For education and revision only. Not individual medical advice; check current NICE/BNF and local guidance before clinical use. Clinical safety and terms.
Public Health: Epidemiology, causation and outbreak investigation
🌍 Epidemiology, causation and outbreak investigation. Epidemiology is the study of the distribution and determinants of health and disease in populations, and its application to control health problems. It provides the numbers behind every public health decision - how common a disease is, who gets it, why, and whether an intervention works. This chapter covers measures of disease frequency and association, epidemiological study designs in practice, infectious disease epidemiology (R₀, herd immunity) and the steps of an outbreak investigation. It complements the Statistics & EBM chapters. Your four questions: How common is it (incidence, prevalence)? Who, where and when (person, place, time)? What is the association, and is it causal? What control measures follow?
Incidence, Prevalence and Rates
Incidence (risk / cumulative incidence) = new cases during a period ÷ population at risk at the start (e.g. 5 per 1,000 per year).
Incidence rate (incidence density) = new cases ÷ total person-time at risk (e.g. per 1,000 person-years) - handles people entering and leaving the population.
Point prevalence = existing cases at a point in time ÷ total population; period prevalence = cases at any time during a period ÷ population.
Prevalence ≈ incidence × average duration (steady state). Prevalence rises when incidence rises or when survival improves (e.g. HIV with antiretroviral therapy; diabetes).
Use: incidence for causes and risk; prevalence for burden and service planning.
Attack rate - cumulative incidence during an outbreak (cases ÷ people exposed); secondary attack rate - cases among contacts of primary cases ÷ susceptible contacts.
Case fatality rate (ratio) = deaths from a disease ÷ cases of the disease (a measure of severity); mortality rate = deaths ÷ population.
Crude rates can mislead when populations differ in age (or other factors) - age is the commonest confounder.
Direct standardisation - apply the age-specific rates of each study population to a single standard population (e.g. the European Standard Population) → age-standardised rates that can be compared directly. Needs age-specific rates.
Indirect standardisation - apply standard (e.g. national) age-specific rates to the study population's age structure to get expected deaths; SMR = observed ÷ expected × 100. SMR >100 = more deaths than expected. Used for small populations or when age-specific rates are unstable. SMRs from different areas should not be compared with each other directly (each uses its own age structure).
Case–control studies; approximates RR for rare outcomes
Attributable risk (risk difference)
Incidence in exposed − incidence in unexposed
Excess risk due to exposure in the exposed
Attributable fraction (exposed)
(RR − 1) ÷ RR
Proportion of cases in the exposed due to the exposure
Population attributable risk
Incidence in population − incidence in unexposed
Excess cases in the whole population
Population attributable fraction (PAF)
(Population incidence − unexposed incidence) ÷ population incidence
Proportion of all cases preventable by removing the exposure
Relative measures describe the strength of association (aetiology); absolute and attributable measures describe public health impact (priority setting).
Example: smoking increases the risk of lung cancer much more (RR about 15–30) than of coronary heart disease (RR about 2), but because CHD is far more common, smoking causes more CHD deaths in absolute terms.
Descriptive epidemiology - person (age, sex, occupation), place (geography, setting), time (trends, seasonality, clusters) - generates hypotheses.
Analytical epidemiology - tests hypotheses: cohort, case–control, cross-sectional, ecological studies (see the EBM chapter for strengths and biases).
Intervention studies - RCTs, cluster RCTs (e.g. community water fluoridation trials), natural experiments (e.g. smoke-free legislation and admissions for myocardial infarction), interrupted time series.
Classic UK examples:
John Snow (1854) - mapping cholera deaths around the Broad Street pump; removal of the handle.
Doll and Hill - case–control (1950) and British Doctors cohort (1951–2001): smoking and lung cancer; smokers die on average about 10 years earlier.
Whitehall studies - social gradient in coronary heart disease among civil servants.
Barker hypothesis - low birth weight and later CVD (developmental origins of health and disease).
Before concluding causation, consider: chance (p-values, CIs), bias (selection, information), confounding, and reverse causation.
Bradford Hill considerations: strength, consistency, specificity, temporality (essential), biological gradient, plausibility, coherence, experiment, analogy (detailed in the EBM chapter).
Necessary and sufficient causes:M. tuberculosis is necessary for TB but not sufficient (host and environmental factors are needed). Most chronic diseases have multiple component causes (Rothman's causal pies).
Proximal and distal causes - smoking is a proximal cause of lung cancer; poverty, marketing and stress are distal ("causes of the causes").
Infectious agent → reservoir → portal of exit → mode of transmission → portal of entry → susceptible host.
Modes: contact (direct, indirect via fomites), droplet (>5 µm, short range - influenza, pertussis), airborne (aerosols - TB, measles, varicella; and SARS-CoV-2 in some settings), faecal–oral (norovirus, hepatitis A, typhoid), vector-borne (malaria, dengue, Lyme disease), blood-borne (HIV, hepatitis B and C), vertical (mother to child - HIV, syphilis, hepatitis B, CMV, GBS).
Breaking the chain: treat the source, hand hygiene and PPE, isolation, environmental cleaning, vector control, safe water and food, immunisation, post-exposure prophylaxis.
