Block 1: Epidemiology |
EPI4-4: THE ODDS RATIO:LINK WITH THE OTHER MEASURES OF EFFECT |
SESSION OBJECTIVE |
At the end of this session you should understand the link between the Odds Ratio and other measures of effect - the prevalence ratio, the incidence proportion ratio and the incidence rate ratio. |
The relationship between the Odds Ratio and the Prevalence Ratio or Incidence Proportion Ration is simpler to understand.
D+ | D- | ||
E+ | a | c | M1 |
E- | b | d | M0 |
N1 | N0 | T |
If the data are based on a prevalence or an incidence study in a closed cohort,the OR can also be worked out a third way.
The prevalence ratio or the incidence proportion ratio of disease in the exposed to that in the unexposed is (a/M1) / (b/M0),
which is equal to (a/a + c)/(b/b + d).
If the prevalence (a+b)/T is small (usually less than 10%) , it can be assumed that a and b are small in relation to c and d and then (schematically)
(a + c) ~ c and (b + d) ~ d,
and the above prevalence ratio is approximately (~) ad/bc which is the Odds Ratio.
D+ | D- | ||
---|---|---|---|
E+ | 53 | 53 | 106 |
E- | 43 | 85 | 128 |
96 | 138 | 234 |
If the data come from either type of study rather than a case control study the Prevalence Ratio or the Incidence Proportion Ratio would be 53 x 128/106 x 43 = 1.49 , compared to the OR of 1.89.
In the above case the two values are not so close, and the reason for this is that the prevalence of the condition is rather high in the population at 96/234 = 41%. If we considered a rarer condition, as in the table below:
D+ | D- | ||
---|---|---|---|
E+ | 53 | 530 | 583 |
E- | 43 | 850 | 893 |
96 | 1380 | 2340 |
If the data come from either type of study rather than a case control study the Prevalence Ratio or the Incidence Proportion Ratio would be 53/583 dividded by 43/893 = 1.89 , which is pretty close to the OR of 53X850/43X530 = 1.98
Here the prevalence is only 4%