Public Health Sciences: Core Concepts and Question Traps
The sciences behind public health
Public health sciences integrate epidemiology, biostatistics, environmental health, and behavioral/social sciences. A CPH candidate must move fluidly between these lenses rather than treating them as isolated subjects.
Epidemiology. Know measures of frequency (incidence vs. prevalence), measures of association (risk ratio, rate ratio, odds ratio), and the distinction between association and causation. Be able to compute and interpret attack rates, case fatality, and attributable risk. Understand study designs-cohort, case-control, cross-sectional, ecologic, randomized trial-and what each can and cannot establish. Recall that ecologic studies compare groups and are vulnerable to the ecologic fallacy, while cross-sectional data cannot establish temporality.
Biostatistics. Understand sampling distributions, confidence intervals, hypothesis testing logic, Type I and Type II error, and the difference between statistical and clinical significance. Know when to use chi-square, t-test, or regression, and how confounding and effect modification differ.
Environmental and social sciences. Recognize exposure pathways, dose-response, and the roles of socioeconomic status, built environment, and health behavior in shaping population risk.
Traps to watch
Questions often swap incidence for prevalence, conflate correlation with causation, or describe a case-control study and ask for a risk ratio (odds ratio is the appropriate measure). Another common trap: presenting a statistically significant result with a trivial effect size and asking whether it matters clinically.
Rule of thumb
When a question gives you a 2x2 table, first ask what is the denominator? Incidence uses the population at risk; prevalence uses the total population. That single question resolves most computational items.
Worked example
In a cohort, 20 of 100 exposed and 10 of 100 unexposed develop disease. Risk ratio = (20/100)/(10/100) = 2.0. If the same data came from a case-control design, you would compute an odds ratio, not a risk ratio.