Page 45 - Journal of Special Operations Medicine - Spring 2014

difference was less than 1 percentage point, indicating even when the “true” survival probability curves are some-
that the ReSTART policy is relatively robust to different what different from what ReSTART and Simple-ReSTART
severity distributions. However, improvement in critical calculations assumed to be. The reader can find more de-
mortality clearly depended on the resource availability. tails on this study in the previously published paper. 8
As the number of ambulances increased, the improve-
ment from using ReSTART declined from a peak of
16.0% mean improvement in scenarios with three am- Discussion
bulances to a minimum 2.9% mean improvement in sce- The need to incorporate information on resource avail-
narios with 14 ambulances (Figure 6). abilities in determining patient priorities for treatment
or transportation in the aftermath of mass-casualty
2
Figure 6 Mean improvement in critical mortality using incidents has been recognized. For example, a recent
ReSTART versus START for scenarios with randomly research effort to standardize triage protocols resulted
distributed severities, by number of ambulances available. in SALT, a guideline that contains four parts: Sort, As-
sess, Lifesaving Intervention, and Treatment/Transport.
18 While there was largely consensus on the first three parts
Average improvement in mortality (percentage) 12 8 6 pectant), the problem of prioritizing patients for treat-
(one important agreement being on the need for triage
16
classes to include immediate, delayed, minor, and ex-
14
ment or transportation in the fourth step was largely left
10
as an open question, which the authors acknowledged
1
required more evaluation. To date, very few specific
ideas have been put forward to address this issue. This
tiple levels. The prioritization decision involves so many
0 4 2 is understandable, as this is a complex problem at mul-
2 4 6 8 10 12 14 16 variables that it is very difficult for a human being to
Number of ambulances
factor in all the details when deciding how to prioritize
patients. A human mind is simply not well equipped for
such a task. On the other hand, mass-casualty events
Similar results were obtained for Simple-ReSTART, are not well-structured events. Each event has its unique
which also provided significantly lower critical mortal- characteristics and emergency response teams typically
ity than START (paired t test, p < .01). Mean decrease have to deal with unexpected developments that derail
in critical mortality was 8.1% for high-acuity distribu- attempts for organization. This makes it very difficult if
tion (95% CI 7.8% to 8.4%, overall range –5.8% to not impossible to model the problem at a level of detail
21.2%), 8.6% for uniform distribution (95% CI 8.2% that is sufficient to rely purely on mathematical real-
to 8.9%, overall range –6.3% to 21.2%), and 8.1% for time solution methods. Therefore, the best chance to
low-acuity distribution (95% CI 7.8% to 8.4%, overall improve the current practice is with a method that uses
range –6.2% to 21.3%). Note that in addition to the a mathematical approach in a way that also recognizes
slightly lower average improvement, the main disadvan- the need to involve human decision makers due to op-
tage of using Simple-ReSTART as opposed to ReSTART erational realities. This is the basic principle behind the
is that in the small number of scenarios where ReSTART development of the prioritization model and ReSTART.
is outperformed by START, Simple-ReSTART results
in an increase in critical mortality compared with Re- The developers of the Sacco Triage Method made an
START. A closer examination of these outliers revealed important contribution by introducing mathematical
that these scenarios tended to be those that fall some- model ing and optimization as a tool to improve pa-
where close to the line that divides the two regions in tient prioritization decisions. We incorporate some ideas
4
Figure 4. Otherwise, the results from Simple-ReSTART from the Sacco Triage Method’s formulation, namely the
are similar to those from ReSTART; in particular, the decline in survival probability with the passage of time.
effect of resource limitations on the percentage improve- However, the Sacco Triage Method has a number of criti-
ment in critical mortality is structurally similar. cal limitations, some of which have already been discussed
in the literature. We believe that the fundamental prob-
While we do not provide details in this paper for brevity, lem with the method is its overreliance on the solution
it is important to note that our sensitivity analysis revealed to a complex mathematical program coupled to a very
that the good performances of ReSTART and Simple- granular patient risk stratification scheme. The developers
ReSTART are fairly robust. This is because we found that make the reasonable argument that START’s use of only
the expected number of survivors is statistically larger un- two different classes for critical patients (immediate and
der ReSTART or Simple-ReSTART than it is under START delayed) does not provide much discriminatory power to

ReSTART: Resource-Based Triage in Mass-Casualty Events 37