Systematic Review Screening and Data Extraction Workflow With Reusable Templates
systematic reviewsstudy screeningdata extractionresearch methodsreview templates

Systematic Review Screening and Data Extraction Workflow With Reusable Templates

RResearchers.site Editorial Team
2026-11-16
8 min read

A reusable workflow for deduplication, screening, disagreement resolution, risk-of-bias preparation, and systematic review data extraction.

A consistent systematic review depends on more than a good search. This workflow shows how to deduplicate records, screen titles and abstracts, document full-text decisions, resolve reviewer disagreements, prepare for risk-of-bias assessment, and extract study data using reusable tables.

Overview

Systematic review screening and data extraction are connected stages, but they require different controls. Screening asks whether a record may meet the eligibility criteria. Extraction records what an included study says, how it was conducted, and which results answer the review question. Mixing these purposes can create inconsistent decisions or premature collection of information from studies that will later be excluded.

Before screening begins, convert the review question into operational eligibility criteria. Define the population, intervention or exposure, comparator, outcomes, study design, publication characteristics, and any language or date limits that genuinely apply. Write examples of both eligible and ineligible studies. A criterion such as “relevant topic” is difficult to apply consistently; a criterion such as “participants must have the specified condition, confirmed by the stated method” is easier to audit.

Keep the search record separate from the screening record. A database-by-database search log can help preserve the history of the evidence-gathering stage; see the systematic literature search strategy and search log guide. Screening software, a spreadsheet, or another review-management tool can all work if the workflow preserves record identifiers, decisions, reasons for exclusion, reviewer assignments, and dates.

The basic sequence is:

  1. Import all search results and preserve the original exports.
  2. Deduplicate records using a documented method.
  3. Pilot the eligibility criteria with both reviewers, if possible.
  4. Screen titles and abstracts against the predefined criteria.
  5. Retrieve and assess potentially eligible full texts.
  6. Resolve disagreements and record one final decision per study.
  7. Prepare the included set for risk-of-bias assessment and extraction.
  8. Lock the extraction form after piloting, then extract and verify data.

Checklist by scenario

Scenario 1: Preparing the screening project

  • Save the search exports in their original format before editing them.
  • Assign a stable record ID that will remain unchanged throughout the review.
  • Record the database, platform, search date, and any export limitations.
  • Define what counts as a duplicate: for example, the same report, a conference abstract and a full article, or multiple records for one study.
  • Decide whether screening will occur independently, sequentially, or through another documented arrangement.
  • Pilot the title-and-abstract form on a small, varied set of records.
  • Revise unclear criteria before formal screening, and preserve the date and reason for the revision.

Reusable screening table

FieldWhat to record
Record IDStable identifier used in every file
Title and citationTitle, authors, year, journal or source
SourceDatabase, registry, website, citation chasing, or other route
Duplicate statusUnique, duplicate, or possible duplicate requiring review
Title/abstract decisionInclude, exclude, or unclear
Full-text decisionInclude or exclude
Exclusion reasonOne predefined, specific reason at full-text stage
Reviewer and dateWho made the decision and when
NotesShort explanation for unusual cases

Scenario 2: Title and abstract screening

Apply the criteria conservatively at this stage. The purpose is usually to remove records that are clearly irrelevant, not to make a final judgment from limited information. If the title and abstract do not provide enough detail to exclude a record confidently, mark it for full-text assessment. Use the same decision labels and exclusion categories for every reviewer.

Do not add new eligibility requirements because a particular record seems inconvenient. If an important ambiguity appears repeatedly, pause and clarify the protocol or decision rule before continuing. Keep a short decision guide with examples, especially for borderline populations, mixed samples, secondary analyses, and multiple reports from the same study.

Scenario 3: Full-text screening

Retrieve the complete report for every record that remains potentially eligible. Record whether the full text was found, unavailable, incomplete, or replaced by a related report. Assess the article against each criterion rather than relying on its abstract or journal description.

Use one primary exclusion reason for each excluded full text, such as wrong population, wrong design, wrong outcome, duplicate study, or insufficient report type. If several reasons apply, record the reason that best explains why the study could not enter the review and retain additional notes where needed. This makes the final study flow easier to reconstruct.

Scenario 4: Resolving reviewer disagreements

First compare the exact criterion and evidence that led to each decision. Many disagreements arise from different interpretations of a word such as “adult,” “controlled,” or “prospective.” If the rule is unclear, consult the written decision guide rather than treating one reviewer’s judgment as the default.

For unresolved cases, use a predefined third-reviewer or consensus process. Record the initial decisions, the final decision, and the reason for the resolution. If the resolution changes an eligibility rule, identify earlier records that may be affected and reassess them. A disagreement log can include the record ID, issue, relevant criterion, final interpretation, and whether the protocol or guidance was updated.

Scenario 5: Building the data-extraction form

Design the extraction form around the review question and planned synthesis. Avoid collecting every detail reported in every article. Excessive fields increase workload and create more opportunities for transcription errors.

