The objective of this study is to investigate whether hospitals known
to be good places to practice nursing have lower Medicare mortality th
an hospitals that are otherwise similar with respect to a variety of n
on-nursing organizational characteristics. Research to date on determi
nants of hospital mortality has not focused on the organization of nur
sing. We capitalize on the existence of a set of studies of 39 hospita
ls that, for reasons other than patient outcomes, have been singled ou
t as hospitals known for good nursing care. We match these ''magnet''
hospitals with 195 control hospitals, selected from all nonmagnet U.S.
hospitals with over 100 Medicare discharges, using a multivariate mat
ched sampling procedure that controls for hospital characteristics. Me
dicare mortality rates of magnet versus control hospitals are compared
using variance components models, which pool information on the five
matches per magnet hospital, and adjust for differences in patient com
position as measured by predicted mortality. The magnet hospitals' obs
erved mortality rates are 7.7% lower (9 fewer deaths per 1,000 Medicar
e discharges) than the matched control hospitals (P = .011). After adj
usting for differences in predicted mortality, the magnet hospitals ha
ve a 4.6% lower mortality rate (P =.026 [95% confidence interval 0.9 t
o 9.4 fewer deaths per 1,000]). The same factors that lead hospitals t
o be identified as effective from the standpoint of the organization o
f nursing care are associated with lower mortality among Medicare pati
ents.