epidemiology

Epidemiology

Study of the distribution and determinants of health-related states or events in specified populations AND the application of this study to the control of health problems

Public health

Science and art of promoting health and extending life at the population level

Descriptive epidemiology

Describes who, what, when, where (person, place, time); no advance hypothesis

Analytical epidemiology

Asks why and how; hypothesis-driven; needs a comparison group

Epidemic

Cases of an illness in excess of normal expectancy in a community or region

Outbreak

An epidemic confined to a localized geographic area

Pandemic

An epidemic spread over a wide region or worldwide

Endemic

Ongoing or constant presence of a disease in a community or group

Sporadic

Occurs infrequently and irregularly

Primary case

The first case in a population

Index case

The first case that comes to the attention of the epidemiologist (not always the primary case)

Case definition

Standard set of criteria so cases are diagnosed and counted consistently; changing it changes the data

Loose vs. strict case definition

Loose catches more cases but includes more false ones; strict is more certain but misses cases

Point source

Single origin of an outbreak (e.g., the Broad Street pump)

Epidemic curve

Graph of cases over time

Frequency distribution

The values a variable can take and the number of people with each value

Spread

Distribution around a central value (range, standard deviation)

Shape

Symmetrical or asymmetrical (skewed left or skewed right)

Bimodal

A distribution with two peaks

Age-specific rate

A rate within one age category

Age-adjusted rate

Rate statistically adjusted so populations with different age structures can be compared

Population structure

Age and sex makeup of a population (Afghanistan is young, the U.S. is older)

Incubation period

Time between exposure and disease (short for foodborne illness, long for lung cancer)

Disease registry

Database of cases of a condition (e.g., SEER for cancer)

Surveillance

Ongoing systematic collection of health data

Descriptive epidemiology = what three things?

Person, Place, Time

Person (descriptive epi)

Who is most at risk: age, sex, race/ethnicity, occupation, SES

Place (descriptive epi)

Geography, birthplace, site of employment, school district

Time (descriptive epi)

Hours to decades; seasonal patterns and long-term trends

The 5 W's of a health question

What, Who, Where, When, Why/How

Count, Divide, Compare

Count cases (numerator); divide by the correct denominator; compare across groups or times

Why does the denominator matter?

Raw counts mislead: Delaware had fewer births than Pennsylvania but a higher birth rate (13.8 vs. 11.4 per 1,000)

Hippocrates (~400 BC)

Explained disease in the natural world; documented person, place, time

John Graunt (1600s)

London haberdasher who found patterns in mortality data (sex, urban/rural, seasonality)

William Farr (1800s)

Began systematic collection of mortality data in Britain

John Snow (1854)

Mapped cholera cases around the Broad Street pump in Soho; identified a point source; handle removed

John Snow key facts

Suspected water not miasma; 600+ deaths; compared water companies; accepted only after germ theory

Death certificates (data source)

Fairly complete, but only fatal cases and cause of death can be questionable

Birth certificates (data source)

Fairly complete with rich data, but content varies by state

Disease registries like SEER (data source)

Good for rare diseases; miss people who don't seek care; reportable diseases vary by state

Lab records (data source)

Quantitative with flexible case definition; no clear denominator; miss people not tested

NORS

National Outbreak Reporting System

Transition model

Boxes are compartments of people; arrows are flow of people over time (Susceptible to Diseased)

Prevalence

Proportion of people with a disease or attribute at a point in time (or period); new plus existing cases; a snapshot

Incidence

Flow of new cases over time from susceptible to diseased

Cumulative incidence

New cases / people initially at risk; closed population, fixed period; a proportion (0 to 1, no units)

Cumulative incidence aliases

Incidence proportion, attack rate

Incidence rate

New cases / total person-time at risk; units 1/time; range 0 to infinity

Person-time (person-years)