CategoryExample fields
Study identityStudy ID, report ID, authors, year, country or setting
DesignDesign type, recruitment method, follow-up period, sample size
ParticipantsEligibility, age range or summary, baseline characteristics
Intervention or exposureDefinition, dose or intensity, timing, delivery context
ComparatorComparator description and differences from the intervention
OutcomesDefinition, measurement tool, time point, reported estimate
AnalysisAnalytic method, adjustment variables, missing-data handling
LimitationsAuthor-stated limitations and reviewer observations
Location in reportPage, table, figure, supplement, or section for each key value

Pilot the form on several studies that differ in design and reporting quality. Check whether two reviewers can locate and interpret each field in the same way. Define how to record “not reported,” “not applicable,” and “unclear”; do not use a blank cell for all three. After piloting, freeze the form for formal extraction, or log any later changes and explain which studies must be revisited.

Scenario 6: Extracting and verifying data

Extract from the complete report and associated supplements, not from the abstract alone. Link multiple reports to one underlying study so that participants are not counted twice. Keep reported values distinct from calculated or converted values, and record the formula or assumption used for any calculation.

For higher-risk fields, use independent duplicate extraction or a targeted verification step. At minimum, verify sample sizes, group labels, outcome definitions, time points, effect estimates, and values that will enter a quantitative synthesis. Resolve discrepancies against the source document and retain an audit note.

What to double-check

  • Deduplication: Check titles, authors, year, DOI, registry identifiers, and study characteristics. A minor title variation may represent the same report, while similar titles may describe separate studies.
  • Study versus report: Distinguish an individual publication from the underlying study. Link protocols, interim reports, primary results, and follow-up reports where appropriate.
  • Eligibility boundaries: Recheck mixed populations, indirect interventions, secondary analyses, and studies with outcomes that are measured differently from the review plan.
  • Missing full texts: Record each retrieval attempt and avoid silently treating an unavailable report as ineligible.
  • Units and time points: Confirm whether measurements use the same scale, whether values are baseline or follow-up, and whether the analysis population matches the extracted sample.
  • Risk-of-bias preparation: Extract the design features and reporting details needed for the chosen assessment approach. Keep risk judgments separate from eligibility decisions.
  • Reference accuracy: Verify article identity, DOI, and report relationships before finalizing the included-study list. A reference audit can help with this stage: use the research paper reference audit checklist.

Common mistakes

  1. Starting without a decision guide. Even experienced reviewers can apply broad criteria differently. Write operational rules and examples before formal screening.
  2. Using vague exclusion reasons. “Not relevant” cannot be audited effectively. Select the specific criterion that failed.
  3. Changing the criteria silently. A revised rule should be dated, explained, and applied consistently to records already screened where necessary.
  4. Extracting only the main article. Supplements, appendices, protocols, and companion reports may contain essential methods or results.
  5. Leaving blanks unexplained. Use controlled labels for not reported, not applicable, and unclear.
  6. Copying numbers without context. A value is incomplete without its group, unit, time point, outcome definition, and source location.
  7. Allowing software defaults to define the method. Tools can organize records, but the review team must define eligibility, duplicate handling, conflict resolution, and audit procedures. For broader tool selection, compare the workflow needs rather than choosing solely by feature count.
  8. Using automated assistance without verification. Summarization or classification tools may support triage, but every inclusion decision and extracted value should be checked against the source. See the guide to evaluating academic AI and research paper summarizers: evaluate these tools safely.

When to revisit

Revisit this workflow before each new review, update, or major training cycle. The screening criteria should be checked whenever the review question, intervention, population, outcome definitions, or eligible study designs change. The extraction form should be reviewed when the planned synthesis changes or when early studies reveal that a necessary field is missing.

Review the software workflow when a platform changes its export format, duplicate detection, permissions, audit trail, or conflict-resolution features. Recheck reporting guidance and local protocol requirements periodically, particularly before finalizing the review. If you are updating an existing review, begin by comparing the previous eligibility criteria, search date, included-study list, and extraction form with the new plan.

Before acting, use this short final checklist:

  • Have the eligibility criteria been written in terms two reviewers can apply?
  • Are original search files, deduplication decisions, and screening dates preserved?
  • Does every excluded full text have a specific reason?
  • Are disagreements and rule changes documented?
  • Has the extraction form been piloted and locked?
  • Are study-level and report-level records linked?
  • Can another researcher trace each important extracted value to its source?
  • Have key fields been verified before synthesis?

A well-documented screening and extraction process does not remove judgment from a systematic review. It makes that judgment visible, repeatable, and easier to correct when the question, evidence base, or review tools change.

Related Topics

#systematic reviews#study screening#data extraction#research methods#review templates
R

Researchers.site Editorial Team

Academic Research Editor

Senior editor and content strategist. Writing about technology, design, and the future of digital media. Follow along for deep dives into the industry's moving parts.