The longer term answer will be 'better gatekeeping' (how can it not be?). But right now the tolerance for throwing the baby out with the bathwater is understandably low. You get to bias towards Type 1 or Type II statistical errors, pick one, but keeping an eye towards proportionate harm as well as historical harm is going to be wise (and hopefully encouraged).
The harms of throwing people off of Medicare are vastly higher than the damage of a few fraudsters. The patients aren't to blame, and shouldn't be victimized by the 'fix'.