End-of-life Health Spending

End-of-living spending is often tagged as a particular source of inappropriate health care, particularly in Medicare.  Most studies estimate that while only about 5% of Medicare beneficiaries die in a year, they account for 21% of spending.  The implication is that a lot of this spending must be wasteful, since the people ended up dying.  A study reported in the journal Science debunks this notion.   (Science Article)   The obvious, commonsense fact, which apparently eludes many academic researchers, is that people who die are very sick, and whether they die in a particular year or not, they are going to incur substantial medical spending.  This is a “backfilling” method.  If instead you take all spending for all beneficiaries for a year and look at how much of it went to people who died, the ones who died only account for 15% of all spending.   The authors attempt to further eliminate the effects of knowing who did die and their costs by developing an algorithm that attempted to predict likely mortality and then look at spending.  The beneficiary sample is divided into a development group, used by the computers to develop a prediction algorithm, and a test group to generate results.  The algorithm shows how hard it is to figure out who is likely to die.  Less than 10% of the beneficiaries who do die in a year had a predicted likelihood of dying of over 50%.  Now maybe it is just a bad algorithm, which the authors acknowledge as a possibility, but I don’t think so because it comes up with the same answer as other similar algorithms.  The importance of this is apparent, because if you don’t have a good sense of who is going to die, how can you decide that you ought to provide less care for them?  I don’t think someone who is really sick is eager to have medical personnel saying, well, you might die so we aren’t going to waste money on health care for you.  The study confirms that what is really tied to high spending is being very sick.  Duh.  But very sick people don’t always, or even frequently, die.  So the idea that we can save a lot of money by limiting care to people about to die is nonsense, unless and until we have a much better way of identifying people who might die.