Article Text
PDF
Abstract
Introduction Triage is a key principle in the effective management of major incidents. There is limited evidence to support existing triage tools, with a number of studies demonstrating poor performance at predicting the need for a life-saving intervention. The Modified Physiological Triage Tool (MPTT) is a novel triage tool derived using logistic regression, and in retrospective data sets has shown optimum performance at predicting the need for life-saving intervention.
Materials and methods Physiological data and interventions were prospectively collected for consecutive adult patients with trauma (>18 years) presenting to the emergency department at Camp Bastion, Afghanistan, between March and September 2011. Patients were considered priority 1 (P1) if they received one or more interventions from a previously defined list. Patients were triaged using existing triage tools and the MPTT. Performance was measured using sensitivity and specificity, and a McNemar test with Bonferroni calculation was applied for tools with similar performance.
Results The study population comprised 357 patients, of whom 214 (59.9%) were classed as P1. The MPTT (sensitivity: 83.6%, 95% CI 78.0% to 88.3%; specificity: 51.0%, 95% CI 42.6% to 59.5%) outperformed all existing triage tools at predicting the need for life-saving intervention, with a 19.6% absolute reduction in undertriage compared with the existing Military Sieve. The improvement in undertriage comes at the expense of overtriage; rates of overtriage were 11.6% higher with the MPTT than the Military Sieve. Using a McNemar test, a statistically significant (p<0.001) improvement in overall performance was demonstrated, supporting the use of the MPTT over the Military Sieve.
Discussion and conclusions The MPTT outperforms all existing triage tools at predicting the need for life-saving intervention, with the lowest rates of undertriage while maintaining acceptable levels of overtriage. Having now been validated on both military and civilian cohorts, we recommend that the major incident community consider adopting the MPTT for the purposes of primary triage.
- ACCIDENT & EMERGENCY MEDICINE
- TRAUMA MANAGEMENT
- HEALTH SERVICES ADMINISTRATION & MANAGEMENT
- Triage
- Major incidents
- Life-saving interventions
Statistics from Altmetric.com
- ACCIDENT & EMERGENCY MEDICINE
- TRAUMA MANAGEMENT
- HEALTH SERVICES ADMINISTRATION & MANAGEMENT
- Triage
- Major incidents
- Life-saving interventions
Key messages
Major incident triage must be rapid, reliable and reproducible irrespective of the provider delivering it.
Existing major incident triage tools perform poorly at predicting need for life-saving intervention.
The MPTT outperforms all existing triage tools at predicting the need for life-saving intervention with the lowest rates of under-triage.
Derived using logistic regression, it’s likely that the MPTT’s performance is the optimum a simple physiological triage tool can have.
Introduction
Major incidents occur worldwide on a regular basis, ranging from natural disasters to terrorist-related incidents, and are brought to our attention through the increasing availability of 24-hour media outlets. Within Europe alone, a number of high profile terrorist atrocities have occurred in the last decade, such as the Paris marauding terrorist firearm attacks (MTFAs) in November 2015, resulting in over 100 fatalities and 300 injured.1 Successful management of the Paris MTFA is attributed to the preparedness of the hospitals, the emergency medical services and the adoption of war surgery principles. The injuries sustained were from high-velocity ballistic weapons, closely mimicking those seen on the battlefield.1
A key principle in the successful management of a major incident, irrespective of mechanism, is the prompt recognition of the critically ill or injured patients and their timely evacuation to the most appropriate facility.2 Triage is the means by which this process is carried out, categorising patients on the basis of their clinical acuity.3 Within a major incident setting, this categorisation is currently most commonly done using an assessment of basic patient physiology, with the intention of detecting any existing physiological instability.4
Within the UK, the Major Incident Medical Management and Support (MIMMS) teaches a two-stage approach to triage — using the Triage Sieve as a rapid assessment on scene and the Triage Sort at casualty clearing stations or in hospitals.2 Alternative methods exist in other countries, including Simple Triage and Rapid Treatment (START) as recommended in the USA and by the United Nations, and CareFlight in use in Australia.5 In response to the coroner’s inquest following the London 7/7 bombing attacks, changes have been proposed to the existing Triage Sieve, to include control of catastrophic haemorrhage and an assessment of patient conscious level.6 These changes would bring civilian practice in line with current UK military teaching (the Military Sieve).7 A comparison of existing triage methods is shown in Table 1.
