- Email: [email protected]
In ‘big bang’ major incidents do triage tools accurately predict clinical priority?: A systematic review of the literature
Injury, Int. J. Care Injured 42 (2011) 460–468
Contents lists available at ScienceDirect
Injury journal homepage: www.elsevier.com/locate/injury
Review
In ‘big bang’ major incidents do triage tools accurately predict clinical priority?: A systematic review of the literature T.M. Kilner a,b, S.J. Brace c,d, M.W. Cooke d, N. Stallard e, A. Bleetman b,f, G.D. Perkins c,* a
Paramedic Sciences, Faculty of Health and Life Sciences, Coventry University, United Kingdom Warwick Medical School, University of Warwick, United Kingdom c Clinical Trials Unit, Warwick Medical School, University of Warwick, United Kingdom d Warwick Medical School and Heart of England NHS Foundation Trust, United Kingdom e Health Sciences Research Institute, Warwick Medical School, University of Warwick, United Kingdom f North West London Hospitals NHS Trust, United Kingdom b
A R T I C L E I N F O
A B S T R A C T
Article history: Accepted 3 November 2010
Introduction: The term ‘‘big bang’’ major incidents is used to describe sudden, usually traumatic, catastrophic events, involving relatively large numbers of injured individuals, where demands on clinical services rapidly outstrip the available resources. Triage tools support the pre-hospital provider to prioritise which patients to treat and/or transport first based upon clinical need. The aim of this review is to identify existing triage tools and to determine the extent to which their reliability and validity have been assessed. Methods: A systematic review of the literature was conducted to identify and evaluate published data validating the efficacy of the triage tools. Studies using data from trauma patients that report on the derivation, validation and/or reliability of the specific pre-hospital triage tools were eligible for inclusion. Purely descriptive studies, reviews, exercises or reports (without supporting data) were excluded. Results: The search yielded 1982 papers. After initial scrutiny of title and abstract, 181 papers were deemed potentially applicable and from these 11 were identified as relevant to this review (in first figure). There were two level of evidence one studies, three level of evidence two studies and six level of evidence three studies. The two level of evidence one studies were prospective validations of Clinical Decision Rules (CDR’s) in children in South Africa, all the other studies were retrospective CDR derivation, validation or cohort studies. The quality of the papers was rated as good (n = 3), fair (n = 7), poor (n = 1). Conclusion: There is limited evidence for the validity of existing triage tools in big bang major incidents. Where evidence does exist it focuses on sensitivity and specificity in relation to prediction of trauma death or severity of injury based on data from single or small number patient incidents. The Sacco system is unique in combining survivability modelling with the degree by which the system is overwhelmed in the triage decision system. The practicalities, training implications, performance characteristics and reliance on computer technology during a mass casualty incident require further evaluation. ß 2010 Elsevier Ltd. All rights reserved.
Keywords: Triage Triage tool Decision algorithm Clinical prioritisation Major incident Mass casualty Pre-hospital Systematic review
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Search methods. . . . . . . . . . . . . . . . . . . . . . . . . Electronic searches . . . . . . . . . . . . . . Search strategy . . . . . . . . . . . . . . . . . Bibliographies and tables of content Inclusion and exclusion criteria . . . . . . . . . . . . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results of search . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
* Corresponding author at: Clinical Trials Unit, Warwick Medical School, University of Warwick, Gibbett Hill Road, Coventry CV4 7AL, United Kingdom. Tel.: +44 024 76150479. E-mail address: [email protected] (G.D. Perkins). 0020–1383/$ – see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.injury.2010.11.005
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
461 461 461 461 461 461 461 461 461
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
Discussion . . . . . . . Conclusions . . . . . . Conflict of interest . Funding . . . . . . . . . References . . . . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
Introduction
Search strategy
The term ‘‘major incident’’ is used to describe any incident where the location, number, severity or type of live casualties requires extra-ordinary resources.1 Major incidents fall broadly into two groups; the ‘big bang’ major incident resulting from a sudden catastrophic event or events with little or no warning, consequently the number of casualties is relatively constant from the time of the incident. Examples of big bang incidents would be a serious transport incident such as the Clapham Junction rail incident in 1988 that resulted in excess of 500 casualties or the 911 terrorist attacks in the USA.14,30 The second type of major incident is the ‘rising tide’ major incident where the precipitating event may not be immediately apparent, but service providers become increasingly aware as casualty/patient numbers increase over time until the demands overwhelm the available resources.9 A recent example of a ‘rising tide’ incident is the international Swine Influenza pandemic41 and severe weather related incidents e.g. flooding. Chemical, biological and nuclear incidents may present with features of both. The initial response to the release of the toxic substance may be similar to a big bang incident, but following this, the delayed effects from the substance may result in a 2nd wave of victims whose presentation may follow more closely the characteristics of a rising tide incident. Triage is a well-established feature of battlefield and major incident casualty management. The concept of triage is thought to have been first described in 1812 by Barron Dominque Jean Larrey, Surgeon-in-Chief to Napoleons’ Imperial Guard.22 The term triage has a number of interpretations in contemporary practice, however for the purpose of this review, triage is considered to be a system for prioritising treatment according to clinical need when the numbers of patients outstrip the resources available to manage them. As triage has evolved, attempts have been made to introduce standardisation into the process, in the form of triage tools and algorithms. Standardisation brings benefits at the incident (where multiple agencies may otherwise use differing systems) and strategically during post-event reviews and research. Whilst existing triage tools attempt to provide a degree of standardisation, the extent to which these tools have been validated is questionable.25 The aim of this review is to identify existing triage tools that may be used for traumatic major incidents and to determine the extent to which their reliability and validity have been assessed.
1 2 3 4 5 6 7 8 9 10 11 12
Methods Search methods Electronic searches Initial searches were executed in March 2009 and again at the end of June 2010. The following electronic databases were searched; The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2009, Issue 2), Cochrane Health Technology Assessment Database (The Cochrane Library 2009, Issue 2), and The Database of Abstracts of Reviews of Effectiveness (DARE) (The Cochrane Library 2009, Issue 2). The Bibliographical databases that were searched were; MEDLINE Ovid (1998–2010, June Week 1), EMBASE Ovid (1998–2010, Week 25) and CINAHL EBSCO (1998–2010).
Triage Triage Triage Triage Triage Triage Triage Triage Triage Triage Triage Triage
. . . . .
[AND] [AND] [AND] [AND] [AND] [AND] [AND] [AND] [AND] [AND] [AND] [AND]
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
461
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
463 467 467 467 467
Sort Sieve START Careflight Major incident Multiple casualt* Mass casualt* Chemical Radi* Nuclear CBRN Tool*
Duplicates were managed by importing all references into an electronic citation manager (EndNote Web v 2.10, Thomson Reuters, PA). Bibliographies and tables of content The bibliographies of appropriate full papers subject to review were examined, as were tables of contents of prehospital and selected emergency journals (Injury, Journal of Trauma-Injury, Infection and Critical Care and Emergency Medicine Journal 1998–2010) to identify further potentially appropriate papers. Inclusion and exclusion criteria The initial search was run by T.K. and then checked and re-run by S.J.B. The outputs of the search strategy were screened by two individuals independently reviewing the title and abstract of all papers for relevance to the scope of this review. Following initial scrutiny the full text of potentially relevant articles were retrieved and reviewed in detail. Studies using data from trauma patients that report on the derivation, validation and/or reliability of the specific pre-hospital triage tools were eligible for inclusion. Purely descriptive studies, reviews, exercises or reports (without supporting data) were excluded. Papers that focussed on tools used in the routine triage of patients in the emergency department were excluded. For pragmatic reasons non-English language papers were excluded as were papers over 10 years old. Papers where studies were based on non-human subjects were also excluded from consideration. Papers were assigned a Level of Evidence (LoE) according to the International Liaison Committee on Resuscitation (ILCOR) definition for evidence appraisal (see Table 1).28 The quality of the papers were recorded as good, fair or poor based on the ILCOR levels of quality.21 Results Results of search The search yielded 1982 papers. After initial scrutiny of title and abstract, 181 papers remained for detailed review of the full paper. From these 11 were identified as relevant to this review (Fig. 1). There were two level of evidence P1 studies,38,39 three level of evidence P2 studies 31,35,36 and six level of evidence P3
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
462 Table 1 ILCOR levels of evidence for prognostic studies.28 C2010 levels of evidence for prognostic studies LOE LOE LOE LOE LOE
P1: P2 P3 P4 P5
Inception (prospective) cohort studies (or meta-analyses of inception cohort studies), or validation of Clinical Decision Rules (CDRs) Follow up of untreated control groups in RCTs (or meta-analyses of follow-up studies), or derivation of CDRs, or rules validated on a split-sample only Retrospective cohort studies Case series Studies not directly related to the specific patient/population (e.g. different patient/population, animal models, mechanical models, etc.)
studies.10,11,18,20,23,24 The two level of evidence P1 studies38,39 were prospective validations of Clinical Decision Rules (CDR’s) in children in South Africa. All the other studies were retrospective CDR derivation, validation or cohort studies. The quality of papers was rated as good (n = 3), fair (n = 7), poor (n = 1). Only one of the papers used patient data from an actual major incident.23 The remainder drew data from single patient incidents or incidents involving a small number of patients. Table 2 summarises the triage tools examined by each paper, the population in which they were evaluated and the level of evidence and quality. Full details of the studies are provided in Supplementary data. The papers examined eight different triage tools (three adult only triage tools, two paediatric tools and three combined adult/paediatric). The characteristics of these tools are summarised in Table 3. In addition three papers reported on the performance of individual physiological parameters. The definition of an adverse outcome varied between studies. Adverse outcomes were described as the need for a specific intervention,24 traumatic death,20,31 severe injury (defined by injury severity scores8,23,38,39) or presence of an immediate life
threatening condition.10,24 The performance of individual physiological tools and multi-parameter triage tools were presented in the papers as sensitivity, specificity, receiver operator curves, odds ratios and likelihood ratios. Sensitivity describes the proportion of patients that are correctly identified that sustain an adverse outcome. Specificity describes the proportion of patients correctly identified to not sustain an adverse outcome.13 Receiver operating characteristics explore the relationship between sensitivity and specificity as a means of determining the predictive value of a test, or in this case an assessment tool. Values for sensitivity are plotted against 1 – specificity, to produce a ‘curve’. The performance of the tool is based on the ‘area under the curve’. In the case of a triage tool the ROC may be comparing the sensitivity and specificity of a tool to predict critical injury. The area under the curve will identify the probability that a critically injured patient will have a higher value than a non-critically injured patient. The ideal tool would detect all the true positives with no false positives and would have an area under the curve of 1. A tool which performs to a standard equivalent to random guess has an area under the curve of 0.5, whilst a useless tool that performs worse than would be achieved
[()TD$FIG] Potentially relevant studies identified from search strategy (n= 1982)
Full papers screened for possible inclusion (n = 181 )
Studies retrieved for detailed evaluation (n = 28)
Studies included from reference lists of reviewed papers and hand sifting (n = 4)
Studies included in the review (n= 11)
Fig. 1. Results of search strategy.
