How to Conduct a Systematic Literature Review (SLR) with PRISMA

PRISMA is a guideline consisting of 27 checklist items and a flow diagram that helps researchers report systematic reviews comprehensively and transparently. The latest version, PRISMA 2020, updates the 2009 version with additions:

• Items 16a-16d: More detailed reporting of search results and study selection
• Item 20: Transparency in handling missing data
• Item 24: Protocol registration (e.g., in PROSPERO) to avoid duplication bias

Why does this matter? Reputable journals like The Lancet, BMJ, and Nature require PRISMA compliance as a publication condition. Without PRISMA, your review risks rejection at the initial screening stage.

The first stage is collecting all potential articles from various databases. The steps are:

• Define databases: Minimum 2-3 major databases such as Scopus, Web of Science, PubMed, IEEE Xplore, or ProQuest depending on your field.
• Create search string: Use combinations of Boolean operators (AND, OR, NOT) with relevant keywords. Example: ("machine learning" OR "deep learning") AND ("healthcare" OR "medical") AND "diagnosis"
• Document results: Record the number of articles from each database, search dates, and the search strings used.
• Remove duplicates: Use Mendeley, Zotero, or the deduplication feature in Rayyan.

Pro Tip: Save the search history from each database. This is crucial for reproducibility and will be requested by journal reviewers.

After removing duplicates, it’s time to screen articles based on title and abstract:

• Create clear inclusion/exclusion criteria before starting screening. Example criteria: publication year (2015-2026), language (English only), article type (peer-reviewed journal), study population.
• Use collaborative tools: Rayyan (free, suitable for teams), Covidence (paid, full features), or systematic review tools in Mendeley.
• Conduct independent screening: If working in a team, at least 2 reviewers must assess each article independently, then discuss disagreements.
• Record exclusion reasons: Every rejected article must have a specific reason (not just “not relevant”).

Articles that pass title/abstract screening are then read in full text for eligibility assessment:

• Read full text carefully: Focus on methodology, population, intervention, and outcomes that align with your research question.
• Use PICOS framework: Population, Intervention, Comparison, Outcome, Study design to assess suitability.
• Quality assessment: Use tools such as Newcastle-Ottawa Scale (for observational studies), Cochrane Risk of Bias (for RCTs), or JBI Critical Appraisal Tools.
• Document all decisions: Every article excluded at this stage must be recorded with its reason.

Articles that pass all previous stages enter the final included studies:

• Data extraction: Create a spreadsheet with columns: author, year, title, country, methodology, sample, main findings, limitations.
• Synthesize findings: This can be narrative synthesis (for heterogeneous data) or meta-analysis (for homogeneous data with combinable effect sizes).
• Identify research gaps: From emerging patterns, determine areas that haven’t been widely studied.
• Write results: Follow the PRISMA checklist structure to ensure no item is missed.