The Utstein template for uniform reporting of data following major trauma

To permit data collection and statistics on major trauma care, in 1999 a working group from the International Trauma Anaesthesia and Critical Care Society (ITACCS) published a recommendation for the Utstein Template for Uniform Reporting of Data following Major Trauma [1]. The template extracted data for the pre-hospital phase, early in-hospital management, and for co-morbidity and outcome. In accordance with the previous Utstein templates, it was commended that data were to be classified as 'Core' (essential) or 'Optional' (supplemental). Despite the intention of facilitating studies to improve the understanding of trauma and trauma care, only a few papers have been published based on the template [2, 3]. This indicates a need for further development, and in particular, a major reduction in the large number (92) of core data variables [1], as well as the addition of more precise definitions of these variables [3].

Trauma registries

Due to the practical difficulties with performing randomised controlled trials in severe trauma cases, valid scientific evidence is often lacking. Systematic prospective registry-based data collection for documenting trauma care is performed by several local, regional and national trauma registries. However, such registries cannot replace randomised clinical trials, but allow for exploration of relationships present in the collected data. The primary aims of these trauma registries are to enable comparative analyses of trauma care and outcome to provide quality improvement and optimal care of the injured patients [4]. The development of a European trauma registry may provide population-based comprehensive data on trauma incidence, epidemiology and trends. Further, it may enable development of regional outcome prediction models (taking special European factors into consideration) and thus set baseline norms for future trauma outcome studies. In Europe, there has been some reluctance to share local and national data, but it is recognised that lessons learned in one area of Europe may be useful for other European states [5]. However, when comparison is conducted, it is important to ensure that the reasons for differences in outcome are due to differences in the quality of trauma care or to differences in trauma systems, and not to variations in population characteristics [6].

TRISS methodology

Over the last two decades, the Trauma and Injury Severity Score (TRISS) method [7, 8], with coefficients for prediction of outcome has been the most commonly used method for comparison of outcome in trauma patients. The TRISS coefficients were originally derived from the United States Major Trauma Outcome Study (US MTOS) [9, 10] but more recently the coefficients have been updated based on patient cases from the National Trauma Data Bank [11]. However, the TRISS method has some limitations, and it has been criticised by many authors [7, 12–21]. Among other things, the TRISS model requires scoring the Revised Trauma Score (RTS) [22] components (Glasgow Coma Scale [GCS] [23], respiratory rate [RR] and systolic blood pressure [SBP]) on admission in the emergency department (ED), and does not take into account co-morbidity. Despite its limitations, TRISS continues to be the most accepted and widely-used tool for comparing trauma outcome in North America and in some parts of Europe.

Comparing and benchmarking European trauma care

In Europe, the UK Trauma Audit and Research Network (TARN) [24], along with the Trauma Registry of the German Society of Trauma Surgery (DGU-TR) [25], represent the largest trauma registries. There has also been a move towards developing a European Trauma Audit and Research Network (EuroTARN) [26], and a core dataset with inclusion and exclusion criteria has been created. Nevertheless, to date, no consensus has been reached between countries on the details and extent of the dataset. A first report from EuroTARN concluded that it is possible to collect data from established trauma registries, and the initial analysis revealed significant international variation [5]. As a continuation of this effort, a European project has been initiated by the DGU-TR, UK TARN and the Scandinavian Networking Group for Trauma and Emergency Management (SCANTEM) [27], for developing a joint European Core Dataset (EuroCoreD) for a future European Trauma Registry.

The 2007 revision of the Utstein template for uniform reporting of data following major trauma

Despite significant efforts [1, 5], comparison of trauma care and outcome within Europe has not yet been carried out in a systematic way, mainly because inclusion criteria, data definitions and coding formats vary significantly between registries, and also because patient selection is not comparable [5, 28]. Further efforts to establish uniform and standardised inclusion and exclusion criteria, as well as a minimum list of core data variables with precise definitions, are essential [3, 5]. In addition, consistent methods of injury scoring need to be agreed upon [4, 29–31]. To address this need for a European consensus, SCANTEM, TARN, DGU-TR and the Italian National Registry of Major Injuries (RITG) [32] carried out a consensus process, concluding with symposia in May and December 2007 at the Utstein Abbey [33], Norway. Selected experts met with the aim of further developing the Utstein Template for Uniform Reporting of Data following Major Trauma. At that time, they defined inclusion and exclusion criteria, and a minimum core dataset with precise definitions. In addition, the aim of the revised template was to develop a standard for comparison of trauma data that was compatible with the large trauma registries in Europe, also adhering to EuroTARN. The template was intended to support the establishment of a European Trauma Registry, promote further development of a European model for outcome prediction and allow European and international trauma auditing and benchmarking.

