How to Systematically Screen Records and Full Texts in Evidence Synthesis

Introduction

In systematic reviews and meta-analyses, identifying relevant studies is not a one-step act of database searching. It is a structured, multistage process of sifting—first through titles and abstracts, and later through full texts. This screening process determines the methodological integrity of the review. Missteps here can lead to missed evidence, inclusion of irrelevant studies, or inconsistent application of criteria—all of which introduce bias.

This guide outlines the full procedure for systematic screening: from post-search record handling to full-text appraisal. It integrates methodological principles with digital tools to optimize efficiency, transparency, and reproducibility.

Part 1: Preparing for the Screening Process

Finalize the Set of Search Records

Before screening begins:

Remove duplicates using citation managers (e.g., EndNote, Zotero) or review platforms (e.g., Rayyan).
Confirm the dataset is clean, containing only unique entries.

Assign Review Authors and Define Roles

Minimum requirement: Two independent reviewers.
Reviewers should ideally be clinically trained, or closely guided by domain experts.
Appoint a third reviewer to resolve disagreements when reviewers diverge in their inclusion decisions.

This triad ensures independence and conflict resolution—a cornerstone of unbiased screening.

Part 2: Title and Abstract Screening

Choose a Screening Platform

Manual screening in citation managers is discouraged due to poor tracking and high error risk.
Rayyan: A web-based platform with real-time collaboration and blinding features. Highly recommended.
ASReview: An AI-assisted tool that prioritizes references based on active learning. Useful for large reviews with limited human resources.

Setup and Workflow

For Rayyan:

Create a new review project.
Upload the deduplicated records.
Add all reviewers as collaborators.
Blind decisions to avoid influencing each other.

For ASReview:

Define stopping rules before screening begins (e.g., stop after reviewing 75% of the dataset or after 12 hours of screening time).
Use only if reviewers are familiar with AI-based prioritization.

Screening Logic

Only screen titles and abstracts at this stage.
Do not open full texts yet.
Look for match to PICO (Population, Intervention, Comparator, Outcome) or DDO (Domain, Determinant, Outcome) structure.

Decision options:

Include (possible match)
Exclude (clearly irrelevant)
Maybe (uncertain—flag for full-text check)

Recording exclusion reasons is optional here but may aid transparency.

Part 3: Documenting with PRISMA Flow

All screening decisions must be tracked in a PRISMA flow diagram:

Number of records identified.
Number removed as duplicates.
Number screened at title/abstract stage.
Number excluded and passed to full-text review.

This flowchart is part of the review’s reproducibility audit trail and must be reported in publications.

Part 4: Full-Text Screening for Eligibility

Retrieve All Full Texts

Use multiple methods to gather full-text PDFs:

Institutional access or VPNs for subscription-based content.
Open-access repositories.
Direct emails to corresponding authors.
Avoid using unauthorized sources for ethical reasons.

Apply Standardized Eligibility Criteria

Set explicit inclusion and exclusion criteria before starting. These criteria should align with the review’s objectives and be piloted if necessary.

Examples:

Exclude non-randomized studies if only RCTs are eligible.
Include only English-language articles (if language restrictions apply).

Use a structured Excel matrix or screening management software to:

Track each article.
Note eligibility decisions.
Record reasons for exclusion at this stage—mandatory for PRISMA reporting.

Reviewer Structure

Again, at least two independent reviewers should assess each full-text.
A third reviewer steps in to resolve conflicts.

Illustrative reasons for exclusion might include:

Wrong population (e.g., pediatric instead of adult)
Insufficient data for extraction
Non-original research (e.g., commentary, editorial)

Save all excluded full-texts with annotation, ensuring an audit trail for transparency.

Part 5: Integrating Decisions into Final PRISMA Flow

The full PRISMA diagram evolves after full-text screening to show:

Number of full-texts retrieved.
Reasons for full-text exclusions.
Final count of studies included in the review.

This visual flow is essential for publication and peer review, and it illustrates the objectivity of study selection.

Conclusion

Systematic screening is both a mechanical and intellectual process. It requires precision, judgment, and transparent collaboration. Digital tools such as Rayyan and ASReview enhance efficiency, but rigor is anchored in well-defined criteria and reproducible procedures.

By approaching screening as a stepwise method—from deduplication to eligibility justification—reviewers can ensure that the final evidence base is both defensible and representative.

Key Takeaways

Begin screening only after deduplication and team role assignment.
Use two independent reviewers, plus a third for conflict resolution.
Title/abstract screening uses PICO/DDO logic without full-text access.
Full-text screening requires recorded exclusion reasons.
Tools like Rayyan and ASReview support efficiency and transparency.
Every decision feeds into the PRISMA flow, which becomes part of the published review.