Describe descriptive epidemiology and its study designs.

Purpose and the Three Axes

• Its procedure runs in a fixed order:
• 1. Define the population to be studied — the defined population or "population at risk", which must be specified in time and place and be large enough
• 2. Define the disease under study — a precise, valid and workable case definition applied uniformly, without which counts from different observers cannot be combined
• 3. Describe the disease by time, place and person
• 4. Measure the disease — in terms of incidence, prevalence, morbidity and mortality
• 5. Compare with known indices — between different populations and subgroups
• 6. Formulate an aetiological hypothesis
• A good hypothesis must specify four things, and this is examined: the population; the cause being considered; the expected outcome; and the dose-response and time-response relationship
• Time distribution — three patterns are described:
• Short-term fluctuations — epidemics, of which the epidemic curve reveals the type
• A point-source (common-source, single-exposure) epidemic gives a sharp rise and rapid fall, with all cases occurring within one incubation period — a food-poisoning outbreak at a wedding being the classic
• A continuous common-source epidemic gives a plateau rather than a peak, lasting as long as the exposure
• A propagated (progressive, person-to-person) epidemic gives a slow rise with successive peaks one incubation period apart, and a gradual decline as susceptibles are exhausted
• The epidemic curve therefore does real work — it indicates the type of epidemic, estimates the incubation period, and points back to the probable time of exposure
• Periodic (cyclical) fluctuations — seasonal trends (measles in winter and spring, diarrhoea in summer and the monsoon, malaria after the rains); and cyclic trends over years, as in the epidemics of measles every 2 to 3 years in the pre-vaccine era, reflecting the accumulation of susceptible children
• Long-term (secular) trends — changes over decades, such as the fall in tuberculosis and rise in coronary disease, which describe the epidemiological transition

Place, Person and the Study Designs

• Place distribution — international variations, which generated some of the most productive hypotheses in medicine (gastric cancer in Japan, colon cancer in the West, Burkitt lymphoma in the African malaria belt); national and regional variations, such as endemic goitre in the sub-Himalayan belt, fluorosis in Andhra Pradesh and Rajasthan, and kala-azar in Bihar; rural versus urban differences; and local clustering
• Migration studies are the most powerful application of place data, because they help separate genetic from environmental causes
• The reasoning — if migrants acquire the disease rates of their new country, the cause is environmental; if they retain the rates of their country of origin, a genetic or deeply cultural factor is implicated
• The classical example is japanese migration to the United States, where gastric cancer rates fell and colon cancer and coronary disease rates rose toward American levels within a generation or two — establishing that these were environmental and largely dietary
• Person distribution — age, which is the most important single variable; sex; ethnic group; marital status; occupation; social class; education; behaviour; stress; and migration
• A cohort effect is not the same as an age effect, and telling them apart is a genuinely difficult point that examiners like
• An age effect means risk changes with age itself; a cohort effect means a particular birth cohort carries an exposure through life
• The illustration — a rise in lung cancer with age in a cross-sectional survey may reflect the fact that older people have smoked for longer, rather than any effect of ageing itself; and as smoking prevalence falls, the pattern shifts with the cohort rather than with age
• The descriptive study designs:
• Case reports and case series — the weakest evidence but historically important as the first signal; the first reports of pneumocystis pneumonia in young men in 1981 were a case series, and announced the AIDS epidemic
• Cross-sectional (prevalence) studies — a snapshot of a population at one point in time, measuring exposure and outcome simultaneously
• Their advantages — quick, relatively cheap, give prevalence for planning, can study several diseases and exposures at once, and are the basis of national surveys such as NFHS
• Their limitations — cannot establish temporality, so cause and effect cannot be separated; vulnerable to reverse causation; unsuitable for rare or short-lived disease; and subject to survival bias
• The classic illustration of their weakness — a cross-sectional survey finding that people who exercise are thinner cannot say whether exercise causes thinness or thinness permits exercise
• Ecological (correlational) studies — comparing populations rather than individuals, using routinely available aggregate data
• They are quick and cheap and are useful for generating hypotheses and studying population-level exposures such as air pollution or legislation, which cannot be measured individually
• But they are subject to the ecological fallacy, and cannot support conclusions about individuals
• The overall limitation of descriptive epidemiology is that it generates hypotheses but cannot test them; that requires analytical study — and saying so completes the answer
• The epidemic curve identifies the type of outbreak and points back to the probable time of exposure
• A single sharp peak means a point source, with all cases falling within one incubation period
• Successive peaks an incubation period apart indicate person-to-person propagation rather than a common source
• A plateau suggests continuing common-source exposure rather than a single contaminated event
• Migration studies separate genes from environment, and are the neatest natural experiment in the subject
• Japanese migrants lost gastric and gained colon cancer, which established both as environmental and largely dietary
• A cohort effect can masquerade as an age effect, as smoking history does in cross-sectional lung cancer data
• Cross-sectional studies cannot establish temporality, which is their defining limitation for causal work
• Exercise and thinness illustrate the difficulty; a snapshot cannot say which came first
• Case series announced the AIDS epidemic, which is why the weakest design still matters as a first signal
• Ecological studies suit population-level exposures such as air pollution and legislation, which cannot be measured individually
• Define the case before counting anything, or observers will not be counting the same disease
• A good hypothesis names population, cause, outcome and dose, which is a standard four-part sub-question
• Age is the most important single person variable, and no analysis proceeds without accounting for it
• Seasonal patterns suggest the mode of transmission, as monsoon diarrhoea and post-monsoon malaria do
• Cyclic measles epidemics reflected susceptible accumulation, every two or three years before vaccination
• Secular trends describe the epidemiological transition, with tuberculosis falling as coronary disease rises
• Endemic goitre follows the sub-Himalayan belt, which is a place distribution with an obvious environmental explanation
• Descriptive studies generate but cannot test hypotheses, which is the sentence that completes the answer
• Prevalence surveys underpin national planning, which is why NFHS is cross-sectional by design
• Local clustering may reveal a common exposure, and is what prompts investigation of a suspected outbreak
• Burkitt lymphoma mapped onto the malaria belt, which generated the hypothesis linking it to Epstein-Barr virus and malaria