Epidemiology and Biostatistics
Epidemiology and Biostatistics tests study design, measures of association, screening, bias, and interpretation of p-values and confidence intervals. Calculations are common, so practice setting up tables.
Core rules
- Risk ratio is for cohort studies; odds ratio is for case-control studies.
- A 95% confidence interval excluding 1.0 for a ratio suggests statistical significance at the 0.05 level.
- Sensitivity detects disease among the diseased; specificity detects non-disease among the non-diseased.
- Confounding is a third variable associated with exposure and outcome, not on the causal pathway.
Worked example
In a cohort, 40 of 100 exposed and 20 of 100 unexposed develop disease. Risk in exposed = 0.40; risk in unexposed = 0.20; risk ratio = 2.0. The interpretation is that exposed people had twice the risk. If this were a case-control study with 40 exposed cases, 60 exposed noncases, 20 unexposed cases, and 80 unexposed noncases, the odds ratio would be (40×80)/(60×20) = 2.67, not the risk ratio.
Traps
- Reporting an odds ratio as if it were a risk ratio in a case-control study.
- Saying a p-value is the probability the null hypothesis is true.
- Ignoring confounding when a crude association looks strong.
- Confusing sensitivity with positive predictive value.