Only iff you can be confident that the occurrence of this trigger word means that a person indeed experienced stillbirth, and wasn't just researching this, reading up on it because of fear, a non-native speaker looking up the definition of the word, etc. Metrics like this are really only weakly correlated with what (you'd think) they're trying to determine. Even targeted advertising is essentially a shotgun approach, as we don't even know what actually makes a person click on an ad.
Also: as long as (probability of losing an user X their expected lifetime value) is lower than (probability of ad hitting a gullible person X money got from the ad), it's more profitable to just display the ad. People all over the thread seem to think that it's only about losing vs. not losing a user. But there's opportunity cost to not losing a user, and frequently it's judged not worth it.
(It's good the author is raising a stink over this, it shifts the ROI calculations a little bit towards being more humane.)