For parts a + b, we used formulas from the VanderWeele interaction tutorial part 2.2. I know this was not assigned to all students, can you confirm whether these were the correct formulas to apply?
Also, we are confused how to decide whether to use the attributable proportion with RR11 in the denominator or (RR11 - 1) in the denominator. What is the difference between these and why would we use one over the other? We did not understand the nuance in his explanation.
Hi Jean,
To answer question 3 in HW6 you only need your previous knowledge about percent attributable risk and the formulas in the slides of effect modification, in particular see Part 3, slide 14 for the formula for AP due to interaction (or Vanderweele, p8), and think about how you can parse out and define the different effects provided in the table. For this question, assume no bias so that you can work with the provided RR’s directly and parse out effects for each cross-classified category. In this question, you only need the general AP formula that you learned in prior classes, not the alternative one with “RR11-1” in the denominator, presented by Vanderweele.
Although it is not what we were asking, if you have already solved this problem using formulas from VanderWeele tutorial part 2.2 with the alternative definition of AP, please indicate so in your HW. The numbers will be slightly different but that is OK, as long as you indicate/explain what you have done. We recommend not to use the VanderWeele formula, since that is not covered in class.
Attributable risk percent (or Attributable Proportion) Attributable risk percent among the exposed. = the proportion of risk among the exposed that is attributable to the exposure = (risk among the exposed - risk among the nonexposed)/risk among exposed = It is the percent (or proportion) of disease in the exposed that could be eliminated if the exposure were eliminated.
The formula for AR% using relative risk is RR-1/RR. To apply this formula, you need to change 1 for the baseline risk. For example, RR11-RR10/RR11.
I hope this explanation makes things clearer. If not, please let us know.
Thanks, Monica. Using VanderWeele's formula, the three calculated attributable proportions add up to 1. With the alternative formula (for example, RR11-RR10/RR11), the sum exceeds 1 (unless I am not calculating the APs correctly). Mustn't the APs add up to 1?
Hi Zara,
You are right, if you sum the AP’s from questions a, b, and c it could exceed 1. However, that is what we are asking you to discuss in part c - Why is that the calculated attributable proportions exceed 1?”
HI all,
The TA's and I reviewed this again today and offer the following further clarification for what we are asking for in question 3:
An equivalent interpretation of what we are asking in q3a is: “Among hypertensive women, what is the proportion of stroke cases that is attributable to OC use?"
For 3b,: “Among OC users, how much of stroke cases is attributable to hypertension?”
In 3c, we’d like you to compute how much of stroke is attributable to the interaction of hypertension and OC use, then contrast this with what you found in a and b.
hope this helps,
JMC
