Statistical Analysis to Quantify the Usefulness of Triage Criteria in Predicting Seriously Injured Patients in Order to Refine Criteria for Activating Level One or Level Two Trauma Calls
- 1. Queen Alexandra Hospital, England
- 2. Department of Anaesthesia and critical care, Univeristy Hospital Southampton, England
- 3. Department of Emergency medicine, Salisbury Hospital, England
ABSTRACT
Background: A trauma triage system uses numerous variables to activate trauma calls and is important for facilitating rapid diagnosis and treatment whilst ensuring optimal resource allocation. This project involved the use of retrospective statistical analysis to quantify the usefulness of various triage criteria in predicting severely injured patients. Following this, the objective was to refine the criteria for activating level one or level two trauma calls in the Wessex Trauma Network (WTN).
Methods: All patients directly admitted to hospitals in the WTN either with or without Trauma Team Activation (TTA) were investigated retrospectively from 2013-2014 (n=4267). Different TTA criteria were analyzed with respect to sensitivity, under-triage, specificity, positive/negative predictive values and overtriage. The current triage criteria were compared to a proposed revised set of criteria.
Results: For the TTA criteria currently in use at UHS, the sensitivity was 87.42% and the over-triage 45.23% when using ISS>15 as a standard of reference for major trauma. In general, physiological criteria were the most powerful predictors of severe injury, with mechanism of injury (MOI) criteria contributing the most to over-triage. The proposed revised set of criteria was retrospectively applied to the data for patients directly admitted to UHS in 2013/14(n=800), resulting in a sensitivity of 85.31% and over-triage of 30.75%.
Conclusion: Findings showed that a simplified revised set of TTA criteria utilizing predominantly physiological and anatomical criteria can safely reduce over-triage without a substantial decrease in sensitivity. There will always be a degree of under-triage, however it is possible to minimize it.
KEYWORDS
Trauma ; Activation ;Criteria ; Triage.
CITATION
Ringrose T, Kellett S, Klein J (2020) Statistical Analysis to Quantify the Usefulness of Triage Criteria in Predicting Seriously Injured Patients in Order to Refine Criteria for Activating Level One or Level Two Trauma Calls. J Trauma Care 4(1): 1033.
ABBREVIATIONS
TARN: Trauma Audit and Research Network) - a national organisation that collects and processes data on moderately and severely injured patients in England and Wales; WTN: Wessex Trauma Network - a regional network comprising all hospitals in the Wessex area along with the region’s ambulance trusts; MTC: Major Trauma Centre – a hospital equipped and staffed to provide care for patients suffering from major traumatic injuries; UHS - University Hospital of Southampton; MOI: Mechanism of Injury – the circumstance in which an injury occurs, for example, high fall, road traffic collision, penetrating injury; GCS: Glasgow Coma Scale– the summation of scores for eye, verbal and motor responses, used to describe the level of consciousness of a person following traumatic brain injury; ED: Emergency Department; ISS: Injury Severity Score - is an anatomical scoring system that provides an overall score for patients with multiple injuries.
INTRODUCTION
In England, trauma is the leading cause of death in people under the age of 40, highlighting the importance for the need of high quality of care from multiple specialists [1]. Major trauma networks are structured groups of personnel and services with the objective of reducing mortality and morbidity following trauma. They have improved the quality of care for severely injured patients since they allow patients to be seen by the most appropriate personnel immediately upon admission.
The involvement of the correct personnel enables hospitals to reach patient care standards and hence generate more income. Regional trauma networks were first introduced in the UK in 2012 facilitating safe and efficient delivery of patients to designated Major Trauma Centres (MTC) in the country, the University Hospital of Southampton (UHS) being one of these. The Wessex trauma network (WTN) includes eight hospitals in the Wessex region and all of the ambulance and air ambulance trusts [2].