The expert panel

The expert panel was comprised of those individuals who were central to developing and managing the largest European trauma registries; the panel included clinicians, database managers and epidemiology experts.

Data variable definition

A data variable should be unambiguously defined (with no misinterpretations) and reasonably simple to register. To meet this requirement, a data variable dictionary should contain information on 'data point number,' 'data point name,' 'descriptive field name,' 'type of data,' 'data point category/value,' 'definition of data point,' 'source of data information' and 'coding guidance.' We based recommended guidelines for data variable definitions on existing trauma registry databases, the Utstein Template for Uniform Reporting of Data following Major Trauma [1], the US National Trauma Data Standard (NTDS) [36] and the Injury Surveillance Guidelines from the World Health Organization (WHO) [37].

Core data variables

A registry should differentiate between data variables that absolutely need to be collected (core data) and the type of additional data that may be desirable (optional data) [1, 37]. The current revision focuses on core data that are considered to be essential for documentation and reporting. We divided the core data into three groups ('Predictive Model,' 'System Characteristic Descriptors' and 'Process Mapping Variables') based on the role of the data variable in a registry.

Predictive model

The predictive model is composed of patient and injury severity variables that are considered to be important for outcome prediction. Predictive models are not determinative; rather, they provide the probability of an outcome for a given patient [38]. Complex models, such as Abbreviated Injury Scale (AIS) [39] derivatives and the RTS, are often used to create such predictive models [38]. Experience from the German and UK trauma registries suggests that there may be better data variables to include in a predictive model than those traditionally used in the TRISS methodology [24, 40–42].

System characteristic descriptors

Data variables in the System Characteristic Descriptor group describe trauma systems. Within Europe, there are large differences in philosophies and structures of trauma care systems, and these data variables should indicate key differences between systems and permit comparisons of the effect of system structure on outcomes.

Process mapping variables

Process mapping variables are intended to describe trauma care at an individual trauma centre (e.g., what happens to a patient after a major trauma); these are used for documentation of the patient journey, care process and care activities.

Specific premises

At present, many trauma registries have difficulty in obtaining data for patients from all involved hospitals when patients are transferred between them; therefore, the expert panel based their consensus on the premise that the core dataset was intended to cover the main hospital where a patient is treated. However, the expert group recommended that all trauma registries develop methods to track patients through the trauma system and that both the primary (local) trauma hospital and the referral trauma hospital record the same set of core data variables. The introduction of a core outcome data variable will secure that the overall effect of the entire trauma system can be measured, even if part of the patient's treatment course is not recorded in detail.

The nominal group technique

Inclusion criteria

NISS is a modification of the Injury Severity Score (ISS) method [43]. ISS is calculated by summing the squares of the highest AIS severity codes in each of the three most severely injured ISS body regions [44]. Hence, ISS will ignore all but the most severe injury in a body region, and often fails to consider worse injuries in other regions of the body [43]. In contrast, NISS is defined as the sum of the square of the three most severe AIS injuries regardless of body region [43]. Several authors have argued for replacing ISS with NISS [43, 45–49]. Osler et al. considered NISS to be easier to calculate and more predictive of survival than the ISS method [43], and a recent study by Lavoie et al. confirmed their findings [46]. NISS will be equal to or greater than ISS for any given patient, and it appears to be a more accurate method for rating severely injured patients [49, 50]; specifically, this is true for patients with multiple head injuries [46]. The increased number of included patients by choosing NISS > 15 instead of ISS > 15 should be seen as an increase in 'sensitivity' without a loss of 'specificity' of an ideal definition of major trauma. An effort should be made to secure that all patients with a NISS > 15 are included, regardless of whether or not the trauma team was activated prior to or upon the patient's arrival at the hospital, and whether or not the patient was admitted to an intensive care unit.

