Scientific Inquiry and Research Design
The Scientific Inquiry and Research Design domain asks you to translate a clinical or population problem into an answerable question and then choose a design that can actually answer it. HS 326 tests the whole chain: paradigm, question type, design, sampling, control of bias, and the limits each design places on inference. Quantitative and qualitative traditions are both fair game.
Match the question to the paradigm
- Questions of effect, frequency, or prognosis are quantitative and need comparative or longitudinal designs.
- Questions of meaning, process, or experience are qualitative; credibility, dependability, confirmability, and transferability replace internal and external validity as quality markers.
- Mixed methods designs are named by sequence: convergent parallel, explanatory sequential, exploratory sequential.
Design families and what each can claim
- RCT: randomization, allocation concealment, blinding where feasible, intention-to-treat analysis; supports causal claims about efficacy under trial conditions.
- Quasi-experimental: interrupted time series, controlled before-after, stepped-wedge without randomization; vulnerable to secular trends and maturation.
- Cohort: incidence and relative risk; case-control: rare outcomes, odds ratios only; cross-sectional: prevalence, never incidence.
- Qualitative: phenomenology for lived experience, grounded theory for theory generation, ethnography for culture, qualitative description for a plain account.
Threats you must name precisely
Selection and allocation bias, attrition, detection bias, performance bias, confounding, regression to the mean, and history. External validity suffers from narrow eligibility criteria, protocol-driven adherence, and single-site sampling.
Worked example
A team asks whether a nurse-led teach-back discharge program reduces 30-day readmission in heart failure. The outcome is common, the intervention is feasible to randomize, and temporality matters, so a parallel-group RCT with concealed allocation, blinded outcome adjudication, a prespecified 30-day readmission outcome, and intention-to-treat analysis is the defensible choice. A case-control design fails here: readmission is not rare, and exposure timing relative to outcome cannot be established cleanly. If randomization is impossible, a stepped-wedge cluster design or an interrupted time series with a stable baseline period is the fallback, and the write-up must acknowledge secular trend as a rival explanation.
Traps that cost marks
- Calling a two-group study experimental when allocation was not random; it is quasi-experimental.
- Confusing crossover with parallel-group designs.
- Claiming randomization removes the need for a sample size calculation or allocation concealment.
- Reporting odds ratios from a cohort as if they were risk ratios.
- Treating qualitative rigor criteria as an absence of validity.