Overview
Most systematic reviews that run into trouble do so early: a question too broad to answer, a search that relies on one database and a handful of keywords, or eligibility criteria that shift once screening begins. By the time a reviewer spots these problems, fixing them usually means re-running the search. Our support is weighted toward the front of the process for exactly that reason.
We work with you as a methodological partner. Depending on what you need, we can draft and test search strategies across databases, set up your screening project in Covidence or Rayyan, act as an independent second screener or data extractor where your institution and target journal permit it, apply the appropriate risk-of-bias tool, and prepare the PRISMA 2020 flow diagram and checklist. Every decision is logged so it can be reported transparently.
If your review includes pooling results statistically, the quantitative synthesis is handled with our meta-analysis support, using the same extraction dataset so nothing is re-keyed. If your question is really about mapping what exists rather than estimating an effect, we will say so and suggest a scoping review instead.
Who it’s for and the problems it solves
This is for you if you are…
- Master's and doctoral students whose thesis includes a systematic review chapter or a stand-alone review paper
- Clinicians, nurses, allied health and public health researchers preparing a review for journal submission
- Education, management and social science researchers adapting systematic methods outside medicine
- Review teams who need a second independent screener or extractor to meet dual-review expectations
- Authors asked by reviewers to redo a search, add a risk-of-bias assessment or report against PRISMA 2020
Common problems we help with
- Your search returned 40,000 hits, or 12, and you cannot tell whether the string or the question is at fault
- You are unsure how to translate a PubMed search with MeSH terms into Embase (Emtree), Scopus or Web of Science syntax
- Screening disagreements between reviewers have stalled and nobody documented the eligibility rules clearly
- Your supervisor asked which risk-of-bias tool you used and why, and you do not have a good answer
- The PRISMA flow diagram numbers do not add up after deduplication and full-text exclusions
What’s included — and what isn’t
Included
- Refining the review question with a structured framework (PICO, PICOS, SPIDER, PEO or PCC as appropriate) and checking whether a similar review already exists or is registered
- Protocol drafting support aligned with PRISMA-P, including eligibility criteria, outcomes, planned synthesis and risk-of-bias approach, ready for PROSPERO, OSF or journal publication
- Search strategy development: concept blocks, synonyms, truncation, proximity operators, controlled vocabulary (MeSH, Emtree, CINAHL Headings) and filters, with a peer-review check against the PRESS checklist
- Translation of the search across databases such as MEDLINE/PubMed, Embase, CINAHL, PsycINFO, ERIC, Scopus, Web of Science and discipline-specific sources, plus a grey-literature and citation-chasing plan
- Deduplication and set-up of a screening project in Covidence, Rayyan or EPPI-Reviewer, with a written screening guide and pilot round
- Title/abstract and full-text screening as an independent second reviewer, with conflict resolution and inter-rater agreement reporting (for example Cohen's kappa)
- Piloted data extraction forms and double extraction or verification of a sample
- Risk-of-bias or quality appraisal using the tool that fits the study designs, such as RoB 2, ROBINS-I, the Newcastle–Ottawa Scale, CASP or JBI checklists
- Narrative synthesis support following SWiM guidance where meta-analysis is not appropriate, and GRADE certainty-of-evidence tables where outcomes allow
- PRISMA 2020 flow diagram, completed checklist and a reproducible search appendix
Not included
- Writing your review's introduction, discussion or conclusions for submission as your own work; we provide methods material and explain results so you can write them
- Being listed as, or substituting for, a named review author without your team's and journal's agreement on contribution and authorship
- Registering the protocol in your name without your review and approval
- Selecting studies or tuning eligibility after seeing results to reach a preferred conclusion
- Promising publication in a particular journal or that a registry will accept the protocol unchanged
What you receive
- Protocol draft and registration pack
- A protocol structured to PRISMA-P items with the fields PROSPERO or OSF will ask for, so you can review, adapt and register it yourself.
- Search strategy file
- Line-by-line strategies for each database with dates searched, platform, hit counts and notes on translation choices, formatted as a PRISMA-S style appendix.
- Screening log and PRISMA flow diagram
- Records identified, deduplicated, screened and excluded at each stage with reasons, reconciled so every number in the PRISMA 2020 diagram can be traced.
- Extraction dataset
- A clean spreadsheet or Covidence export of study characteristics and outcome data, with a codebook explaining every field.
