Introduction Secondary triage at a major incident allows for a more detailed assessment of the patient. In the UK, the Triage Sort (TSO) is the preferred method, combining GCS, systolic BP (SBP) and RR to categorise Priority 1 casualties. The Shock Index (SI) is calculated by dividing HR by SBP (HR/SBP). This study examines whether SI is better at predicting need for life-saving intervention (LSI) following trauma than TSO.
Methods A prospective observational study was undertaken. Physiological data and interventions performed in the Emergency Department and operating theatre were prospectively collected for 482 consecutive adult trauma patients presenting to Camp Bastion, Afghanistan, over a 6-month period. A patient was deemed to have required LSI if they received any intervention from a set described previously.
Results Complete data were available for 345 patients (71.6%). Of these, 203 (58.8%) were gold standard P1, and 142 (41.2%) were non-P1. The TSO predicted need for LSI with a sensitivity of 58.6% (95% CI 51.8% to 65.4%) and specificity of 88.7% (95% CI 83.5% to 93.9%). Using an SI cut-off >0.75 provided greater sensitivity of 70.0% (95% CI 63.6% to 76.3%) while maintaining an acceptably high (although lower than TSO) specificity of 74.7% (95% CI 67.5% to 81.8%). At this SI cut-off, there was evidence of a difference between TSO and SI in terms of the way in which patients were triaged (p<0.0001).
Discussion Our study showed that a SI >0.75 more accurately predicted the need for LSI, while maintaining acceptable specificity. SI may be more useful than TSO for secondary triage in a mass-casualty situation; this relationship in civilian trauma should be examined to clarify whether these results can be more widely translated into civilian practice.
Project registration number RCDM/Res/Audit/1036/12/0050.
- Accident & Emergency Medicine
- Education & Training (see Medical Education & Training)
- Trauma Management
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
- Accident & Emergency Medicine
- Education & Training (see Medical Education & Training)
- Trauma Management
The Triage Sort (TSO) is taught by MIMMS for the purposes of secondary triage at Major Incidents. We find that it has a sensitivity of 58% at predicting a priority one patient as defined by the need for Life Saving Intervention (LSI).
The Shock Index (SI), calculated by diving the patient's heart rate (HR) by systolic blood pressure (SBP), has a sensitivity of 70% at predicting a priority one patient.
The increase in sensitivity gained by using the SI over the TSO is statistically significant, p<0.0001.
Triage is not a new concept. It has been developing since the Napoleonic Wars, and has undergone several different incarnations: initially, a senior surgeon would separate off those who either had trivial wounds or unsurvivable ones, largely on the basis of anatomical injury pattern, to focus on those who needed intervention to be saved.1 ,2 More recently, during the Falklands Conflict, a combination of experienced review of the anatomical injuries alongside physiological measurements was used.3 Triage has, on occasions, been used to identify those patients who could be treated and then returned to fight in the battles to maintain the fighting strength. After the bombing of a British military medical facility, two Emergency Physicians designed the Major Incident Medical Management and Support course, MIMMS, which formalised major incident management and set a UK standard for triage; a system based on simple physiological parameters that can be taught to minimally experienced clinicians, and aims to identify those patients most in need of medical attention.4 MIMMS teaches a two-stage approach: primary triage using the Triage Sieve is conducted at the incident scene; secondary triage, performed at the casualty clearing station, or in a safe environment, uses the Triage Sort (TSO).5 The TSO allocates patients to a category based on a combined score of the GCS, Systolic BP (SBP) and the RR. Internationally, there are several other triage systems used, including the Simple Triage and Treatment (START) in the USA, and Careflight in Australia.
As such incidents are unpredictable, a prospective study is almost impossible to conduct. Studies looking at major incidents fall into two distinct groups, each with their own limitations. The first is a retrospective database analysis of actual major incident data (such as the 7 July 2005 bombings), which is hampered by the fact that accurate data capture at the scene of a disaster is often difficult.6 The second option is to use individual trauma registry data. The latter assumes that patient physiology will behave in the same way regardless of whether they present singly or as part of a cohort in a major incident. Only one study has collected this data prospectively and that was in a paediatric population.7 The TSO is derived from the Triage–Revised Trauma Score8 and was designed to detect those who would die from their trauma, as these were thought to represent those who must be sickest. This may not actually be the best basis for triage, as Wilson, and over a century later, Baxt pointed out,2 ,9 as there is no point directing resources to those who cannot be saved. Other work has aimed to identify those with an Injury Severity Score (ISS) >15 (often used as the definition of major trauma) as this group correlates well with mortality, but also contains most patients who require significant specialist intervention to achieve the best outcomes.10 It is almost a given in such circumstances that ideal care may not be achieved, and the priority must be to save life. Consequently, Baxt introduced the concept of requirement for a life-saving intervention (LSI), and subsequently authors such as Garner, Wallis and Horne further developed this definition.9 ,11–13
Alternatives to the BP have been looked at in the search for the most accurate, yet simple cardiovascular assessment of significant trauma. With physiological variables in isolation having limited predictive value, a combination of variables has been looked at instead.14–16 Shock Index (SI), calculated by dividing a patient's HR by their SBP (HR/SBP), was first described in 1967, and has been discussed in the literature for many years, with strong links between ISS >15, massive transfusion requirement and mortality.17 ,18 Even with apparently stable vital signs, an elevated SI has been shown to be associated with critical illness and shock.19 The literature shows strong correlations between a SI of 0.9 and 1.4 and mortality and ISS >15. More importantly, in the context of LSIs, it has also been shown to correlate with requirement for massive transfusion and intensive care requirement.20 ,21
The aim of this study was to identify whether the SI was a suitable alternative to the TSO for the purposes of secondary triage. The objectives were to identify the sensitivities and specificities of various cut-offs of SI, allowing for the identification of the optimum SI cut-off, and to compare this against the TSO at predicting need for LSI.