An independent audit by the ‘Trauma Audit and Research Network’ (TARN) showed that following the introduction of MTC’s, 1 in 5 patients who would have died from severe injuries are now surviving, a 20 per cent improvement [1]. Trauma triage systems seek to provide this level of care to those requiring it whilst preventing the inappropriate use of financial and personnel resources. The activation of a multi-disciplinary team is an integral part of the trauma care system and has been shown to improve outcomes. The current Adult Major Trauma Team Activation System at the University Hospital Southampton, and in fact all other hospitals in the network, work on a two-tier response to trauma calls. The generation of a level one trauma call will alert a full trauma team response, whereas a level two trauma call for less severe trauma, will alert the Emergency Department trauma team alone. Hence, the primary reason for stratifying is to minimise the unnecessary disruption to other emergency work taking place. For every level one trauma call that is made, orthopaedic, general and anaesthetic specialists along with numerous other healthcare professionals, must attend therefore causing inevitable disruption and impact to other oncall workload.
The Injury Severity Score (ISS) is a complex scale used to assess the severity of trauma. It was designed by the ‘Abbreviated Injury Scale (AIS) committee of the Association for the Advancement of Automotive Medicine’ using consensus data. A major trauma or a poly-trauma is classified by the ISS exceeding a score of 15 [3-5].
Trauma team activations (TTA) are based upon information passed to the Emergency department in a pre-alert and aid the department in deciding which call to put out (level one or two). There are a series of criteria that activate a level one trauma call involving physiological, anatomical and mechanism of injury, criteria of which are numerous (e.g.: GCS <10 or Road traffic accident etc). If pre-hospital information meets at least one of the TTA criteria then the trauma team will be activated [6]. Several studies have investigated the criteria for TTA ( 8-15) . However it has often been challenging to define an optimal set of criteria since there are many factors which play a determining role: pre-morbidity, mechanism of injury, the extent of injury, patient heterogeneity) [7-16]. Many systems use a variation of the recommended field triage criteria in the guidelines of the ‘American College of Surgeons - Committee on Trauma’, including physiological, mechanism of injury and anatomical criteria. Mechanism of injury criteria allows identification of additional severely injured patients that may not display physiological signs so hence would not have been picked up otherwise. However the use of MOI criteria has a tendency to over-triage and incorrectly identify patients as severely injured [17]. Ideal trauma team activation criteria should work effectively to initiate the appropriate level of trauma response by activating either a level one or level two trauma call. It is therefore important that the criteria is sensitive in order to minimise under triage but also specific in order to minimise over triage also. The use of triage activation criteria brings about a trade off between over-triaging (activating a level one trauma call when not entirely necessary) and under-triaging (not activating the call when the patient would have needed or benefited from it). ‘The American College of Surgery – Committee on Trauma’ state in their guidelines that in general, priority should be given to decreasing under-triage since it can result in preventable mortality or morbidity resulting from delays in definitive care. Under-triaging can delay diagnosis and treatment of severely injured patients and hence worsen prognosis and increase mortality. There will always be an element of over-triaging which is very disruptive, especially in smaller hospitals with fewer staff and resources. The predominant problem with over-triaging is resource allocation, since members of the multi-disciplinary trauma team will be diverted away from the work load of their day to day specialities, which can in turn compromise the care of other patients and increase hospital costs. Unfortunately, attempts to decrease over-triage are likely to come at the cost of an increase in under-triage. Therefore it is necessary that trauma centres find a balance between utilising scarce resources and the requirement for acute care in severely injured patients [7-11,17].
A repeating trend amongst other studies assessing efficacy of activation criteria in predicting major trauma is that Mechanism of injury(MOI) criteria has a lower predictive value and sensitivity, contributing the most to over-triage and an unnecessary use of resources [7,9,10,12]. However, this does not rule out the possibility that some individual MOI criteria might possess significant predictive power alone [9]. A study conducted by the American college of surgeons discovered that physiological criteria was the most powerful at predicting severely injured patients (ISS >15) [10]. On the contrary, multiple studies have found that the frequently used physiological variable of heart rate is not a powerful predictor of severely injured patients. It is likely that this is due to the large number of associating factors that can influence heart rate (e.g.: level of fitness, fever, alcohol use, etc [12,19].
