To start a medical review, define a focused question, register a protocol, and plan search, screening, and extraction steps using PRISMA.
Starting well saves months later. This guide gives a clear path from blank page to a polished draft. You’ll turn a loose idea into a tight plan, set transparent methods, and build the backbone for a paper editors can trust.
What A Medical Review Is And Isn’t
A review gathers and appraises existing studies to answer a specific question. It doesn’t invent new experiments. It maps what is known, how strong that evidence is, and where gaps remain. Your job is to be precise, transparent, and fair.
Common Review Types And When To Use Them
Pick a format before you write a single line. Each type answers a different need and sets expectations for scope, methods, and reporting.
| Type | Purpose | When It Fits |
|---|---|---|
| Scoping Review | Maps concepts, definitions, and study types. | Early fields, broad questions, method mapping. |
| Narrative Review | Explains a topic with expert synthesis. | Background overviews with clear structure. |
| Systematic Review | Answers a focused question with a pre-set method. | Clinical or policy decisions need structured evidence. |
| Meta-analysis | Pools effect sizes statistically. | Similar trials with extractable outcomes. |
| Rapid Review | Streamlined methods for time-sensitive needs. | Guidance under tight timelines. |
| Umbrella Review | Synthesises published reviews. | Many reviews exist on related angles. |
Starting A Medical Literature Review: First Steps
Lock the topic and audience. A clinical journal, a health policy outlet, and a basic science venue expect different depth and tone. Write a one-sentence aim that names population, intervention, comparator, and outcomes where relevant.
Craft A Sharp Question With PICO/PEO
Use a template to avoid vague aims. PICO suits interventional topics: Population, Intervention, Comparator, Outcome. PEO suits exposure topics: Population, Exposure, Outcome. Draft a few versions, then pick the tightest one that still matters to readers.
Check The Field For Saturation
Scan top journals and preprint servers to see whether a full synthesis already exists. If it does, adjust scope, outcomes, or time window. Cite the gap you plan to fill. This step prevents redundant work and helps your pitch land.
Write And Register A Protocol
A protocol guards against bias and scope drift. It lists aims, search sources, keywords and subject headings, inclusion and exclusion criteria, screening flow, quality tools, and planned synthesis. Register on a public registry when possible. Add the link in your manuscript.
Build A Search You Can Defend
Your search must be reproducible. Use at least two major databases and one grey source. Combine controlled vocabulary terms with free-text synonyms. Save strings and dates. Pilot the search and check that known sentinel papers appear; if they don’t, refine terms.
Pick Databases And Sources
Core options include MEDLINE, Embase, CENTRAL, Web of Science, Scopus, and CINAHL, with trial registries and preprint servers as needed. Add subject repositories for subfields. Document each source, platform, and date searched.
Design Search Strings
Blend MeSH/Emtree terms with text words. Use proximity operators where supported. Balance recall and precision so you don’t drown in noise or miss key trials. Save each string in a text file and in the appendix.
Set Clear Eligibility Criteria
Define inclusion and exclusion rules up front. State study designs, settings, age ranges, comparators, outcome measures, language limits, and time frames. Pre-plan how you’ll handle preprints and conference abstracts.
Screen, Extract, And Appraise
Move from records to studies in a controlled flow. Screen titles and abstracts in pairs, resolve disagreements with a third reviewer when needed, then review full texts. Keep a log of reasons for exclusion at the full-text stage.
Use A PRISMA Flow Diagram
Track numbers from identification through inclusion. Report counts for records found, deduplicated, screened, full texts assessed, and final studies. The diagram becomes a figure in the draft and a transparency signal for readers. See the PRISMA 2020 checklist for items your report should cover.
Extract Data With A Piloted Form
Build a form that captures design, sample size, setting, intervention or exposure, comparators, outcomes, follow-up, effect estimates, and notes. Pilot the form on three to five papers, adjust fields, then extract in pairs.
Assess Bias With Validated Tools
Match the tool to the design: RoB-2 for randomised trials, ROBINS-I for non-randomised studies, QUADAS-2 for diagnostic accuracy, AMSTAR-2 for reviews. Two assessors work in parallel and reach consensus with a third when views differ.
Synthesis: From Notes To Findings
Choose narrative, tabular, or quantitative synthesis based on study similarity. When pooling effects, predefine the model, effect measure, and heterogeneity thresholds. When studies are too diverse, stick to structured narrative with tables and figures that carry the weight.
Plan Subgroups And Sensitivity Checks
State subgroup rules in the protocol. Run leave-one-out checks, compare fixed and random models, and examine small-study effects. Report any shifts in direction or size and explain likely drivers.
