Epidemiology and Biostatistics
Core concepts
Master measures of disease frequency (incidence, cumulative incidence, prevalence and their interrelationships), measures of association (risk ratio, odds ratio, rate ratio, attributable risk), and study designs (cohort, case-control, cross-sectional, RCT, ecologic). Understand validity: random and systematic error, selection/information/confounding bias, effect modification, and the criteria for causal inference. Know screening test properties-sensitivity, specificity, predictive values, likelihood ratios-and how prevalence shifts predictive values while sensitivity and specificity stay fixed. For biostatistics, be fluent in sampling distributions, standard error, confidence intervals, hypothesis testing (type I/II error, power), common tests (t-test, chi-square, ANOVA, non-parametric), correlation versus regression, and survival analysis basics.
Traps
- Confusing incidence (new cases, needs follow-up time) with prevalence (all existing cases).
- Assuming predictive value is a fixed test property-it depends on prevalence.
- Reporting odds ratio as if it were a risk ratio in common outcomes.
- Mixing up confounding (a nuisance to control) with effect modification (a real difference to report).
- P-value misinterpretation: it is not the probability the null is true.
Worked example / rule of thumb
Sensitivity = TP/(TP+FN); specificity = TN/(TN+FP). Positive predictive value = TP/(TP+FP). If a test has 90% sensitivity and 95% specificity and prevalence is 10% in 10,000 people: TP=900, FN=100, FP=450, TN=8,550; PPV=900/(900+450)=67%. Drop prevalence to 1% and PPV falls to about 15% despite unchanged sensitivity/specificity. Rule of thumb: screen rare diseases only with highly specific tests, and always compute PPV using the local prevalence.