Lecture 2B Flashcards
What is descriptive epidemiology?
Taking a sample that you can infer back to the population with
What is analytical epidemiology?
Comparing “like for like” samples. Very hard to do as samples rarely identical. Two populations different age, income demographic etc
Epidemiological study designs?
See lecture for diagram.
Broadly observational and experimental.
Observational can be descriptive or analytical.
Descriptive involves taking a population and looking at what’s happening in population eg ecological and cross-sectional surveys
Analytical involves taking two populations and comparing eg case control and cohort studies
Ecological studies?
Done at present moment. Count cases that occur within groups.
Pick an exposure eg motorway and outcome eg heart disease at group level.
Ecological fallacy a problem. Making inferences about individuals from a group
Ecological association or fallacy?
Falsely inferring individual level association from group level association
Cross-sectional studies
Done on individual level at present.
Getting prevalence
Problems with responder bias
What is an issue for every study type?
Chance (random error)
Case-control study
Always retrospective
Find cases and define controls. Compare and contrast case and control groups and exposures that potentially caused cases.
Controls should reflect population but also be comparable with cases.
Case control problems in hospitals, take a case of lung cancer on respiratory ward and a control that smokes. Might conclude that smoking doesn’t cause lung cancer.
Susceptible to recall bias, might try find something to explain their condition
Cannot use rate ratio
Cohort studies?
Take two groups that are outcome free. Break into those that are exposed and those not exposed. Look at outcomes and incidence rates for both groups.
Problems with loss to follow up- sick don’t go to clinic etc. People who stay different to those who leave. Problem with confounding
Odds ratio is popular
What is non-differential misclassification?
Bias but is random and so doesn’t have major effect on outcome
Incidence rate: count per person time at risk?
Count number times occurred in sample. Divide by total number of sample years. Multiply by 1000 for number of person years in one thousand with disease.
Acronym for study design?
PICO
Population
Intervention or exposure
Comparison or control
Outcome of interest?
Analysis of studies
Ecological groups
Cross sectional individuals
Case control only odds ratio
Cohort rate or odds ratio