Total time each person was observed and at risk, added across everyone

Closed population

No one enters or leaves; use cumulative incidence

Open population

People enter and leave; use incidence rate

Lost to follow-up

Participant leaves the study; contributes time only until they left

Morbidity

Burden of disease on a population

Mortality

Burden of fatal disease on a population

Cumulative mortality

Fatal cases / people initially at risk (cumulative incidence where the event is death)

Mortality rate

Fatal cases / person-time at risk

Case fatality (CFR)

Fatal cases / total cases; the denominator is cases, not the population

Prevalence formula

prevalent cases / population

Cumulative incidence formula

new cases / # initially at risk

Incidence rate formula

new cases / person-time

Case fatality formula

deaths / # cases

Prevalence vs. incidence memory trick

Prevalence = Pool of existing plus new cases (snapshot); Incidence = Inflow of new cases only

Prevalence depends on…

Both incidence AND duration; high prevalence can mean many new cases or long survival

Prevalence is often used for…

Chronic diseases (long duration, onset hard to date)

Which measure uses which denominator?

Initially at risk = cumulative incidence; person-time = incidence rate; whole population at a time point = prevalence; total cases = case fatality

A person is lost to follow-up at 2 years. Person-time contributed?

2 person-years

When does a person stop contributing person-time?

At first of: disease, loss to follow-up or death, or end of study

Assumption of person-time calculation

Probability of disease is constant over the study period (not always valid)

Population-level person-years shortcut

Mid-year population approximates person-years for a one-year period

Multivitamin example

468 / 1,150 = 40.7% prevalence

Hepatitis B infants example

27 / 308 = 8.8% prevalence; time scale is age

Hepatitis B students example

39 / 750 = 5.2% cumulative incidence over 4 years

Diabetic women example

72 / 1,862 = 38.7 deaths per 1,000 person-years

AIDS 2003 example

44,232 / 290,809,777 x 100,000 = 15.21 per 100,000 person-years

XDR-TB example

52 / 53 = 98.1% case fatality

CFR slide example

10,000 deaths / 50,000 cases = 20%

Kaplan-Meier

Survival method showing proportion still disease-free over time

Censored

Person who leaves observation without the event (lost to follow-up or end of study)

Kaplan-Meier calculation

New survival = previous survival x (1 - events / number at risk); cumulative incidence = 1 - survival

Hazard rate

Instantaneous risk of the event

Cox proportional hazards

Regression-based model that produces a hazard ratio

Hazard ratio (HR)

Hazard in exposed / hazard in unexposed; COVID-diabetes study HR = 1.17 (adjusted)

Incidence rate vs. hazard

Incidence rate estimates the mean hazard over all person-time

Surveillance bias

Exposed people get more testing, so more outcomes are detected

Why match on sex, age, and test date (COVID study)?

These relate to both infection and diabetes risk, so matching reduces confounding

Why exclude prior diabetes (COVID study)?

To study incident diabetes; people must be disease-free at the start

Why exclude diabetes or death within 30 days (COVID study)?

Avoids reverse causation and people not truly at risk

COVID-diabetes study limitations

Unmeasured confounding (obesity, activity), surveillance bias, differential health care use, untested unexposed people

Why can disease frequency change without a change in risk?

Case definition change, case detection, surveillance policy, new tests or criteria, population change, awareness, random events

7 question types in epidemiology

Diagnostic, Causes, Treatment, Prognosis, Health promotion, Surveillance, Disparities

10 Essential Public Health Services groups

Assessment, Policy development, Assurance, with Equity at the center

Ecologic study

Uses aggregate group-level data, not individual data

Case report / case series

Describes one patient or a small group with similar diagnosis

Cross-sectional study

Exposure and outcome measured at one point in time; gives prevalence

Cohort study

Start with exposure, follow for outcome; observational

Prospective cohort

Enroll now and follow forward in time

Retrospective cohort

Uses existing records; exposure and outcome have already happened

Case-control study

Start with outcome (cases and controls) and look back at exposure

Randomized controlled trial (RCT)

Investigator randomly assigns the exposure

Randomization

Chance assigns exposure, balancing known and unknown factors