- Risk-of-bias tables and figures
- Domain-level judgments with supporting quotes, summary plots (for example traffic-light and summary bar charts) and a note on how bias informs the synthesis.
- Methods reporting notes
- Plain-language notes explaining each methodological decision so you can write and defend your methods section.
How it works
Question and feasibility check
We refine the question, run scoping searches to estimate the yield and check PROSPERO, OSF and journals for overlapping reviews before you commit.
Protocol and search build
We draft the protocol with you and build the search in one primary database first, testing it against a set of known relevant papers.
Search execution and deduplication
Strategies are translated and run across the agreed databases, results are exported to a reference manager and deduplicated, and all dates and counts are logged.
Screening and extraction
After a calibration round on a sample, screening proceeds in two stages with conflicts resolved by discussion or a third reviewer, followed by piloted extraction.
Appraisal, synthesis and reporting
We apply the agreed risk-of-bias tool, support synthesis and GRADE where relevant, and assemble the PRISMA 2020 diagram, checklist and appendices.
Standards, tools and appraisal instruments
- PRISMA 2020 and extensions
- The 27-item PRISMA 2020 checklist and flow diagram for reporting, PRISMA-P for protocols and PRISMA-S for reporting the literature search. Updates to several extensions are in development, so we check the PRISMA website at the start of each project.
- Cochrane Handbook
- The Cochrane Handbook for Systematic Reviews of Interventions (version 6.5 at the time of writing) guides our approach to searching, bias assessment and synthesis for intervention reviews, even when the review is not a Cochrane review.
- Protocol registration
- PROSPERO for eligible health-related reviews with a health outcome (relaunched on a new platform in February 2025, with every author now confirming the record), OSF Registries for other disciplines, or a published protocol.
- RoB 2
- Cochrane's revised tool for randomized trials, assessed per outcome across five domains, with variants for cluster and crossover trials.
- ROBINS-I and ROBINS-E
- For non-randomized studies of interventions and of exposures. ROBINS-I version 2 has been released in draft with restructured domains, so we agree with you which version to use and report it explicitly.
- Other appraisal tools
- Newcastle–Ottawa Scale, JBI critical appraisal checklists, CASP checklists, QUADAS-2 for diagnostic accuracy and MMAT for mixed-study reviews, matched to your included designs.
- GRADE
- Rating certainty of evidence per outcome (risk of bias, inconsistency, indirectness, imprecision, publication bias) and building summary-of-findings tables, including GRADEpro GDT where helpful.
- Screening software
- Covidence, Rayyan, EPPI-Reviewer or ASReview for screening, with reference management in EndNote, Zotero or Mendeley. We work within whatever your institution licenses.
What makes a search strategy reproducible
A good search is built from concepts, not keywords. Each concept in your question (for example the population and the intervention) becomes a block of free-text synonyms combined with OR, alongside the database's controlled vocabulary, and the blocks are then combined with AND. Outcomes are often left out of the search deliberately, because they are poorly described in abstracts.
Each database has its own syntax and thesaurus, so a PubMed string using MeSH terms and [tiab] field tags has to be rewritten, not pasted, for Embase via Ovid or for Scopus. We record every line and its hit count so another team could run the same search and get the same records.
- Test the string against known relevant papers before running it everywhere
- Use truncation (child*) and proximity operators (adj3, NEAR/3) deliberately, not by default
- Document limits such as date, language and study-design filters and justify each one
- Plan grey literature, trial registries and citation chasing, not only bibliographic databases
Dual screening and why agreement matters
Two people screening independently reduces the chance that relevant studies are missed and makes eligibility decisions defensible. Before full screening starts we run a calibration batch, compare decisions, sharpen ambiguous criteria and record the changes. Agreement statistics and the handling of conflicts then go into your methods section.
If your program expects you to screen alone, we can instead verify a random sample of your decisions and report that honestly as a limitation rather than implying full dual review.
Choosing the right risk-of-bias tool
There is no universal quality score. Randomized trials, cohort studies, qualitative studies and diagnostic accuracy studies are vulnerable to different biases, and tools are built around those differences. Summing items into a single quality score is discouraged in current guidance; judgments are made domain by domain and then used in the synthesis, for example through sensitivity analysis or GRADE downgrading.
