Student & Academic
How to run a systematic literature review — a practical guide for graduate students and researchers in health sciences, education, and social sciences who need to follow the formal SLR protocol, from writing a search strategy to the PRISMA flow diagram to synthesis.
A systematic literature review (SLR) is not a thorough narrative literature review. It is a formal research methodology with specific requirements for reproducibility, transparency, and documentation. The difference matters: a narrative review reflects the researcher's expert judgment about which sources are relevant; a systematic review follows a pre-specified protocol designed to minimize selection bias and produce a reproducible, auditable record of how the literature was identified and assessed.
Systematic reviews are required or strongly preferred in health sciences, clinical medicine, epidemiology, education research, and social policy research — any field where the quality of a review of evidence is itself a research product, not just background for a paper.
The formal requirements of a systematic review include: a documented search protocol (which databases, which search terms, which date ranges); explicit inclusion and exclusion criteria for sources; a multi-stage screening process (title/abstract screening, full-text screening); a quality assessment for each included study; a structured data extraction; and a synthesis that acknowledges the methodological variation across included studies.
The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) framework, updated in Page et al. (2021) in the BMJ, provides the standard reporting checklist and flow diagram that most journals in health sciences and many in social sciences require when publishing systematic reviews.
This guide covers how to run a systematic literature review at the graduate student level.
The protocol must be written before the review, not derived from it: In a narrative review, the researcher finds the literature and then describes their approach. In a systematic review, the protocol — the search strategy, inclusion criteria, and extraction template — must be documented before the review begins, and the review must follow the protocol. This prevents a common bias: retrospectively including studies that support a preferred conclusion and excluding those that don't.
Search comprehensiveness is a formal requirement: A systematic review must search all relevant databases for a field — not just the ones that are most convenient. In health sciences, this typically means MEDLINE (PubMed), Embase, and the Cochrane Library at minimum; additional databases (CINAHL, PsycINFO, Web of Science) depending on the topic. Grey literature (clinical trial registries, government reports, conference proceedings) may also be required. An incomplete search is a methodological flaw that reviewers and examiners will identify.
Screening happens in stages, with documented decisions: Each source found in the database search must be screened twice: once at the title/abstract level (usually faster), once at the full-text level (slower). Both stages require documented inclusion/exclusion decisions. Excluded sources at the full-text stage must be listed with reasons for exclusion.
Quality assessment is required: Each included study must be assessed for methodological quality using a validated tool. Common tools include the Cochrane Risk of Bias tool (for randomized trials), ROBINS-I (for non-randomized intervention studies), and the Newcastle-Ottawa Scale (for observational studies). The quality assessment doesn't necessarily exclude studies — but it informs how much weight their findings are given in the synthesis.
Before beginning a review, write a protocol. Registering the protocol (at PROSPERO for health research, or the Open Science Framework) is standard practice and signals rigor to reviewers and examiners.
SYSTEMATIC REVIEW PROTOCOL
1. Research question: [The formal PICO(T) or PCC framework statement]
Population (who): [...]
Intervention/Exposure/Concept (what): [...]
Comparison (vs. what, if applicable): [...]
Outcomes (what are you measuring): [...]
Time (if applicable): [...]
2. Search strategy
Databases: [List all databases to be searched]
Search terms and Boolean operators:
Database 1: ([Term1] OR [Term2]) AND ([Term3] OR [Term4])
Database 2: [...]
Date limits: [From year X to year Y, or no date limit — and why]
Language limits: [English only, or all languages + translation — and why]
3. Inclusion criteria
Study types: [RCT / observational / qualitative / mixed — and why]
Population: [Who must be in the study for inclusion]
Intervention/exposure: [What must be present]
Outcomes: [What outcomes must be measured]
Publication type: [Peer-reviewed only / grey literature included / etc.]
4. Exclusion criteria
[What excludes a study at each stage]
5. Screening process
Stage 1 — Title/abstract: [Who screens, how conflicts are resolved]
Stage 2 — Full-text: [Who screens, how conflicts are resolved]
[For rigorous reviews: two independent screeners with inter-rater reliability calculated]
6. Data extraction template
[What data will be extracted from each included study: sample, intervention, outcomes, effect sizes, quality assessment]
7. Quality assessment tool
[Which validated tool will be used and why]
8. Synthesis approach
[Narrative synthesis / meta-analysis / thematic synthesis — and criteria for choosing]
The PRISMA flow diagram documents the number of records at each stage of the review: records identified, records screened, records assessed for eligibility, studies included, and reasons for exclusion at each stage.
PRISMA 2020 FLOW
Records identified through database searches: [N]
+ Records from other sources (grey literature, citation searching): [N]
Total records identified: [N]
After duplicate removal: [N records]
Stage 1 — Title/abstract screening:
Records screened: [N]
Records excluded: [N] (reason: did not meet inclusion criteria)
Records assessed for full-text: [N]
Stage 2 — Full-text screening:
Records assessed: [N]
Records excluded: [N] (with reasons)
Studies included in review: [N]
(Included in qualitative synthesis: [N])
(Included in meta-analysis: [N], if applicable)
This diagram is required for PRISMA-compliant reporting and is expected in dissertations and publications using the SLR methodology.
The PICO(T) framework (Population, Intervention, Comparison, Outcomes, Time) forces specificity in the research question. This specificity determines which studies are included and which aren't — it's the foundation of the protocol.
