Review of Existing Scoring Systems for Massive Blood Transfusion in Trauma Patients: Where Do We Stand?

Abstract

Background: 

Uncontrolled bleeding is the main cause of the potential preventable death in trauma patients. Accordingly, we reviewed all the existing scores for massive transfusion posttraumatic hemorrhage and summarized their characteristics, thus making it easier for the reader to have a global view of these scores—how they were created, their accuracy and to which population they apply.

Methods: 

A narrative review with a systematic search method to retrieve the journal articles on the predictive scores or models for massive transfusion was carried out. A literature search using PubMed, SCOPUS, and Google scholar was performed using relevant keywords in different combinations. The keywords used were “massive transfusion,” “score,” “model,” “trauma,” and “hemorrhage” in different combinations. The search was limited for full-text articles published in English language, human species and for the duration from January 1, 1998 to November 30, 2018.

Results: 

The database search yielded 295 articles. The search was then restricted to the inclusion criteria which retrieved 241 articles. Duplicates were removed and full-texts were assessed for the eligibility to include in the review which resulted in inclusion of 24 articles. These articles identified 24 scoring systems including modified or revised scores. Different models and scores for identifying patients requiring massive transfusion in military and civilian settings have been described. Many of these scorings were complex with difficult calculation, while some were simple and easy to remember.

Conclusions: 

The current prevailing practice that is best described as institutional or provider centered should be supplemented with score-based protocol with auditing and monitoring tools to refine it. This review summarizes the current scoring models in predicting the need for MT in civilian and military trauma. Several questions remain open; i.e., do we need to develop new score, merge scores, modify scores, or adopt existing score for certain trauma setting?

INTRODUCTION

Traumatic injury is one of the leading causes of death worldwide, representing approximately 10% of global deaths, especially in young population ^{(1, 2)}. Uncontrolled bleeding is the main cause of the potential preventable death in trauma patients. Moreover, 80% of deaths in the operating room and half of the deaths within 24 h post-trauma are related to exsanguination and coagulopathy ⁽³⁾. Although there is no consensus on the best definition of massive transfusion (MT), it is generally defined in the literature as transfusion of ≥10 units of packed red blood cells (PRBCs) in 24 h and it does not include the other blood products needed. A resuscitation strategy that aims at delivering blood products at a fixed ratio to support hemostasis is referred to as hemostatic resuscitation. It is an effective treatment option in massively bleeding patients. It is required in 3% of general trauma patients for hemostatic resuscitation; however, MT in trauma centers is generally unplanned and needs a large amount of blood products ^{(3, 4)}. Patients who require aggressive hemostatic resuscitation could consume 70% of all transfused blood ⁽³⁾. In addition, there are many steps in a complicated journey for preparing blood for transfusion; it ought to be processed and delivered to the clinical area in short time and in a sustained manner. Therefore, preplanning and coordination between transfusion service/blood bank and emergency department staff is crucial. Early identification of the right patient who needs MT at the right time and well-organized management are vital to attain the most optimal outcome ^(5–8). MT protocol (MTP) refers to a standardized management approach to facilitate rapid supply of massive amounts of blood designed for reducing blood product use and mortality. MTP activation triggers include persistent hemodynamic instability, active bleeding necessitating surgical or radiologic intervention, need for immediate blood transfusion in the trauma bay, and use of a scoring system ⁽³⁾.

While there are a number of scoring systems and algorithms for the prediction of MTP use in civilian or military trauma patients ^{(5, 6, 9)}; there is no agreement or consensus for a specific, perfect scoring system. Of note, not all the scores were generated or designed for critically injured patients as some of them were developed on non-severely injured patients. The existence of multiple scoring systems suggests that there is no single scoring system that is perfect. The clinical decision to activate MTP is not easy. This activation process is often done very early and may result in wastage of blood products with a substantial impact on the cost. To try and reduce these issues, multiple predicting scores have been proposed to identify patients in need of MT. Keeping up with so many scores is not easy (or even confusing), and deciding which score to adopt is also quite difficult. Therefore, we decided to review all the existing scores and summarize their characteristics, thus making easier for the reader to have a global view of these scores—how they were created, their accuracy and to which population they apply. This narrative review could serve as a platform for seeking a consensus for a specific score in a specific trauma setting.

