That's a good point -- so it's not all about survival rate.

Yvain's point doesn't prove that conclusion*. A baby who has a low chance of surviving the year is going to have some parents going "If zhe does survive, what then? I mean, will zhe ever recover, what if it recurs? I can't keep going through this, especially if it's just gonna end up with me grieving in 5 years time!"

*(I still tend to agree with the conclusion, because of things like your second paragraph, but Yvain's point is actually somewhat irrelevant to the conclusion)

They've experienced <1 year of the baby, and there's been at least 1 near-fatal occurence; that's a high, and worrying, frequency. The image that gives is that, by age 12, the child will have been through several (>4) such events. Too many.

A 12-year-olds parents have had 12 years, and ONE nigh-fatal incident. The child is >50% of the way to reproductive age. That's a very different situation.

Maybe it's about how easy it is to imagine that the child (or fetus) never existed? It's pretty damn obvious that a twelve-year-old existed. A baby, less so, and a fetus, hardly at all.

I suspect this plays a large part, combined with the sunk cost fallacy.