Preventive Medicine and Public Health Flashcards

A public health researcher is evaluating a screening program for a rare disease (prevalence 0.5%) using a test with 95% sensitivity and 90% specificity. After widespread implementation, clinicians report that most positive results are turning out to be false positives. Which statistical concept BEST explains this phenomenon?

Answer: Low positive predictive value due to low disease prevalence

When disease prevalence is very low (0.5%), even a test with high sensitivity and specificity will have a low positive predictive value (PPV). Using Bayes' theorem: with 90% specificity, 10% of true negatives test positive. In a population of 100,000 with 500 true cases, the test yields 475 true positives but ~9,950 false positives — giving a PPV of only ~4.6%. This is the base rate fallacy applied to clinical screening.

During an outbreak investigation, a cohort study of foodborne illness at a catered event reveals the following: attack rate among those who ate the shrimp = 72%, attack rate among those who did NOT eat the shrimp = 68%. The relative risk is 1.06. What is the MOST appropriate conclusion?

Answer: Shrimp is unlikely to be the causative vehicle; another food item should be investigated

A relative risk of 1.06 is very close to 1.0, indicating essentially no increased risk from eating shrimp. Despite the high absolute attack rates (suggesting a widespread exposure), the nearly identical rates in both groups means shrimp is not the differentiating factor. Classic outbreak investigation methodology requires identifying a food with a substantially elevated relative risk (typically >2.0) to implicate it as the vehicle. Another food item consumed by nearly everyone — regardless of shrimp consumption — is the more likely culprit.

A country implements a mandatory fortification program adding folic acid to all flour products. Five years later, epidemiologists observe a paradoxical increase in the reported incidence of neural tube defects (NTDs). Which of the following is the MOST likely explanation for this finding?

Answer: Improved case ascertainment and surveillance infrastructure capturing previously missed diagnoses

Paradoxical increases in disease incidence following public health interventions are classically explained by Will Rogers phenomenon or, more commonly here, improved case ascertainment. When a program is implemented, it often comes with enhanced surveillance, better diagnostic criteria, and improved reporting systems. Previously missed NTD cases (stillbirths, early fetal losses, or unreported live births) begin to be captured. This surveillance artifact must be ruled out before concluding that an intervention is ineffective or harmful.

An occupational health physician is evaluating a cluster of mesothelioma cases among workers at a shipyard. The latency period between first asbestos exposure and disease onset averages 35–40 years. The shipyard closed in 1985. Under the Bradford Hill criteria, which criterion is MOST challenged when attributing causation solely based on the biological plausibility of asbestos-mesothelioma exposure in this cohort?

Answer: Temporality, because exposure records from the closed shipyard may not definitively establish exposure preceded disease in all individual cases

While asbestos is firmly established as the leading cause of mesothelioma, the Bradford Hill criterion of temporality — that exposure must precede disease — can be methodologically difficult to prove in individual cases when the source facility closed decades ago. Employment records may be incomplete, exposure duration and intensity may be unknown, and workers may have had multiple subsequent asbestos exposures. Coherence and biological plausibility are actually well-established for asbestos-mesothelioma; strength is strong (RR >20 for heavy exposure); specificity is less applicable as a modern criterion.

A health department conducts a randomized controlled trial of a hypertension management app. The intention-to-treat (ITT) analysis shows no significant reduction in cardiovascular events, but a per-protocol analysis shows significant benefit. Dropout in the intervention arm was 40%, predominantly among the sickest patients. What is the MOST accurate interpretation?

Answer: The ITT analysis is likely correct; the per-protocol result is biased by selective dropout of high-risk participants who stopped using the app

This scenario illustrates informative censoring and healthy adherer bias. When the sickest patients disproportionately drop out of the intervention arm, the per-protocol analysis compares the healthiest intervention users to all control participants — a fundamentally biased comparison. The ITT analysis preserves randomization and reflects real-world effectiveness including non-adherence, making it the more conservative and regulatory-standard estimate. The discrepancy between ITT and per-protocol should raise concern about selection bias in the per-protocol analysis, not confidence in the per-protocol result.

A public health agency compares infant mortality rates between two counties. County A (urban, population 500,000) has a crude infant mortality rate of 6.2 per 1,000 live births. County B (rural, population 80,000) has a crude rate of 4.8 per 1,000. After age-standardization to the state population, County A's rate drops to 4.1 and County B's rises to 6.9. What phenomenon BEST explains this reversal?

Answer: Simpson's paradox arising from confounding by the age distribution of mothers, with County B having proportionally more high-risk teenage and advanced-age pregnancies masked by a favorable crude rate

This is a textbook example of Simpson's paradox in epidemiology. County B's apparently better crude infant mortality rate is confounded by its age distribution of mothers. If County B has a higher proportion of mothers in the low-risk 20–34 age group (common in rural agricultural communities with stable young families), the crude rate will appear favorable. However, within each age stratum of mothers, County B's outcomes are worse. Standardization to the state population — which has a different age-of-mother distribution — removes this confounding and reveals County B's true elevated risk. This is why crude rates must never be used for inter-population comparisons without standardization.