00:01
There are lots of reasons why it can be very difficult to measure health outcomes.
00:05
One of the basic ones is something called survivor bias.
00:10
And this is where if you were to survey a group of people and ask them, how did you fare in a certain situation, the only people available to survey are people who made it out of that situation.
00:23
The famous example is if you surveyed every soldier coming home from world war ii and asked them, did you die in the war? obviously, everyone says no.
00:32
And then the conclusion you draw from that is, well, there were no fatalities in the war.
00:38
So in the same way, you can have a survivor bias, if you will, when you're trying to measure health outcomes.
00:46
Another one is trying to actually develop a good way to have counterfactuals.
00:54
What are counterfactuals? that's essentially when you ask the question, what if we didn't do x? what if we did y instead? what would have happened? well, that's essentially a guess.
01:06
There's no way to prove what exactly would have been the other outcome.
01:11
For example, if i can give you drug a or drug b, and i give you drug a, we have results from that.
01:18
But then i want to measure what would have happened with result b.
01:22
Well, we don't know that because we didn't do it.
01:25
So now we're guessing as opposed to observing an actual fact.
01:30
Another example is that everyone's different, right? so because of that, the results you have may not be the results someone else has, but both could be optimal.
01:49
In other words, let's say you recover from an illness in five days, whereas i take six days to recover...