Measures of Association Flashcards
Definition of association & Ho hypothesis
A statistical quantity that gives an indication of the magnitude (strength) of association between exposure & outcome
Ho :
- there is no association between exposure & outcome
Types of measures of association (2)
- Relative Risk (RR)
- Risk ratio
- Rate ratio
(( usually cohort study )) - Odds Ratio (OR)
(( usually case-control study ))
Range of values for RR & OR
0 to infinity
If RR or OR = 1
There is no association between exposure & outcome
If RR or OR > 1
& phrasing in word
- positive association
- exposure associated with increased risk of outcome
Those who are exposed have __ times the risk of developing the outcome compared to those who are unexposed
If RR or OR < 1
& phrasing in word
- negative association
- exposure associated with decreased risk of outcome or potentially protective effect
Those who are exposed have __ % the risk of developing the outcome compared to those who are unexposed
OR
Those who are exposed have __ % less risk of developing the outcome compared to those who are unexposed
Risk Ratio
- use
- formula
- used in cohort study
- same length of follow up
Risk Ratio = (cumulative incidence in exposed / cumulative incidence in unexposed)
= (a/b+c) / (c/c+d)
Cumulative incidence
- proportion who develop the outcome within a specified period (new cases)
Rate Ratio
- use
- formula
- used in cohort study
- different lengths of follow up
- consider person-years / person-time
eg total number of years contributed by all participants
Rate Ratio = (incidence rate in exposed / incidence rate in unexposed)
= (a/PTe) / (c/PT0)
Incidence rate
- rate at which new cases of outcome occur in a population at risk for the outcome
Odds Ratio (OR)
- use
- formula
- typically used in case-control study (cos RR cannot be calculated directly)
- can also be cross-sectional study or cohort study (less common)
- can estimate RR when outcome is rare (eg <10%)
OR = (odds that a case was exposed / odds that a case was not exposed)
= (ad/bc)
Why use OR > RR in case-control study
- RR cannot be calculated directly
- subjects are classified on basis of cases
- compared on basis of exposure
- cannot calculate incidence in the exposed or unexposed groups
Why OR can predict RR when outcome is rare?
- a ««_space;b
- c ««_space;d
Hence, RR = (a/a+b) / (c/c+d)
= (a/b) / (c/d)
= (ad/bc)
= OR
Multiple Exposure Groups
Need to define a reference group
Why is there a need to select a reference group in studies with multiple exposure groups?
How to select reference group?
- for comparison with other exposure groups
- select the least exposed group as reference group
