Preventive Medicine and Public Health Flashcards

A county health department implements a universal lead screening program for all children aged 1–2 years. After one year, the prevalence of elevated blood lead levels (≥5 µg/dL) appears to have increased compared to pre-program data. Which of the following best explains this phenomenon?

Answer: The screening program identified previously undetected cases, inflating apparent prevalence

This is a classic example of 'surveillance artifact' or 'ascertainment bias.' Before universal screening, only symptomatic or high-risk children were tested, so many true cases went undetected. The denominator (all children screened) is now larger, but so is the numerator (newly identified cases). The apparent rise in prevalence reflects improved case-finding, not a true increase in disease burden. This phenomenon is commonly seen when new or expanded screening programs are introduced.

A researcher conducts a prospective cohort study and calculates an attributable risk percent (AR%) of 60% for smoking and lung cancer. A policy maker asks what this statistic means for public health planning. Which interpretation is most accurate?

Answer: 60% of lung cancer cases in smokers are attributable to smoking itself

Attributable risk percent (AR%) — also called etiologic fraction among the exposed — represents the proportion of disease in the exposed group that is attributable to the exposure. It equals (RR−1)/RR × 100. An AR% of 60% means that 60% of lung cancer cases occurring among smokers would be prevented if smoking were eliminated. Reducing population-level incidence by 60% would require the population attributable risk percent (PAR%), which also accounts for prevalence of exposure in the population.

During an outbreak investigation of Legionnaires' disease at a hotel, the attack rate among guests who used the rooftop pool was 18%, while among guests who did not use the pool it was 4%. The relative risk is 4.5 and the chi-square p-value is 0.003. The investigators then stratify by floor of residence and find that both the pool-users and non-pool-users on floors 8–10 (nearest the cooling tower) have similarly elevated attack rates. What does this stratified analysis most strongly suggest?

Answer: The cooling tower, not the pool, is the likely primary source of the outbreak

When stratification by floor shows that proximity to the cooling tower — not pool use — is the common factor driving elevated attack rates in both exposed and unexposed groups, it suggests the pool association was confounded by proximity to the cooling tower. Legionella typically colonizes cooling towers and aerosolizes through the ventilation system, reaching nearby floors regardless of pool use. The stratified analysis points to the cooling tower as the primary reservoir, a critical reorientation of the investigation.

A new rapid diagnostic test for tuberculosis is evaluated in a high-prevalence correctional facility population (prevalence = 30%) and found to have sensitivity 88% and specificity 92%. When the same test is deployed in a low-prevalence refugee screening program (prevalence = 1%), which outcome is most predictable without recalculating?

Answer: The positive predictive value will be substantially lower in the refugee program

Sensitivity and specificity are intrinsic test characteristics that remain constant across populations (assuming the same test conditions). However, predictive values are heavily dependent on disease prevalence. As prevalence drops from 30% to 1%, the PPV falls dramatically — a positive result becomes far more likely to be a false positive. In contrast, NPV rises (true negatives dominate). This is a critical concept for population-based screening: a test that performs well in high-prevalence settings may generate an unacceptable burden of false positives in low-prevalence settings.

A public health team applies the Bradford Hill criteria to assess causality between particulate matter (PM2.5) exposure and cardiovascular mortality. The association is strong (RR 1.8), consistent across 40 studies, biologically plausible, and temporally correct. However, the dose-response relationship shows a threshold effect below 12 µg/m³ rather than a monotonic gradient. How does this threshold finding affect the causality assessment?

Answer: It weakens but does not refute causality, as dose-response is one of nine criteria and not required for causation

Bradford Hill explicitly stated that none of his nine viewpoints (including biological gradient/dose-response) is individually necessary or sufficient to establish causality. A threshold effect — where risk increases only above a certain exposure level — is scientifically plausible for many toxicologic mechanisms (e.g., exceeding a compensatory physiologic reserve). The absence of a simple monotonic gradient weakens the biological gradient criterion but must be weighed against the other strongly positive criteria. Bradford Hill himself cautioned against mechanical application of the criteria as a checklist.

A community-level randomized controlled trial (cluster RCT) evaluates a workplace wellness intervention targeting hypertension control. The unit of randomization is the employer (30 employers randomized), but analysis is performed at the individual level (n=4,200 employees) without accounting for clustering. What is the primary statistical consequence of this analytical error?

Answer: Standard errors will be underestimated, inflating the type I error rate

In cluster RCTs, individuals within the same cluster (here, the same employer) are more similar to each other than to individuals in other clusters — captured by the intraclass correlation coefficient (ICC). Ignoring clustering and analyzing as if all individuals were independent inflates the effective sample size. This underestimates standard errors, narrows confidence intervals, and produces falsely small p-values, dramatically increasing the type I error (false positive) rate. The design effect (DEFF = 1 + (m−1)×ICC, where m = cluster size) must be applied to correct this. This error has led to retractions of cluster trial publications.