence level, applicability, indirectness, and limitations.
  7. The reference list has been verified.
  8. A colleague can re-run the search from your artifacts alone.

If any item is missing, the update is a draft. Cochrane's emphasis on documenting the search process and rerunning searches before publication is the underlying reason: the record is what makes the result reproducible and defensible.

FAQs

You can, but you should label it a rebuild rather than an update. Without the original string, you cannot confirm that old and new results are comparable, and you cannot show which terms changed. Reconstruct the eligibility criteria and sources from the old results, document the missing string as a limitation, and build the new string from the criteria. The next update will then be cheaper because you will have a saved string to re-run.