The aim of this report is to use retrospective statistical analysis to quantify the usefulness of various triage criteria in predicting seriously injured patients. This is in order to refine criteria for activating level one or level two trauma calls within the Wessex Trauma Network (WTN), with the overall intention being to lead to an improvement in the efficiency of trauma care within the region. Compared to many of the other similar studies, the sample size for this project is comparatively several times larger at 4,267 patients. A two tiered trauma response system has been developed and implemented in the University Hospital of Southampton, whereas some of the similar studies took place in only one tiered trauma systems.
MATERIALS AND METHODS
This study involved retrospective statistical analysis of all patients directly admitted to hospitals in the WTN in 2013 and 2014 (n=4267) , both with a pre-alert and without a prealert. Data was requested from the Trauma Audit and Research Network database regarding patients admitted to the hospitals in the WTN in 2013/14 and their corresponding ISS’s and injuries. Additional data on these specific patients was also obtained through the NHS server via a secure login. This involved recording relevant information from the patient notes such as physiological observations both pre hospital and in the emergency department.
The various TTA criteria being statistically analysed were retrospectively applied to the cohort of patients to see how many of the patients admitted with particular ISS’s fulfilled or activated these criteria either at the scene of the incident or in the emergency department. A table was constructed (Figure 1)
Figure 1 Table of the data showing the number of patients that fulfilled each TTA criteria
This table contains the data obtained following interpretation of the data sent from TARN and using the secure NHS database. The table shows all of the patients directly admitted to hospitals in the WTN in 2013/14, the ISS bracket they fall under and the number of patients that fulfilled each TTA criteria. The table also specifies as to whether a pre-alert was made to the hospital beforehand or not. See appendices for enlarged version of the table.
formatting the number of patients that activated each individual criteria being examined, the ISS bracket they fell under, and whether the hospital was pre-alerted in each instance or not.
There were some limitations in regards to the accessibility of certain information in the data collection process. For certain activation criteria, namely the physiological and intubation criterion, data for 2014 was only available to access for University Hospital Southampton due to login restrictions. It is relevant to mention that the UHS comprises a significant proportion of the patient population in this study. Therefore this limitation would not affect the validity of the results as it merely means that for certain criteria the sample size is partially decreased. There were also several criteria for which only 2013 data was available, these being multiple blunt injuries, spinal cord injury and crushed/ degloved/amputated limbs. However this was not an issue with all remaining criteria for which data was available from all of the hospitals in the WTN for both 2013 and 2014.
Following full collection of the data and the completion of the table comprising all patients admitted with ISS ranging from 1 to 65, it was then possible to begin conducting statistical tests on the data. Numerous outcomes were calculated for each criteria including sensitivity, under triage, specificity, over triage and positive and negative predictive values (Figure 2).
Figure 2 The statistical outcomes calculated for each individual criterion Sensitivity = a/ (a+c), under-triage = 1- sensitivity = c/a+c, specificity = d/ (b+d), PPV = a/ (a+b), over-triage = 1-PPV = b/ (a+b), NPV = d/(c+d). Specificity and NPV are highlighted since they were decided to be of limited value to the study but have been included for completeness.
Specificityis defined as the probability of no TTA amongst patients with minor injury. Whilst ‘Negative Predictive Value (NPV)’ refers to the probability of minor injury amongst all patients that did not activate a pre-alert. The outcomes remain in the table for completeness, however it was decided that the ‘specificity’ and ‘NPV’ were of limited value since most patients with minor injury are never even considered for the activation of the trauma team.
The findings were then utilised with the aim of creating a more refined and optimal set of TTA criteria. This new refined set of criteria was then applied retrospectively to the same TARN and NHS data from 2013/14, in order to observe how many severely injured patients would have been missed had this revised set of data been in use.