1801 papers were excluded as not relevant following review of title, and where available, abstract
153 excluded as did not fulfil inclusion/exclusion criteria
21 studies excluded from review on inclusion/exclusion criteria
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
463
Table 2 Summary of studies reporting on derivation, validation or reliability of triage tools included in the review. Study ID
Title
Triage tools considered
Population
Outcomes
LOE
Quality
Garner 2001
Comparative analysis of multiple-casualty incident triage algorithms
Adults presenting to trauma centre by ambulance n = 1144
Severe injury
LOE/P3
Fair
Gebhart 2007
START triage: does it work?
START Modified START Triage sieve CareFlight triage START
Survival
LOE/P3
Fair
Hong 2008
Does the simple triage and rapid treatment method appropriately triage patients based on trauma injury severity score?
START
Adults presenting to level II trauma centre n = 355 Adults >65 years n = 100
LoE/P3
Fair
Husum 2003
Respiratory rate as a prehospital triage tool in rural trauma.
LoE/P3
Poor
Kahn 2009
Does START Triage work? An outcomes assessment after disaster
Physiologic Severity Score (Modified Revised Trauma Score) START
Agreement between novice and expert against AIS Mortality
Modified Baxt criteria
LoE/P3
Fair
Leach 2008
Do outcome measures for trauma triage agree?
LoE/P3
Fair
Newgard 2010
A critical assessment of the out-of-hospital trauma triage guidelines for physiologic abnormality Precise formulation and evidence-based application of resource-constrained triage
Non-procedural outcome and ICU admission ‘v’ emergent procedure Mortality or LOS >2 days
LoE/P2
Good
Survival
LoE/P2
Fair
Survival probability ‘v’ ISS and RTS Triage score ‘v’ ISS (16+), NISS (16+) and Garner criteria T1 urgent priority ‘v’ non-urgent (not T1) Using ISS (16 + ), NISS(16 + ) and Garner criteria
LoE/P2
Fair
LoE/P1
Good
LoE/P1
Good
Sacco 2005
American College of Surgeons Committee on Trauma – Trauma Triage Algorithm American College of Surgeons field triage STM and START
Sacco 2007
A new resource constrained triage method applied to victims of penetrating injury
STM and START
Wallis 2006a
Validation of the paediatric triage tape
Paediatric triage tape
Wallis 2006b
Comparison of paediatric major incident primary triage tools
Paediatric triage tape Careflight JumpSTART START
by random guess would have an area under the curve of >0.5. These are summarised in Table 4 (adult single parameter), Table 5 (adult multi-parameter triage tools) and Table 6 (paediatric single and multi-parameter tools). Discussion In the context of a major incident where the number of patients outstrips the available resources a triage tool must be able to provide a means of rapidly sorting patients based on clinical priority, moderated by the capacity of the system to manage those patients. Triage tools need to be able to distinguish between patients in need of immediate treatment, those where treatment can be delayed and those that do not require medical treatment. If the health system is overwhelmed, identification of those with unsurvivable injuries, or with a small chance of survival (relative to the resources required for their care) will allow resources to be prioritised to those with a better chance of survival. The balance between the sensitivity and specificity of the tools is likely to be influenced by the degree of ‘overwhelmedness’ of the emergency response. Only one of the triage tools identified in this review (the Sacco Triage Model (STM)) takes into consideration the degree of overwhelmedness.35,36 The STM is a computer based triage system that uses mathematical modelling to consider evidence-based survival probabilities of patients, estimated deterioration rates, timing and availability of transport, treatment resources and the number and physiology of victims. The system uses physiological information based on respiratory rate, pulse and motor response to predict survivability. For optimal performance the system requires
Adults >14 years with penetrating trauma n = 737 Persons from train crash that went to ED n = 148 Adults with level II trauma n = 474
>15 years injured adults n = 6259 Patients with blunt trauma n = 76459 Patients with penetrating trauma n = 7274 Children <13 years with acute injury n = 3461 Children <13 years with acute injury n = 3461
real time communication between those assessing and triaging victims and the control computer. In this way the computer system can dynamically apply priorities to victims based on numbers, severity of injury and treatment/transport capacity. Using retrospective data from trauma registries, Sacco and colleagues compared STM to START and found improved estimates of survivorship in both blunt, penetrating and military age trauma victims.35,36 The feasibility of this promising approach requires further validation in the mass casualty situation. The implications of using a computer based system and the training requirements are beyond the remit of this review. The remaining studies are significantly limited by the fact that they are static in that they did not take into consideration the degree by which the system is overwhelmed. Garner et al. compared START, CareFlight and Triage Sieve approaches using retrospective data from 1144 trauma patients.10 They found that START and CareFlight outperform the Triage Sieve methodology and individual elements from the tools in terms of sensitivity and specificity. Kahn et al. retrospective analysis of a commuter train crash involving 262 people during which the START tool was used to categorise patient acuity.23 The complexities of an incident are difficult to reproduce within a simulated casualty exercise and so the data from this study allows for more realistic evaluation. However, in applying the START triage tool, Kahn found a substantial over triage in all categories; immediate, delayed and minor with 79 out of 148 patients being over triaged.23 The sensitivity for the delayed and minor categories in predicting significant life threatening conditions were 39.1% and 45.8%, respectively, which the authors attribute to the application of the tool.
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
464
Table 3 Summary of triage tools identified in the literature using the search strategy. Triage tool
Std/MCa
Age
Variables/components
Score/outcome
American College of Surgeons field triage5
Std
Adults & children
Physiological Anatomical Mechanism Others e.g. age or co-morbidities
Multiple parameter trigger A stepwise approach is taken through the decision tool, yes to any question denotes transfer to Level 1 Trauma facility
Triage sieve1
MC
Adults
Ability to walk Breathing Respiratory rate Capillary refill or pulse rate
Multiple parameter trigger PI, P2, P3, P4
Triage sort1
MC
Adults
GCS (coded value) Respiratory rate (coded value) Systolic BP (coded value)
Composite score PI, P2, P3, P4
START3
MC
Adults
Ability to walk Breathing Respiratory rate Pulse Ability to follow commands
Multiple parameter trigger Minor, immediate, delayed
Careflight10
MC
Adults & children
Ability to walk Ability to obey commands Breathing Radial pulse
Multiple parameter trigger delayed, urgent, immediate, unsalvageable
Paediatric triage tape17
MC
Children
Ability to walk Breathing Respiratory rate Capillary refill or pulse rate
Multiple parameter trigger PI, P2, P3, P4
MC
Children 1–8 years
Ability to walk Breathing (5 ventilations if apnoeic with pulse) Respiratory rate Pulse AVPU
Multiple parameter trigger Minor, immediate, delayed
MC
Adults & children
Respiratory rate Pulse rate Motor response Resource availability
Composite score Triage strategy based on resource availability Survival probabilities
JumpSTART
34
Sacco triage method (STM)28,29
a
Std – standard triage, MC – mass casualty triage.