A research question for a systematic review is more specific than for a narrative review: not "what are the effects of mindfulness on mental health?" but "what is the effect of mindfulness-based interventions, compared to treatment as usual, on anxiety symptoms in college students, as measured by validated anxiety scales, in studies conducted between 2010 and 2024?"
The search strategy is a set of Boolean search queries — combinations of search terms connected by AND, OR, and NOT — tailored to each database's indexing vocabulary.
Most health sciences databases use Medical Subject Headings (MeSH) for indexing. The search strategy should include both MeSH terms and free-text keywords, combined with OR (to capture synonyms) and then combined with other concepts using AND.
Example:
("mindfulness" OR "mindfulness-based stress reduction" OR "MBSR" OR "mindfulness meditation" OR "mindfulness-based cognitive therapy")
AND
("anxiety" OR "anxiety disorders" OR "generalized anxiety" OR "GAD")
AND
("college students" OR "university students" OR "undergraduate" OR "graduate students")
This search is run separately in each required database. Record the number of results from each database before combining and removing duplicates.
Upload all records to a screening tool (Covidence, Rayyan, or a spreadsheet). Screen each record at the title/abstract level: include if the study appears to meet inclusion criteria, exclude if it clearly doesn't, tag as uncertain if you can't tell from the abstract alone.
Uncertain records advance to full-text screening by default. Speed matters here: title/abstract screening for 500+ records typically takes several hours; use a consistent inclusion/exclusion checklist for each decision.
Obtain full text for all records that passed Stage 1 screening. For each, apply the full inclusion/exclusion criteria. For each excluded record, document the reason (wrong population, wrong intervention, wrong outcomes, wrong study design, etc.). This exclusion list with reasons is published with the review.
For each included study, extract the specified data using the extraction template from the protocol: study design, sample characteristics, intervention description, outcomes measured, effect sizes, and any quality flags.
Consistent extraction is critical: the same information must be extracted from each study in the same way for synthesis to be valid.
Apply the appropriate quality assessment tool to each included study. Document the assessment for each domain of risk/quality, with the rationale for each judgment. These assessments inform the synthesis: findings from higher-quality studies are given more weight.
If the included studies are methodologically similar enough, a meta-analysis (statistical pooling of effect sizes) may be possible. If not — which is common — a narrative synthesis describes what the studies found, organized by outcome or sub-group, with reference to the quality assessment.
The Cochrane Handbook for Systematic Reviews (Higgins et al., 2022) provides detailed guidance on synthesis approaches for different types of evidence.
Research question (PICO): In college students (P), do mindfulness-based interventions (I), compared to waitlist/treatment as usual (C), reduce anxiety symptoms (O) as measured by validated scales, in studies published 2012-2024 (T)?
Databases searched: PubMed, PsycINFO, Cochrane Library, CINAHL — resulting in 847 records before duplicate removal.
After duplicate removal: 614 records screened
Stage 1 screening: 78 records passed to full-text (536 excluded: wrong population, wrong intervention, review/protocol articles, case studies)
Stage 2 screening: 18 studies included (60 full texts excluded: outcomes not measured with validated scales [22], no control/comparison group [14], not mindfulness-based [12], full text not accessible [6], other [6])
PRISMA flow summary: 847 identified → 614 deduplicated → 78 full texts → 18 included
Quality assessment: Cochrane Risk of Bias tool; 6 studies low risk, 8 moderate risk, 4 high risk
Synthesis: Narrative — too much methodological heterogeneity (different MBI types, different anxiety scales, different populations) for meta-analysis; organized by intervention type and student population
Running a systematic literature review is a formal research methodology, not a more thorough version of a standard literature review. The protocol, the documented search strategy, the staged screening, the quality assessment, and the PRISMA flow diagram are all required elements — not optional enhancements. For graduate students in health sciences, education research, and social policy, the systematic review is often the dissertation's or thesis's primary methodological contribution. Learning the formal protocol early, before the review begins rather than during it, prevents the most common systematic review failure mode: discovering mid-review that the protocol wasn't documented and the search wasn't comprehensive enough to meet publication or examination standards.
More WebSnips articles that pair well with this topic.
How to build flashcards from your reading — a practical guide for students who want to convert reading notes into high-quality flashcards that actually produce long-term retention, rather than low-quality cards that take time to make and time to review without producing learning.
How to organize PDFs for research — a practical guide for students and researchers who want a systematic, searchable, and durable PDF organization system that remains useful when a collection grows from 20 files to 400, and that makes it easy to find any source without remembering the filename.
How to prepare a conference presentation from notes — a practical guide for students and early-career researchers who want to transform their research notes into a focused, compelling conference talk or poster, without losing the nuance of the underlying work or overloading their audience.
How to read faster without losing comprehension — a practical guide for students who have more assigned reading than they can finish, and want to increase their effective reading rate through legitimate strategies grounded in cognitive science, not speed-reading myths.
How to write a dissertation literature chapter — a practical guide for PhD students and graduate researchers who want to write a literature chapter that makes a sustained argument about the field, rather than a survey that summarizes source after source without building toward a coherent claim.
How to avoid plagiarism with good note-taking — a practical guide for students who want a note-taking system that clearly distinguishes quotes from paraphrases from their own ideas, so attribution is automatic rather than a last-minute panic before submission.