Part III: Select a Research Design

You have a research question. You’ve thought carefully about what you want to know and why it matters. Now comes the hard part: figuring out how to answer it.

Research design isn’t about following a recipe or ticking boxes on a methods checklist. It’s about making tradeoffs—between rigor and feasibility, between depth and breadth, between the question you want to answer and the question you actually can answer given your constraints.

The perfect design often doesn’t exist. What exists is the best design you can implement given your resources, ethical constraints, timeline, and context. The goal is to maximize the validity of your claims while acknowledging their limits.

Every design has limitations. Randomized trials are powerful for establishing causality, but they’re often impossible to conduct for policy questions, unethical for harmful exposures, and impractical in crisis settings. Observational studies can tackle questions trials can’t, but they struggle with confounding. Qualitative methods reveal depth and meaning that numbers miss, but they face questions about generalizability. No single approach answers every question, and choosing poorly means either failing to answer your question at all or answering it incorrectly.

WHAT THIS PART COVERS

The chapters ahead walk you through the major families of research design used in global health, each suited to different types of questions and constraints:

Randomized Controlled Trials represent our most powerful tool for establishing causal effects. When oral rehydration therapy was tested against intravenous fluids in Bangladesh in the 1970s, it was randomized trials that proved this simple, cheap intervention could save lives as effectively as hospital-based treatment. That evidence has since saved an estimated 50 million children. But randomization isn’t always possible, ethical, or even necessary—and understanding when it is and isn’t appropriate matters as much as knowing how to do it.

Quasi-Experimental Designs emerge from the recognition that randomization often isn’t on the table. Policy evaluations can’t always be randomly assigned. Disease outbreaks don’t respect experimental protocols. The credibility revolution in social science has given us rigorous methods—difference-in-differences, regression discontinuity, interrupted time series—that can approximate experimental conditions when true experiments are impossible.

Observational Studies have been the backbone of epidemiology since John Snow traced cholera to the Broad Street pump. Cross-sectional surveys, case-control studies, and cohort designs each offer different strengths: speed versus depth, efficiency for rare outcomes versus clear temporal sequences. When Ebola swept through West Africa, researchers didn’t have the luxury of randomized designs—they needed answers now.

Qualitative and Mixed Methods address the questions numbers can’t answer. You might show statistically that an intervention increases clinic attendance, but you won’t understand why women walk past that clinic to reach a traditional birth attendant three villages away without qualitative depth.

Other Designs round out your toolkit: N-of-1 studies, mathematical modeling, synthetic controls, economic evaluations, and diagnostic accuracy studies.

As you work through these chapters, you’ll notice a recurring theme: the most important research questions in global health rarely have a single “correct” design. A cholera outbreak, an earthquake, a new vaccine — each generates questions that demand different methods. The skill isn’t memorizing which design goes where. It’s learning to match the design to the question, the context, and the constraints you actually face, and being honest about what your chosen design can and cannot tell you.