Evidence from the analysis of single physiological parameters found that level of consciousness (ROC AUC 0.86) systolic BP (ROC AUC 0.87) and respiratory rate (ROC AUC 0.86) were good predictors of death.20 By contrast and as an illustration of the importance of the definition of what the adverse outcome is, respiratory rate in isolation was a poor predictor of the need for specific intervention with a ROC AUC of 0.5.10 These three physiological variables form the cornerstone of most triage systems. Indeed when Newgard and colleagues attempted to develop a decision rule for application after basic physiological assessment, the resulting tool was limited by unacceptable low sensitivity and specificity (72% and 69%, respectively).31 Additional consideration for mass casualty incidents are the complexity of the tool and the ease of application need to be weighed up against the tool’s ability to prioritise the patients correctly. All of the identified multiple casualty triage tools utilise physiological parameters as a basis for stratifying patients by anticipated clinical priority. The CareFlight triage tool is the only identified tool which, whilst based upon physiological parameters does not require any measurement, for example obeys commands rather than a measured Glasgow Coma Score.10 Whilst there is a paucity of literature examining the validity of mass casualty triage tools in general, the body of evidence in respect of paediatric mass casualty triage tools is particularly meagre.12 Most of the triage tools adopt an approach using physiological parameters as a means of identifying those who are at risk of death or critical injury. In the adult patient determining ranges for physiological values suitable for all adults requires a single scale for each parameter. In children normal physiological
values change with age and as a consequence a triage tool based on a single range for each physiological parameter would not be appropriate. The Paediatric Triage Tape attempts to address this by providing age specific algorithms, using height as a means of estimating age. Other tools such as START and JumpSTART have a single cut off point of 8 years with the former suitable for children over 8 years old and the latter for children aged 1–8 years. The Careflight tool has the advantage of applicability to both adults and children in a single algorithm as it seeks to identify the presence or absence of a physiological function and not on a specific measure of that function. Wallis and Carley examined the performance of four paediatric triage tools using data from 3461 paediatric patients as a means of predicting the need for critical intervention as well as severity of injury.38,39 The analysis revealed poor levels of sensitivity in each of the identified paediatric triage tools. Although the tools were good at identifying children with an immediate priority the line between the urgent and delayed categories becomes more indistinct. It is acknowledged that there is an element of over triage in children due to the variation in physiological variables with age and size and this defines the need for a validated tool that doesn’t over triage children which may lead to already stretched clinical services becoming overwhelmed. Based on this review, the evidence supporting the use of triage tools has a number of limitations. The first is the absence of a standardised outcome to allow comparisons between different studies. The ideal outcome is the identification of a patient with potentially survivable injuries in need of urgent medical intervention to prevent death or severe morbidity. As a practical example a
Table 4 Summary of data analysis by component (adult). Sensitivity
Specificity
14.8% 25.2%
95.3% 95.3%
72.6%
ROC AUC
Other
Prediction of major trauma (ISS >14)
737
Husum 2006
Prediction Prediction Prediction Prediction Prediction Prediction Prediction Prediction Prediction Prediction Prediction Prediction
1144 1144 737 737 1144 1144 737 737 1144 1144 1144 355
Garner 2001 Garner 2001 Husum 2006 Husum 2006 Garner 2001 Garner 2001 Husum 2006 Husum 2006 Garner 2001 Garner 2001 Garner 2001 Gebhart 2007
99.2% 0.72 (95% CI 0.67–0.77) 0.87 (95% CI 0.84–0.89) 0.76 (95% CI 0.73–0.79)
36.3% 33.3%
93.2% 91.8%
Total coded score = <1 0.57 Total coded score = >2 0.85
Total coded score = <1 Specificity 0.96 Total coded score = >2 0.63
Paper Garner 2001 Garner 2001 Garner 2001 Husum 2006 Husum 2006 Husum 2006
0.77 (95% CI 0.73–0.8)
0.64 (95% CI 0.58–0.7) Total coded score = 1 Positive predictive value 0.4 Negative predictive value 0.98 Total coded score = >2 Positive predictive value 0.08 Negative predictive value 0.99
of of of of of of of of of of of of
need for specific intervention need for specific intervention need for specific intervention trauma death major trauma (ISS >14) trauma death
N=
0.5 (95% CI 0.43–0.56) 0.86 (95% CI 0.83–0.88) 0.78 (95% CI 0.75–0.81) 0.9 (95% CI 0.88–0.92)
96.2%
of of of of of of
1144 1144 1144 737 737 737
Prediction Prediction Prediction Prediction Prediction Prediction
0.85 (95% CI 0.81–0.9) 0.86 (95% CI 0.84–0.89) 0.71 (95% CI 0.68–0.74) 30.4%
Indicator
need for specific intervention need for specific intervention trauma death major trauma (ISS >14) need for specific intervention need for specific intervention trauma death major trauma (ISS >14) need for specific intervention need for specific intervention need for specific intervention trauma death
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
Parameter Respiratory rate >29 Respiratory rate <10 >29 Respiratory rate Respiratory rate Respiratory rate Respiratory rate after fluid resuscitation in the field Respiratory rate after fluid resuscitation in the field GCS motor score <6 GCS motor score Level of consciousness Level of consciousness Systolic BP <80 mmHg Systolic BP Systolic BP Systolic BP Capillary refill >2 s Heart rate >120 Heart rate Respiratory rate (<30 coded 1 point = >30 coded 0 points) Pulse (Present coded 1 point, absent coded 0 points) GCS (15 coded score 1, <15 coded score 0 points)
465
Newgard 2010 2504 NPV 0.64 (95% CI 0.61–0.67)
The original START tool included capillary refill as an assessment of circulation, this was later changed to assessment of the radial pulse.
0.72 (95% CI 0.70–0.74)
a
PPV 0.77 (95% CI 0.74–0.79) 0.69 (95% CI 0.67–0.72)
45% (95% CI 37–54) 45% (95% CI 37–54) 82% (95% CI 75–88)
Triage sieve (capillary refill) Triage sieve (heart rate) CareFlight Physiologic Severity Score (Respiratory rate, systolic BP, Level of Consciousness) Physiologic Severity Score (Respiratory rate, systolic BP, Level of Consciousness) American College of Surgeons field triage (ACSCOT)
0.82 (95% CI 0.79–0.85)
0.93 (95% CI 0.91–0.94) 89% (95% CI 87–91) 88% (95% CI 86–90) 96% (95% CI 94–97)
45.8% (95% CI 36.7–55.2) START (Green/minor)
89.3% (95% CI 71.8–97.7)
39.1% (95% CI 19.7–61.5) START (Yellow/delayed)
11.9% (95% CI 5.3–22.2)
Mortality or LOS> 2 days
Husum 2006 737 Prediction of major trauma (ISS >14)
Garner 2001 Garner 2001 Garner 2001 Husum 2006 1144 1144 1144 737
Kahn 2009 162
Kahn 2009 162
1144 1144 162
Subjects Indicator
Prediction of need for specific intervention Prediction of need for specific intervention Prediction of immediate life threatening conditions (Baxt Criteria) Prediction of immediate life threatening conditions (Baxt Criteria) Prediction of immediate life threatening conditions (Baxt Criteria) Prediction of need for specific intervention Prediction of need for specific intervention Prediction of need for specific intervention Prediction of trauma death 85% (95% CI 78–90) 84% (95% CI 76–89) 100% (95% CI 15.8–100)
Specificity Sensitivity Parameter
START (capillary refill)a START (palpable pulse) START (Red/immediate category)
Table 5 Summary of data analysis by triage tool (adult).
86% (95% CI 84–88) 91% (95% CI 89–93) 77.3% (95% CI 67.1–85.5)
ROC AUC
Other
Odds ratio 35 (95% CI 21–61) Odds ratio 52 (95% CI 31–90) Positive likelihood ratio 4.4 (95% CI 3.0–6.5) Positive likelihood ratio 0.44 (95% CI 0.26–0.75) Positive Likelihood ratio 4.3% (95% CI 1.4–12.7) Odds ratio 7 (95% CI 4–10) Odds ratio 6 (95% CI 4–10) Odds ratio 99 (95% CI 56–176)
Paper
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
Garner 2001 Garner 2001 Kahn 2009
466
patient with severe bleeding in need of haemorrhage control would be potentially salvageable. By contrast the same patient with a catastrophic brain injury would be non-salvageable. In addition, in a time critical injury, the longer that treatment is delayed, the smaller the chance of survival will become. The studies identified in this review used death,11,20,31,35,36 a high injury severity score,10,8,38,39 prediction of need for a specific intervention,23 hospital admission of greater than 2 days31 or presence of life threatening criteria.24 This variability in outcome definitions limits the extent to which comparisons of tools performance can be made between papers. This point is illustrated by Leach et al. who showed poor agreement between injury severity scores >15 and need for emergent resuscitation (kappa 0.31) and emergent surgery (kappa 0.15) as outcome measures in patients with major trauma.24 The closest studies to defining true ‘‘survivability’’ were made by Garner et al. and Sacco.10,36 Garner et al. modified the Baxt and Upenieks criteria to identify a cohort of critically ill patients with potentially survivable injuries.2,10 Patients were defined on the basis of any one of the following criteria: (1) non-orthopaedic operative procedure within 6 h of admission with positive operative findings which would have been life threatening if not treated, (2) fluid resuscitation > 1 L to maintain BP > 90 mm Hg, (3) invasive central nervous system monitoring with either CT evidence of intracranial haemorrhage or raised intracranial pressure, (4) requirement for definitive airway or mechanical ventilation and (5) tension pneumothorax. Patients that subsequently died were excluded. Sacco et al. used mathematical modelling based on respiratory rate, pulse and motor response to calculate survivability scores for patients. When applied to data from 76,459 blunt trauma patients this approach outperformed injury severity score and revised trauma score prediction.35 These approaches have the advantage in that they appear to capture the sense of both severity of injury and implied survivability. Of the studies considered in this review, only the Wallis studies used prospectively collected data based on field determination of triage categories.38,39 Yet even this study was limited by the fact that in most situations triage decisions would have been limited to single or small number of victims at any one time. Retrospective analyses of patient records or registries are limited by the fact that they do not take into consideration how the triage tool performs in practice. This is important as error rates in translating simple physiological variables into a severity score can run as high as 30%.33 In addition, error rates may increase when the system is under strain as may be the case during a mass casualty incident.37 Thus the real performance characteristics of all of the various triage tools in true mass casualty incidents are unknown. The application of effective and accurate triage is also influenced by the preparation and education surrounding the tools and processes to be used. Each mass casualty incident has its own unique characteristics hence preparing the healthcare community has become an expanding and increasingly complex area to ensure cohesive responses to all threats.9 It is also unclear if performing disaster drills is actually effective in improving response skills of healthcare providers,19,40 as there are few objective data within the literature assessing pre- and post-drill knowledge8 and its retention. Despite the increasing likelihood of a big bang major incident will involve chemical, biological, radiological or nuclear (CBRN) element, the literature review failed to reveal the existence of a triage tool validated for use in such circumstances. The triage tools for conventional major incidents are derived from the physiological response to trauma and cannot be extrapolated to CBRN incidents. The range of physiological responses and ability to prognosticate will differ following the release of lung-damaging, blood, blister, and nerve agents.6 Additional gaps in the evidence
