Module 5
Lesson 4
Hi and welcome back!
In the previous lesson I have introduced one measure of disproportionality, the “relative reporting
ratio”. In this lesson I will introduce two additional common measures of disproportionality. They
might come up in text books or in the literature, so we thought it would be important for you to hear
about them, to understand how they're computed and also how they differ from each other.
Let's reuse the database example from the previous lesson:
The database contains 100 reports in total. We are interested in a specific drug and a specific
adverse event. I have reorganized the database so we can highlight the reports that are important to
compute the expectation. We have the reports with the drug, the reports with the event, the reports
with both and the reports with neither. We observe three reports with both the drug and the event.
Recall that for estimating the number we would expect if the drug and the adverse event are
unrelated, the relative reporting ratio used all reports in the database to compute the frequency of
the adverse event. That frequency was 8%, giving an expected number of 0.8. The ”relative reporting
ratio” was thus three reports divided by 0.8, giving a value of 3.75.
Another measure of disproportionality is the “proportional reporting ratio”. Like the “relative
reporting ratio” the “proportional reporting ratio” is an observed-to-expected ratio but the expected
number is counted differently than for the relative reporting ratio. Let me show you!
In the case of the “proportional reporting ratio” we only use the reports without the drug to
estimate the frequency of the adverse event and in our case it becomes 5/90, which is about 6%. So,
since we have 10 reports with the drug, if the drug and the adverse event are completely unrelated,
we expect to see 6% out of those 10 reports having both the drug and event and that is 0.6 reports.
So, our expectation for the “proportional reporting ratio” becomes 0.6 and the value becomes 3
divided by 0.6 which is 5.
Yet another measure of disproportionality is the “reporting odds ratio”. Like its name indicates, the
“reporting odds ratio” is a ratio of odds. The odds for the adverse events in the subset of reports
having the drug divided by the odds for the event in the subset of reports not having the drug.
So, concretely in our example it becomes 3/7 for the reports having the drug divided by 5/85 for the
reports not having the drug so the value of the reporting odds ratio is thus 7.3.
At this point you might start to worry! Three different measures of disproportionality? What if they
point to different directions? Some being above one while the other being below one. Well, rest
assured! it can be proven mathematically that the three of them always point to the same direction.
Either they are all above one or they're all below one.
And finally, you should know that for more realistic scenarios where the drug and the adverse event
represent a small portion of the database, all three measures of disproportionality that I have shown
you will have very similar values.