EPID 503 – Class 12 Cohort Study Design
Hierarchy of Study Designs Establish Causality Generate Hypotheses Gold standard for establishing causality Randomized Controlled Trials Cohort Studies Case Control Studies Cross-sectional Studies Ecological Studies Good place to start investigating a research question
Recall – Randomized Trials Study Population RANDOMIZATION Exposed to factor Not exposed to factor Event No event Source Population
Thought Question Why do we not always do RCTs if they are one of the best designs?
Randomized Trials Not Always Appropriate
Cohort Study Design Represents Another Way Key features: Start with exposed and unexposed and observe over time to see who develops disease Participants must be “at risk” of developing outcome Study Population Self Selection Not exposed Exposed Event No event Event No event
What are the Strengths of This Design? Study Population Self Selection Not exposed Exposed Event No event Event No event
What are the Strengths of This Design? Can estimate incidence Know that exposure preceded disease Can study rare exposures Can study multiple outcomes Study Population Self Selection Not exposed Exposed Event No event Event No event
What are the Strengths of This Design? Study Population Gold Standard of Observational Designs Self Selection Not exposed Exposed Event No event Event No event
What are the Limitations of This Design? People self-select exposures so exposed and unexposed may differ with respect to important characteristics Cannot study rare diseases Prospective design is expensive Prospective design can require long follow-up Study Population Self Selection Not exposed Exposed Event No event Event No event
How to Get Around Long-Follow Up?
How do we analyze and summarize data from a cohort?
Disease Yes No Total a b a + b c d c + d a + c b + d a + b+ c+ d Exposure What measure can we calculate here?
Disease Yes No Total a b a + b c d c + d a + c b + d a + b+ c+ d Exposure Cumulative Incidence exp = a/(a+b) = Risk Ratio Cumulative Incidence no exp = c/(c+d)
Disease Yes Person-Time a PTexp No c PTunexp Total a + c Total PT Exposure What measure can we calculate here?
Disease Yes Person-Time a PTexp No c PTunexp Total a + c Total PT Exposure Incidence Rate exp = a/PTexp = Rate Ratio Incidence Rate no exp = c/PTunexp
Example Ratio Interpretation Risk Ratio (or Rate) Ratio = 1.1 Among the population studied during the time frame of interest, the risk (or rate) of disease in exposed persons was 1.1 times that of unexposed persons. Among the population studied during the time frame of interest, the risk (or rate) of disease in exposed persons was 10% higher than that in unexposed persons.
Confidence Intervals Used to assess the statistical significance of an estimate
Interpreting Confidence Intervals Offers information about our confidence in the point estimate The wider they are the less confident we are Confidence intervals get more narrow the greater the sample size (holding all else equal) Based on hypothetical random replication of the data