Did a high-profile program really slash hospital spending? Or was it a cautionary tale of 'regression to the mean'?