Epidemiology and Biostatistics
What Epidemiology and Biostatistics tests
This domain tests the ability to read a study and compute the numbers that drive clinical decisions: incidence and prevalence, measures of association, test performance, bias, confounding, confidence intervals and p-values. The exam tends to use small 2 × 2 tables and short abstracts.
Core rules
- Incidence = new cases divided by population at risk over time; prevalence = all cases divided by population at a point. Prevalence is approximately incidence × duration.
- Cohort studies give relative risk = [a/(a+b)] / [c/(c+d)]; case-control studies give odds ratio = (a/c) / (b/d) = ad/bc.
- Sensitivity = a/(a+c); specificity = d/(b+d); positive predictive value = a/(a+b); negative predictive value = d/(c+d). Predictive values change with prevalence; sensitivity and specificity do not.
- A 95% confidence interval excludes the null when it does not contain 1 for ratios or 0 for differences. A p-value is the probability of the observed data or more extreme results given the null hypothesis, not the probability that the null is true.
- Bias types: selection, information or measurement, and recall. Confounding is a third factor associated with both exposure and outcome; it is handled by randomisation, restriction or matching at design, and by stratification or multivariable modelling at analysis.
Worked example
In 1,000 people with prevalence 10%, sensitivity 90% and specificity 95%: 100 are diseased, of whom 90 test positive and 10 negative. Of 900 non-diseased, 45 test positive and 855 negative. PPV = 90/135 = 67%; NPV = 855/865 = 99%. Reason: a positive test in a low-prevalence population still carries a one-in-three chance of being wrong.
Traps
Computing PPV from sensitivity and specificity without prevalence; treating a small p-value as proof of a large or important effect; using an odds ratio as if it were a risk ratio when the outcome is common; ignoring loss to follow-up.