Comparison of existing major incident triage tools
There is limited evidence to support existing triage methods, and repeated studies have demonstrated limited performance of triage tools at predicting the priority 1 (P1) patient,8 9 if defined as a casualty who is in need of a life-saving intervention.3 The Modified Physiological Triage Tool (MPTT) was derived from a military cohort using logistic regression methodology with the purpose of identifying those in need of life-saving intervention.10 Previous studies have successfully validated its use within both a military and civilian environment, and demonstrated superior performance characteristics when compared with existing triage methods.10 11
The aim of this study was to perform a prospective evaluation of the MPTT, a novel physiological triage tool.
Materials and methods
Physiological data and interventions performed within the emergency department and operating theatre at Camp Bastion, Helmand Province, Afghanistan, were prospectively collected for all adult patients (>18 years) with trauma between March and October 2011. Patient demographics and injury mechanism were not prospectively recorded. In order to provide patient characteristics, a separate, retrospective analysis of the Joint Theatre Trauma Registry (JTTR) was performed for the study period.
Only patients with complete in-hospital physiological data (HR, RR, GCS and systolic BP (SBP)) were included. In order to prevent bias and increase statistical power, outliers defined a priori as RR >45 beats per minute (bpm), HR >170 and SBP >206 mm Hg, in keeping with the MPTT derivation study, were removed.10
Interventions were recorded as free text on a separate data collection sheet by study investigators. Patients were subsequently categorised as gold standard P1 if they received one or more life-saving interventions from a previously defined list.3
Patients were triaged using existing triage tools (Triage Sieve, Military Sieve, Modified Military Sieve, START and CareFlight) and the MPTT against the gold standard definition of the P1 patient. In keeping with previous work, in order to classify patients with military triage tools, START and CareFlight, a surrogate of GCS<13 was used to define the unconscious patient and an SBP of 90 mm Hg was taken to represent the presence of a palpable pulse.8 10 11 Patients were assumed to be non-ambulant for the purposes of triage tool classification.
Performance was measured using a combination of sensitivity, specificity, undertriage (1-sensitivity) and overtriage (1-positive predictive value), with 95% CIs calculated for each. For triage tools with similar performance, a McNemar test with Bonferroni calculation was applied to determine statistically significant difference in performance.12
This study was registered as a service evaluation with the Royal Centre for Defence Medicine (project number RCDM/Res/Audit/1036/12/0050). Additionally the study received approval from the Human Research Ethics Committee of the University of Cape Town, the primary institution of the lead author (reference 285/2013).
Results
During the study period, of 497 patients who presented to the emergency department, 29 (5.8%) were excluded as the data recorded were not sufficient to determine P1 status. A further 107 (21.5%) were removed due to incomplete physiological data and 4 (0.2%) were physiological outliers (HR >170 bpm n=2, RR >45 bpm n=2). The final data set for analysis consisted of 357 patients (71.8%), of whom 59.9% (n=214) were P1 (Figure 1).
Patient participation flow diagram. P1, priority 1.
The JTTR included 458 patients during the study period, of whom the overwhelming majority were male, with blast being the predominant mechanism of injury. In keeping with blast being the predominant mechanism, extremity injuries (lower>upper) occurred most frequently (Table 2).
Study characteristics
The MPTT demonstrated the greatest sensitivity of all existing triage tools (83.6%, 95% CI 78.0% to 88.3%), which was an absolute increase of 14.9% and 19.6%, respectively, over the existing Military Sieve and the Modified Military Sieve, respectively, giving an undertriage rate for the MPTT of 16.4%. By contrast, of all existing civilian triage tools, START had the greatest sensitivity (57.5%, 95% CI 50.6% to 64.2%) with an undertriage rate of 42.5%, demonstrating an absolute decrease in sensitivity of over 25% when compared with the MPTT. The existing UK civilian Triage Sieve performed poorly with sensitivity and specificity of 46.7% (95% CI 39.9% to 53.7%) and 88.1% (95% CI 81.6% to 92.9%), respectively.