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
467
Table 6 Summary of data analysis by triage tool (child). Triage tool
Sensitivity
Specificity
Indicator
Subjects
Paper
Paediatric triage tape
ISS > 15 37.8% (95% CI 32.7–42.5) NISS > 15 26.1% (95% CI 23–28) Intervention (Garner) 41.5% (95% CI 36.8–45.6)
Prediction of severity of injury (Injury Severity Score, and New Injury Severity Score) and the need for specific intervention
3461
Wallis 2006
START (>8 years)
ISS > 15 31.3% (95% CI 21.5–42.8) NISS > 15 22.3% (95% CI 15.6–30.7) Intervention (Garner) 39.2% (95% CI 29.3–50) ISS > 15 Sensitivity 3.2% (95% CI 1.3–7.5) NISS > 15 2.4% (95% CI 1–5) Intervention (Garner) 0.8% (95% CI 0.1–4.1) ISS > 15 48.4% (95% CI 43.4–52.8) NISS > 15 31.5% (95% CI 28.5–34.1) Intervention (Garner) 46% (95% CI 41.2–50.2)
ISS > 15 98.6% (95% CI 98.3–98.8) NISS > 15 98.5% (95% CI 98.5–99.1) Intervention (Garner) Specificity 98.9% (95% CI 98.6–99.2) ISS > 157 7.9% (95% CI 77.3–78.9) NISS > 15 77.3% (95% CI 76.6–78.3) Intervention (Garner) 78.7% (95% CI 77.9–79.5) ISS > 15 97.8% (95% CI 97.7–98) NISS > 15 97.8% (95% CI 97.6–98) Intervention (Garner) 97.7% (95% CI 97.6–97.8)
Prediction of severity of injury (Injury Severity Score, and New Injury Severity Score) and the need for specific intervention
3461
Wallis 2006
Prediction of severity of injury (Injury Severity Score, and New Injury Severity Score) and the need for specific intervention
3461
Wallis 2006
Prediction of severity of injury (Injury Severity Score, and New Injury Severity Score) and the need for specific intervention
3461
Wallis 2006
JumpSTART (1–8 years)
CareFlight
ISS > 15 98.8% (95% CI 98.6–99.1) NISS > 15 99% (95% CI 98.7–99.3) Intervention (Garner) 98.9% (95% CI 98.6–99.1)
based are around the use of triage tools as a basis for the deferral of treatment because the level of resource required by an individual would cause an unacceptable impact on other individuals. Triage tools are used to determine which patients to transfer to a trauma centre in a number of healthcare systems. These triage tools are usually more complex than those used for major incidents, where accuracy of prediction is more important than the timeliness to complete the assessment. Despite their increased complexity (e.g. inclusion of mechanism of injury, anatomical injury patterns, physiological response, patient demographics) no single triage system has been shown to be 100% accurate.4,32 In the setting of a major incident, the ability to reliably exclude patients that do not require urgent treatment is important if the limited resources are to avoid being overwhelmed.15,16 The limitations to existing triage systems were recently reviewed by a multidisciplinary committee in the US. The conclusions from their review in terms of the paucity of evidence of validated systems were similar to ours.27 This group went on to use a consensus approach to develop a new triage guideline (the sort assess, life-saving interventions, treatment/transport (SALT) tool), which is proposed to standardised mass casualty triage processes across the US. Early evaluations of this tool in simulated incidents suggest that the SALT mass casualty triage system can be applied quickly and confidently in the field. It appears to be safe with relatively little under triage at the cost of significant over triage.6,7,26 Like many of the other tools it does not take into consideration the degree by which the system is overwhelmed. There remains a need to examine the effect of this tool on patient outcomes. Conclusions There is limited evidence for the validity of existing triage tools which are proposed for use in a big bang major incident. Where evidence does exist it appears to focus on the sensitivity and specificity in relation to either the prediction of trauma death or the prediction of severity of injury based on data from single or small number patient incidents. None of the traditional triage tools take account of the degree by which the health resources are overwhelmed. In our opinion, the Sacco triage system appears the
most promising in respect to its inclusion of resources allocation and its unique calculation of survivability index. The practicalities, training implications, performance characteristics and reliance on computer technology during a mass casualty incident all require further evaluation. Conflict of interest There are no conflicts of interests. Funding This study is part of work funded by the Emergency Preparedness Clinical Leadership Advisory Group, Department of Health. The funders had no involvement in the study design: collection, analysis and interpretation of data; the writing of the manuscript or the decision to submit the manuscript for publication. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.injury.2010.11.005. References 1. Advanced Life Support Group. Major incident medical management and support, 2 ed., London: BMJ Books; 2002. 2. Baxt WG, Upenieks V. The lack of full correlation between the Injury Severity Score and the resource needs of injured patients. Ann Emerg Med 1990;19:1396–400. 3. Benson DO, Keonig KL, Schultz CH. Disaster triage: START then SAVE - a new method of dynamic triage for victims of a catastrophic earthquake. Prehosp Disaster Med 1996;11:117–24. 4. Carron PN, Taffe P, Ribordy V, et al. Accuracy of prehospital triage of trauma patients by emergency physicians: a retrospective study in western Switzerland. Eur J Emerg Med 2010 Aug 19. 5. Committee on Trauma, American College of Surgeons, Resources for the optimal care of the injured; 2006. 6. Cone DC, MacMillan DS, Parwani V, Van Gelder C. Pilot test of a proposed chemical/biological/radiation/nuclear-capable mass casualty triage system. Prehosp Emerg Care 2008;12(April–June (2)):236–40. 7. Cone DC, Serra J, Burns K, et al. Pilot test of the SALT mass casualty triage system. Prehosp Emerg Care 2009;13(October–December (4)):536–40. 8. Cooke MW, Brace SJ. Training for disaster. Resuscitation 2010;81:788–9.
468
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
9. Department of Health. NHS Emergency Planning Guidance 2005; 2005. http:// www.dh.gov.uk/en/Publicationsandstatistics/Publications/PublicationsPolicy AndGuidance/DH_4121072 (accessed 12.10.10). 10. Garner A, Lee A, Harrison K, Schultz CH. Comparative analysis of multiplecasualty incident triage algorithms. Ann Emerg Med 2001;38:541–8. 11. Gebhart ME, Pence R. START triage: does it work? Disaster Manage Resp (DMR) 2007;5:68–73. 12. Gausche-Hill M. Pediatric disaster preparedness: are we really prepared? J Trauma 2009;67(2 Suppl.):S73–6. 13. Grimes DA, Schultz KF. Uses and abuses of screening tests. Lancet 2002;359: 881–4. 14. Hidden A. Investigation into the Clapham Junction Railway Accident; 1989. http://www.railwaysarchive.co.uk/documents/DoT_Hidden001.pdf (accessed 12.10.10). 15. Hirshberg A, Frykberg ER, Mattox KL, Stein M. Triage and trauma workload in mass casualty: a computer model. J Trauma 2010 [Epub ahead of print]. 16. Hirshberg A, Scott BG, Granchi T, et al. How does casualty load affect trauma care in urban bombing incidents? A quantitative analysis. J Trauma 2005;58(April (4)):686–93. discussion 694–5. 17. Hodgetts TJ, Hall FJ, Maconochie I. Paediatric triage tape. Prehosp Immed Care 1998;2:155–9. 18. Hong R, Sierzenski PR, Bollinger M, et al. Does the simple triage and rapid treatment method appropriately triage patients based on trauma injury severity score? Am J Disaster Med 2008;3:265–71. 19. Hsu EB, Jenckes MW, Catlett CL, et al. Effectiveness of hospital staff masscasualty incident training methods: a systematic literature review. Prehosp Disaster Med 2004;19:191–9. 20. Husum H, Gilbert M, Wisborg T, et al. Respiratory rate as a prehospital triage tool in rural trauma. J Trauma – Injury Infect Crit Care 2003;55:466–70. 21. International Liaison Committee on Resuscitation: Defining Quality of Evidence.doc. http://mc.manuscriptcentral.com:80/societyimages/ilcor/Defining%20 Quality%20of%20Evidence.doc (accessed 12.05.09). 22. Iserson KV, Moskop JC. Triage in medicine, part I: concept, history, and types. Ann Emerg Med 2007;49(3):275–81. 23. Kahn CA, Schultz CH, Miller KT, Anderson CL. Does START triage work? An outcomes assessment after a disaster. Ann Emerg Med 2009;54(3):424–30. 430.e1. 24. Leach SR, Swor RA, Jackson RE, et al. Do outcome measures for trauma triage agree? Prehosp Emerg Care 2008;12:467–9. 25. Lerner EB, Schwartz RB, Coule PL, et al. Mass casualty triage: an evaluation of the data and development of a proposed national guideline. Disaster Med Public Health Preparedness 2008;2(Suppl. 1):S25–34.
26. Lerner EB, Schwartz RB, Coule PL, Pirrallo RG. Use of SALT triage in a simulated mass-casualty incident. Prehosp Emerg Care 2010;14(January–March (1)): 21–5. 27. Lerner EB. Studies evaluating current field triage: 1966–2005. Prehosp Emerg Care 2006;10:303–6. 28. Morley P. Evidence evaluation worksheets: the systematic reviews for the evidence evaluation process for the 2010 International Consensus on Resuscitation Science. Resuscitation 2009;80:719–21. 29. National Audit Office. Facing the challenge: NHS emergency planning in England, HC 36. London: National Audit Office; 2002. 30. neverforget911.org. Statistics and numbers from the September 11th, 2001 attacks 2002. http://neverforget911.org/statistics.aspx (accessed 12.10.10). 31. Newgard CD, Rudser K, Hedges JR et al. The ROC investigators. A critical assessment of the out-of-hospital trauma triage guidelines for physiologic abnormality journal of trauma-injury infection & critical care, vol. 68(2); February 2010, pp. 452–462. 32. Osen HB, Bass RR, Abdullah F, Chang DC. Rapid discharge after transfer: risk factors, incidence, and implications for trauma systems. J Trauma 2010;69(September (3)):602–6. 33. Prytherch DR, Smith GB, Schmidt P, et al. Calculating early warning scores – a classroom comparison of pen and paper and hand-held computer methods. Resuscitation 2006;70(August (2)):173–8. Epub 2006 Jun 27. 34. Romig LE. Pediatric triage. A system to JumpSTART your triage of young patients at MCIs. J Emerg Med Services 2002;27:52–8. 35. Sacco WJ, Navin DM, Fiedler KE, et al. Precise formulation and evidence-based application of resource-constrained triage. Acad Emerg Med 2005;12(August (8)):759–70. 36. Sacco WJ, Navin DM, Waddell II RK, et al. A new resource-constrained triage method applied to victims of penetrating injury. J Trauma 2007;63(August (2)):316–25. 37. Smith AF, Oakley RJ. Incidence and significance of errors in a patient ‘track and trigger’ system during an epidemic of Legionnaires’ disease: retrospective case note analysis. Anaesthesia 2006;61:222–8. 38. Wallis LA, Carley S. Comparison of paediatric major incident primary triage tools. Emerg Med J 2006;23:475–8. 39. Wallis LA, Carley S. Validation of the paediatric triage tape. Emerg Med J 2006;23:47–50. 40. Williams J, Nocera M, Casteel C. The effectiveness of disaster training for health care workers: a systematic review. Ann Emerg Med 2008;52:211–22. 41. World Health Organisation. Global alert and response – what is the pandemic (H1N1) 2009 virus?; 2009. http://www.who.int/csr/disease/swineflu/ frequently_asked_questions/about_disease/en/index.html (accessed 14.10.09).
