It is useful to try to quantify processes, even when quantification is challenging. Or even when it is impossible with a reasonable degree of accuracy.
We know that longevity, assuming a "regular" life (no 2 packs of cigarettes a day, no handle of vodka a week, moderate exposure to the sun, some exercise, etc.), is strongly influenced by our genes, over which we have very little control, so far.
Assuming, for a 60-year-old woman, a gene-determined baseline probability of reaching age 90 in good health of 70% -- which, to simplify things, we take as observed frequencies, i.e., out of 1,000 women with the same set of genes, 700 reach age 90 -- how much behavior or activity x, y, z would move the needle in one direction or the other?
Since each man or woman is an island, we cannot know. But you can start thinking and pulling out reasonable numbers, based on observational studies, some intuitions, some quasi-experiments.
And I would say that continuing to work until 90 can move the needle in one direction or the other by a maximum of 5% (so, taken the 70% as baseline, we get 65-75%, because for some, like me, working is not exactly hell but at least purgatory, for others it might be the "purpose" that keeps them "alive").
Let's assume that for the woman taken on as a study subject in this comment, continuing to work for, say, 15 hours a week until the end of her life would increase her life span by two years (from 88 to 90 or from 90 to 92). But this entails other costs, less time spent with her husband or friend or children or grandchildren or dogs, which may be costs or benefits to her, since she does not like young children very much after all.
Thinking in terms of probabilities, costs and benefits is always helpful.