The MPTT had the lowest specificity (51.0%, 95% CI 42.6% to 59.5%) and with it the highest rate of overtriage (28.1%), an absolute increase of 11.6% compared with the existing Military Sieve. Using a McNemar test with Bonferroni correction, statistically significant differences were recorded between both the MPTT and the Military Sieve (χ2= 83.012, p<0.001) and the MPTT and the Modified Military Sieve (χ2=64.015, p<0.001). Figure 2 summarises the performance accuracy of the triage tools in their ability to predict the need for life-saving intervention, with a sensitivity analysis displayed in table 3.
Triage tool performance.
Sensitivity analysis
Discussion
In this study we have prospectively validated the MPTT in a military environment. A key principle of major incident triage is that irrespective of the provider using it, it must be rapid, reproducible and reliable. This is the third study demonstrating the MPTT’s improved performance at predicting the need for life-saving intervention when compared with existing triage tools.10 11
There is a paucity of evidence surrounding major incident triage in both the adult and paediatric population, with an increasing number of studies questioning the evidence base and demonstrating limited performance of existing triage tools.8 9
Based on the existing Triage Sieve, the MPTT (table 1) was developed to provide an evidence-based physiological triage tool for use as the primary triage method at a major incident. Its components were derived using logistic regression, demonstrating the optimum individual physiological thresholds at predicting need for a life-saving intervention from a previously defined list.3 10 Analysis and validation studies on both military and civilian trauma registries have demonstrated superior performance by the MPTT in terms of sensitivity and minimising undertriage, although at the expense of poorer but still acceptable levels of specificity and overtriage.10 11 Because of the nature of its derivation, the performance demonstrated by the MPTT in this study is likely to represent the optimum performance that a physiological triage tool can have at predicting the need for life-saving intervention.
The need for life-saving intervention as an outcome measure for triage tool performance is well documented, and it is a more appropriate measure than the Injury Severity Score (ISS).13 Unlike the ISS, it represents an assessment of immediate acuity and not a retrospectively defined assessment of anatomical injury.9 Previous studies have demonstrated a lack of correlation between the ISS and need for life-saving intervention.13 14 Indeed, the majority of our patients did not sustain, by definition, ‘major trauma’ (ISS>15), but approximately 60% received a life-saving intervention and were considered P1.
Our study has demonstrated that the MPTT has the greatest sensitivity at predicting the need for life-saving intervention within a prospective military cohort, yielding the lowest rate of undertriage. However, this comes at the expense of specificity and overtriage, which at 51.0% and 28.1%, respectively, was the highest observed in this study.
To date there are no guidelines by which to benchmark the performance of our major incident triage tools. Within the field triage setting for single trauma patients, it is recommended that overtriage and undertriage be kept to a maximum of 35% and 5%, respectively.15 While this is achievable using a multistage field triage algorithm, incorporating physiological, anatomical and mechanistic assessments, it is unlikely to be achievable by a single physiological algorithm alone. All triage tools within our analysis satisfied this overtriage threshold. With an undertriage rate of 16.4%, the MPTT clearly exceeds the 5% threshold, but is far superior to the highest performing existing method, START (42.5%).
Using a military data set with which to validate a triage tool for use in civilian major incidents is a limitation of this study. Previous studies (including the civilian validation of the MPTT) report an overwhelming preponderance of blunt trauma, with falls <2 m the most common mechanism of injury affecting an older population (median age 61.4 years), with an almost equal predilection for gender (55.6% male).11 16 While this is representative of the epidemiology of individual patient injury, worldwide the incidence of MTFAs is increasing; high-velocity weapons and improvised explosive devices are producing similar patterns of injury to those encountered on the battlefield in Afghanistan.1
While the MPTT has demonstrated improved performance compared with existing methods in both the military and civilian environments, we acknowledge that major incidents are inherently different, with no two being the same. The ability to provide an effective response to a major incident is multifactorial, with the location, number of casualties involved and the resources available being key features.
Mortality has previously been shown to increase in a linear relationship as overtriage increases.17 While an overall scene overtriage rate of 64% was tolerated following the London 7/7 bombings, it must be recognised that this was within an urban environment, with a developed major trauma system.18 We would speculate that it is unlikely at this level to be tolerated in the setting of a rural environment or an immature major trauma system.
