The type of clinical question determines which study designs can produce a valid answer. Therapy questions ask whether an intervention changes an outcome; the strongest design is the randomized controlled trial (RCT), because random assignment creates groups that are comparable at the start, and the trial follows them forward to observe outcomes. Harm questions ask whether an exposure causes a bad outcome; because it is usually unethical to assign people to a harmful exposure, the strongest designs are observational, particularly cohort studies that follow exposed and unexposed people forward, and case-control studies that start from the outcome and look backward at exposure. Diagnosis questions ask how well a test distinguishes diseased from non-diseased people; they require a cross-sectional design in which the test and a reference standard are applied to the same patients. Prognosis questions ask what will happen to patients with a condition over time; they require a cohort followed forward. Case series, which describe a group of patients without a comparison group, can generate hypotheses but cannot answer comparative questions about therapy or harm.
Learn MotionHow Doctors Judge Whether a Medical Study Can Be Trusted
The Clinical Question and Why Study Design Follows From It
Which Design Can Answer Which Question
2 / 3
The map is not a ranking of prestige; it is a statement about what each design can structurally produce. A therapy question needs a comparison between groups that were similar before treatment, and only random assignment reliably produces that, so the randomized trial is the match. A harm question usually cannot be answered by random assignment because you cannot ethically expose people to a suspected harm, so the answer has to come from observing what already happened, either forward in a cohort or backward in a case-control study. A diagnosis question needs the test and the reference standard applied to the same patients at the same time, which is a cross-sectional structure. A prognosis question needs patients followed forward over time, which is a cohort. Notice where case series sits: it describes patients but has no comparison group, so it can suggest a hypothesis but cannot answer whether one treatment beats another.
0:00 / 0:00