Physiological data (prehospital when available, and on arrival in hospital), and interventions performed within the emergency department (ED) and operating theatre, were prospectively collected for consecutive adult (>18 years) trauma patients presenting to the ED at Camp Bastion, Afghanistan, between March and September 2011, and who met trauma team activation criteria (Table 1).
A data sheet (shown below) was used to record the hospital number, date of injury, physiological parameters (RR, HR, GCS), presence of palpable pulse and SBP. Glasgow Motor Score (specifically the ability to obey commands) was also recorded. Injury mechanism was not recorded (Table 2).
The gold standard used to define a Priority One (P1) patient was requirement of one or more LSI from a predefined list or death within the department. This list was derived through a modified Delphi process of deployed Military Consultants.13 The interventions are listed in Figure 1.
Only patients where data was entered for interventions undertaken (including ‘none’) were included in the study. This included prehospital and in-hospital interventions. All patients receiving a LSI, or who died in department, were classified as P1, and all others as non-P1. Timing of interventions was not recorded. For surgical procedures, the authors (JV and SH) determined case-by-case which were time-critical.
The TSO was applied to the in-hospital physiology, dividing casualties into P1 and non-P1. Unless a GCS was specified, patients were assumed to have been intubated for a reduced conscious level. Post-intubation RRs were not used for calculation of the TSO. Similarly, the SI, using in-hospital physiology again, was used to categorise patients as either P1 or non-P1, using a range of SI cut-off values (0.4 to 1.0 in 0.05 increments). In order to allow a comparison of classifications using TSO and SI to be made, only patients for whom the TSO and SI could be calculated were included in the analysis.
Triage of patients using the TSO and the SI was compared with the gold standard, and sensitivities and specificities were estimated with 95% CIs. Receiver Operator Curves (ROCs) were calculated for the SI cut-offs, from which an appropriate SI cut-off yielding acceptable sensitivity and specificity was identified. A 2×2 table was generated to compare this triage tool to the TSO and a McNemar test was applied to test for a difference between the tools (Figure 2).
The study was registered as a service evaluation with the Royal Centre for Defence Medicine (project number RCDM/Res/Audit/1036/12/0050). The data collected also included variables for assessment of the prehospital triage tools as part of a separate study.
Of the 482 patients who presented to the emergency department during the study period, 33 were excluded due to the authors (JV and SH) being unable to determine either P1 or non-P1. Of the remaining 449, there was sufficient hospital physiological data to allow classification to P1 or non-P1 using the TSO, and SI was recorded for 345 (71.6%). Of these, 203 (58.8%) were P1 and 142 (41.2%) were non-P1.
The TSO correctly identified 119 P1 and 126 non-P1 patients, yielding a sensitivity and specificity of 58.6% (95% CI 51.8% to 65.4%) and 88.7% (95% CI 83.5% to 93.9%), respectively (Tables 1 and 3).
P1 patients had a median SI of 0.93 (IQR 0.71–1.26) compared with 0.61 (IQR 0.52–0.75) in the non-P1 group.
Sensitivities and specificities for classification of casualties to P1 or non-P1, for different SI cut-offs, ranging from 0.40 to 1.00 are given in Table 2, and presented graphically in Figures 1 and 2 (Table 4).
Triaging based on a SI cut-off value of 0.75 provided greater sensitivity than that achieved by the TSO (70.0% vs 58.6%), while maintaining an acceptably high (although lower than TSO) specificity (74.7% vs 88.7%) (Figures 3 and 4).
Comparison of the TSO and SI (SI >0.75 meaning classification as P1) as triage tools using McNemar's test showed strong evidence of a difference between these tools (p<0.0001) (Tables 3 and 5).
In situations where many casualties present simultaneously and resources are limited, clinicians need to be able to triage effectively. Clinical acumen alone has been shown to be insensitive at predicting serious injury—potentially as low as 47%.22 The TSO (widely used in the UK for second-line triage at casualty clearing stations and hospitals) only has a sensitivity of 59% for those patients needing LSI in this military population. The SI was considerably more sensitive than the TSO (70.0% vs 58.6%) at a cut-off of 0.75, although with a lower specificity (74.65% vs 88.7%).
