AtlasSession 3L8: Study designp.31–48, 102–103
Case-control studies
- Study designs primer
- Classification
- Cohort studies
- Case-control studies
Fig. 1
Hernia recurrence after mesh repair: 1,000 cases vs 1,000 controls, lightweight mesh in the operation notes (patients)
Fig. 1: detective work backwards. We start today with 1,000 with recurrence and 1,000 with no recurrence, then an arrow sweeps back through the operation notes to see who had lightweight mesh. We chose 1 without for each with, so 50 in every 100 of our sample had a recurrence: that tells us nothing about how often it really happens.
- Our sample with a recurrence
- 50 in every 100
- No recurrence for each recurrence
- 1
- How often it really happens?
- can’t tell: we chose the split
Study Timeline · First look
Level 1 · First look · Lecture 8 · p.31–48, 102–103
Understand starting from the outcome and looking back for causes.
Start from cases and controls, compute the odds of exposure in each, and explain why risk can't be calculated (unless nested) and why recall bias and reverse causality threaten the design.
Detective work (the step you are on)
Start with burgled houses and untouched ones, then check which had windows left open.
A detective starts with burgled houses and untouched ones, then checks which had windows left open. Here we start with hernias that came back and hernias that didn’t, then look back for the mesh used.