module 1 lesson 4 Flashcards
Ecological study design
data representing entire population used to evaluate an entire population
Why is study design important
The purpose of a study design is to organize the collection of relevant information to provide strong evidence for a causal relationship between a factor of interest and the development of a health outcome.
Counterfactual outcomes
the ideal experiment:
people are subject to exposure and followed for a period of time and calculate the incidence proportion (risk for exposure - Re)
Now turn back time so the same people are in the original state at the same time prior to exposure. Follow for the same period of time and caculate incidence proportion (risk for unexposed - Ru)
A causal effect of exposure is defined as Re-Ru
Four major reasons study result does not represent a true effect:
- lack of temproality
- lack of validity due to confounding
- lack of validity due to bias
- lack of validity due to chance
How can statistical analysis assist in establishing association versus effect
- stat analysis can est association, but not necessarily true effect
-can help removing confounding effects and dealing with chance
NOT helpful in dealing with biases and lack of temporality
Categories of study design
Observational: prospective, retrospective
Experimental: randomized parallel, randomized crossover, cluster randomized parallel, quasi-experimental
Case control- cumulative, density, case-cohort
cross sectional
Ecological
Observational
Natural experiments - investigator does not control exposure
study groups often not comparable
**research on harmful factors must use this approach
Prospective study
study starts in real time and entire study cohort is at risk at the start of the study
Retrospective
study starts after the health outcome has occurred; entire study cohort is “at risk”some time in the past
Strengths of observational study
- good evidence is needed for a risk factor
- exposure is rare
- desire accurate measure of exposure
- little known about exposure
Limitations of observational study
relatively expensive
- lengthy
- difficult to recruit
- loss to follow up
Experimental cohort
investigator manipulates exposure
usually subjects randomly assigned to groups
used only when tx is beneficial
BEST results
Types of experimental study designs
- Randomized Parallel
- Randomized cross over
- Cluster randomized parallell
- Quasi experimental trial
Randomized parallel
individuals are randomized to experimental groups
All groups followed in parallel over time
Most effective
Randomized cross over
individuals randomly experience all group treatments in sequence
efficient for some treatment effect combinations
requires tx and effect short lived
Cluster randomized parallell
collections of individuals randomized to experimental groups
Allows trials to be conducted in restricted settings
Quasi experimental trial
non-random assignment of individuals or colldections of individuals
Basic sequence of data collection in case control study
- identification of individual with health outcome
- individual without health outcome are identified
- presence of factor of interest determined for cases and controls at an appropriate time in past
Strengths of case control
- desirable if health outcome is rare
- more efficient, less expensive, shorter in duration
- desirable if little is known about risk factors for the outcome
- preferred if follow up is difficult
Limitations of case control
can’t directly estimate risk - only risk ratio
exposure based on recall or past record (bias)
est temporality
Cross sectional study design
Assess prevalence of population characteristics at a “point in time”
Major limitations: lacks temporal relationship
odds ratio used as measure of association
often based on survey data
inexpensive and quick
Sampling in cross sectional studies
probability or representative sample -sampling scheme that provides accurate measures for the target population
-simple random
-stratified random
systematic random
Nonpropability or convenience sample - only representative of sampled population
Ecological study design
data representing entire population
exposure status and outcomes status are single values applied to entire population
Ecological fallacy
when based on aggregate data, may not represent association that exists on individual basis. Other factors may be responsible.