Exclusion criteria

Using NISS > 15 as a single inclusion criterion will include some patients that are at high risk of confounding data analysis. To remove such patients from the analysis, a set of exclusion criteria was defined. The expert panel recommended excluding first hospital admissions more than 24 hours after the injury (e.g., prolonged search and rescue missions), patients declared dead before hospital arrival, or those with no signs of life (pupillary response, spontaneous ventilation, presence of carotid pulse, measurable or palpable blood pressure, extremity movement, or cardiac electrical activity) [51] upon hospital arrival and those having no response to hospital resuscitation. In addition, it was recommended that asphyxias, drowning and burns should be excluded (Table 2).

Pre-hospital deaths should be excluded for practical reasons, since in some countries patients declared dead in the pre-hospital setting are transported directly to the morgue; whereas in other countries, they are admitted to hospital. All patients who arrive in the ED with spontaneous circulation should be included, even if they have had a period of cardiac arrest before being admitted or if they die in the ED.

Asphyxia, drowning and burns are sufficiently different from blunt and penetrating injuries to require other datasets, and need to be considered separately. The UK National Burn Injury Database [52] is currently in use specifically for this purpose. Although the AIS 2005 edition has codes for asphyxia and drowning, such injuries were not included in earlier AIS editions, making comparisons across versions more difficult. In some (but not all) countries, major burn patients are sent to dedicated burn unit hospitals, thereby confounding comparisons. Burn patients should be excluded if the burn represents the predominant injury, or if a patient is treated in a specialised burn unit. In such patients, outcome is determined by factors other than those suggested in this paper. Including burn patients will not represent a sufficient number of patients to report on; hence, burn-related injuries will add little power to the predictive model.

Predictive model variables

Age is an independent predictor of survival after trauma [53, 54]. While the original TRISS model operates with only two age categories, current predictive models utilise different age groups, and we therefore recommend reporting the patient's nominal age (continuous) at the time of injury, in years without decimals, and always rounding down. Patients under one year of age should be reported with one decimal number (e.g., six months is 0.5).

Gender is recommended as a core data, since some studies have reported no association between gender and mortality after traumatic injury [55]; whereas others have found age-specific associations between male gender and outcome [56–58].

An evaluation of type of injury (blunt versus penetrating trauma) is useful for determining which patients are candidates for surgical haemostasis [59], and is essential in the TRISS model [8]. The previous Utstein document recommended that for cases involving both blunt and penetrating injuries, the predominant type of injury should be recorded [1]. The expert panel defined the dominating injury as the one with the highest AIS score. In the rare event of a patient having both blunt and penetrating traumas with the same AIS severity score, penetrating trauma is defined as the predominant injury.

The significance of the mechanism of injury (MOI) in prediction of trauma and outcome is, to a large extent, undetermined [60]. The MOI should be of value for epidemiology or subgroup analysis, and should be described in categories with reasonable prevalence rates. The International Classification of Diseases, 10th revision (ICD-10) [61], chapter XX, External causes of morbidity and mortality (V01-Y98), was initially examined for the purpose of the template; however, it was found to be too detailed, with too many injury codes. Therefore, the expert panel developed a reduced set of categories, which should make data collection easier. The set still enables the analysis of important subgroups, and since it is compatible with the ICD-10 codes, it will allow future category expansion if required.

In the ICD, most injuries can be grouped into two dimensions: intent and mechanism [62]. 'Intention of injury' provides information about the role of the human intent of an injury. The included list of categories is based on the ICD-10 codes, and is selected by the expert panel since it covers most injury intentions.

Pre-hospital cardiac arrest after trauma has dismal outcomes [67–70]; it was therefore chosen as a core data variable.

