Study designs Flashcards
Key features of sample
Unbiased
Representative
Precise
Descriptive epidemiology
Sampling to infer back on population
hard to achieve perfect sample to represent population
Analytical epidemiology
Compare like for like samples
hard to achieve similar samples
Ecological study How?
Snapshot in time
Count cases by GROUPS (eg cases within a country)
Define characteristics (exposure and outcome)
Decide data involved - categorical or continuous
Measurement scales ecological study
Discrete/Qualitive - nominal (distinct categories) or ordinal (ordered categories)
Continuous/Quantitive - interval (meaningful differences) vs ratio
Issues for ecological studies
Definition of characteristics
Measurement variation
Confounding (ecological fallacy: falsely inferring individual association from group level association)
Chance (random error)
What is ecological fallacy?
falsely inferring individual association from group level association
Cross sectional studies
Snapshot of now Individual level (can be a series of studies over time)
Cross sectional survey, who are we generalising to?
Population
What population can w get access to CSS
Study population - GP practice?
How can we get access to population CSS
Sampling frame (census)
Who is in your study CSS?
The sample
Issues for cross sectional studies?
Sampling bias
Responder/participant bias
Chance - random error
Case control study
Find cases (eg diseased)
Find controls (eg not diseased)
Look back to past to find exposures
Compare
Issues with case control study
Selection bias (controls should reflect population and be comparable to cases)
Information bias (poor recall, missclassification)
Confounding
Chance - random error
Case control stud odds ratio
Odds ratio is not rate ratio
looking back in time, NOT future
Cohort study concept
Find exposed and unexposed (OUTCOME FREE INDIVIDUALS)
Group according to exposure
WAIT
Look to future to compare and contrast outcomes
Issues with cohort study
Loss from follow up (differential loss/survivor bias)
Information bias (misclassification)
Confounding
Chance - random error
Measurement cohort study
Incidence rate
Odds ratio OR rate ratio (case control is just odds)
Describing a study
PICO
P - population
I - intervention / exposure
C - comparison / control
O - outcome
Descriptive epidemiology studies
Ecological - group analysis
Cross-sectional - individual analysis
Analytical epidemiology studies
Case control - odds ratio
Cohort - odds ratio or rate ratio
Time for studies and cost
LOW -
Ecological study
Medium:
Cross sectional
Case control
High:
Cohort
Bias present for ecological
Sampling
Confounding (high)
Random error
Bias present for cross sectional survey
Sampling
Recall
Confounding
Random error
