Research tutorial

1. Frame a useful question

Define the problem, population, setting, and outcome or experience you want to understand. A quantitative question might compare outcomes or estimate an association; a qualitative question might explore how people experience a process.

For an intervention question, PICO can help organize population, intervention, comparison, and outcome. Example: among nursing students, how does one teaching approach compare with another for a specified learning outcome? A question does not establish that the intervention is effective.

2. Find and document evidence

List the main concepts and synonyms before searching. In PubMed, combine related alternatives with OR and distinct concepts with AND. Review Search Details to see how the query was interpreted. Record the database, exact search, search date, and eligibility criteria so someone else can reproduce the search.

Read the full methods and results rather than relying on an abstract. Check corrections or retractions and explain any date or language limits. A useful starting paper is not a complete systematic search.

References: PubMed User Guide

3. Match the design to the question

Randomized trials, cohort studies, case-control studies, cross-sectional surveys, qualitative studies, and mixed-methods designs answer different kinds of questions. Distinguish random sampling from random assignment: the first concerns how people enter the sample; the second concerns how an intervention is allocated.

Identify selection bias, confounding, measurement limitations, attrition, and the population to which findings might apply. An observed association alone does not establish causation. For qualitative work, explain recruitment, context, data collection, analytic decisions, and researcher reflexivity.

4. Plan ethics, measurement, and data handling

Before recruitment or use of identifiable information, consult the institution’s research oversight process to determine applicable review and consent requirements. Do not self-declare an activity exempt simply because it seems low risk.

Document recruitment, consent, confidentiality, access controls, and retention plans. Choose measures with evidence for the intended population, language, and purpose. Check the instrument owner’s permissions, version, and scoring instructions; a questionnaire found online is not automatically free to reuse or validated for a new setting.

References: OHRP foundational training

5. Specify and interpret the analysis

Define outcomes, comparisons, missing-data handling, and planned analyses before inspecting the results. Justify sample size for the design. Check distributions, outliers, independence, and any clustering or repeated measurements before selecting a method.

Report estimates with uncertainty and explain practical meaning. A p-value does not give the probability that the null hypothesis is true or the size of an effect. Distinguish confirmatory analyses from exploratory follow-up work and address multiple testing when relevant.

Choose a statistical method →

References: ASA statement on p-values

6. Report transparently and appraise critically

State what was planned, what was done, and what changed. Report participant flow, missing observations, effect estimates, uncertainty, and limitations. Separate results from interpretation, disclose funding and conflicts, and share data or code only when consent and governance allow.

Use the reporting guideline that fits the design: CONSORT for randomized trials, STROBE for observational studies, and PRISMA for systematic reviews. A reporting checklist improves transparency; it is not proof that a study is unbiased or well designed.

References: CONSORT reporting guidance · STROBE reporting guidance · PRISMA reporting guidance