Cohorts, Case-Control Studies & Prognosis
Read longitudinal and outcome-sampled research with attention to time, denominators and follow-up.
Quick revision ↓ Medical Dictionary
01Cohort foundations
A cohort requires a defined source population, a meaningful time origin and consistent outcome ascertainment. For incidence, participants must be at risk of the outcome at the relevant starting point. Describe how exposure changes during follow-up and how losses are handled. A prospective label does not ensure comparable groups; clinicians and patients can select exposures in ways associated with future outcomes.
02Choosing case-control participants
Cases should arise from a clearly defined source population. Controls should represent the exposure distribution in that source population under the study's sampling scheme, rather than simply being unusually healthy volunteers. Ascertain exposure comparably in cases and controls. An odds ratio is a standard measure in conventional case-control analysis; what it estimates in the underlying population depends on how controls were sampled.
03Person-time and censoring
People can contribute different lengths of observation. Person-time accumulates their time at risk. In survival analysis, right censoring means the exact event time after the last observed point is unknown, not that an event never occurs. Standard survival estimates rely on assumptions about censoring; if people leave because their unobserved prognosis is worse, an apparently reassuring survival curve may be misleading.
04Survival curves and hazard ratios
A Kaplan-Meier curve estimates survival over time while accounting for censoring under its assumptions. Check numbers at risk because late estimates can rest on very few people. A hazard is an instantaneous event rate among those still at risk. A hazard ratio is not automatically a risk ratio at a fixed time, and a single proportional-hazards summary may be unsuitable if relative hazards change substantially.
05Prognosis and prediction
A prognosis study estimates future outcomes for people at a defined disease stage. A prediction model combines characteristics to estimate individual risk. Discrimination concerns separating people with and without the outcome; calibration concerns agreement between predicted and observed risks. A model can rank people well yet systematically overestimate their risks. External validation and evidence of clinical usefulness are needed before routine use in a new setting.
High-Yield Revision
- Define the source population and start of follow-up.
- Censored does not mean event-free forever.
- Good discrimination does not guarantee good calibration.
Common Pitfalls
- Do not select controls simply because they are exceptionally healthy.
- Do not interpret a hazard ratio as a fixed-time risk ratio without justification.
References & Further Reading
Use the full methods and relevant chapters when appraising a paper. These sources support further study; the examples and questions above are original teaching exercises.
- Oxford CEBM — Critical appraisal tools
- Cochrane Handbook — Methods for evidence synthesis
- CDC — Analysing and interpreting epidemiological data
- ASA — Statement on p-values and statistical significance
- Oxford CEBM — Likelihood ratios
- Cochrane — GRADE and summary of findings
- WMA — Declaration of Helsinki, 2024 revision
- EQUATOR — Reporting guideline library
- NIH OBSSR — Mixed-methods research resources
- Institute for Healthcare Improvement — Model for Improvement