METHODS

A systematic search method was used to retrieve the journal articles on the predictive scores or models for MT. A literature search was conducted using PubMed, SCOPUS, and Google scholar database. The keywords used for literature search were ”massive transfusion,” “score,” ”model,” ”trauma,” and “hemorrhage” in different combinations. The search was limited to human studies for the period between January 1, 1998 and November 30, 2018. Only full-text articles published in English language were included.

RESULTS

Initially, the literature search using the relevant keywords resulted in retrieval of 295 articles. After following the inclusion criteria, the search yielded 241 articles indexed in MEDLINE. Duplicates were removed and then the full-texts were assessed for eligibility for inclusion in the review. Finally, selected studies yielded 24 models or scores including modified or revised scores (Fig. 1). The variables used in each model and their utility in predicting the need for blood transfusion are described as follows.

Fig. 1: Flow-chart for review design.

Scores from civilian trauma

Assessment of blood consumption (ABC score)

The ABC is a popular model developed based on a retrospective study among civilian trauma patients (n = 596) in a level 1 trauma center ⁽⁹⁾. Patients who survived at least 30 min after arrival in trauma center and received any blood transfusions during hospitalization were included. The study used non-laboratory and non-weighted parameters based on clinical experience. The parameters comprise penetrating mechanism of trauma, systolic blood pressure (SBP) < 90 mm Hg at ED, heart rate (HR) >120 bpm at ED, and positive result on focused assessment for the sonography of trauma (FAST) test (Table 1 ). The presence of any of these components adds one point to the total score which ranges from zero to four. Sensitivity of ABC score ≥ 2 was 75% and specificity was 86% for predicting MT. Cotton et al. validated the score in three different trauma centers with similar MT rate (15%) and Krumeri et al. in a rural level 1 trauma center, both were retrospective studies ^{(10, 11)}. The limitations were reported particularly in the heavily blunt trauma population and in the elderly.

Table 1: Studies used in the analysis

Table 1 (Continued): Studies used in the analysis

Trauma-associated severe hemorrhage (TASH score)

The TASH Score was developed based on the analysis of 6,044 severely injured blunt trauma patients’ data obtained from the multicenter civilian trauma registry database run by the German Society for Trauma Surgery (TR-DGU) ⁽¹²⁾. TASH score was measured using seven independent but weighted variables such as gender, SBP, HR, Hb, FAST test, base excess (BE), and extremity or pelvic fractures ⁽¹²⁾. The range of scores was 0 to 28 where each point paralleled to a risk for MT in percent. The high performance of the score was evident while having an increased AUROC of 0.905. MT was required in more than 90% cases with TASH score >16. Maegele et al. ⁽¹³⁾ revalidated the TASH score by analyzing the TR-DGU data of 5,834 patients during the period 2004 to 2007, retrospectively. Most of these parameters can be obtained within 15 min in high-quality centers (or centers with high financial resources); however, it might not be the case in centers with limited resources. Thus, the need for laboratory testing and difficulties in memorizing and mathematical calculation are limitations for this scoring system.