For patients intubated before arrival at the hospital, the ED GCS is not attainable, and this could potentially result in large amounts of missing data [18]. For this reasons, the pre-hospital, pre-interventional GCS should be recorded, thus permitting its inclusion in future prediction models. Some studies have shown that the pre-hospital GCS score correlate well with the arrival GCS score [71, 72], and that the pre-hospital GCS score is a strong predictor of outcome [73]. A change in the GCS from the field to hospital arrival is highly predictive of outcome [72, 74]. In the RTS, GCS carries the greatest weight [22], and the motor component of GCS appears to contain most of the predictive power of the total GCS score [75]. This component needs further assessment, something that may be facilitated by reporting both the pre-hospital pre-interventional GCS motor component as well as the first recorded GCS motor component upon arrival in the ED/hospital.

Chan et al. found that even in the case of normal SBP upon arrival in the ED, out-of-hospital hypotension was a clinical predictor of severe injury [76]; it was also a strong predictor of the need for emergency surgery, according to Lipsky et al. [77]. Others have documented an increased risk of intra-abdominal injury in blunt trauma patients with hypotension in the hospital [78, 79]. Franklin et al. reported 16% mortality in patients with a period of pre-hospital hypotension but with stable vital signs on ED presentation, vs. 27% mortality for patients with normal pre-hospital SBP who developed hypotension in the ED [80]. Changes in the SBP values, from the pre-hospital pre-interventional value to the SBP value on hospital arrival, could give valuable information about physiologic derangements.

Missing RR values for patients arriving in the hospital (patients intubated before arrival) is the most common cause of a lack of RTS values [81]. Collecting the pre-hospital, pre-interventional RR as well as allowing the use of coded values instead of precise values may compensate for this [18]. In addition, continued documentation of RR for later determination of its predictive power is highly recommended. Creating alternative predictive models may also necessitate the continuous collection of RR values.

Admission arterial base excess (BE) is a predictor of mortality after trauma [40, 82–84]; it is also considered an indicator of haemodynamic instability, high transfusion requirement, and an indicator of metabolic and coagulatory decompensation in trauma patients [40, 85]. Other results indicate that arterial BE is a predictor of intra-abdominal injury [78, 79], and the European guidelines for management of bleeding following major trauma have recommended arterial base deficit as a sensitive test to estimate and monitor the extent of bleeding and shock [59]. Kroezen et al. showed that replacing RTS by base deficit as a measure of physiological disturbance could predict mortality as well as RTS in the TRISS model [86]. The expert panel recommends recording the first arterial BE after arrival in the hospital.

There is a high frequency of established coagulopathy in multiple-injured patients upon their arrival in the ED [87, 88]. Coagulation abnormalities presenting early after trauma have been found to be an independent predictor of mortality [87–90], and the prognostic value of measuring coagulation parameters upon ED admission has been documented [89]. The first measured International Normalized Ratio (INR), obtained within the first hour after hospital arrival, was chosen as a core data for measuring coagulation abnormalities.

The term 'intensive care unit' (ICU) is variously defined across the world, and in some referral hospitals, many patients are intubated upon arrival or transferred back to a local hospital while still intubated and ventilated. This makes the calculation of length of stay (LOS) in the ICU difficult. The revised template recommends recording the total number of days spent on a mechanical ventilator, rather than ICU LOS, as a measure of resource consumption. Furthermore, these data permit the use of the concept of ventilator-free days to quantify morbidity after trauma.

As outcome measures, the expert panel suggest hospital LOS, discharge destination, Glasgow Outcome Scale (GOS) [91] score at hospital discharge and 30-day mortality. The GOS contains information about morbidity, and this data collection represents an increased focus on information regarding morbidity. In the original US MTOS, "end of acute care" was used as an outcome measure [9]; however, 30-day mortality is considered to be a more fixed endpoint than "end of acute care/hospital discharge" or "end of somatic care," which will vary depending on the transfer and rehabilitation policies of an individual system [92]. Using 30-day mortality is consistent with the previous Utstein recommendations [1]. Death occurring later than 30 days after injury is more likely to be caused by other conditions, such as pre-existing disease [81]. An analysis of 69,650 patient admissions from the TARN showed that 4.8% of the patients died within 93 days of admission [41]. Of these, only 9% died later than 30 days after admission; these were mainly patients with a low ISS (< 9) and aged > 65 years. In a Scandinavian trauma registry, 4.6% of the deaths occurred later than 30 days after injury [92], whilst data from the German trauma registry (cases with NISS ≥ 16) indicate that 4.9% of the patients that died did so later than 30 days after injury [personal communication with the DGU-TR]. Survival status of all patients at a single given endpoint is needed, and 30-day mortality was chosen to represent this information.

