I'm not sure that's survivorship bias, since you can measure when dependents (or emancipated dependents) drop off from their support systems. The measure isn't just those that report back "I'm doing fine" but those that are tracked. There are a number of organizations that track these statistics, although I feel like the nordic countries seem to follow the strictest rigour. It's a null hypothesis to show that foster children (who are more likely to have mental disorders - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3922948/, et al) do better with caring parental figures. They can only degenerate without stability.