Contents lists available at ScienceDirect
Injury journal homepage: www.elsevier.com/locate/injury
Review
In ‘big bang’ major incidents do triage tools accurately predict clinical priority?: A systematic review of the literature T.M. Kilner a,b, S.J. Brace c,d, M.W. Cooke d, N. Stallard e, A. Bleetman b,f, G.D. Perkins c,* a
Paramedic Sciences, Faculty of Health and Life Sciences, Coventry University, United Kingdom Warwick Medical School, University of Warwick, United Kingdom c Clinical Trials Unit, Warwick Medical School, University of Warwick, United Kingdom d Warwick Medical School and Heart of England NHS Foundation Trust, United Kingdom e Health Sciences Research Institute, Warwick Medical School, University of Warwick, United Kingdom f North West London Hospitals NHS Trust, United Kingdom b
A R T I C L E I N F O
A B S T R A C T
Article history: Accepted 3 November 2010
Introduction: The term ‘‘big bang’’ major incidents is used to describe sudden, usually traumatic, catastrophic events, involving relatively large numbers of injured individuals, where demands on clinical services rapidly outstrip the available resources. Triage tools support the pre-hospital provider to prioritise which patients to treat and/or transport first based upon clinical need. The aim of this review is to identify existing triage tools and to determine the extent to which their reliability and validity have been assessed. Methods: A systematic review of the literature was conducted to identify and evaluate published data validating the efficacy of the triage tools. Studies using data from trauma patients that report on the derivation, validation and/or reliability of the specific pre-hospital triage tools were eligible for inclusion. Purely descriptive studies, reviews, exercises or reports (without supporting data) were excluded. Results: The search yielded 1982 papers. After initial scrutiny of title and abstract, 181 papers were deemed potentially applicable and from these 11 were identified as relevant to this review (in first figure). There were two level of evidence one studies, three level of evidence two studies and six level of evidence three studies. The two level of evidence one studies were prospective validations of Clinical Decision Rules (CDR’s) in children in South Africa, all the other studies were retrospective CDR derivation, validation or cohort studies. The quality of the papers was rated as good (n = 3), fair (n = 7), poor (n = 1). Conclusion: There is limited evidence for the validity of existing triage tools in big bang major incidents. Where evidence does exist it focuses on sensitivity and specificity in relation to prediction of trauma death or severity of injury based on data from single or small number patient incidents. The Sacco system is unique in combining survivability modelling with the degree by which the system is overwhelmed in the triage decision system. The practicalities, training implications, performance characteristics and reliance on computer technology during a mass casualty incident require further evaluation. ß 2010 Elsevier Ltd. All rights reserved.
Keywords: Triage Triage tool Decision algorithm Clinical prioritisation Major incident Mass casualty Pre-hospital Systematic review
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Search methods. . . . . . . . . . . . . . . . . . . . . . . . . Electronic searches . . . . . . . . . . . . . . Search strategy . . . . . . . . . . . . . . . . . Bibliographies and tables of content Inclusion and exclusion criteria . . . . . . . . . . . . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results of search . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
* Corresponding author at: Clinical Trials Unit, Warwick Medical School, University of Warwick, Gibbett Hill Road, Coventry CV4 7AL, United Kingdom. Tel.: +44 024 76150479. E-mail address: [email protected] (G.D. Perkins). 0020–1383/$ – see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.injury.2010.11.005
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
461 461 461 461 461 461 461 461 461
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
Discussion . . . . . . . Conclusions . . . . . . Conflict of interest . Funding . . . . . . . . . References . . . . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
Introduction
Search strategy
The term ‘‘major incident’’ is used to describe any incident where the location, number, severity or type of live casualties requires extra-ordinary resources.1 Major incidents fall broadly into two groups; the ‘big bang’ major incident resulting from a sudden catastrophic event or events with little or no warning, consequently the number of casualties is relatively constant from the time of the incident. Examples of big bang incidents would be a serious transport incident such as the Clapham Junction rail incident in 1988 that resulted in excess of 500 casualties or the 911 terrorist attacks in the USA.14,30 The second type of major incident is the ‘rising tide’ major incident where the precipitating event may not be immediately apparent, but service providers become increasingly aware as casualty/patient numbers increase over time until the demands overwhelm the available resources.9 A recent example of a ‘rising tide’ incident is the international Swine Influenza pandemic41 and severe weather related incidents e.g. flooding. Chemical, biological and nuclear incidents may present with features of both. The initial response to the release of the toxic substance may be similar to a big bang incident, but following this, the delayed effects from the substance may result in a 2nd wave of victims whose presentation may follow more closely the characteristics of a rising tide incident. Triage is a well-established feature of battlefield and major incident casualty management. The concept of triage is thought to have been first described in 1812 by Barron Dominque Jean Larrey, Surgeon-in-Chief to Napoleons’ Imperial Guard.22 The term triage has a number of interpretations in contemporary practice, however for the purpose of this review, triage is considered to be a system for prioritising treatment according to clinical need when the numbers of patients outstrip the resources available to manage them. As triage has evolved, attempts have been made to introduce standardisation into the process, in the form of triage tools and algorithms. Standardisation brings benefits at the incident (where multiple agencies may otherwise use differing systems) and strategically during post-event reviews and research. Whilst existing triage tools attempt to provide a degree of standardisation, the extent to which these tools have been validated is questionable.25 The aim of this review is to identify existing triage tools that may be used for traumatic major incidents and to determine the extent to which their reliability and validity have been assessed.
1 2 3 4 5 6 7 8 9 10 11 12
Methods Search methods Electronic searches Initial searches were executed in March 2009 and again at the end of June 2010. The following electronic databases were searched; The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2009, Issue 2), Cochrane Health Technology Assessment Database (The Cochrane Library 2009, Issue 2), and The Database of Abstracts of Reviews of Effectiveness (DARE) (The Cochrane Library 2009, Issue 2). The Bibliographical databases that were searched were; MEDLINE Ovid (1998–2010, June Week 1), EMBASE Ovid (1998–2010, Week 25) and CINAHL EBSCO (1998–2010).
Triage Triage Triage Triage Triage Triage Triage Triage Triage Triage Triage Triage
. . . . .
[AND] [AND] [AND] [AND] [AND] [AND] [AND] [AND] [AND] [AND] [AND] [AND]
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
461
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
463 467 467 467 467
Sort Sieve START Careflight Major incident Multiple casualt* Mass casualt* Chemical Radi* Nuclear CBRN Tool*
Duplicates were managed by importing all references into an electronic citation manager (EndNote Web v 2.10, Thomson Reuters, PA). Bibliographies and tables of content The bibliographies of appropriate full papers subject to review were examined, as were tables of contents of prehospital and selected emergency journals (Injury, Journal of Trauma-Injury, Infection and Critical Care and Emergency Medicine Journal 1998–2010) to identify further potentially appropriate papers. Inclusion and exclusion criteria The initial search was run by T.K. and then checked and re-run by S.J.B. The outputs of the search strategy were screened by two individuals independently reviewing the title and abstract of all papers for relevance to the scope of this review. Following initial scrutiny the full text of potentially relevant articles were retrieved and reviewed in detail. Studies using data from trauma patients that report on the derivation, validation and/or reliability of the specific pre-hospital triage tools were eligible for inclusion. Purely descriptive studies, reviews, exercises or reports (without supporting data) were excluded. Papers that focussed on tools used in the routine triage of patients in the emergency department were excluded. For pragmatic reasons non-English language papers were excluded as were papers over 10 years old. Papers where studies were based on non-human subjects were also excluded from consideration. Papers were assigned a Level of Evidence (LoE) according to the International Liaison Committee on Resuscitation (ILCOR) definition for evidence appraisal (see Table 1).28 The quality of the papers were recorded as good, fair or poor based on the ILCOR levels of quality.21 Results Results of search The search yielded 1982 papers. After initial scrutiny of title and abstract, 181 papers remained for detailed review of the full paper. From these 11 were identified as relevant to this review (Fig. 1). There were two level of evidence P1 studies,38,39 three level of evidence P2 studies 31,35,36 and six level of evidence P3
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
462 Table 1 ILCOR levels of evidence for prognostic studies.28 C2010 levels of evidence for prognostic studies LOE LOE LOE LOE LOE
P1: P2 P3 P4 P5
Inception (prospective) cohort studies (or meta-analyses of inception cohort studies), or validation of Clinical Decision Rules (CDRs) Follow up of untreated control groups in RCTs (or meta-analyses of follow-up studies), or derivation of CDRs, or rules validated on a split-sample only Retrospective cohort studies Case series Studies not directly related to the specific patient/population (e.g. different patient/population, animal models, mechanical models, etc.)
studies.10,11,18,20,23,24 The two level of evidence P1 studies38,39 were prospective validations of Clinical Decision Rules (CDR’s) in children in South Africa. All the other studies were retrospective CDR derivation, validation or cohort studies. The quality of papers was rated as good (n = 3), fair (n = 7), poor (n = 1). Only one of the papers used patient data from an actual major incident.23 The remainder drew data from single patient incidents or incidents involving a small number of patients. Table 2 summarises the triage tools examined by each paper, the population in which they were evaluated and the level of evidence and quality. Full details of the studies are provided in Supplementary data. The papers examined eight different triage tools (three adult only triage tools, two paediatric tools and three combined adult/paediatric). The characteristics of these tools are summarised in Table 3. In addition three papers reported on the performance of individual physiological parameters. The definition of an adverse outcome varied between studies. Adverse outcomes were described as the need for a specific intervention,24 traumatic death,20,31 severe injury (defined by injury severity scores8,23,38,39) or presence of an immediate life
threatening condition.10,24 The performance of individual physiological tools and multi-parameter triage tools were presented in the papers as sensitivity, specificity, receiver operator curves, odds ratios and likelihood ratios. Sensitivity describes the proportion of patients that are correctly identified that sustain an adverse outcome. Specificity describes the proportion of patients correctly identified to not sustain an adverse outcome.13 Receiver operating characteristics explore the relationship between sensitivity and specificity as a means of determining the predictive value of a test, or in this case an assessment tool. Values for sensitivity are plotted against 1 – specificity, to produce a ‘curve’. The performance of the tool is based on the ‘area under the curve’. In the case of a triage tool the ROC may be comparing the sensitivity and specificity of a tool to predict critical injury. The area under the curve will identify the probability that a critically injured patient will have a higher value than a non-critically injured patient. The ideal tool would detect all the true positives with no false positives and would have an area under the curve of 1. A tool which performs to a standard equivalent to random guess has an area under the curve of 0.5, whilst a useless tool that performs worse than would be achieved
[()TD$FIG] Potentially relevant studies identified from search strategy (n= 1982)
Full papers screened for possible inclusion (n = 181 )
Studies retrieved for detailed evaluation (n = 28)
Studies included from reference lists of reviewed papers and hand sifting (n = 4)
Studies included in the review (n= 11)
Fig. 1. Results of search strategy.
