Build, review, and export reproducible literature searches
This guide explains how to prepare a focused search question, configure PICOS inputs and filters, generate PubMed search strings, review search steps, retrieve hits and article details, translate strategies for Embase and Cochrane, and hand records off to screening.
What AccuSearch does
AccuSearch is a literature-search workspace for systematic and scoping reviews. You define a research question or structured PICOS criteria, review AI-prepared search strings, run the strategy against PubMed, retrieve article metadata, and export search evidence for review documentation and screening.
The tool is designed around reviewable search strategy development. Generated searches are drafts that should be checked, edited when necessary, and re-run when the search scope changes.
| What you provide | What AccuSearch prepares | What you review or export |
|---|---|---|
| Research question or PICOS inputs | Structured PubMed search steps | Step descriptions and query strings |
| Outcome types and filters | Population, intervention, outcome, and filter blocks | Hit counts for each step and final query |
| Reviewed PubMed strategy | Article detail retrieval and translations | Search strategy, hits, articles, Embase, and Cochrane exports |
The complete AccuSearch workflow
Each stage depends on the stage before it. If you change the form or search strings after generating results, review the downstream hits and article details again.
- 1
Define the search question
Start from a research question or enter PICOS fields manually.
- 2
Configure scope and filters
Select outcome types, publication types, dates, languages, species, geography, and other filters intentionally.
- 3
Generate PubMed search strings
Create numbered search steps that describe terms, filters, combinations, and the final query.
- 4
Review and edit the strategy
Check synonyms, field tags, boolean logic, warnings, and # references before running the search.
- 5
Retrieve hits and articles
Get PubMed hit counts, then retrieve article metadata from the final search.
- 6
Translate, export, or hand off
Generate Embase and Cochrane strategies, export outputs, or send article records into screening.
Choose manual PICOS or AI-assisted autofill
AccuSearch can start from a manually completed form or from a research question that is used to draft the form fields. The autofill mode is helpful when a user has a clear question but has not yet converted it into search inputs.
| Input method | Best used when | User responsibility |
|---|---|---|
| Manual PICOS | The review team already knows the population, intervention, outcomes, and filters. | Enter the fields directly and keep each concept focused. |
| AI-assisted autofill | The team has a narrative research question and wants a structured starting point. | Review every drafted field before generating search strings. |
Write inputs that produce focused search blocks
The form separates the search question into fields that can become distinct query blocks. Keeping each field narrow makes generated terms easier to review and easier to combine.
| Field | Purpose | Example |
|---|---|---|
| Search title | A readable project name for the saved search. | Metastatic pancreatic cancer first-line therapies |
| Population | The disease, condition, population, or setting to retrieve. | Metastatic pancreatic cancer |
| Intervention | The treatment, exposure, technology, or comparator focus. | FOLFIRINOX, gemcitabine, nab-paclitaxel |
| Other outcomes | Additional concepts that should shape outcome search blocks. | Overall survival, progression-free survival, adverse events |
| Outcome types | Preset evidence families that add domain-specific outcome logic. | Clinical efficacy, safety outcomes, economic outcomes |
Use filters intentionally
Filters can make a search more precise, but they can also exclude relevant records. A filter should be selected because it matches the protocol, not because it makes the hit count smaller.
| Filter family | What it controls | Review guidance |
|---|---|---|
| Text availability | Abstract, free full text, and full text availability. | Use when the review protocol limits sources by accessible text type. |
| Study design | Clinical trials, randomized trials, observational evidence, pragmatic trials, pharmacokinetic studies, and related designs. | Use design filters when the intended evidence type is narrow and explicit. |
| Publication type | Meta-analyses, systematic reviews, reviews, case reports, guidelines, conference abstracts, and more. | Use to target or exclude document classes that match the review objective. |
| Language, sex, species | Language choices, human or animal records, and sex-specific filters. | Use only when these are true eligibility boundaries. |
| Geography and date | Countries, regions, predefined publication windows, or custom dates. | Use to match epidemiology, market, jurisdiction, or recency requirements. |
Review generated search strings before running hits
Search generation produces numbered PubMed steps. Each step has a description and a query. Some steps introduce terms, while later steps combine earlier steps into broader strategy blocks or the final search.
| Step type | What it does | Example pattern |
|---|---|---|
| Concept step | Searches a single topic such as disease, intervention, or outcome. | pancreatic cancer terms in title/abstract and MeSH |
| Filter step | Applies publication type, date, language, species, or text availability logic. | English language or custom publication date |
| Combination step | Combines earlier steps by reference. | #4 AND #7 |
| Final step | Represents the query used for final hit and article retrieval. | Population AND intervention AND selected outcome blocks |
Understand step numbers and # references
PubMed strategies often build a final query from earlier blocks. In AccuSearch, #4 means "use the query from Step 4." A search string such as #4 AND #7 combines two existing steps.
