I alluded to this earlier today, but just to be clear: I'm glad that CBO is willing to do long-term budget forecasts, but in most cases they're not worth a whole lot.
At least with Social Security, we're only dealing with basic demographic stuff: birth rates, mortality rates, immigration. Still, there are huge error bars involved; policy, economics, and technology could easily make those projections wind up way off by changing the basic demographics.
But with Medicare, we're not just talking about those things: we're also talking about all the things that could affect health care over a long period of time. Kevin Drum pointed out one side of that yesterday: perhaps technological breakthroughs will make medical costs in the coming decades sort of like computer costs in the last decade. Could happen! On the other hand, maybe medical costs will continue spiraling upwards. And maybe we'll all start living to 150 and spend the last half of that consuming massive amounts of medical care. Who knows? Don't forget that "medical costs" include a whole bunch of only marginally related markets, whether it's for drugs, or labor (doctors, nurses), or insurance, or medical devices, or hospitals...we really don't know how any of that is going to change in the long run.
In my view, projecting any of this beyond, say, twenty years is just pure guesswork. Yes, it's better to have the guesswork than not, but I'd be very, very cautious about employing major policy changes based on those guesses.
The other half of this is that to the extent the health care cost problems really are important over the long run (and, indeed, I think they probably are) it's really a health care problem, not primarily at least a federal budget problem (as CEPR's Dean Baker explained yesterday). Now, if Paul Ryan's plan to convert Medicare into a voucher program yielded large savings in overall health care costs -- not just federal budget costs -- then we would have something to talk about. But there seems to be a lot of evidence that it won't.