Start Here

Write the question before choosing the design

Write one sentence that identifies the population, the main exposure/intervention or phenomenon, the comparator when relevant, and the outcome. A vague topic such as “diabetes in our hospital” is not yet a research question.

Check whether the answer already exists

Search recent reviews, guidelines and major primary studies. Record what is known, what remains uncertain, and why answering the question locally or with a new synthesis would add value.

Choose the study design that matches the question

• Description of one/few unusual patients → case report or case series.
• Prevalence or association measured at one period → cross-sectional study.
• Start from an outcome and look backwards for exposure → case-control study.
• Start from an exposure/group and observe outcomes over time → cohort study.
• Assign an intervention randomly → randomised controlled trial.
• Synthesize existing studies with explicit reproducible methods → systematic review; add meta-analysis only when quantitative pooling is appropriate.
• Map a broad field, concepts or evidence gaps → scoping review.
• Explore experiences, meanings or processes → qualitative study.

Build the team and assign roles

Identify a supervisor or principal investigator, clinical content expertise, methods/statistics expertise, data access support and any required information specialist. Agree authorship expectations early, but do not promise authorship solely for seniority or access.

Write the protocol before touching the study data

Define objectives, design, population, eligibility criteria, variables, outcomes, sample-size rationale, data sources, analysis, privacy safeguards and dissemination plan. For evidence syntheses, define the search, screening and extraction methods before screening begins.

Obtain the required governance determinations

Send the protocol through the authorised institutional route. Human-participant research, identifiable clinical data, specimens and prospective recruitment commonly require ethics and governance review. Audit/QI and some publication-only activities may use a different route, but the project team should not self-declare exemption when institutional review is required.

Create a data-management plan

Specify exactly what data will be collected, where it comes from, who may access it, how identifiers will be separated, how files will be named/versioned, where the key will be stored, retention rules and how data will be destroyed or archived. Collect only variables needed for the stated aims.

Lock the analysis plan

Define the primary outcome, secondary outcomes, effect measures, descriptive statistics, planned models/tests, confounders, subgroup analyses, missing-data approach and sensitivity analyses before examining final results.

Conduct the study exactly as approved

Track deviations, amendments, withdrawals, missing data and protocol changes. Do not quietly change eligibility criteria or outcomes because the initial results are inconvenient. Obtain approval for amendments when required.

Report using the correct guideline

Use CARE, STROBE, CONSORT, PRISMA, STARD, COREQ/SRQR, SQUIRE or another design-specific guideline. Reporting guidelines do not replace good study design; use them while planning so required information is captured prospectively.

Prepare the manuscript and submit responsibly

Agree authorship, select a journal within scope, complete declarations, report approvals and consent accurately, include registration numbers where applicable, and respond to peer reviewers point-by-point.

Close the loop

Complete any institutional study closure requirements, update trial/review registrations if applicable, archive the approved study record, and add the verified publication to Research Pulse.