Determining the number of people who die in a hospital or whose death may have some connection to hospital care seems like it could be fairly straightforward, but it is actually somewhat controversial. With the advent of pay-for-performance and value-based purchasing, hospital mortality measures have assumed substantial financial import. Researchers take a look at two primary approaches to determining hospital mortality in a study published in the Annals of Internal Medicine. (Annals Article) Specifically, the researchers wanted to know if there were meaningful differences in rates of death in a hospital and for the period of time including 30 days after hospitalization. The looked at deaths for Medicare fee-for-service patients with heart attack, heart failure and pneumonia diagnoses for the period 2004-2006. As is usually the case, the rates were risk-adjusted for a variety of factors to assure maximum comparability.
Particularly in the Medicare system of DRG reimbursement, hospitals may have incentives to either transfer patients or attempt to minimize their stay, so mortality calculations that look only in-hospital mortality rates may be misleading. For example, while lengths of stay for these diagnoses have decreased and in-hospital death rates have declined, rates of death for the 30 day period after discharge have increased. The authors compared that kind of in-hospital calculation with one that included the 30 days after hospitalization. It could be argued that hospitals have little responsibility for events occurring after the patient leaves the hospital, but mortality following a hospitalization may at least partly reflect care in the hospital, as well as discharge and followup planning. Consistent with other prior research the researchers found that there were both absolute and comparative differences in using the two methods.
The average in-hospital and 30-day mortality rates were 10.8% and 16.1% for heart attack, 5.2% and 11.2% for heart failure and 6.4% and 12.2% for pneumonia. So there was a significant difference between the two calculations. From 8% to 15% of the hospitals had a different performance classification for the two different measures. And lower length-of-stay was associated with lower in-hospital mortality but not lower 30 day mortality. It would appear that it would be best not to use the in-hospital mortality rate or to in some way convert that rate to a per day statistic. It also may be that the DRG reimbursement method is encouraging hospitals to move patients out, in some cases to the detriment of their outcomes.