PES6251 Chap.5 Case-control Designs and Odds Ratios
Case-control Designs and Odds Ratios
Look backward from outcome
Case-control studies select cases and controls by outcome status and compare prior exposure. They are efficient for rare outcomes, but validity depends on controls representing exposure in the population that produced the cases. Exposure odds among cases are divided by exposure odds among controls to obtain an odds ratio.
Because investigators choose case and control counts, sampled proportions do not estimate population risk directly.
Define cases before searching for causes
Case criteria should identify the same outcome consistently across participants and avoid depending on exposure knowledge.
A broad or shifting definition mixes different conditions, while detection through one healthcare or sporting system may select cases with particular access. State the time, place and diagnostic route that generated cases.
Controls must come from people who could have met that definition and would have been detected through the same route if the outcome had occurred.
Make exposure retrieval symmetrical
Use the same look-back period, records and interview prompts for cases and controls. Cases may remember suspected causes more intensely, and interviewers who know the hypothesis may probe them differently.
Records reduce recall only when coverage and accuracy are comparable. A longer exposure window for one group creates more opportunities to find an exposure and biases the comparison even if every recorded item is correct.
Keep odds separate from risk
The sampled numbers of cases and controls are set by design, so their outcome proportions cannot recover population risk.
Calculate exposure odds within each group, form the odds ratio and interpret it using odds language. Under a justified rare-outcome condition the value may approximate a risk ratio, but that is an additional argument. Matching can improve control of selected confounders, yet analysis must respect the matched structure and cannot repair an unsuitable control source.
What this chapter covers
- 01
Selection by outcome status
- 02
Source-population control logic
- 03
Comparable exposure windows
- 04
Recall and interviewer bias
- 05
Odds ratio by exposure odds
- 06
Matching and matched analysis
Worked example · free
Compute exposure odds correctly
- 2Calculate exposure odds among cases.
- 2Calculate exposure odds among controls.
- 4Form and interpret the odds ratio.
Key terms
- Case
- A participant meeting the study's defined outcome criteria.
- Control
- A participant representing exposure in the source population that produced the cases.
- Exposure Odds
- The number exposed divided by the number unexposed within a comparison group.
- Odds Ratio
- Exposure odds among cases divided by exposure odds among controls.
- Recall Bias
- Systematic exposure-reporting differences arising because outcome experience changes memory or search for causes.
Case-control Designs and Odds Ratios FAQ
Why are controls the key design decision?
Controls should estimate exposure in the population that could have produced the cases. If they come from another setting with different training, healthcare access or record quality, exposure differences may reflect recruitment rather than outcome status. Convenience alone is not a defensible control-selection rule.
Can an odds ratio be described as risk?
Not automatically. Case-control researchers select case and control counts, so sampled outcome proportions do not estimate population risk. Interpret the odds ratio as exposure odds among cases relative to controls. Under appropriate rare-outcome conditions it may approximate a risk ratio, but that needs justification rather than assumption.
How does recall bias change an odds ratio?
Cases may search for possible causes more intensely and report past exposure more completely than controls. That differential recall can exaggerate an association. Records created before outcome, identical prompts and interviewer blinding can reduce the problem, provided exposure completeness is comparable across both groups.
Assessment move
Start every case-control critique with the control source, then inspect case definition and exposure measurement. Build a labelled two-by-two table before using the cross-product. Finish by stating whether the most plausible bias would exaggerate, dilute or unpredictably change the odds ratio. Ask whether every control could have become and been detected as a case through the same system.
Compare the exposure look-back window, record coverage and interview prompts across groups. After calculating exposure odds separately, interpret the odds ratio without converting sampled case proportions into population risk. If matching was used, identify the matched factor, explain why it was chosen and check that the proposed analysis preserves the matched structure.
Working through Case-control Designs and Odds Ratios in PES6251? Sia is AskSia’s AI Statistics tutor — ask any PES6251 Case-control Designs and Odds Ratios 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.
