Cohort and Case Control Studies Flashcards
What are the steps in creating a retrospective cohort study?
- ID cohort (ex: CRC)
- Collect predictor (age, sex, BMI, etc.)
- Measure the outcome? (Diagnosis date, path exam)
Name one strength of a retrospective cohort study
- Not expensive and/or time consuming
- Can study rare outcomes
Name one weakness of a retrospective cohort study
- Can’t go back to investigate something interesting
- Can’t ask for more information
- High chance for confounders and bias
Name 3 strengths of a prospective cohort study
- Risk and rate disease estimate over time
- Rare exposures
- Temporal relationship between exposure and outcome
- Natural history of disease
- Lower potential for bias vs case-control
- Avoid recall and selection bias
- Results more conclusive than case-control
- Longer is stronger
What are the steps of creating a prospective cohort study?
- Assemble the cohort
- Measure predictor variables and potential confounders for all participants
- Follow-up the cohort and measure outcomes
Name 3 types of cohort studies.
- Occupational
- Convenience
- Geographic
- Disease/Procedure
- Prepresentative (NHANES)
What are 3 challenges of large cohort studies?
- Requirements of resources and manpower
- Huge database mgmt
- Followup op problems
- Exposure information
- Data quality?
- Collection of biological samples?
- Expensive and time-consuming, especially prospective
- Cannot directly infer causal relation
What are the two types of descriptive observational studies?
- Case reports/Case series (Clinical)
2. Cross-sectional (Epi)
What are the 3 types of analytical observational studies?
- Cohort
- Case-control
- Ecological
What is information bias?
AKA observation or measurement bias, when key information/outcomes are measured, collected, or interpreted inaccurately.
For which type of study is incidence the best report of risk?
Cohort studies
What is the formula for incidence (risk)?
# ppl get disease/ #ppl at riskxtime at risk
What are the outcome variable types of a cohort study?
Cumulative incidence Incidence rate (per person time)
What are the adjusted measures of association for cumulative incidence?
Relative risk (direct), Standardized incidence ratio (indirect) Hazard ratio (Cox model) Rate ratio (Poisson regression)
How are participants selected in a retrospective case/control study?
- Same underlying population
- Represent people who would have had the same chance of being selected as cases if they had the outcome
- Standard diagnostic criteria
- Eligibility criteria
What are the 5 essential features of retrospective case-control study design?
- Directionality (outcome-exposure)
- Timing (case ascertainment can be retro or prospective)
- Rare or new disease
- Challenging
- Population- vs. hospital-based (more common)
A nested case-control study is a _ study within a _ study.
A case-control study within a prospective cohort study.
How can a nested case-control study be prospective?
Using incidence density sampling, within a prospective cohort, with specimens and exposures collected prospectively prior to disease (no recall bias). RR can be estimated without the need for rare disease assumption.
How are case-cohort study cases and non-cases selected?
From within parent cohort, no matching, either at baseline or length of followup
What statistics can be calculated in a case-cohort study?
odds ratio, risk ratio, logistic regression
must weight the probabilities when stratifying
What are advantages of the case-cohort study?
efficient (only if more than one endpoint), relatively inexpensive, flexible, reduces selection bias (same population), multiple outcomes within the subcohort, can calculate person-time risk
What are disadvantages of the case-cohort study?
Often complex analysis, increased info bias if subcohort established after baseline, may have too many censoring to be robust, not for longitudinal exposure changes
When is a case-crossover study most appropriate (what exposure/outcome)?
When investigating an intermittent exposure with immediate and transient effect on risk related to an abrupt outcome.
How are case-crossover study cases and non-cases selected?
Cases serve as their own controls at another time point
