Applying quantitative econometric tools to assess investment in NIH (or any other investment) is in itself not unreasonable, but is potentially dangerously misleading for non-experts, including media, the public, and policy-makers.
This study seems a poster child for valuing precision of estimates and parameters, over other more important measures. Most informed people would agree that there are worse outcomes than death. Problems like autism, schizophrenia, and severe intractable chronic pain are largely ignored in these computations.
Authors attempt to justify their use of YLL, but while this is reasonable for initial exploration of the possible value of the approach, the necessary qualification is far less apparent and conspicuous to the reader than I think it should be.