Medicare is now reporting actual risk-adjusted mortality rates for pneumonia, MI, and heart failure. The topic must be important because NPR’s "Talk of the Nation" spent 30 minutes interviewing Don Berwick and me about it — on the day of Hillary’s speech nonetheless!
Here are a few observations about the new Centers for Medicare & Medicaid Services initiative, some of which I made on the NPR broadcast:
1. First, I see this as a healthy and inevitable trend. We are moving away from a singular focus on process measures – which have real advantages (no need for case-mix adjustment; they can be measured at the time of care) but are too narrow and game-able – and toward blending in reports of outcomes. There are now two kinds of outcomes on the CMS site: the mortality rate reports join patient experience surveys, which were added earlier this year.
2. Second, the science of case-mix adjustment, while still imperfect, is finally good enough for government work (which is, after all, what this is). The brains behind the methodology come from a team led by my old UCSF colleague Harlan Krumholtz, a world-class cardiology outcomes researcher at Yale. And the science will drastically improve as we transition to computerized medical records, which will allow more of the data (both the outcome data and the case-mix adjustment variables) to be drawn from clinical notes rather than billing records.
3. Third, I think the researchers and CMS made several good decisions about how to run and present the reports. Such as…
- When a patient is transferred from one hospital to another (such as from a community hospital to an academic medical center like UCSF), any death is attributed to the first hospital. A decision otherwise would have created an incentive for higher-level hospitals to refuse transfers of patients who seem likely to die.
- The reported mortality rates are 30-day rates, not in-hospital rates. This does a couple of useful things. First, it removes an incentive for gaming the discharge decision for dying patients (some patients are best allowed to die in the hospital rather than being hustled out to a hospice or home for their last few days). More importantly, a 30-day rate causes hospitals to focus on post-discharge care, stepping up our heretofore-wimpy efforts to ensure a good transition to home. (Parenthetically, there is also a lot of interest in switching to “bundled payments” for “episodes of care” rather than DRG payments for the hospitalization only… yet another strategy designed to push hospitals to improve their discharge process). All of this is part of a macro-trend to de-silo the hospitalization and catalyze a new focus on post-discharge care. This is a healthy change.
- Finally, while reporting the raw mortality rates and their statistical precision, CMS chose to place hospitals in only three buckets: “average,” “below average,” and “above average.” This means that those who want “rankings” may be disappointed by the HospitalCompare data. For example, of 9 acute care hospitals in San Francisco, none had mortality rates for any of the three diseases that were deemed significantly different than the national average. Don called this conservative reporting strategy the product of “a little bit of timidity” on CMS’s part, but I was more supportive – I think it would be scientifically irresponsible to rank a hospital as “below average” when its results aren’t statistically different from the mean. (I’ll be less charitable if, over time, persistent outliers are not branded as such.)
From a user-friendliness standpoint, the site now includes an interactive map that allows you to drill in on hospitals in a city, and a pretty good compare function that allows one to contrast the performance of several hospitals at a time. We’re not quite at the Consumer Reports, “Let’s Compare the Lexus to the Infiniti,” level of sophistication and user interface, but it has gotten much more snazzy and accessible.
That said, I continue to believe that the real value of these reports comes not from consumer choice – most people are going to go to their local hospital, particularly in an emergency. Instead, the reports create tension for change inside hospitals – nobody wants to look bad, and nothing focuses a CEO’s or hospital board’s attention more than seeing publicly reported data that appears to indicate that you’re killing more people than you should. By “turning the lights on” (in Don’s words) in hospital care, the CMS mortality reports should promote more vigorous efforts to improve quality and safety.
I call that progress.