Grade The Certainty Of Evidence
Use GRADE to rate each outcome across risk of bias, inconsistency, indirectness, imprecision, and publication bias. Present a summary of findings table with plain-language takeaways for the main outcomes.
Write The Draft: Section By Section
Follow journal norms so editors can scan fast. Keep methods reproducible, results crisp, and claims balanced.
Title And Abstract
State the question and design in plain terms. Mirror the structure of the paper in a structured abstract. Include databases searched and the end date. Skip hype and keep adjectives tight.
Introduction
Set context in three short paragraphs: what is known, why the question matters, and the specific aim. End with a one-sentence objective that mirrors your protocol.
Methods
Report protocol registration, eligibility rules, databases and dates, full search strings, screening process, data items, bias tools, effect measures, and synthesis plan. Place long strings and forms in the appendix or a repository link. For scope and question framing, the Cochrane Handbook chapter on scope is a clear benchmark.
Results
Lead with the PRISMA flow, then study characteristics, risk of bias, and the main effect estimates. Use figures for forest plots and funnel plots when suitable. Keep the text to the point; let tables carry details.
Discussion
Open with the main finding in one line. Compare with the best prior syntheses, give plausible reasons for any differences, and state practical meaning for clinicians or policy teams. Lay out limits from methods or evidence and note where research should go next.
References And Data
Use a reference manager and a consistent style. Share extraction sheets and code in a repository when possible. Data access builds trust and invites reuse.
Quality And Reporting Checkpoints
Use checklists to keep the draft tight and transparent. Run through PRISMA for reporting, bias tools for study appraisal, and GRADE for certainty. Ask a librarian or information specialist to review the search before submission.
Ethics And Registration
Most reviews don’t need ethics board review since they use public data, yet some journals request a statement. If you plan to contact authors for extra data, state that plan. Register protocols on public platforms when allowed.
Style, Tone, And Readability
Keep sentences short and active. Use plain words where you can. Define statistical terms once, then use them consistently. Write for a reader who knows the field but has limited time.
Tools That Speed Up The Work
Use a citation manager, reference style files, and screening software. Rayyan, Covidence, EPPI-Reviewer, and similar tools improve throughput and audit trails. Keep backups and export decision logs for your appendix.
Ethical Authorship And Roles
Set roles at the start. Decide who leads the question, who crafts search strings, who screens, who extracts, and who writes. Use a contributorship model so readers see who did what.
Typical Timeline And Milestones
Time varies by scope and team size. A tight topic with clear outcomes can move from protocol to submission in a few months, while broad syntheses take longer. Break the work into small sprints and close each stage before moving on.
| Stage | Main Outputs | Est. Duration |
|---|---|---|
| Scoping & Question | Aim, PICO/PEO, target journal list. | 1–2 weeks |
| Protocol & Registration | Methods, eligibility, plan link. | 1–2 weeks |
| Search & Screening | Saved strings, PRISMA counts. | 3–6 weeks |
| Extraction & Appraisal | Data sheets, bias ratings. | 3–5 weeks |
| Synthesis & Writing | Figures, tables, draft. | 3–4 weeks |
| Revision & Submission | Checklist pass, cover letter. | 1–2 weeks |
Common Pitfalls To Dodge
These traps burn time and invite desk rejections. Plan to avoid them from day one.
Vague Or Moving Question
A drifting aim wrecks screening and synthesis. Freeze wording in the protocol and use it across the draft.
One-Database Searches
Single-source searches miss key trials. Use at least two large databases plus a trial registry and a preprint source when relevant.
No Bias Appraisal
Skipping bias tools weakens claims. Pick the right tool per design and have two people rate each study.
Unclear Data Handling
State how you managed missing data, split groups, or duplicated cohorts. Log any contact with authors and how responses changed extraction.
A Simple Starter Template
Copy this skeleton into your document and adapt fields to your topic. It keeps sections in the right order and prompts you to record decisions.
Working Title
“Effect of [intervention] on [outcome] in [population]: a systematic review.”
Aim
To assess the effect of [intervention] versus [comparator] on [primary outcome] in [population].
Protocol
Registry link; databases; end date; eligibility rules; primary and secondary outcomes.
Search
Full strings for each database; filters used; manual search plan.
Screening
Number of reviewers; conflict resolution plan; tools used.
Extraction
Data items; pilot test plan; storage location.
Appraisal
Tools by design; consensus process; training notes.
Synthesis
Effect measures; models; subgroup and sensitivity plan; narrative structure.
Outputs
Figures, tables, flow diagram; repository links.
Final Checks Before You Submit
Read the draft aloud, trim repeats, and tighten verbs. Confirm numbers match across text, tables, and figures. Re-run the PRISMA checklist and a spell check. Ask a domain peer to read for clarity and a librarian to assess the search.