PES6251 Chap.4 Cohort Designs and Risk Measures
Cohort Designs and Risk Measures
Follow exposure toward outcome
Cohort studies group people by exposure and compare later outcomes. Prospective cohorts define exposure and follow-up before outcomes occur; retrospective cohorts reconstruct the same time order from existing records. Incidence requires an initially at-risk denominator, and relative risk divides exposed risk by unexposed risk.
Hazard ratios incorporate event timing and should not be treated as simple risk ratios without checking model assumptions.
Define entry and follow-up consistently
Participants should be outcome-free at the point risk begins and observed over comparable periods.
If one exposure group enters later, is monitored more intensively or already includes prevalent outcomes, the apparent incidence contrast can be distorted. Prospective work can standardise these decisions in advance, while retrospective work must recover them from records created for another purpose.
State the index date, eligibility window and outcome-detection process for each group.
Read the ratio with its direction attached
Compute risk separately for exposed and unexposed participants before dividing. A ratio above one indicates greater observed risk in the numerator group, a ratio below one indicates lower observed risk and one indicates equality.
Reversing exposure labels produces the reciprocal and answers a different question. The interpretation must name the exposure, outcome and period; it must not silently become a causal conclusion when exposure was selected rather than assigned.
Account for uncertainty and disappearance
Confidence intervals show precision and whether the data remain compatible with materially different effects.
Attrition deserves equal attention because missing outcomes may depend on exposure and prognosis. Compare loss and reasons by group, inspect whether remaining participants differ at baseline and test how plausible missing values would change the result. A perfectly computed relative risk can still describe a biased cohort if the people who remain observable are selected by the outcome process.
What this chapter covers
- 01
Prospective cohort design
- 02
Retrospective cohort reconstruction
- 03
Incidence and at-risk denominators
- 04
Relative risk calculation
- 05
Hazard ratios and event time
- 06
Loss to follow-up and confounding
Worked example · free
Calculate and qualify relative risk
- 3Calculate each group risk.
- 2Divide exposed by unexposed risk.
- 3Interpret without claiming causation.
Key terms
- Cohort
- A group defined by exposure or another shared characteristic and observed for later outcomes.
- Incidence
- The occurrence of new outcomes among people initially at risk during a stated period.
- Relative Risk
- Risk in the exposed group divided by risk in the unexposed group.
- Hazard Ratio
- A comparison of instantaneous event rates across follow-up under a specified time-to-event model.
- Attrition Bias
- Distortion arising when loss to follow-up differs by exposure and outcome risk.
Cohort Designs and Risk Measures FAQ
What is the main difference between prospective and retrospective cohorts?
Prospective cohorts record exposure and follow participants forward under procedures planned by the researchers. Retrospective cohorts use existing records to reconstruct exposure and later outcomes. Both start with exposure groups, but prospective collection offers more measurement control while retrospective work is faster and inherits record limitations.
What does a relative risk of two mean?
It means observed risk in the exposed group is twice the risk in the unexposed group over the stated follow-up. The interpretation must name the exposure, outcome and period. It does not by itself show that exposure caused the outcome because confounding, selection and measurement error may remain.
Why should loss to follow-up be reported by exposure group?
Different attrition can break comparability. If high-risk exposed participants leave more often, the remaining exposed group may appear healthier and the risk ratio can be biased downward. Report numbers and reasons for leaving, compare baseline characteristics and consider sensitivity analyses rather than treating missing outcomes as neutral.
Assessment move
Practise with labelled two-by-two tables and compute exposed and unexposed risks on separate lines. After every ratio, add an uncertainty sentence naming confounding or attrition. Compare prospective and retrospective examples by measurement control, follow-up length, record completeness and feasibility rather than treating one label as automatically stronger.
For each example, state the index date, confirm that participants were initially at risk, and draw the route from exposure classification to outcome detection. Recalculate the ratio after reversing the exposure labels so you can see why direction must accompany the number. Then compare losses between groups and write a short sensitivity argument about how missing outcomes could exaggerate or dilute the observed association.
Working through Cohort Designs and Risk Measures in PES6251? Sia is AskSia’s AI Statistics tutor — ask any PES6251 Cohort Designs and Risk Measures question and get a clear, step-by-step explanation grounded in how PES6251 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