Cincinnati individual transfusion triggers (CITT) triggers

Callcut et al. ⁽⁶⁾ analyzed data of 170 trauma patients presented to a Level 1 trauma center based on ED records, operative logs, and blood transfusion data from arrival to surgical intervention. All patients who required immediate major operative interventions were included in the study. This study inquired the utility of five transfusion triggers such as SBP <90 mm Hg, Hb <11 g/dL, temperature <35.5 C, INR >1.5, and base deficit (BD) ≥ 6 in civilian trauma settings. These individual triggers contributed to the predictive value of transfusion differently. Each trigger had high sensitivity and specificity for the need of MT, for example, the INR had 95% specificity and 92% NPV. The study demonstrated that INR > 1.5 indicates transfusion of 10 units of blood within 6 h in a patient requiring surgical intervention. INR [OR 11.3 (95% CI 2.7–47)] and SBP [OR 8.5 (95% CI 3.4–21)] were highly predictive while temperature was least predictive [OR 3.4 (95% CI 1.5–7.7)]. The study concluded that these triggers have differential predictive values; had more utility as individually weighted criterion and were early dependable indicators of the MT requirement, as well as the quantity of transfusion volumes in civilian trauma settings. When equal weighting approach was followed, 31% of those who had five triggers had PRBC transfusion.

Massive transfusion score (MTS) and revised MTS

The prospective, observational, multicenter, major trauma transfusion (PROMMTT) study aimed at identifying associations between blood transfusion during active resuscitation and patient outcomes ⁽⁴⁾. The study included 1,245 adult trauma patients survived at least 30 min after admission and received a minimum of 1 unit of RBC transfusion within 6 h of admission. In patients who received transfusions of at least 3 units during the first 24 h after admission, higher plasma and platelet ratios were associated with decreased mortality. In addition, the subsequent risk of death among survivors at 24 h post-admission was not associated with plasma or platelet ratios.

Callcut et al. studied the patients enrolled in the PROMMTT study to evaluate eight unique triggers derived from CITT and ABC studies ^{(5, 6, 9)}. These triggers were INR >1.5; SBP <90 mm Hg; Hb <11 g/dL; BD≥6; FAST (+ve); HR≥120 bpm; penetrating trauma and temperature <35.5C. The original MTS was proposed by excluding the temperature from these unique triggers, but revised MTS excluded HR, FAST, and mechanism of trauma ⁽⁵⁾.

MTS was found to be both sensitive and reliable in foreseeing the need of MT at 6 h as well as at 24 h with the increase in number of positive triggers. In addition, the negative predictive value (NPV) improved when triggers in combinations were used compared to the use of individual triggers alone. Studies demonstrated that most trauma patients with hemorrhage require a major share of their blood products by 6 h; consequently, those patients were reassessed at 6 h to make a decision on need of continued transfusion ^{(4, 14)}. Patients with < 2 positive triggers (MTS<2) were highly unlikely to receive a MT by 6 h. Failure to stabilize the MTS triggers by hour 6 implies a poor prognosis. In addition, MTP patients with normalized revised MTS_6hour components did not require any further blood products in the initial 24 h of care. Therefore, the revised MTS_6hour extends the utility of MTS from early identification of patients in need for MT, to a decision-making tool along with clinical gestalt whether to continue the transfusion or not after 6 h. Revised MTS was not only a better predictor of MT when compared to ABC score; it was also a powerful predictor of mortality; for instance, for each additional positive trigger at hour 4 or 6, the odds of death for trauma patients at 28 days increased 2-fold ⁽⁴⁾.

Prince of Wales Hospital (PWH) score

This model was developed based on a retrospective study conducted in a regional trauma center ⁽¹⁵⁾. Data of civilian trauma patients with moderate to severe injuries (ISS ≥ 9) with age ≥ 12 years who attended the trauma center was obtained from PWH Trauma Registry. The study included 1,891 patients in whom 92 patients required ≥ 10 units RBC within 24 h. The study demonstrated seven variables that were easily measured in the ED and significantly predicted the need for MT. These variables include HR ≥ 120/min; SBP ≤90 mm Hg; GCS ≤8; displaced pelvic fracture; CT scan or FAST positive; BD >5 mmol/L; Hb ≤7 g/dL; and Hb 7.1–10 g/dL. In PWH score, weighted scores were assigned to each variable according to their adjusted odds ratio. The overall correct classification for predicting MT requirement was 96.9% at a cutoff of ≥6 points, while the MT incidence at this value was 82.9%. The AUROC for predicting MT requirement using this model was 0.889.