1801 papers were excluded as not relevant following review of title, and where available, abstract
153 excluded as did not fulfil inclusion/exclusion criteria
21 studies excluded from review on inclusion/exclusion criteria
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
463
Table 2 Summary of studies reporting on derivation, validation or reliability of triage tools included in the review. Study ID
Title
Triage tools considered
Population
Outcomes
LOE
Quality
Garner 2001
Comparative analysis of multiple-casualty incident triage algorithms
Adults presenting to trauma centre by ambulance n = 1144
Severe injury
LOE/P3
Fair
Gebhart 2007
START triage: does it work?
START Modified START Triage sieve CareFlight triage START
Survival
LOE/P3
Fair
Hong 2008
Does the simple triage and rapid treatment method appropriately triage patients based on trauma injury severity score?
START
Adults presenting to level II trauma centre n = 355 Adults >65 years n = 100
LoE/P3
Fair
Husum 2003
Respiratory rate as a prehospital triage tool in rural trauma.
LoE/P3
Poor
Kahn 2009
Does START Triage work? An outcomes assessment after disaster
Physiologic Severity Score (Modified Revised Trauma Score) START
Agreement between novice and expert against AIS Mortality
Modified Baxt criteria
LoE/P3
Fair
Leach 2008
Do outcome measures for trauma triage agree?
LoE/P3
Fair
Newgard 2010
A critical assessment of the out-of-hospital trauma triage guidelines for physiologic abnormality Precise formulation and evidence-based application of resource-constrained triage
Non-procedural outcome and ICU admission ‘v’ emergent procedure Mortality or LOS >2 days
LoE/P2
Good
Survival
LoE/P2
Fair
Survival probability ‘v’ ISS and RTS Triage score ‘v’ ISS (16+), NISS (16+) and Garner criteria T1 urgent priority ‘v’ non-urgent (not T1) Using ISS (16 + ), NISS(16 + ) and Garner criteria
LoE/P2
Fair
LoE/P1
Good
LoE/P1
Good
Sacco 2005
American College of Surgeons Committee on Trauma – Trauma Triage Algorithm American College of Surgeons field triage STM and START
Sacco 2007
A new resource constrained triage method applied to victims of penetrating injury
STM and START
Wallis 2006a
Validation of the paediatric triage tape
Paediatric triage tape
Wallis 2006b
Comparison of paediatric major incident primary triage tools
Paediatric triage tape Careflight JumpSTART START
by random guess would have an area under the curve of >0.5. These are summarised in Table 4 (adult single parameter), Table 5 (adult multi-parameter triage tools) and Table 6 (paediatric single and multi-parameter tools). Discussion In the context of a major incident where the number of patients outstrips the available resources a triage tool must be able to provide a means of rapidly sorting patients based on clinical priority, moderated by the capacity of the system to manage those patients. Triage tools need to be able to distinguish between patients in need of immediate treatment, those where treatment can be delayed and those that do not require medical treatment. If the health system is overwhelmed, identification of those with unsurvivable injuries, or with a small chance of survival (relative to the resources required for their care) will allow resources to be prioritised to those with a better chance of survival. The balance between the sensitivity and specificity of the tools is likely to be influenced by the degree of ‘overwhelmedness’ of the emergency response. Only one of the triage tools identified in this review (the Sacco Triage Model (STM)) takes into consideration the degree of overwhelmedness.35,36 The STM is a computer based triage system that uses mathematical modelling to consider evidence-based survival probabilities of patients, estimated deterioration rates, timing and availability of transport, treatment resources and the number and physiology of victims. The system uses physiological information based on respiratory rate, pulse and motor response to predict survivability. For optimal performance the system requires
Adults >14 years with penetrating trauma n = 737 Persons from train crash that went to ED n = 148 Adults with level II trauma n = 474
>15 years injured adults n = 6259 Patients with blunt trauma n = 76459 Patients with penetrating trauma n = 7274 Children <13 years with acute injury n = 3461 Children <13 years with acute injury n = 3461
real time communication between those assessing and triaging victims and the control computer. In this way the computer system can dynamically apply priorities to victims based on numbers, severity of injury and treatment/transport capacity. Using retrospective data from trauma registries, Sacco and colleagues compared STM to START and found improved estimates of survivorship in both blunt, penetrating and military age trauma victims.35,36 The feasibility of this promising approach requires further validation in the mass casualty situation. The implications of using a computer based system and the training requirements are beyond the remit of this review. The remaining studies are significantly limited by the fact that they are static in that they did not take into consideration the degree by which the system is overwhelmed. Garner et al. compared START, CareFlight and Triage Sieve approaches using retrospective data from 1144 trauma patients.10 They found that START and CareFlight outperform the Triage Sieve methodology and individual elements from the tools in terms of sensitivity and specificity. Kahn et al. retrospective analysis of a commuter train crash involving 262 people during which the START tool was used to categorise patient acuity.23 The complexities of an incident are difficult to reproduce within a simulated casualty exercise and so the data from this study allows for more realistic evaluation. However, in applying the START triage tool, Kahn found a substantial over triage in all categories; immediate, delayed and minor with 79 out of 148 patients being over triaged.23 The sensitivity for the delayed and minor categories in predicting significant life threatening conditions were 39.1% and 45.8%, respectively, which the authors attribute to the application of the tool.
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
464
Table 3 Summary of triage tools identified in the literature using the search strategy. Triage tool
Std/MCa
Age
Variables/components
Score/outcome
American College of Surgeons field triage5
Std
Adults & children
Physiological Anatomical Mechanism Others e.g. age or co-morbidities
Multiple parameter trigger A stepwise approach is taken through the decision tool, yes to any question denotes transfer to Level 1 Trauma facility
Triage sieve1
MC
Adults
Ability to walk Breathing Respiratory rate Capillary refill or pulse rate
Multiple parameter trigger PI, P2, P3, P4
Triage sort1
MC
Adults
GCS (coded value) Respiratory rate (coded value) Systolic BP (coded value)
Composite score PI, P2, P3, P4
START3
MC
Adults
Ability to walk Breathing Respiratory rate Pulse Ability to follow commands
Multiple parameter trigger Minor, immediate, delayed
Careflight10
MC
Adults & children
Ability to walk Ability to obey commands Breathing Radial pulse
Multiple parameter trigger delayed, urgent, immediate, unsalvageable
Paediatric triage tape17
MC
Children
Ability to walk Breathing Respiratory rate Capillary refill or pulse rate
Multiple parameter trigger PI, P2, P3, P4
MC
Children 1–8 years
Ability to walk Breathing (5 ventilations if apnoeic with pulse) Respiratory rate Pulse AVPU
Multiple parameter trigger Minor, immediate, delayed
MC
Adults & children
Respiratory rate Pulse rate Motor response Resource availability
Composite score Triage strategy based on resource availability Survival probabilities
JumpSTART
34
Sacco triage method (STM)28,29
a
Std – standard triage, MC – mass casualty triage.
