You are on page 1of 7

The “MERIT” Trial of Medical Emergency Teams in Australia:

An Analysis of Findings and Implications for the

100,000 Lives Campaign

100,000 Lives Campaign

We invite you to join a Campaign to make health care safer and more effective — to
ensure that hospitals achieve the best possible outcomes for all patients. IHI and
other organizations that share our mission are convinced that a remarkably few
proven interventions, implemented on a wide enough scale, can avoid 100,000
deaths between January 2005 and July 2006, and every year thereafter. Complete
details on the web at http://www.ihi.org/IHI/Programs/Campaign/.

This document is in the public domain and may be used and reprinted without permission provided
appropriate reference is made to the Institute for Healthcare Improvement.
The “MERIT” Trial of Medical Emergency Teams in Australia: An
Analysis of Findings and Implications for the 100,000 Lives Campaign

In the June 18, 2005, issue of The Lancet, the pioneering Australian clinical investigators
who first described in detail the use of “Medical Emergency Teams” (METs)—the
intervention that we call “Rapid Response Teams” in the 100,000 Lives Campaign—
report on a “cluster-randomised trial” of the MET system in 23 Australian hospitals. The
study attempted to determine “whether the MET system could reduce the incidence of
cardiac arrests, unplanned admissions to intensive units, and deaths among patients in
general hospital wards.” (MERIT Study Investigators. Introduction of the medical
emergency team (MET) system: a cluster-randomised controlled trial. The Lancet. 2005;
365:2091-2097.)

This is a bold study, and a highly commendable example of an attempt at systematic,


scientific evaluation of a complex socio-technical innovation. The study report is
extremely well-written, and both the report and the MERIT investigators performed a
cluster randomized trial, one of the strongest possible study designs to evaluate the
impact of quality improvement interventions at the hospital level. Their study design was
robust, and their analytic plan was sound and sophisticated. They considered and
adjusted for correlations in patient-level data and inter-institutional variability, and they
appropriately chose to perform an “intent-to-treat” analysis for their pre-specified
primary outcome (cardiac arrest, unplanned ICU admissions, and unexpected death). The
study report is extremely well-written and is refreshingly candid about potential
limitations of the work. This openness of the study team allows for significant lessons to
be learned from its ongoing work.

The researchers summarize their main findings as follows: “The MET system greatly
increases emergency team calling, but does not substantially affect the incidence of
cardiac arrest, unplanned ICU admissions, or unexpected death.”

Both prior to and after this publication, key IHI faculty and Campaign staff have been in
frequent communication with the MERIT study investigators to understand what they
have learned, what their current interpretations are, and what the implications of their
study may be for the 100,000 Lives Campaign. The following summarizes some of our
findings to date.

MERIT STUDY DESIGN:

• The study randomized 23 hospitals into two groups: 12 received training


in the MET process and were asked to implement the program; 11 did not
receive training and were asked to delay introduction of a MET system during
the study period.

• A baseline period of two months preceded a training and implementation


period of four months, followed by a study period of six months.

2
• Outcome variables assessed were (a) cardiac arrests (i.e., no palpable
pulse) without a prior “do-not-resuscitate” order, (b) unplanned ICU
admissions, and (c) “unexpected” deaths (i.e., deaths without a prior DNR
order). In addition, (d) a “composite” score was calculated as the sum of
these three events. For all four events, (a) - (d), the actual rate used was
“events per 1000 admissions” and the population studied included patients
admitted to inpatient general wards (including the coronary care unit, but not
the ICU, ORs, or EDs).

• Of the 23 hospitals in the study, 17 were teaching hospitals. Bed sizes


ranged from 182 to 457.

PRIMARY FINDINGS:

• During the study period, MET hospitals were more likely than control
hospitals to call emergency teams for help (3.1 calls per 1000 admissions vs.
8.7 calls; (p = 0.0001)).

• Mean crude event rates (per 1000 admissions) were as follows during the
baseline and study periods:

CONTROL CONTROL MET MET


HOSPITALS HOSPITALS HOSPITALS HOSPITALS
BASELINE STUDY BASELINE STUDY
PERIOD PERIOD
(A) Cardiac 2.61 1.64 1.60 1.31
Arrest
(B) Unplanned 5.29 4.96 4.68 4.19
ICU Admission
(C) Unexpected 1.61 1.18 1.65 1.06
Death
Any Primary 7.07 5.86 6.58 5.31
Outcome
(A, B, or C)

3
• In the overall study—among the 23 hospitals—unexpected deaths fell by
over 30% between baseline and study periods, as shown in the graphs below.
The graphs show three data points: the baseline period of 2 months; the 4-
month implementation period when 12 of the hospitals were prepared for
introduction of the MET system; and the 6-month study period.

6
Control Hospitals MET Hospitals
6

5 5
Unexpected Deaths

Unexpected Deaths
4 4

3 3

2 2

1 1

0 0

Baseline Impl Study Baseline Impl Study

• In patients with primary outcome events, recording of physiologic


variables included in specifications for triggering MET calls was erratic. For
example, among patients with primary events, a record of vital signs—blood
pressure, heart rate, and respiratory rate—within 15 minutes before the event
was missing in 62% of cases.

• MET calls did not occur reliably when appropriate physiological triggers
occurred. For example, in patients who had unplanned ICU admission and
whose appropriate physiological variables were measured more than 15
minutes before the event and met criteria for triggering a MET call, MET calls
actually occurred only 9% of the time in control hospitals and only 30% of the
time in the MET hospitals.

