No. If you're 30, they might (or might not) group you with other 30 year olds, if that's important, and run this model against you all as a group. But you don't punch in "he has 50 years left to live" in the model. The model doesn't know any of that, it just knows your transaction history, and it works better when the population of histories are more similar than dissimilar. Though it's also very effective when run against a large population too.

Basically, these models can pretty much predict whether you are pathological or good one. I know that models also include things like if you are a pet owner, number of pets, siblings, number of disputes in cc, etc.