Today John Halamka pointed me to what I was looking for. The explanation comes from Don Berwick:
Hospitals grab at least $1 billion in extra fees for emergency room visits | The Center for Public Integrity:
... Dr. Donald Berwick, the immediate past administrator of the Centers for Medicare and Medicaid Services (CMS), which administers the Medicare program, said a small portion of the billing increase is likely caused by outright fraud, but in the majority of cases hospitals are legally boosting profits by targeting the vulnerabilities of Medicare’s payment system. “They are learning how to play the game,” Berwick said about the hospitals....
... Berwick, the former CMS head, said patients haven’t changed. What’s changed is the aggressiveness of how hospitals bill. “They are smart,” Berwick said. “If you create a payment system in which there is a premium for increasing the number of things you do or the recording of what you do, well, that’s what you’ll get.”...Don't be fooled by his background leading CMS. Berwick has a long record in health reform, and an unimpeachable reputation. He got one year in CMS before the GOP got rid of him. He's telling the truth.
The deeper story goes like this:
- In the 1990s reformers tried to come up with a fair way to reimburse for the work physicians did, particularly 'cognitive' work vs. procedural work. In part they wanted to to equalize the playing field between medicine and surgery. This was a horrendous task even before the AMA got their hands on it.
- By the time the AMA was done a new kind of accounting system was created to track what doctors did. It was called "Evaluation and Management" coding, which looks to the uninitiated like a set of 10 or so "CPT Code" (also AMA controlled).
- This introduction of E&M codes changed medicine -- for the worse. Immediately. I won't bore you with the details, but basically doctors worked to the accounting system instead of focusing on improving patient care. Accounting matters.
- Four years later proceduralists complained and the E&M codes got much worse. At this point they were almost impossible to understand. There was supposed to be a usability test but it never happened. Somewhat better, but even more complex, codes were stalled in 1999.