RESULTS AND DISCUSSION
Following statistical analysis it can be observed that in general, physiological criteria tend to have a higher sensitivity and hence are more powerful predictors of severe injury. However, it should be noted that there are certain physiological criteria which are not as powerful predictors of major trauma: ‘ED respiratory rates’ and ‘ED/pre hospital GCS 10-12’ are amongst these (Figure 3).
Figure 3 Tables depicting the performance of various TTA criteria being analysed The table demonstrates the performance of the various TTA criteria at predicting severe injury (ISS >15) using all patients admitted to hospitals in the WTN in 2013/14
In overview, both MOI and Anatomical criteria tended to be less sensitive predictors however there were exceptions to this trend, the most significant being the ‘penetrating injury’ and ‘Long bone fractures in multiple limbs’ criterion. Penetrating injury in fact turned out to be the most powerful of all of the TTA criteria in terms of sensitivity (90.91%), and ability to predict severe injury. However it must be noted that this is of limited significance in a region such as Wessex, since the incidence of penetrating injury is relatively low in comparison to other mechanisms of trauma. Penetrating injury in fact also turned out to be the TTA criteria that contributed the most to over-triage (62.96%). Under triage
tended to follow the same trend with the most sensitive criteria contributing the least to under-triage. Therefore the findings demonstrate that MOI/Anatomical criteria are likely to result in a greater proportion of severely injured patients not being correctly identified as such, which can result in devastating and fatal outcomes. However the advantage of also incorporating MOI/Anatomical criteria is that despite higher levels of undertriage when taken independently, they do identify some degree of patients that may have been missed by just physiological criteria. Hence the criteria can still be of use but only when taken in combination with physiological criteria. Although, it is relevant to mention that with the use of MOI and anatomical criteria, often come higher rates of over-triage, the implications of which have been previously discussed.
The positive predictive value can be defined as the probability of the patient in question actually being severely injured (ISS>15) when the criteria is activated (pre-alert). Being the converse of this, over-triage is hence defined as ‘1-PPV’. In general, MOI criteria tended to contribute the most to over-triage (Penetrating injury and Crushed/Degloved/Amputated limb contributing the most). Therefore on the whole, anatomical and physiological criteria tended to contribute less so to over-triage with the criteria of ‘Flail chest’ in fact contributing 0% to over-triage since the mere occurrence directly implies severe injury
This study reports that MOI criteria in general have a lower predictive power than physiological and anatomical criteria in terms of ability to predict severe injury. This theme seems to be consistent with the results of previously conducted studies [7,9,10,12]. Another common theme amongst other studies is that MOI criteria tend to contribute the most to over-triage. This is re-enforced by this study with ‘Penetrating Injury’ and ‘Crushed/ degloved/amputated limb’ being the two criteria contributing the most to over-triage, with over-triage rates of 62.96% and 57.14% respectively (Figure 3).
The findings of the study showed that the mechanism of injury criteria of ‘Penetrating injury’ in fact turned out to be the criteria with both the greatest predictive power and also the largest rate of over-triage. It is significant to mention that the ‘Penetrating injury’ criteria is likely to be more relevant to other geographical regions, which have a greater incidence of penetrating injury, for example the inner cities of major urban areas. Penetrating injury can encompass a somewhat widespread variety of injuries, many of which include more minor stab wounds or air powered rifle wounds. Therefore, it would be advisable that a more specific and localised criteria would be more useful. This perhaps would be achieved by including the anatomical location of the penetrating injury, e.g.: to the chest/vital organ zone on the torso..
Patterns of trauma can vary significantly between different regions and other studies have even shown that trauma patterns can still be variable amongst hospitals in the same region [8]. Hence this could affect the applicability of the proposed revised set of trauma activation criteria in relation to other regions.