- Randomized trials: RoB 2
- Non-randomized intervention studies: ROBINS-I (confirm version)
- Observational exposure studies: ROBINS-E or Newcastle–Ottawa Scale
- Qualitative studies: CASP or the JBI qualitative checklist
- Diagnostic accuracy: QUADAS-2
When a systematic review is not the right design
Some questions are better served by other evidence syntheses. If the literature is too heterogeneous to compare, a scoping review may be more honest. If you need an answer within weeks for a policy decision, a rapid review with transparent shortcuts may fit. We raise these options at the feasibility stage so your time goes into the design that fits your question.
What the work can look like
Illustrative example — fictional data
Illustrative example: school-based physical activity and adolescent anxiety
A fictional doctoral candidate wants to review randomized and non-randomized trials of school-based physical activity programs for anxiety symptoms in 11–18 year olds. Her first PubMed search used three keywords and returned 214 records. The numbers below are invented to show how the process is documented.
Search rebuilt
Three concept blocks (adolescents; school setting; physical activity) combined MeSH terms with free-text synonyms. Tested against eight known trials, the new string retrieved all eight. Across MEDLINE, Embase, PsycINFO, ERIC and SPORTDiscus it returned 6,480 records, 4,112 after deduplication.
Screening
A 200-record calibration round produced kappa = 0.58; two criteria were clarified (definition of 'school-based' and minimum program length) and a second round reached 0.81. Title/abstract screening excluded 3,947 records; 165 full texts were assessed and 23 studies were included.
Appraisal
Fourteen randomized trials were assessed with RoB 2 and nine non-randomized studies with ROBINS-I. Most trials were rated 'some concerns' because outcome assessors knew group allocation, which later informed a GRADE downgrade for risk of bias.
Illustrative example with invented figures, not a real client project. Real searches, yields and decisions depend on your question and databases.
What you’ll need to provide
- Your draft review question and the reason the review is needed
- Any existing protocol, supervisor guidance or target journal instructions
- Access details for the databases your institution subscribes to, or agreement to use the ones we can access
- A small set of papers you know should be included, used to test the search
- Your reviewer team arrangements: who screens, who resolves conflicts and who approves decisions
You don’t need everything ready to ask for a quote — tell us what you have and we’ll say what’s missing.
What affects price and turnaround
Price depends on
- Breadth of the question and expected number of records after deduplication
- Number of databases and whether grey literature, trial registries and citation chasing are included
- Which stages you want support with, from protocol only to the full pipeline
- Whether we act as a second screener or extractor, and for how many records or studies
- Number of included studies requiring risk-of-bias assessment and the tools involved
- Whether GRADE tables or meta-analysis are needed
Turnaround depends on
- Search yield: screening time scales directly with the number of unique records
- Speed of team decisions on eligibility questions and conflict resolution
- Full-text retrieval, which depends on your library access and interlibrary loans
- Whether the protocol must be registered before screening starts
Every project is quoted individually. How we price our work.
Questions people ask before starting
Can you be the second reviewer for my screening?
Often, yes. Many teams need an independent second screener or extractor. Whether that is appropriate depends on your institution's rules and your target journal's authorship and contribution policies, so we agree up front how our role will be described, typically in the acknowledgments or contributor statement.
Does PROSPERO accept my review?
PROSPERO registers systematic reviews with a health-related outcome and is strict about registering before data extraction begins. Reviews outside health, and many scoping reviews, are usually registered on OSF Registries instead. We check eligibility against PROSPERO's current guidance at the start, because the platform was relaunched in 2025 and its forms and workflow changed.
Which databases do I need to search?
It depends on your field. Health reviews typically need at least MEDLINE and Embase plus a subject database such as CINAHL or PsycINFO; education and business reviews lean on ERIC, Scopus, Web of Science, Business Source or similar. One database is rarely enough for a systematic review. We recommend a set and explain the reasoning.
What if my supervisor wants a different risk-of-bias tool?
We discuss the options with the reasoning for each and follow your supervisor's decision where it is defensible. If a requested tool does not fit your study designs, we explain why in writing so you can raise it with them.
Can AI tools speed up screening?
Active-learning tools such as ASReview, and prioritization features in some screening platforms, can reduce the number of records read manually. They need a transparent stopping rule and must be reported, and some journals and supervisors remain cautious. PRISMA is developing guidance on reporting AI use, so we document any tool use carefully.
I only need help with the search. Is that possible?
Yes. Search strategy development and translation is one of the most common single-stage requests, and it is the stage where early input saves the most time later.
Will you write the review?
For student work, no: you write the review and we supply the methods material, logs, tables and explanations. For professional research teams we can contribute to methods and results reporting, with our role declared according to the journal's policy.