CM7.{5,8} | CM7.{5,8} | Study Designs, Association and Bias — Summary & Reflection

KEY TAKEAWAYS

Epidemiological study designs range from descriptive (case reports, case series, cross-sectional, ecological — hypothesis-generating, no causal inference) to analytical observational (cohort — prospective/retrospective — and case-control) to experimental (RCT). Cohort studies measure incidence in exposed and unexposed groups; the summary measure is the Relative Risk (RR) = [a/(a+b)] / [c/(c+d)]. Case-control studies compare exposure odds in cases vs controls; the summary measure is the Odds Ratio (OR) = ad/bc. OR approximates RR when disease incidence is <5–10% (the rare disease assumption). Association is established by statistical significance; causation additionally requires Bradford Hill criteria — particularly temporality (the only necessary criterion), strength, consistency, and biological gradient. Threats to valid inference include: selection bias (Berkson's, healthy worker effect, loss-to-follow-up); information bias (recall bias in case-control studies, observer bias); and confounding (a third variable associated with both exposure and outcome, not an intermediate). Confounding is controlled by randomisation (design), restriction, matching, stratification, or multivariate analysis.

REFLECT

You are asked to design a study to investigate whether prolonged mobile phone use (>4 hours/day) is associated with hearing loss in young adults in India. (a) Which study design would you choose and why? (b) What is the main bias you would need to guard against in your chosen design? (c) Name one potential confounder that could distort your results and describe how you would control for it at the design stage. Write three to four sentences addressing these three points, drawing on the distinctions between study designs and sources of systematic error covered in this module.