The statistical analysis in this study involved including age as a criterion in order to observe if there was any significant trend related to age as a factor in the trauma triage protocol. The findings indicate that there is an inverse correlation between sensitivity and age, since sensitivity decreases as age increases. Conversely there was a positive relationship between age and the rate of over-triage, since over-triage increases as age increases. It is believed that perhaps these trends may be due to symptoms often being somewhat more covert and masked in the elderly population due to the greater likelihood of existing co-morbidities.
In order to minimise any increase in under-triage that might result from the removal of certain criteria, only the specific criteria with both a low predictive power and a high contribution to over-triage were removed in the formation of the revised set of activation criteria. It is relevant to mention that although information recorded in the ED can be useful for comparison, it is only pre hospital observations that are of significance for inclusion in the revised TTA criteria. This is since only information known prior to the patient arriving at the hospital can be used to decide whether a level one trauma call should be activated or not
The findings from the statistical analysis were used and several modifications to the revised set of TTA criteria were made. ‘Flail chest’ along with the top 6 most sensitive pre hospital physiological criteria were taken as an initial sample to be retrospectively applied to the 2013/14 patient data. This therefore included ‘flail chest’, ‘intubated pre hospital’, ‘pre hospital adult systolic BP 130’ and ‘pre hospital resp rate 30’. As discussed previously, the criterion of ‘flail chest’ has an over-triage rate of 0%, so although it may not have a particularly high sensitivity, it will not contribute to over-triage so is hence worth including. This is likely to be the case since it would be rare for a flail chest to occur in isolation; hence it is very probable that other traumas will have occurred simultaneously.
Following retrospective application to a sample of 134 severely injured patients directly admitted to UHS in 2013/14, it was found that only 90 (67%) of these would have been correctly identified as severely injured had this set of TTA criteria been in use. Consequently this prompted some alterations to be made to the proposed criteria (See figure 4).
Figure 4 Tables depicting the retrospective application and performance of the revised set of criteria The figure illustrates the numerous stages involved in refining the proposed TTA criteria. The first and the second proposed sets were initially applied retrospectively to a relatively small patient sample, it was clear that these sets were not sensitive enough, hence prompting further modifications to be made. All proposed sets of criteria were applied to TARN data for patients directly admitted to UHS in 2013/14.
Amongst these modifications were the adjustment of the parameters of certain activation criteria (Resp rate 25, Pulse rate 120). As shown in Figure 4, implementation of these changes, raised the sensitivity of the proposed criteria to 76% (133 of 174 severely injured patients being identified).
However, as depicted in Figure 6, the TTA criteria currently in use in UHS had a sensitivity of 87.42% and over-triage rate of 45.23% over 2013/14. Consequently with this in consideration, several modifications were implemented, namely the inclusion of ‘pre hospital O2 saturation on air 110).
It is relevant to mention that there were some limitations in relation to the retrospective application of the revised criteria. Due to restrictions regarding the accessibility of data, the criteria could only be applied to those patients directly admitted to UHS (n=800).
CONCLUSION
The final proposed set of TTA criteria was retrospectively applied to patients directly admitted to UHS in 2013/14 (n=800). The revised criteria would have yielded a sensitivity of 85.31% and an over-triage rate of 30.75% had it been in use. The performance of the revised set of TTA criteria was then compared to the current criteria in use, as shown in Figure 5
Figure 5 Comparison of the performance of the current TTA criteria and the proposed set of criteria The tables depict the performance of the two sets of TTA criteria when utilizing ‘all patients admitted to UHS in 2013/14’ as the patient sample in both instances.
The simplified TTA criteria significantly reduced the rate of overtriage from 45.23% to 30.75%. Although the rate of under-triage increased from 12.58% to 14.69%, the increase is not substantial and is likely over-shadowed by the benefits of such a significant reduction in the rate of over-triage. Unfortunately given the nature of trauma, there will always be some level of under-triage, however it is possible to minimise it. Utilising these findings, a simplified revised set of TTA criteria using predominantly physiological and anatomical criteria can act to safely reduce over-triage without a substantial increase in under-triage