Choosing Between Observational Designs
Core summary
Choosing the right observational design depends on your research question, the rarity of the disease, available resources, and time constraints. This lesson provides a practical decision framework for selecting between cross-sectional, case-control, cohort, and registry-based designs.
Detailed explanation
Detailed explanation
The decision process starts with your research question. If you need prevalence data or a snapshot of associations, use a cross-sectional study. If the disease is rare or has a long latency, and you need to explore multiple exposures efficiently, choose a case-control study. If you need to establish temporal sequence, measure incidence, and quantify risk, a cohort study is appropriate — prospective for maximum control, retrospective for speed and cost savings. If you have access to a large database or registry and your variables of interest are captured in it, a database study maximizes sample size and efficiency. Important considerations include feasibility (time and budget), ethical constraints (can you ethically follow exposed individuals?), available data sources, and the level of evidence your audience expects. No design is universally best — each answers a different type of question with different trade-offs. Your proposal should explicitly state why you chose your design over alternatives.
Clinical example
A resident wants to study whether a new physiotherapy protocol reduces post-stroke disability. The outcome is common (stroke disability), the exposure is specific (new protocol), and temporal sequence matters. A retrospective cohort comparing patients before and after protocol introduction would be efficient. A cross-sectional study cannot establish sequence; a case-control study would be awkward because the outcome is not rare.
Research example
When investigating the link between a rare occupational exposure (e.g., asbestos in a specific factory) and mesothelioma (a rare cancer), a case-control study is the most efficient design. Both the exposure and outcome are rare, making a cohort study impractical (you would need to follow millions of workers). The case-control design lets you start with the few mesothelioma patients and look backward.
Knowledge check
Q1. If you want to study a rare disease with a long latency period, which design is MOST efficient?
Q2. When is a cross-sectional study preferable to a cohort study?
Q3. What should you include in your proposal when justifying your study design choice?