Research Question

A research question is a clear, focused, and answerable question that defines the purpose and direction of a study, determining its design, methodology, data collection, and analysis. In evidence synthesis, research questions are often structured using frameworks such as PICO (Population, Intervention, Comparator, Outcome) for intervention reviews or PEO (Population, Exposure, Outcome) for questions about exposures, which are usually answered by observational studies.

The research question is one of the most important elements of a study. It determines what will be measured, how it will be measured, and what conclusions can be drawn. A vague or poorly formulated question leads to unfocused methods, ambiguous results, and limited impact. A precise, well structured question guides every subsequent decision, from study design and sample size to statistical analysis and interpretation. Grant reviewers evaluate the clarity and significance of the research question as a primary indicator of a proposal's merit, and journal editors look for a clearly stated question that the study's methods are equipped to answer.

Developing a research question begins with identifying a topic of interest and reviewing the existing literature to understand what is already known and what gaps remain. The question is then refined from a broad topic into a specific, answerable question.^[1] The PICO framework is the most widely used structuring tool for clinical questions: it specifies the Population of interest, the Intervention or exposure being studied, the Comparator (what the intervention is compared against), and the primary Outcome.

A well formulated research question has several characteristics. It is specific enough to be answerable within a single study. It is feasible given available resources, timeline, and ethical constraints. It is relevant to the field, addressing a genuine gap in knowledge or practice. It is novel, contributing new information, which can include confirming, refuting or extending earlier findings, rather than repeating what is already settled. And it is ethical, answerable without exposing participants to unacceptable risk. Hulley and colleagues' FINER criteria cover similar ground: a good question is feasible, interesting, novel, ethical and relevant.^[2]

Research questions can be classified by their intent. Descriptive questions ask what exists (what is the prevalence of a condition?). Relational questions ask whether variables are associated (is smoking associated with lung cancer?). Causal questions ask whether an intervention produces an effect (does the drug reduce mortality?). Each type calls for a different research design.

In systematic reviews, the research question must be defined with sufficient precision to guide the literature search, study selection, data extraction, and synthesis.^[3] Overly broad questions produce unmanageable reviews; overly narrow questions may find too few eligible studies.^[3]

Formulate the research question before selecting the study design, writing the protocol, or collecting any data. Revisit and refine the question iteratively as the literature review reveals what is already known. For systematic reviews, state the finalized question in the protocol, and register the protocol (in PROSPERO, for reviews with a health-related outcome) or publish it. In confirmatory research the question should not change after data collection begins unless the change is documented transparently; modifying it post hoc introduces the risk of data driven revisions that compromise the study's integrity.

A team interested in reducing hospital readmissions for heart failure refines their broad interest into a PICO structured question: "In adults hospitalized for acute heart failure (P), does a pharmacist led discharge counseling program (I), compared to standard discharge instructions (C), reduce 30 day all cause readmission (O)?" This question directly informs the study design (a randomized trial), the primary outcome (30 day readmission), the sample (hospitalized heart failure patients), and the comparator (standard care).