Research Methods and Study Design
What the exam tests
Research methods and study design tests framing questions, choosing designs, controlling bias and confounding, defining endpoints, sampling, randomization, blinding, and reporting guidelines. Expect questions on cohorts, case-control, cross-sectional, RCTs, and evidence synthesis, plus protocol and ethics requirements.
Core rules
- The design must fit the question: incidence and natural history need cohorts; rare outcomes may need case-control; prevalence needs cross-sectional; causal claims need strong design and assumptions.
- Pre-specify eligibility, exposure, outcome, confounders, and analysis; avoid outcome-driven definitions.
- Randomization addresses measured and unmeasured confounding at allocation; blinding reduces ascertainment and performance bias.
- Report using CONSORT for trials, STROBE for observational studies, PRISMA for systematic reviews.
- IRB approval and informed consent apply unless a valid waiver is granted.
Worked example
A team asks whether an early follow-up call reduces 30-day readmission. A cohort comparing called versus not-called patients is vulnerable to indication bias: sicker patients may be less likely to receive calls. A randomized trial with concealed allocation, intention-to-treat analysis, and a pre-specified readmission definition is stronger. If randomization is impossible, use a target trial framework, adjust for confounders, and avoid causal language beyond the design. The reasoning is that the comparator and assignment mechanism determine what bias control is credible.
Traps
- Calling an association causal because the p-value is small.
- Using immortal time or post-baseline criteria that misclassify exposure.
- Switching primary outcomes after seeing data.
- Ignoring consent or waiver requirements for identifiable data.