Evidence from the analysis of single physiological parameters found that level of consciousness (ROC AUC 0.86) systolic BP (ROC AUC 0.87) and respiratory rate (ROC AUC 0.86) were good predictors of death.20 By contrast and as an illustration of the importance of the definition of what the adverse outcome is, respiratory rate in isolation was a poor predictor of the need for specific intervention with a ROC AUC of 0.5.10 These three physiological variables form the cornerstone of most triage systems. Indeed when Newgard and colleagues attempted to develop a decision rule for application after basic physiological assessment, the resulting tool was limited by unacceptable low sensitivity and specificity (72% and 69%, respectively).31 Additional consideration for mass casualty incidents are the complexity of the tool and the ease of application need to be weighed up against the tool’s ability to prioritise the patients correctly. All of the identified multiple casualty triage tools utilise physiological parameters as a basis for stratifying patients by anticipated clinical priority. The CareFlight triage tool is the only identified tool which, whilst based upon physiological parameters does not require any measurement, for example obeys commands rather than a measured Glasgow Coma Score.10 Whilst there is a paucity of literature examining the validity of mass casualty triage tools in general, the body of evidence in respect of paediatric mass casualty triage tools is particularly meagre.12 Most of the triage tools adopt an approach using physiological parameters as a means of identifying those who are at risk of death or critical injury. In the adult patient determining ranges for physiological values suitable for all adults requires a single scale for each parameter. In children normal physiological
values change with age and as a consequence a triage tool based on a single range for each physiological parameter would not be appropriate. The Paediatric Triage Tape attempts to address this by providing age specific algorithms, using height as a means of estimating age. Other tools such as START and JumpSTART have a single cut off point of 8 years with the former suitable for children over 8 years old and the latter for children aged 1–8 years. The Careflight tool has the advantage of applicability to both adults and children in a single algorithm as it seeks to identify the presence or absence of a physiological function and not on a specific measure of that function. Wallis and Carley examined the performance of four paediatric triage tools using data from 3461 paediatric patients as a means of predicting the need for critical intervention as well as severity of injury.38,39 The analysis revealed poor levels of sensitivity in each of the identified paediatric triage tools. Although the tools were good at identifying children with an immediate priority the line between the urgent and delayed categories becomes more indistinct. It is acknowledged that there is an element of over triage in children due to the variation in physiological variables with age and size and this defines the need for a validated tool that doesn’t over triage children which may lead to already stretched clinical services becoming overwhelmed. Based on this review, the evidence supporting the use of triage tools has a number of limitations. The first is the absence of a standardised outcome to allow comparisons between different studies. The ideal outcome is the identification of a patient with potentially survivable injuries in need of urgent medical intervention to prevent death or severe morbidity. As a practical example a
Table 4 Summary of data analysis by component (adult). Sensitivity
Specificity
14.8% 25.2%
95.3% 95.3%
72.6%
ROC AUC
Other
Prediction of major trauma (ISS >14)
737
Husum 2006
Prediction Prediction Prediction Prediction Prediction Prediction Prediction Prediction Prediction Prediction Prediction Prediction
1144 1144 737 737 1144 1144 737 737 1144 1144 1144 355
Garner 2001 Garner 2001 Husum 2006 Husum 2006 Garner 2001 Garner 2001 Husum 2006 Husum 2006 Garner 2001 Garner 2001 Garner 2001 Gebhart 2007
99.2% 0.72 (95% CI 0.67–0.77) 0.87 (95% CI 0.84–0.89) 0.76 (95% CI 0.73–0.79)
36.3% 33.3%
93.2% 91.8%
Total coded score = <1 0.57 Total coded score = >2 0.85
Total coded score = <1 Specificity 0.96 Total coded score = >2 0.63
Paper Garner 2001 Garner 2001 Garner 2001 Husum 2006 Husum 2006 Husum 2006
0.77 (95% CI 0.73–0.8)
0.64 (95% CI 0.58–0.7) Total coded score = 1 Positive predictive value 0.4 Negative predictive value 0.98 Total coded score = >2 Positive predictive value 0.08 Negative predictive value 0.99
of of of of of of of of of of of of
need for specific intervention need for specific intervention need for specific intervention trauma death major trauma (ISS >14) trauma death
N=
0.5 (95% CI 0.43–0.56) 0.86 (95% CI 0.83–0.88) 0.78 (95% CI 0.75–0.81) 0.9 (95% CI 0.88–0.92)
96.2%
of of of of of of
1144 1144 1144 737 737 737
Prediction Prediction Prediction Prediction Prediction Prediction
0.85 (95% CI 0.81–0.9) 0.86 (95% CI 0.84–0.89) 0.71 (95% CI 0.68–0.74) 30.4%
Indicator
need for specific intervention need for specific intervention trauma death major trauma (ISS >14) need for specific intervention need for specific intervention trauma death major trauma (ISS >14) need for specific intervention need for specific intervention need for specific intervention trauma death
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
Parameter Respiratory rate >29 Respiratory rate <10 >29 Respiratory rate Respiratory rate Respiratory rate Respiratory rate after fluid resuscitation in the field Respiratory rate after fluid resuscitation in the field GCS motor score <6 GCS motor score Level of consciousness Level of consciousness Systolic BP <80 mmHg Systolic BP Systolic BP Systolic BP Capillary refill >2 s Heart rate >120 Heart rate Respiratory rate (<30 coded 1 point = >30 coded 0 points) Pulse (Present coded 1 point, absent coded 0 points) GCS (15 coded score 1, <15 coded score 0 points)
465
Newgard 2010 2504 NPV 0.64 (95% CI 0.61–0.67)
The original START tool included capillary refill as an assessment of circulation, this was later changed to assessment of the radial pulse.
0.72 (95% CI 0.70–0.74)
a
PPV 0.77 (95% CI 0.74–0.79) 0.69 (95% CI 0.67–0.72)
45% (95% CI 37–54) 45% (95% CI 37–54) 82% (95% CI 75–88)
Triage sieve (capillary refill) Triage sieve (heart rate) CareFlight Physiologic Severity Score (Respiratory rate, systolic BP, Level of Consciousness) Physiologic Severity Score (Respiratory rate, systolic BP, Level of Consciousness) American College of Surgeons field triage (ACSCOT)
0.82 (95% CI 0.79–0.85)
0.93 (95% CI 0.91–0.94) 89% (95% CI 87–91) 88% (95% CI 86–90) 96% (95% CI 94–97)
45.8% (95% CI 36.7–55.2) START (Green/minor)
89.3% (95% CI 71.8–97.7)
39.1% (95% CI 19.7–61.5) START (Yellow/delayed)
11.9% (95% CI 5.3–22.2)
Mortality or LOS> 2 days
Husum 2006 737 Prediction of major trauma (ISS >14)
Garner 2001 Garner 2001 Garner 2001 Husum 2006 1144 1144 1144 737
Kahn 2009 162
Kahn 2009 162
1144 1144 162
Subjects Indicator
Prediction of need for specific intervention Prediction of need for specific intervention Prediction of immediate life threatening conditions (Baxt Criteria) Prediction of immediate life threatening conditions (Baxt Criteria) Prediction of immediate life threatening conditions (Baxt Criteria) Prediction of need for specific intervention Prediction of need for specific intervention Prediction of need for specific intervention Prediction of trauma death 85% (95% CI 78–90) 84% (95% CI 76–89) 100% (95% CI 15.8–100)
Specificity Sensitivity Parameter
START (capillary refill)a START (palpable pulse) START (Red/immediate category)
Table 5 Summary of data analysis by triage tool (adult).
86% (95% CI 84–88) 91% (95% CI 89–93) 77.3% (95% CI 67.1–85.5)
ROC AUC
Other
Odds ratio 35 (95% CI 21–61) Odds ratio 52 (95% CI 31–90) Positive likelihood ratio 4.4 (95% CI 3.0–6.5) Positive likelihood ratio 0.44 (95% CI 0.26–0.75) Positive Likelihood ratio 4.3% (95% CI 1.4–12.7) Odds ratio 7 (95% CI 4–10) Odds ratio 6 (95% CI 4–10) Odds ratio 99 (95% CI 56–176)
Paper
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
Garner 2001 Garner 2001 Kahn 2009
466
patient with severe bleeding in need of haemorrhage control would be potentially salvageable. By contrast the same patient with a catastrophic brain injury would be non-salvageable. In addition, in a time critical injury, the longer that treatment is delayed, the smaller the chance of survival will become. The studies identified in this review used death,11,20,31,35,36 a high injury severity score,10,8,38,39 prediction of need for a specific intervention,23 hospital admission of greater than 2 days31 or presence of life threatening criteria.24 This variability in outcome definitions limits the extent to which comparisons of tools performance can be made between papers. This point is illustrated by Leach et al. who showed poor agreement between injury severity scores >15 and need for emergent resuscitation (kappa 0.31) and emergent surgery (kappa 0.15) as outcome measures in patients with major trauma.24 The closest studies to defining true ‘‘survivability’’ were made by Garner et al. and Sacco.10,36 Garner et al. modified the Baxt and Upenieks criteria to identify a cohort of critically ill patients with potentially survivable injuries.2,10 Patients were defined on the basis of any one of the following criteria: (1) non-orthopaedic operative procedure within 6 h of admission with positive operative findings which would have been life threatening if not treated, (2) fluid resuscitation > 1 L to maintain BP > 90 mm Hg, (3) invasive central nervous system monitoring with either CT evidence of intracranial haemorrhage or raised intracranial pressure, (4) requirement for definitive airway or mechanical ventilation and (5) tension pneumothorax. Patients that subsequently died were excluded. Sacco et al. used mathematical modelling based on respiratory rate, pulse and motor response to calculate survivability scores for patients. When applied to data from 76,459 blunt trauma patients this approach outperformed injury severity score and revised trauma score prediction.35 These approaches have the advantage in that they appear to capture the sense of both severity of injury and implied survivability. Of the studies considered in this review, only the Wallis studies used prospectively collected data based on field determination of triage categories.38,39 Yet even this study was limited by the fact that in most situations triage decisions would have been limited to single or small number of victims at any one time. Retrospective analyses of patient records or registries are limited by the fact that they do not take into consideration how the triage tool performs in practice. This is important as error rates in translating simple physiological variables into a severity score can run as high as 30%.33 In addition, error rates may increase when the system is under strain as may be the case during a mass casualty incident.37 Thus the real performance characteristics of all of the various triage tools in true mass casualty incidents are unknown. The application of effective and accurate triage is also influenced by the preparation and education surrounding the tools and processes to be used. Each mass casualty incident has its own unique characteristics hence preparing the healthcare community has become an expanding and increasingly complex area to ensure cohesive responses to all threats.9 It is also unclear if performing disaster drills is actually effective in improving response skills of healthcare providers,19,40 as there are few objective data within the literature assessing pre- and post-drill knowledge8 and its retention. Despite the increasing likelihood of a big bang major incident will involve chemical, biological, radiological or nuclear (CBRN) element, the literature review failed to reveal the existence of a triage tool validated for use in such circumstances. The triage tools for conventional major incidents are derived from the physiological response to trauma and cannot be extrapolated to CBRN incidents. The range of physiological responses and ability to prognosticate will differ following the release of lung-damaging, blood, blister, and nerve agents.6 Additional gaps in the evidence