For injury severity description, the expert panel recommends recording all AIS codes. It has previously been shown that a comparison of survival for trauma registries that use different AIS editions is potentially possible [93]. A difference in registrars and different levels of AIS training probably represents a greater problem than problems with using different AIS versions. However, registries are recommended to use the same AIS edition as a uniform way of coding; hence, the newest available version should be used at all times. At present, this is the AIS 2005 edition.

System characteristics descriptors

The expert panel considered the time between when the alarm call is answered (at the emergency dispatch centre) and when the patient arrives at the first hospital to be an important system characteristic descriptor (this variable can also be useful for mapping the entire process of pre-hospital rescue) and recommend that this interval be reported.

Recording the highest level of competence of the pre-hospital care providers was regarded by the expert panel as an important measure for describing the pre-hospital system, and for cross-border comparisons of pre-hospital trauma care and outcome. The highest available level of the providers may vary somewhat in Europe, but the revised template's categorisation of level of provider is based on the levels proposed by McSwain [94], since these levels will encompass most pre-hospital systems.

Pre-hospital intubation is an important parameter that represents pre-hospital advanced life support (ALS) [95]. There is an ongoing debate on the use and role of ALS [96–98] measures in the out-patient management of trauma victims, and such information should be made available for possible inclusion in future survival prediction models. The original US TRISS model was derived from a dataset that excluded trauma patients who were intubated out-of-hospital [9]. However, these patients constitute a significant proportion of European trauma victims; hence, the expert panel recommends including them. In a study by Arbabi et al., early field intubation was associated with a decreased risk of fatal outcome, as compared to ED intubation [71]. In their study, intubation status was also an independent predictor of fatal outcome, after adjusting for ISS, SBP, mechanism, age and ED-GCS [71].

The type of transportation delivering the patient to the hospital is an important descriptor of the pre-hospital trauma care system, but the use of a helicopter vs. ground ambulance remains controversial [99, 100], and assessment is confounded by differences in various EMS and HEMS systems. Since transportation type has been so widely debated, this is recommended as core data to be collected. The data variable does not cover what is a commonly used transport combination in some parts of Europe; ground ambulance to the local hospital and fixed-wing or rotary-wing transfer to the regional trauma centre. The data variable was developed to cover the type of transportation delivering the patient to the reporting hospital.

A formal trauma team is an essential part of an organised trauma system [104]. Activation of the trauma team is an organised initial response to a trauma [104, 105] with the primary goal of securing fast and efficient treatment of severely injured patients [106–108]. Information on whether the trauma patient was met by an activation of the trauma team was recommended as vital for describing a trauma system.

The revised template recommends recording the highest level of in-hospital care for the trauma patient in the main trauma hospital (Table 4).

Process mapping variables

The "Chain of Survival" concept emphasises that all time-sensitive interventions must be optimised to maximise the chance of patient survival [109]. The expert panel recommends recording the time between when the emergency call is answered (at the emergency call centre) and when the first medical provider arrives at the patient. This core data represent parts of the first link in that chain, and is an important measure of the quality of the pre-hospital EMS system.

As an overall marker of the efficiency of patient treatment (including resuscitation, diagnostics and surgery), the expert group suggested to consider the time required to achieve normal arterial BE. A good evidence basis does not exist for this recommendation, however, if BE worsens after arrival, prognosis worsens as well [110]. The BE should be measured regularly after hospital arrival, and the template advises to document the initial measurement of BE immediately after hospital arrival, and in cases of abnormal values, to document the time in hours until normalisation.

The efficiency of the initial in-hospital management is assessed by the time from hospital admission until the time marked on the first CT scan images. This data represents the time required to perform key in-hospital diagnostic tests.

The time to the first key EI should be recorded; this measure represents how quickly an urgent intervention fundamental to the treatment and stabilisation of a patient's specific injuries [105] is performed. This core data measures the efficiency of the trauma system in the initial phase. The time elapsed between injury and EI should be minimised.