18.3 Epidemiology Measures & Population Health

18.3 Epidemiology Measures & Population Health

Quick Answer: Incidence = new cases over time; prevalence = existing cases at a time (≈ I × duration). Attack rate = ill / exposed in an outbreak. Herd immunity threshold ≈ 1 − 1/R0. Primary/secondary/tertiary prevention map to before disease / early detection / limit disability. p-values ≠ effect size; power = 1 − β; CIs show precision and null exclusion.

This section ties classical epidemiology measures to prevention, outbreak response, disparities framing, and the inferential statistics tested alongside biostats on CBSE/PBLI-style items.

Incidence vs Prevalence

Relationship (steady state, low disease): Prevalence ≈ incidence × average duration.

Worked example 1 — Incidence and prevalence

City of 100,000 with no disease at t = 0. Over 1 year, 500 new cases of a chronic condition develop; none die or leave; all remain diseased.

• Annual cumulative incidence ≈ 500/100,000 = 0.5%
• End-of-year prevalence ≈ 500/100,000 = 0.5% (first year)

If average duration becomes 10 years with steady 500 new cases/year, prevalence approaches 500 × 10 / 100,000 = 5%. Improved treatment that prolongs life with disease increases prevalence even if incidence is flat—classic exam trap.

Worked example 2 — Person-time rate

200 patients followed: 50 for 1 year, 100 for 2 years, 50 for 4 years without the event until 12 events occur.

Person-years = 50·1 + 100·2 + 50·4 = 50 + 200 + 200 = 450
Incidence rate = 12/450 = 0.0267 per person-year (≈ 2.67 per 100 person-years)

Mortality Rates and Related Metrics

CFR ≠ cause-specific mortality. CFR uses cases in the denominator; mortality uses population.

Worked example 3 — CFR vs mortality

Population 1,000,000; 2,000 myocardial infarctions; 400 MI deaths in a year.

• CFR = 400/2000 = 20%
• Cause-specific MI mortality = 400/1,000,000 = 0.04% per year (40 per 100,000)

Attack Rate

Attack rate is cumulative incidence in a clearly defined exposed group during an outbreak window:

Attack rate = number who became ill / number exposed (at risk)

Often expressed as percent. Food-specific attack rates compare foods at a picnic.

Worked example 4 — Outbreak table

Risk ratio = 0.75/0.20 = 3.75. Potato salad is the leading hypothesis for the vehicle. Secondary attack rate among household contacts uses new cases among contacts / susceptible contacts.

Herd Immunity Threshold (Concept)

Basic reproduction number R0 = average secondary cases from one infectious person in a fully susceptible population.
Effective reproduction number Rt falls as immunity rises.

With homogeneous mixing and solid immunity, herd immunity threshold (HIT):

HIT ≈ 1 − 1/R0

Worked example 5

If R0 = 5, HIT ≈ 1 − 1/5 = 0.80 (80%) immune needed so each case produces <1 secondary case on average.
If R0 = 15 (highly transmissible agent), HIT ≈ 93%.
Imperfect vaccines require higher coverage: roughly HIT / vaccine effectiveness.

Real populations are not perfectly mixed; clustering and waning immunity complicate estimates—but the formula is standard exam material.

Outbreak Investigation Steps

A practical sequence (order may overlap):

1. Prepare / verify diagnosis — confirm the disease is real (lab, clinical criteria); rule out pseudo-outbreaks.
2. Confirm outbreak — cases exceed expected baseline.
3. Case definition — person, place, time, clinical/lab criteria (possible/probable/confirmed).
4. Find and count cases — active surveillance, line list.
5. Descriptive epidemiology — time (epidemic curve), place (spot map), person (age, sex, exposures).
6. Generate hypotheses — source, mode, vehicle, risk factors.
7. Test hypotheses — analytic epidemiology (cohort/case-control attack rates), environmental/lab testing.
8. Implement control and prevention — remove source, isolate, vaccinate, educate—often start control before all analytics finish when risk is high.
9. Communicate findings — public, clinicians, agencies.
10. Maintain surveillance — ensure outbreak ends; evaluate response.