MIMMS teaches a progressive method of triage, with secondary triage (using the Triage Sort) providing a more detailed assessment of the patient within a permissive environment. We support this concept and believe that initial primary triage should be as sensitive as possible in order to reduce the number of patients undertriaged. Once within the casualty clearing station or other permissive environment, secondary triage can occur, allowing for refinement of the triage process and a reduction in patients overtriaged. The Triage Sort as it exists currently shows limited performance for predicting the need for life-saving intervention19; we speculate that future work may be required in order to refine the Triage Sort, and support the consideration of including an assessment of anatomical injury, mechanism of injury and clinical acumen.
Limitations
There are clear limitations to our study, a number of which are in common with major incident research in general. We have chosen to perform the prospective validation of the MPTT on a military population, after the tool was derived on a military cohort. While the MPTT has previously been successfully validated on a large civilian cohort, we acknowledge that our population is unlikely to be fully representative of the population involved in a civilian major incident.11
Our prospective data collection did not record details of mechanism of injury or demographics. In order to provide these study characteristics, we performed a retrospective search of the JTTR for the study period.
Despite the data being collected prospectively, 140 cases (28.2%) were removed from our analysis due to incomplete recording of physiological data.
Incomplete data capture has been previously described following a major incident, with rates as high as 38.0% recorded following the London 7/7 bombings.9 Within both the major incident setting and our operational setting, human factors are likely to explain some of the data loss — individuals working in austere, high-pressured environments with multiple patients. During our study period, a number of personnel changes would have occurred, and this may have also contributed to a reduction in data collection. We acknowledge that these factors in conjunction may in turn lead to a form of selection bias.
In keeping with previous work, as the presence of radial pulse was not a recorded variable, for the purposes of classification using START and CareFlight, an SBP of 90 mm Hg was adopted as a surrogate to represent the presence of palpable pulse. We recognise this as a limitation and accept that a number of individuals will still have a radial pulse at an SBP of less than 90 mm Hg. Despite a number of other triage studies using both higher and lower values, 90 mm Hg was used due to its correlation with increased mortality following both blunt and penetrating trauma.5 9 20 21 We acknowledge that BP measurement during primary major incident triage is inappropriate, but in the absence of a recorded variable a surrogate is required for the purposes of comparative validation.
Additionally we have made the assumption that all patients within the study are non-ambulant. While the principal means of patient transport is through helicopter transfer, supporting this assumption to an extent, we acknowledge that a number of helicopter patients are ambulant and are transported seated. Additionally some patients could have self-presented if they were injured close to the site of the medical treatment facility. In a true major incident setting, these patients would be triaged as P3, walking wounded with no physiological assessment performed until arriving at a casualty clearing station where they would undergo secondary triage.2
In an ideal setting, triage tools should be evaluated in the situations in which they are designed to function, that is, a major incident. Owing to the unpredictable nature of major incidents, this is largely impractical and so we frequently turn to the retrospective analysis of major incidents or trauma registries to perform research, both of which have limitations. Numbers of seriously injured patients at a major incident may be low — there were only eight at the Royal London Hospital following the London 7/7 bombings — and are confounded by missing data (only four of the seriously injured had available records).9 By comparison, trauma registries have a greater number of patients, although still with the challenge of missing data. However due to most registries having inclusion criteria, we are unable to represent the population at a major incident who are P3 or have minor injuries. This in turn will have an effect on our ability to reliably comment on the specificities we report in the analysis of our triage tools.
Conclusion
In summary we present a prospective validation of the MPTT, the first evidence-based physiological triage tool for use in the major incident setting. The findings of this study demonstrate that the MPTT outperforms all other existing triage tools at predicting the need for life-saving intervention, with the lowest rates of undertriage while maintaining acceptable levels of overtriage. Having now been validated on both military and civilian cohorts, we recommend that its use be considered for the purposes of major incident primary triage.
References
Footnotes
Contributors JV and SH conceived the study. SH collected the data. JV conducted the analysis, supervised by JES. JV drafted the manuscript, and all authors contributed substantially to its revision. JV takes responsibility for the paper as a whole.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.