Epidemiology and Biostatistics
Core concepts
Master measures of disease frequency (incidence vs prevalence; attack rate; mortality vs case fatality) and measures of association (risk ratio, odds ratio, attributable fraction). Know when each study design applies: cross-sectional (prevalence, no temporality), cohort (incidence, RR, temporality), case-control (OR, rare outcomes), and experimental designs (randomization, blinding).
Understand sensitivity, specificity, PPV, NPV, and how prevalence affects predictive values. Be able to compute and interpret confidence intervals and recognize confounding, bias (selection, information, misclassification), and effect modification. Know screening vs diagnostic testing, herd-level vs individual-level measures, and sample size determinants (expected effect, power, alpha, variance). Understand outbreak investigation steps and surveillance types (passive, active, sentinel, syndromic).
Traps
- Reporting prevalence as if it were incidence in a cohort.
- Confusing PPV (probability of disease given positive test) with sensitivity (probability of positive test given disease).
- Using odds ratio as if it estimates risk directly when outcome is common.
- Ignoring herd-level clustering when analyzing individual animal data.
Worked rule of thumb
PPV depends on prevalence. With sensitivity 90% and specificity 95%: at 10% prevalence, PPV ≈ 0.90×0.10 / (0.90×0.10 + 0.05×0.90) = 0.09/0.135 ≈ 0.67. At 1% prevalence, PPV ≈ 0.009/(0.009+0.0495) ≈ 0.15. Same test, dramatically different predictive value-so a positive result in a low-prevalence herd is weak evidence.
For study design, ask: Does the design measure incidence (cohort) or start from outcome (case-control)? That determines whether RR or OR is appropriate.