Case-control studies Flashcards
What is a case control study?
Studies which compare exposure level in a group of individuals who have a specific disease (case) and in another group of individuals without the disease (controls)
- Observe to see what happened
What is the definition of a case?
Some one with the disease or problem - a case
What is the definition of a control?
Must be individuals who would have been cases if had developed outcome
Where are sources that can be used as controls?
Community based controls (preferable) :
- Electoral register
- Random digit dialling
- Primary care controls
- Spouse or family controls
Other disease controls:
- Hospital controls
- Cancer registry controls
How is exposure measured?
- Measurement is performed retrospectively
- Data reported by subject / from records
What are strengths of case-control studies?
- Relatively quick ( no long periods of follow
up) - Efficient for rare outcomes
- Allows examination of >1 exposure for 1
outcome - No loss to follow up bias
What are weaknesses of case control studies?
- Temporal sequence unknown –> reverse
causality (cannot be sure exposure came
before outcome) - Inefficient for rare exposures
- Prone to bias (selection and measurement -
recall and interviewer) - Confounding - other factors may explain
association - Can not calculate incidence rate
What is the definition of a confounder?
A third variable that provides alternative explanation for observed association between exposure and outcome
What can a regression model be used for?
Can be used to assess the strength of association between an exposure and outcome.
What are the main types of regression models?
Continuous outcome –> Linear regression model –> Regression coefficient summary statistic
Binary outcome –> Logistic regression model –> Odds ratio summary statistic
What is the equation for odds ratio?
Odds Ratio (OR) = odds in exposed group /
odds in unexposed group
What are outcome must an odds ratio have?
Outcome must be binary
How can you calculate the confidence interval for odds ratio?
Can calculate sample odds ratio from the data - but its the population rather than the sample
that is of interest
- So can estimate population odds ratio using
sample odds ratio and use Confidence
Interval to show precision of the estimate of
population odds ratio
What do the P values mean for odds ratio?
Smaller p values = stronger evidence against the null hypothesis
P values should be interpreted in terms of strength of evidence, not statistical significance
What should you use to draw a conclusion?
A combination of the p value and confidence interval should be used to draw conclusions