Epidemic curve shapes: point source (single peak), continuous common source (prolonged plateau), propagated (serial peaks at incubation intervals).

Levels of Prevention

Primordial prevention (sometimes tested) targets underlying social/environmental conditions that give rise to risk factors (e.g., urban design reducing default sedentary life).

Exam trap: treating severe disease complications is tertiary, not primary—even if heroic.

Risk Factors vs Determinants of Health

• Risk factor: measurable attribute/exposure associated with increased probability of disease (hypertension → stroke; smoking → lung cancer). May be causal or marker.
• Determinants of health: broader set of personal, social, economic, and environmental factors that influence health status—income, education, housing, racism/discrimination, access to care, built environment, health behaviors, genetics.

Health disparities (basic-science framing on CBSE): systematic, potentially avoidable differences in health outcomes between groups defined by social advantage. Mechanisms taught at the interface of biology and environment include chronic stress physiology (HPA/sympathetic activation), differential toxic exposures, nutrition access, infectious exposure density, and unequal treatment access—not "biological race as destiny." Disparities questions often ask which intervention level or structural factor best explains a gradient after individual risk factors are listed.

Statistical Significance vs Clinical Significance

Worked example 6 — Significant but trivial

Drug lowers systolic BP by 0.5 mm Hg; n = 100,000; p = 0.001. Statistically significant, clinically negligible for individuals. Conversely, a 15 mm Hg drop in a small pilot with p = 0.08 may be clinically promising but not "significant" at α = 0.05—need more power, not dismissal of the effect size.

Type I Error, Type II Error, and Power

Power = 1 − β = probability of detecting a true effect of a specified size.

Power rises with: larger effect size, larger sample size, higher α, lower variance, more precise measurements, one-sided tests (when appropriate).

Worked example 7 — Interpreting β

Study designed with 80% power (β = 0.20) at α = 0.05 to detect ARR of 5%. If the true ARR is 5%, there is still a 20% chance of a nonsignificant result. A negative underpowered study does not prove no effect.

Confidence Intervals

A 95% confidence interval (CI) for a mean difference or RR is a range from a procedure that, in repeated sampling, covers the true parameter 95% of the time (frequentist teaching). Practical exam uses:

1. Width → precision (narrower = more precise, often larger n)
2. Whether null is inside → for difference, null is 0; for ratio, null is 1. If 95% CI for RR is 0.6–0.9, excludes 1 → consistent with statistical significance at α = 0.05 (two-sided) under standard correspondence
3. Clinical bounds → even if CI excludes null, if entire CI is within a trivial effect range, clinical importance remains doubtful

Worked example 8 — CI reading

RR = 1.20; 95% CI 0.95–1.52 → not statistically significant at 0.05; compatible with modest harm, null, or modest benefit—imprecise.
RR = 0.70; 95% CI 0.65–0.76 → precise protective association; still check absolute risks for decision-making.

Integration for CBSE Vignettes

• Chronic disease more prevalent despite flat incidence → longer duration/survival.
• Outbreak picnic → attack rates by food; highest AR and AR ratio drive hypothesis.
• "What percent must be immune?" → 1 − 1/R0.
• Vaccine before exposure → primary prevention; mammography → secondary; post-stroke rehab → tertiary.
• Tiny p with tiny ARR → emphasize clinical vs statistical significance.
• Nonsignificant trial with n = 20 → suspect low power, not proof of equivalence.
• RR 0.5 (0.4–0.6) → significant and relatively precise; still compute NNT when rates given.

Population measures describe burden and spread; prevention levels organize interventions; inferential statistics police how we claim knowledge. Together they complete the 4–6% biostats/epidemiology band on the CBSE.