There are clear limitations to this study—it was conducted using patients who have largely sustained either blast or penetrating trauma in a military operational environment. These findings may not translate to a civilian population suffering largely blunt trauma injuries. There are also issues studying a military population on deployment. There is a preponderance of young males, and levels of fitness are high, but the trauma may be compounded by other variables, such as dehydration. It is unlikely that these findings will wholly translate to the civilian population, especially as SI has been shown to be slightly less reliable in the elderly.23 Incomplete or ambiguous data capture resulted in 137 patients (28.4%) being excluded. The authors collecting the data largely attribute this to periods of high intensity because of operational tempo impeding data collection.
Second, SBP was not recorded prehospital, and the TSO and SI were calculated using data captured following arrival in the ED. The authors recognise that some patients included could have had an LSI performed by the medical emergency response team (MERT), which may not have been captured. If the casualty was intubated on MERT, then this was picked up by this data collection system. Blood transfusion criteria, however, might be missed—the amount transfused in the hospital could be too low to trigger the gold standard if the patient had already received large amounts of blood products prehospital. However, it is unlikely that a patient will have required a massive transfusion prehospital without need for one of the other LSI on arrival.
The use of a list of LSI as a gold standard is still open to debate. While the concept is well accepted, the exact definition of ‘life-saving’ is still contested. The list used was derived by expert consensus from a group of clinicians with extensive military trauma experience,13 and so, realistically represents interventions that might be appropriate during a major incident. The use of individual patients to test a major incident triage tool has been challenged in the past, as the tool is not being used in the circumstances for which it was designed. However, it makes no biological sense that the physiology of consecutive patients from individual incidents should be different from the physiology of patients with the same injuries who happen to present all at once. There is, of course, no way within this study to assess the impact of resource limitation (eg, number of surgeons available to operate) on triage categorisation, but this limitation is currently common to almost all triage systems. While one system24 has attempted to build-in these constraints to tailor triage to the incident, there is no military equivalent at present. The reduction in specificity may be an issue in medical treatment facilities where the capacity is low; while a more sensitive tool will identify more of those needing intervention, dropping specificity means that the number of false positives alongside those genuinely in need will also increase. In the context of the numbers of casualties historically generated and the size and proximity of UK hospitals, we feel that this trade-off is still favourable. The specificity would be decreased still further if the same system were applied to all casualties whether walking or not. By identifying only stretcher cases, we have excluded the P3 walking wounded category from our calculations. However, these would preferably be managed as a separate stream at the casualty clearing station and the hospital. In these P3 areas, it is reasonable.
Previous studies looking at SI in civilian cohorts have consistently identified correlations with need for some LSI (eg, blood transfusion).25 However, most previous work has tended to focus on higher cut-offs, such as 0.9 or 1.0.26 While correlations with intervention are strong in this range, the cut-offs are likely to be too high to represent a meaningful threshold for intervention, especially in a multiple casualty scenario—in one series, the mortality of patients with SI >1.0 was 40%.17 Historically, triage research (including the work by Champion that led to the TSO) has focussed on highlighting those who will die—ultimately a poor use of critical resources. In one population, SI of >1.1 was the optimal predictor of death, but for major trauma (ISS >15) it was 0.71, for ICU ≥1 day it was 0.77, and for blood transfusion ≥2 units it was 0.85.21 A recent large Australian study of civilian blunt trauma patients showed that of those with a SI between 0.6 and 0.8, 15% still required transfusion in the first 4 h.27 Our data supports a cut-off around 0.75 as being the ideal combination of suitable sensitivity (70.0%) and acceptable specificity (74.6%) for any LSI.
SI is harder to measure than the other commonly used variables, such as HR. Most people would need a calculator to work it out, unless they have preprinted sheets with a table of SI values for given HRs and BP. This limits its effectiveness as a forward triage tool. However, at a Casualty Clearing Station, or at a hospital awaiting an ASHICE/ATMIST call, it could prove beneficial. It is also reasonable to expect that, if adopted widely, manufacturers of monitors would be able to show a calculated SI alongside BP measurements.
These findings show that in a military population, the SI is a more appropriate secondary triage tool than the existing UK method, the TSO. A cut-off of 0.75 predicts the need for a LSI with a sensitivity of 70.0%, vs 58.6% for the TSO.
Contributors JV designed the study, analysed the data, and produced the first draft. SH collected the data prospectively for the purposes of validating a primary triage tool. He subsequently reviewed and edited the first and second drafts. SB performed the statistical analysis for the work and reviewed each draft. JES reviewed and edited the second draft.
Competing interests JV, SH, JES are all serving members of the HM Armed Forces.
Ethics approval Data was collected prospectively with permission from ADMEM. The study was registered as a service evaluation with the Royal Centre for Defence Medicine (project number RCDM/Res/Audit/1036/12/0050).
Provenance and peer review Not commissioned; externally peer reviewed.