Key terms:incubation period (infection to symptoms), latent period (infection to infectiousness), infectious period, carrier (asymptomatic infectious person - e.g. typhoid, hepatitis B, MRSA, meningococcus).
Basic reproduction number (R₀) - the average number of secondary cases from one case in a fully susceptible population. Depends on contact rate, transmission probability per contact and duration of infectiousness.
Infection
Approximate R₀
Measles
12–18
Pertussis
12–17
Varicella
10–12
Mumps, rubella
4–7
SARS-CoV-2 (original strain)
2.5–3 (higher for later variants)
Seasonal influenza
1.2–1.5
Effective reproduction number (Rt or Re) - the actual number at a given time, reduced by immunity and control measures. Rt <1 = the epidemic shrinks.
Herd immunity threshold = 1 − 1/R₀. Measles (R₀ 15) → 1 − 1/15 ≈ 93–95% must be immune - hence the 95% MMR coverage target. Influenza (R₀ 1.5) → about 33%.
Vaccine effectiveness = (attack rate in unvaccinated − attack rate in vaccinated) ÷ attack rate in unvaccinated.
Herd immunity protects those who cannot be vaccinated (infants, immunosuppressed people) - a reason vaccine hesitancy affects everyone.
An outbreak is two or more linked cases, or a greater than expected number of cases in a place and time; a cluster is an aggregation of cases that may or may not be linked.
Steps (they overlap, and control measures start early):
Confirm the outbreak - compare with expected numbers; exclude reporting artefacts.
Verify the diagnosis - clinical and laboratory confirmation; typing (whole genome sequencing).
Establish a case definition - person, place, time and clinical/laboratory criteria (confirmed, probable, possible).
Case finding - active search for additional cases.
Descriptive epidemiology - epidemic curve, maps, characteristics of cases.
Generate hypotheses about source and transmission.
Test hypotheses - analytical studies: cohort study for a defined group (e.g. all wedding guests - attack rates by food eaten, RR) or case–control study for a community outbreak (OR).
Control measures - remove the source, exclude infected food handlers, isolation, prophylaxis, vaccination, communication.
Communicate - outbreak control team, public, professionals; final report and lessons learned.
Outbreak control team (OCT) - led by the UKHSA health protection team consultant, with environmental health officers (local authority), microbiology, infection control, and communications.
Exclusion from work/school - e.g. food handlers and healthcare workers with gastroenteritis until 48 hours after symptoms stop; specific exclusion for STEC and typhoid needs microbiological clearance.
Surveillance = "information for action": the ongoing systematic collection, analysis, interpretation and dissemination of health data.
Notifiable diseases - under the Health Protection (Notification) Regulations 2010, registered medical practitioners must notify the local health protection team (UKHSA) on clinical suspicion - do not wait for laboratory confirmation; urgent cases by telephone. Laboratories separately notify causative agents.
Other surveillance: laboratory reporting (SGSS), sentinel GP surveillance (RCGP Research and Surveillance Centre - influenza-like illness), syndromic surveillance (NHS 111, emergency department attendances), hospital infection surveillance (MRSA, C. difficile, E. coli bacteraemia), wastewater surveillance, Yellow Card (adverse drug reactions), cancer registries, death registration.
Passive (routine reporting - cheap, underestimates) versus active (seeking cases - more complete, resource-intensive) surveillance.
Wedding cohort: 60 ate chicken, 30 became ill (attack rate 50%); 40 did not eat chicken, 4 became ill (10%). RR = 0.5 ÷ 0.1 = 5 - chicken is the likely vehicle.
Herd immunity: R₀ 5 → threshold 1 − 1/5 = 80%. If a vaccine is 90% effective, coverage needed = 0.8 ÷ 0.9 = 89%.
Vaccine effectiveness: attack rate 20% in unvaccinated and 2% in vaccinated → VE = (0.20 − 0.02) ÷ 0.20 = 90%.
SMR: 150 observed deaths, 120 expected → SMR = 125 (25% more deaths than expected).
Attributable fraction in the exposed: RR 4 → (4 − 1) ÷ 4 = 75% of cases among the exposed are due to the exposure.
Scenario. Twenty residents of a care home develop vomiting and diarrhoea over 5 days, with new cases appearing every 1–2 days among residents and staff.
Answer: A propagated outbreak (person-to-person), most likely norovirus. Inform the UKHSA health protection team, isolate cases, close affected areas to admissions, enhance cleaning with chlorine-based agents (alcohol gel is less effective), use soap-and-water hand hygiene, and exclude symptomatic staff until 48 hours symptom-free.
Scenario. A GP suspects measles in an unvaccinated 4-year-old. What should she do before laboratory confirmation?
Answer: Notify the health protection team immediately by phone on clinical suspicion, arrange testing (oral fluid or PCR), advise isolation until 4 days after rash onset, and identify vulnerable contacts (infants, pregnant women, immunosuppressed people) who may need post-exposure prophylaxis (MMR or immunoglobulin).
Scenario. Diabetes prevalence has risen over 10 years while incidence has been stable. Why?
Answer: Prevalence ≈ incidence × duration - people with diabetes are living longer (better survival), and improved detection may also contribute.