I recently wrote up a short report with Toby Ord on how we ought to deal with future gains to health when doing cost-benefit analysis for Giving What We Can, and in our other projects. Typically, cost benefit analyses discount future costs and gains at somewhere between 3-7% every year into the future. We propose that while this makes sense for financial assets, it is quite inappropriate for dealing with health improvements and other similar welfare gains. The main reasons put forward for using a discount rate to lower the importance of the future relative to the present are:
Opportunity costs are a sound reason to discount the value of an organisation's financial returns, because they can always be invested and turned into a larger amount in the future. However, this does not apply to health, quality of life, or financial gains which go to other parties. A person who is healthier earlier in time might be able to use that health to earn extra money sooner, and invest it for longer.  However, any such return will go to that individual, rather than the charity or government deciding how to spend their money to deliver the greatest benefit. This failure of program recipients to reinvest returns in valuable activities rather than short-term consumption spending is noted in Poor Economics (2012) as a prime reason why programs haven't lifted people out of poverty at the expected rates. To account for this, such benefits should be treated quite differently. Other benefits to quality of life may not even be possible to reinvest in this way in principle.
Another reason to discount is that far future benefits are more speculative, and changes to the world in the meantime can disrupt your project or make it irrelevant. For example, a vaccine development project that hopes to deliver a vaccine in a few decades faces a higher risk of being defunded or the disease in question disappearing, than does a similar project that expects to deliver a vaccine in a matter of years. This is a good reason to discount future benefits and costs, but the appropriate rate will vary dramatically depending on what you are looking at, and will not necessarily be the same every year into the future.
A third reason offered is that in practice we must avoid infinities, or at least very large numbers, in our results. The details are in the full version, but the bottom line is that a) a proper model will not usually produce infinities or very large numbers b) where one does, this is actually informative, and finally c) a discount rate does not necessarily prevent enormous numbers anyway, for instance if population growth exceeds your discount rate.
The final reason offered is that we should apply a discount rate to account for 'time preference': things in the future are inherently less important. This is the hardest point to debate because people do not agree on the appropriate method for settling disputes on such moral questions. However, Toby and I strongly disagree with discounting due to pure time preference. Some of the reasons we offer to doubt the wisdom of this approach are:
While you may not find all of these points persuasive, I hope that they collectively will give you reason to doubt that pure time preference is appropriate, especially at the high rate at which it is often applied today. Our approach would be to apply a discount rate to financial assets in the normal way, model the opportunity costs and uncertainty that applies to non-financial returns explicitly, and not apply any rate of time preference. As always, I look forward to revising our position in response to your constructive feedback.
 Health earlier in life does allows you to 'invest' in your future health through exercise and so on, but this is a small effect, and only applies within the same life, not between lives.