Editing and deleting referenced steps
- When you edit a step that later steps reference, review the dependent steps after saving.
- Combination steps cannot be deleted directly because other logic may depend on them.
- When an eligible step is deleted, remaining steps are renumbered and references are remapped.
- If deleting a step causes a dependent combination to collapse, the UI previews affected steps before applying the change.
Edit, add, or upload reviewed search strings
Search strings can be refined directly in the page or by downloading an Excel template, editing it externally, and uploading the revised version. This is useful when a search specialist needs to adjust syntax or apply a strategy agreed outside the application.
| Action | Use it when | Downstream effect |
|---|---|---|
| Edit step | A generated query needs targeted syntax, term, or field-tag changes. | Dependent steps may need review. |
| Add new step | The strategy needs an extra concept, filter, or combination block. | Review the final step before retrieving hits. |
| Delete step | A non-combination step is not needed. | References are remapped and affected steps are previewed. |
| Upload Excel | The team edits search strings outside AccuSearch. | Uploaded steps replace current steps and should be treated as manually reviewed. |
Retrieve and review PubMed hit counts
Hit retrieval runs the reviewed PubMed strategy and returns hit counts for each search step. The final search step provides the hit count and session information needed to retrieve article details.
| Before getting hits | Why it matters |
|---|---|
| No invalid PubMed warnings remain. | Invalid formatting can prevent reliable execution. |
| The final combination step matches the intended strategy. | Article retrieval uses the final search result. |
| Manual edits have been reviewed. | Hit counts should reflect the current query, not an older version. |
Retrieve article details after hits are ready
Article retrieval uses the final PubMed search information to fetch records and display them in a table. Retrieved fields can include PMID, title, abstract, authors, journal, year, DOI, PMCID, keywords, MeSH terms, and PubMed or PMC links.
| Article field | How users use it |
|---|---|
| Title and abstract | Review topic relevance and prepare screening records. |
| Authors, journal, year | Identify publication context and citation details. |
| PMID, DOI, PMCID | Trace records and link articles across downstream workflows. |
| Keywords and MeSH terms | Inspect indexing and terminology coverage. |
Translate PubMed search strings for Embase and Cochrane
After PubMed search steps exist, AccuSearch can generate platform-specific versions for Embase and Cochrane. The translated steps remain reviewable and may include warnings when a PubMed filter or field mapping does not transfer cleanly.
| Translation behavior | Meaning for review |
|---|---|
| Boolean combination steps are preserved. | A step such as #1 AND #2 remains a combination of earlier translated steps. |
| Field tags are mapped to platform syntax. | Title/abstract logic is adapted for Embase or Cochrane conventions. |
| Unsupported filters can produce warnings. | Review warnings before using translated strategies outside AccuSearch. |
| Translated steps can be reviewed and edited. | Search specialists can adjust syntax before export. |
Export search strategies, hits, and article details
AccuSearch exports each major output independently. Export the strategy for documentation, hits for search accounting, and article details for screening or evidence-management workflows.
| Export | What it contains |
|---|---|
| Search strategies | Step numbers, descriptions, and search strings, with translated strategies when available. |
| Hits | Step-level PubMed hit counts and the final-search summary. |
| Articles | Retrieved article metadata, identifiers, abstracts, and indexing terms. |
Use AccuSearch results for screening
When AccuSearch is linked to a workflow, retrieved articles can be passed into AccuScreener as screening records. The handoff uses available identifiers such as PMID, DOI, article ID, and title-year combinations to reduce duplicate records.
Designing a reliable search strategy
| Do | Avoid |
|---|---|
| check_circle Keep the population field focused on the disease, condition, or setting. | cancel Put interventions, outcomes, and filters all into the population field. |
| check_circle Review generated synonyms and field tags before retrieving hits. | cancel Assume generated PubMed syntax is final without search-specialist review. |
| check_circle Use filters only when they reflect the protocol. | cancel Add filters only to reduce the hit count. |
| check_circle Treat # references as dependencies between search steps. | cancel Edit a referenced step without checking later combination steps. |
| check_circle Get fresh hits after changing search strings. | cancel Use old hit or article results after manual edits. |
| check_circle Review translation warnings before using Embase or Cochrane strategies. | cancel Assume every PubMed filter maps cleanly to another platform. |
Before using results downstream
- check The research question or PICOS inputs match the protocol.
- check Outcome types are selected intentionally.
- check Filters are eligibility-driven, not only hit-count-driven.
- check Generated search steps have been reviewed for terms, syntax, and boolean logic.
- check Invalid PubMed warnings are resolved before retrieving hits.
- check Manually edited steps and dependent # references have been checked.
- check Fresh hits were retrieved after the latest search-string changes.
- check Article details and exports come from the intended final search.