- 1What descriptive designs can and cannot tell usIn June 1981 a report described five young men in Los Angeles with a rare pneumonia, and case reports, case series and surveillance went on to establish how a new disease spread long before any trial was possible. What descriptive designs can and cannot establish, with microcephaly after Zika in Pernambuco, Ebola contact tracing, polio surveillance in India and IDSR in sub-Saharan Africa.7 min
- 2Cross-sectional studies and the problem of timingCross-sectional studies measure exposure and outcome at the same point in time. A worked dysentery outbreak in rural Tanzania, from a surveillance alert to a decommissioned well, then exclusive breastfeeding in Ethiopia’s pastoral regions, why a cross-section cannot say which came first, and how duration links prevalence and incidence.5 min
- 3Case-control studies: starting from the outcomeCase-control studies start with the outcome and look back at exposures. A matched case-control study of cholera during active conflict in Yemen, selecting controls from the population that produced the cases, matching and overmatching, why a matched design needs a matched analysis, recall bias, and nested case-control designs.6 min
- 4Cohort studies: following people forward in timeCohort studies start with exposure and follow people forward to see who develops the outcome. Prospective and retrospective cohorts, with ART outcomes in KwaZulu-Natal and opportunistic infections in Ethiopia drawn from medical records, what cohorts can establish that case-control studies cannot, loss to follow-up, and exposures that change over time.7 min
- 5Controlling confounding in observational dataWhen we cannot randomize, we can control for confounders in the design and in the analysis. Outdoor air pollution and children’s respiratory health in Delhi, then stratification and regression, propensity scores and target trial emulation, inverse probability of treatment weighting, TMLE, and the limit every one of them shares: unmeasured confounding.8 min
- 6Choosing and reading an observational designChoosing an observational design starts with the question, then the outcome, the timeline, the budget and the exposure. Why three studies from the chapter fit their designs, constraints beyond the textbook factors, and what to check first when reading a case-control, cohort or cross-sectional study, including what the authors did not measure.7 min