Buist et al have concluded in their paper that a medical emergency
team provides a 50% reduction in the incidence of and mortality from
unexpected in hospital cardiac arrest. [1] However the data presented do
not clearly support this assertion.
The authors have mentioned the Hawthorne effect, and quite clearly
temporal changes in health care provision over the 4 years will have
created a significant bias in the outcomes under study, no attempt to
account for these has been made. Comparisons with other similar hospitals
would have been advantageous, over corresponding periods of time, both
within Victoria and nationally. This data could have been used as a
predictor variable in the regression equation.
The patient population itself has clearly changed with significant
differences in diagnostic related group (DRG) codings and the method of
recruitment into the hospital population. With global trends in
decreasing length of stay, the data presented in table 1 showed a
significnat increase over the 4 years, however the marked difference in
standard deviation between the two groups (14.8 vs. 6.3)is either an error
in the table or evidence of a greatly skewed variation between the two
years under study. It is unclear wheter using the DRGs as a predcitor term
in the rergression equation is adequate to eliminate this confounder.
Temporal changes in overall health care would seem likely to account for
the change seen in in-hospital mortality, which are in some part the
'activities associated' with the medical emergency team alluded to by the
authors
The number of deaths increased during the study, but the number of
cardiac arrests decreased, the authors do not explain how the 346 in
hospital deaths in 1999 not due to unexpected cardiac arrest differed from
the 307 in 1996, though this difference is not significant. Further data
is also required to explain whether the team designating a patient as 'do
not resusciate', i.e. they died but not from unexpected cardiac arrest,
has had a large impact on the outcome. In 1999, there were 47 unexpected
cardiac arrest deaths, yet 40 patients died after the arrival of the team
and have been coded into other groups, a total of 87 deaths compared to 73
deaths in 1996. This differential misclassifcation bias seems to acount
for the whole impact of the change in the study outcome of deaths from
unexpected cardiac arrest between 1996 and 1999.
It is always dificult to prove that service improvements actually
improve something in a statistically recordable manner, without an
appropriate comparator group. Such a group is not provided in this paper
and leaves the authjors conclusions open to question.
{1} Buist MD, Moore GE, Bernard SA, Waxman BP, Anderson JN, Nquyen
TV. Effects of a medical emergncye team on reduction of incidence of and
mortlaity from unexpected cardiac arrest in hospital:ppreliminary study.
BMJ 2002;324:387-90
Rapid Response:
Comments on paper
Editor
Buist et al have concluded in their paper that a medical emergency
team provides a 50% reduction in the incidence of and mortality from
unexpected in hospital cardiac arrest. [1] However the data presented do
not clearly support this assertion.
The authors have mentioned the Hawthorne effect, and quite clearly
temporal changes in health care provision over the 4 years will have
created a significant bias in the outcomes under study, no attempt to
account for these has been made. Comparisons with other similar hospitals
would have been advantageous, over corresponding periods of time, both
within Victoria and nationally. This data could have been used as a
predictor variable in the regression equation.
The patient population itself has clearly changed with significant
differences in diagnostic related group (DRG) codings and the method of
recruitment into the hospital population. With global trends in
decreasing length of stay, the data presented in table 1 showed a
significnat increase over the 4 years, however the marked difference in
standard deviation between the two groups (14.8 vs. 6.3)is either an error
in the table or evidence of a greatly skewed variation between the two
years under study. It is unclear wheter using the DRGs as a predcitor term
in the rergression equation is adequate to eliminate this confounder.
Temporal changes in overall health care would seem likely to account for
the change seen in in-hospital mortality, which are in some part the
'activities associated' with the medical emergency team alluded to by the
authors
The number of deaths increased during the study, but the number of
cardiac arrests decreased, the authors do not explain how the 346 in
hospital deaths in 1999 not due to unexpected cardiac arrest differed from
the 307 in 1996, though this difference is not significant. Further data
is also required to explain whether the team designating a patient as 'do
not resusciate', i.e. they died but not from unexpected cardiac arrest,
has had a large impact on the outcome. In 1999, there were 47 unexpected
cardiac arrest deaths, yet 40 patients died after the arrival of the team
and have been coded into other groups, a total of 87 deaths compared to 73
deaths in 1996. This differential misclassifcation bias seems to acount
for the whole impact of the change in the study outcome of deaths from
unexpected cardiac arrest between 1996 and 1999.
It is always dificult to prove that service improvements actually
improve something in a statistically recordable manner, without an
appropriate comparator group. Such a group is not provided in this paper
and leaves the authjors conclusions open to question.
{1} Buist MD, Moore GE, Bernard SA, Waxman BP, Anderson JN, Nquyen
TV. Effects of a medical emergncye team on reduction of incidence of and
mortlaity from unexpected cardiac arrest in hospital:ppreliminary study.
BMJ 2002;324:387-90
Competing interests: No competing interests