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
467
Table 6 Summary of data analysis by triage tool (child). Triage tool
Sensitivity
Specificity
Indicator
Subjects
Paper
Paediatric triage tape
ISS > 15 37.8% (95% CI 32.7–42.5) NISS > 15 26.1% (95% CI 23–28) Intervention (Garner) 41.5% (95% CI 36.8–45.6)
Prediction of severity of injury (Injury Severity Score, and New Injury Severity Score) and the need for specific intervention
3461
Wallis 2006
START (>8 years)
ISS > 15 31.3% (95% CI 21.5–42.8) NISS > 15 22.3% (95% CI 15.6–30.7) Intervention (Garner) 39.2% (95% CI 29.3–50) ISS > 15 Sensitivity 3.2% (95% CI 1.3–7.5) NISS > 15 2.4% (95% CI 1–5) Intervention (Garner) 0.8% (95% CI 0.1–4.1) ISS > 15 48.4% (95% CI 43.4–52.8) NISS > 15 31.5% (95% CI 28.5–34.1) Intervention (Garner) 46% (95% CI 41.2–50.2)
ISS > 15 98.6% (95% CI 98.3–98.8) NISS > 15 98.5% (95% CI 98.5–99.1) Intervention (Garner) Specificity 98.9% (95% CI 98.6–99.2) ISS > 157 7.9% (95% CI 77.3–78.9) NISS > 15 77.3% (95% CI 76.6–78.3) Intervention (Garner) 78.7% (95% CI 77.9–79.5) ISS > 15 97.8% (95% CI 97.7–98) NISS > 15 97.8% (95% CI 97.6–98) Intervention (Garner) 97.7% (95% CI 97.6–97.8)
Prediction of severity of injury (Injury Severity Score, and New Injury Severity Score) and the need for specific intervention
3461
Wallis 2006
Prediction of severity of injury (Injury Severity Score, and New Injury Severity Score) and the need for specific intervention
3461
Wallis 2006
Prediction of severity of injury (Injury Severity Score, and New Injury Severity Score) and the need for specific intervention
3461
Wallis 2006
JumpSTART (1–8 years)
CareFlight
ISS > 15 98.8% (95% CI 98.6–99.1) NISS > 15 99% (95% CI 98.7–99.3) Intervention (Garner) 98.9% (95% CI 98.6–99.1)
based are around the use of triage tools as a basis for the deferral of treatment because the level of resource required by an individual would cause an unacceptable impact on other individuals. Triage tools are used to determine which patients to transfer to a trauma centre in a number of healthcare systems. These triage tools are usually more complex than those used for major incidents, where accuracy of prediction is more important than the timeliness to complete the assessment. Despite their increased complexity (e.g. inclusion of mechanism of injury, anatomical injury patterns, physiological response, patient demographics) no single triage system has been shown to be 100% accurate.4,32 In the setting of a major incident, the ability to reliably exclude patients that do not require urgent treatment is important if the limited resources are to avoid being overwhelmed.15,16 The limitations to existing triage systems were recently reviewed by a multidisciplinary committee in the US. The conclusions from their review in terms of the paucity of evidence of validated systems were similar to ours.27 This group went on to use a consensus approach to develop a new triage guideline (the sort assess, life-saving interventions, treatment/transport (SALT) tool), which is proposed to standardised mass casualty triage processes across the US. Early evaluations of this tool in simulated incidents suggest that the SALT mass casualty triage system can be applied quickly and confidently in the field. It appears to be safe with relatively little under triage at the cost of significant over triage.6,7,26 Like many of the other tools it does not take into consideration the degree by which the system is overwhelmed. There remains a need to examine the effect of this tool on patient outcomes. Conclusions There is limited evidence for the validity of existing triage tools which are proposed for use in a big bang major incident. Where evidence does exist it appears to focus on the sensitivity and specificity in relation to either the prediction of trauma death or the prediction of severity of injury based on data from single or small number patient incidents. None of the traditional triage tools take account of the degree by which the health resources are overwhelmed. In our opinion, the Sacco triage system appears the
most promising in respect to its inclusion of resources allocation and its unique calculation of survivability index. The practicalities, training implications, performance characteristics and reliance on computer technology during a mass casualty incident all require further evaluation. Conflict of interest There are no conflicts of interests. Funding This study is part of work funded by the Emergency Preparedness Clinical Leadership Advisory Group, Department of Health. The funders had no involvement in the study design: collection, analysis and interpretation of data; the writing of the manuscript or the decision to submit the manuscript for publication. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.injury.2010.11.005. References 1. Advanced Life Support Group. Major incident medical management and support, 2 ed., London: BMJ Books; 2002. 2. Baxt WG, Upenieks V. The lack of full correlation between the Injury Severity Score and the resource needs of injured patients. Ann Emerg Med 1990;19:1396–400. 3. Benson DO, Keonig KL, Schultz CH. Disaster triage: START then SAVE - a new method of dynamic triage for victims of a catastrophic earthquake. Prehosp Disaster Med 1996;11:117–24. 4. Carron PN, Taffe P, Ribordy V, et al. Accuracy of prehospital triage of trauma patients by emergency physicians: a retrospective study in western Switzerland. Eur J Emerg Med 2010 Aug 19. 5. Committee on Trauma, American College of Surgeons, Resources for the optimal care of the injured; 2006. 6. Cone DC, MacMillan DS, Parwani V, Van Gelder C. Pilot test of a proposed chemical/biological/radiation/nuclear-capable mass casualty triage system. Prehosp Emerg Care 2008;12(April–June (2)):236–40. 7. Cone DC, Serra J, Burns K, et al. Pilot test of the SALT mass casualty triage system. Prehosp Emerg Care 2009;13(October–December (4)):536–40. 8. Cooke MW, Brace SJ. Training for disaster. Resuscitation 2010;81:788–9.
468
T.M. Kilner et al. / Injury, Int. J. Care Injured 42 (2011) 460–468
9. Department of Health. NHS Emergency Planning Guidance 2005; 2005. http:// www.dh.gov.uk/en/Publicationsandstatistics/Publications/PublicationsPolicy AndGuidance/DH_4121072 (accessed 12.10.10). 10. Garner A, Lee A, Harrison K, Schultz CH. Comparative analysis of multiplecasualty incident triage algorithms. Ann Emerg Med 2001;38:541–8. 11. Gebhart ME, Pence R. START triage: does it work? Disaster Manage Resp (DMR) 2007;5:68–73. 12. Gausche-Hill M. Pediatric disaster preparedness: are we really prepared? J Trauma 2009;67(2 Suppl.):S73–6. 13. Grimes DA, Schultz KF. Uses and abuses of screening tests. Lancet 2002;359: 881–4. 14. Hidden A. Investigation into the Clapham Junction Railway Accident; 1989. http://www.railwaysarchive.co.uk/documents/DoT_Hidden001.pdf (accessed 12.10.10). 15. Hirshberg A, Frykberg ER, Mattox KL, Stein M. Triage and trauma workload in mass casualty: a computer model. J Trauma 2010 [Epub ahead of print]. 16. Hirshberg A, Scott BG, Granchi T, et al. How does casualty load affect trauma care in urban bombing incidents? A quantitative analysis. J Trauma 2005;58(April (4)):686–93. discussion 694–5. 17. Hodgetts TJ, Hall FJ, Maconochie I. Paediatric triage tape. Prehosp Immed Care 1998;2:155–9. 18. Hong R, Sierzenski PR, Bollinger M, et al. Does the simple triage and rapid treatment method appropriately triage patients based on trauma injury severity score? Am J Disaster Med 2008;3:265–71. 19. Hsu EB, Jenckes MW, Catlett CL, et al. Effectiveness of hospital staff masscasualty incident training methods: a systematic literature review. Prehosp Disaster Med 2004;19:191–9. 20. Husum H, Gilbert M, Wisborg T, et al. Respiratory rate as a prehospital triage tool in rural trauma. J Trauma – Injury Infect Crit Care 2003;55:466–70. 21. International Liaison Committee on Resuscitation: Defining Quality of Evidence.doc. http://mc.manuscriptcentral.com:80/societyimages/ilcor/Defining%20 Quality%20of%20Evidence.doc (accessed 12.05.09). 22. Iserson KV, Moskop JC. Triage in medicine, part I: concept, history, and types. Ann Emerg Med 2007;49(3):275–81. 23. Kahn CA, Schultz CH, Miller KT, Anderson CL. Does START triage work? An outcomes assessment after a disaster. Ann Emerg Med 2009;54(3):424–30. 430.e1. 24. Leach SR, Swor RA, Jackson RE, et al. Do outcome measures for trauma triage agree? Prehosp Emerg Care 2008;12:467–9. 25. Lerner EB, Schwartz RB, Coule PL, et al. Mass casualty triage: an evaluation of the data and development of a proposed national guideline. Disaster Med Public Health Preparedness 2008;2(Suppl. 1):S25–34.
26. Lerner EB, Schwartz RB, Coule PL, Pirrallo RG. Use of SALT triage in a simulated mass-casualty incident. Prehosp Emerg Care 2010;14(January–March (1)): 21–5. 27. Lerner EB. Studies evaluating current field triage: 1966–2005. Prehosp Emerg Care 2006;10:303–6. 28. Morley P. Evidence evaluation worksheets: the systematic reviews for the evidence evaluation process for the 2010 International Consensus on Resuscitation Science. Resuscitation 2009;80:719–21. 29. National Audit Office. Facing the challenge: NHS emergency planning in England, HC 36. London: National Audit Office; 2002. 30. neverforget911.org. Statistics and numbers from the September 11th, 2001 attacks 2002. http://neverforget911.org/statistics.aspx (accessed 12.10.10). 31. Newgard CD, Rudser K, Hedges JR et al. The ROC investigators. A critical assessment of the out-of-hospital trauma triage guidelines for physiologic abnormality journal of trauma-injury infection & critical care, vol. 68(2); February 2010, pp. 452–462. 32. Osen HB, Bass RR, Abdullah F, Chang DC. Rapid discharge after transfer: risk factors, incidence, and implications for trauma systems. J Trauma 2010;69(September (3)):602–6. 33. Prytherch DR, Smith GB, Schmidt P, et al. Calculating early warning scores – a classroom comparison of pen and paper and hand-held computer methods. Resuscitation 2006;70(August (2)):173–8. Epub 2006 Jun 27. 34. Romig LE. Pediatric triage. A system to JumpSTART your triage of young patients at MCIs. J Emerg Med Services 2002;27:52–8. 35. Sacco WJ, Navin DM, Fiedler KE, et al. Precise formulation and evidence-based application of resource-constrained triage. Acad Emerg Med 2005;12(August (8)):759–70. 36. Sacco WJ, Navin DM, Waddell II RK, et al. A new resource-constrained triage method applied to victims of penetrating injury. J Trauma 2007;63(August (2)):316–25. 37. Smith AF, Oakley RJ. Incidence and significance of errors in a patient ‘track and trigger’ system during an epidemic of Legionnaires’ disease: retrospective case note analysis. Anaesthesia 2006;61:222–8. 38. Wallis LA, Carley S. Comparison of paediatric major incident primary triage tools. Emerg Med J 2006;23:475–8. 39. Wallis LA, Carley S. Validation of the paediatric triage tape. Emerg Med J 2006;23:47–50. 40. Williams J, Nocera M, Casteel C. The effectiveness of disaster training for health care workers: a systematic review. Ann Emerg Med 2008;52:211–22. 41. World Health Organisation. Global alert and response – what is the pandemic (H1N1) 2009 virus?; 2009. http://www.who.int/csr/disease/swineflu/ frequently_asked_questions/about_disease/en/index.html (accessed 14.10.09).