• Although the number of primary outcome events was lower in MET than
in control hospitals (i.e., -0.26 events per 1000 admissions, a difference of
about 4% relative to the baseline rate of 6.82 events per 1000 admissions for
the combined hospital groups), the 95% confidence interval for this difference
was very wide (-2.449 to 1.921) relative to the baseline rate.

4
ANALYSES:

• The substantial decrease in unexpected deaths in both control and


treatment hospitals suggests that such a decrease can be achieved in a
relatively short period of time.

• The majority of patients did not have sufficient physiological monitoring


in the general medical ward setting to allow for reliable triggering of MET
responses.

• Although the MET hospitals did increase their rate of emergency calls,
such calls did not occur reliably, even when monitoring occurred and MET
criteria were satisfied. Thus, the study suggests incomplete implementation of
the MET system.

• The empirical rate of MET calls in the study hospitals—8.7 per 1000
admissions—is below the rate of calls that IHI faculty believe they see in
effective RRT implementation in our US hospitals, which is close to 15 calls
per 1000 admissions. In addition, the MERIT study’s authors tell us that the
original, effective MET implementation in two initially successful hospitals as
reported in prior Australian literature led to call rates of 23 and 25 per 1000
admissions—three times the rate in the MET hospitals in this trial.

• The investigators reported to us that most hospitals in Australia were


already moving toward the MET process when the MERIT study began,
making it difficult to find even 23 hospitals that had not yet implemented the
MET system. This information also suggests that even the control hospitals
may have begun before or during the study period to put some elements of the
MET system and awareness into place. In addition, several of the
experimental hospitals did not, in fact, implement the MET system during the
study period.

• Of the population of 23 hospitals, 17 were teaching hospitals, raising


questions about the applicability of METs in academic vs. community hospital
settings. Implementing METs in academic centers may take longer because of
the complexity of the academic environment, and may require different
strategies than in non-academic settings.

• The study design was powered to have a 90% chance of detecting a 30%
decline in the primary events, assuming a baseline rate of 30 events per 1000
admissions. The baseline event rate was far lower—6.82 events per 1000—
and a recalculation shows that 90% power to detect a 30% decline in outcome
events would have required 100 hospitals, rather than 23. In addition, the
100,000 Lives Campaign RRT target is a 10% reduction in deaths, a level of
improvement that the MERIT study had very low power to detect.

5
IMPLICATIONS FOR THE 100,000 LIVES CAMPAIGN:

• Despite the robust cluster randomized trial design of the MERIT study, the
investigators ran into number of issues that tend to arise in this type of study,
including (a) incomplete and inconsistent implementation of the MET process
in experimental hospitals, (b) high proportion of teaching hospitals in the
study, (c) possible unmeasured adoption of early *or rapid response system*
response processes by the control hospitals, (d) low rates of monitoring of
relevant physiological variables in general ward patients, and (e) extremely
low experimental power to detect improvements of the magnitude sought in
the Campaign (fewer events than expected, greater inter-institutional
variability in event rates than anticipated). These issues, which are discussed
candidly by the investigators, may have contributed to their inability to show a
significant effect of the MET intervention.

• Faced with these issues, it is appropriate—in fact, highly desirable—to


explore secondary analyses that might deepen understanding of the study and
develop hypotheses to inform future research. Such exploratory analyses are
especially important given the most stunning finding of the MERIT study: the
dramatic and statistically significant 30% reduction in mortality in
intervention and control hospitals over the short study period (see Figure).
Something happened in these hospitals that saved lives. Perhaps it was partial
implementation of MET-like behaviors in the control hospitals; perhaps there
were unmeasured interventions to reduce deaths. The Figure also
demonstrates a profound reduction in the variance in mortality rates by the
end of the study, precisely what would be expected with a successful quality
improvement intervention. While some of this reduction in variance may be
explained by the statistical issues, such as the longer study period compared to
the baseline period, this is a potentially important finding.

• Regardless of the reasons for the 30% reduction in mortality documented


in the MERIT study, the implications are clear: dramatic reductions in
mortality are achievable within the timeframe of the 100K Campaign.

• Although the MERIT study in The Lancet was technically negative in its
findings comparing treatment to control hospitals in an “intention to treat”
analysis, limitations in both the conclusiveness and the generalizability of the
MERIT study (that is, limitations in both internal and external validity) leave
us undeterred in our confidence that the RRT process, well-implemented, can
reduce in-hospital mortality. Prior published reports, though not of
randomized trials, in addition to ongoing time-series analysis from
participating hospitals, remain strongly supportive of the effectiveness of
RRTs in at least some circumstances.

6
• The implementation issues embedded in the MERIT study emphasize the
need for the 100,000 Lives Campaign hospitals to attend carefully to the
training, support, encouragement, and monitoring of the RRT process, itself.
Good target rates of RRT calls may be in the range of 15 to 25 per 1000
admissions to ensure proper uptake of the RRT process.

• Monitoring of patient status and vital signs among general ward inpatients
may be a vulnerable part of the RRT process. The RRT cannot be called if no
one is noticing the patients’ changing status. Classical monitoring methods
and schedules may not be up to the task of properly supporting and triggering
the RRT capability. This is an important area for further investigation and
development.

• The dramatic reduction of unanticipated deaths over time in all 23 subject


hospitals in the MERIT study suggests that changes of that magnitude are, one
way or another, actually achievable among inpatients in at least some
hospitals. Based on our experience and the literature as a whole, IHI believes
that RRTs are an important and highly promising intervention for reducing
deaths and other untoward events due to “failure to rescue”—that is, for the
prompt identification and treatment of patients whose clinical condition is
spiraling downwards.