Vandromme score

Vandromme et al. ⁽¹⁶⁾ developed a predictive model for MT based on a study among civilian trauma patients admitted to the level 1 trauma center in the duration between 2005 and 2007 in the USA. The study added blood lactate (BL) variable along with other variables derived from military trauma studies to predict the need for significant transfusion ^(17–19). The authors included 306 patients in their study duration and additional data of 208 patients included in the duration 2007 to 2008 as the validation cohort. MT was defined as in the previous studies as administration of at least 10 units of PRBCs within 24 h of admission. Clinical variables used to develop the model were: BL ≥ 5 mmol/L, HR >105 bpm, INR >1.5, Hb ≤11 g/dL, and SBP <110 mm Hg. The need for ≥ PRBCs within 24 h of admission included ≥ 3 positive clinical measures (sensitivity: 53%, specificity: 98%, PPV: 33%, NPV: 99%). The PPV was found increasing with numbers of positive measurements. The multivariate model with four of five positive clinical predictors led to overall high AUC (≥0.91). The inconsistency between the relatively high specificity and NPV in contrast to the relatively low PPV and sensitivity, or the utility in correctly classifying patients with MT requirement was attributed to the low MT rate witnessed in this study (2.4%), as PPV is greatly influenced by the prevalence of the disease in a specified population ⁽²⁰⁾.

Trauma induced coagulopathy clinical score (TICCS)

The TICCS was developed to distinguish patients who need damage control resuscitation (DCR) based on a prospective observational study among 82 trauma patients ⁽²¹⁾. The score ranges from 0 to 18 using variables such as general severity (2 points if the patient oriented to a resuscitation room rather than a regular ED room); SBP < 90 mm Hg equals 5 points; and extent of body injury (2 points for the torso, abdominal or the pelvic ring region, 1 point for head or each extremity). This score was intended to use in the prehospital settings since it was based only on collected easily clinical parameters. A cut-off value of 10 has 100% sensitivity (95% CI: 53.9–100) and 96% specificity (95% CI: 88.2–99.2). The PPV and NPV were 73% and 100% respectively ⁽²¹⁾.

Moore model

Moore et al. ⁽²²⁾ developed a model for the need for MT based on a multicenter retrospective study among major trauma patients without severe brain injury. Two out of three among the total 383 patients sustained a blunt trauma and nearly a quarter of patients received MT. MT was defined as PRBC transfusion volume >3,000 mL during the first 6 h after admission. Stepwise multivariate logistic regression analysis showed the best-fit model including ISS ≤25 = 0 points; ISS >25 = 1 point, minimum pH on first hour of arrival, and minimum SBP during first hour after arrival. The performance of the score was expressed as AUROC of 0.804.

Baker model

The Baker model was developed by combining physiologic scores and ISS to determine the need for blood transfusion among trauma patients on arrival to the ED ⁽²³⁾. This model was based on retrospective data of 654 trauma patients presented to a level I trauma center in the USA over a 6-month period. The investigators identified four risk factors for transfusion after injury such as SBP <90, HR >120, GCS <9, and high-risk injury including ventral chest trauma between the midclavicular lines, abdominal trauma with diffuse tenderness, survival of a motor vehicle crash (MVC) in which another occupant died, motor vehicle ejection, or penetrating torso trauma. All four risk factors on arrival had 100% transfusion rate, three risk factors had 68%, two risk factors had 42%, one risk factor corresponds to 12% transfusion rate, and absence of all risk factors had a 2% transfusion rate. The SBP <90 mm Hg had highest relative risk for transfusion.

The emergency room transfusion score (ETS)

Ruchholtz et al. ⁽²⁴⁾ developed ETS based on multivariate logistic regression analysis of data obtained retrospectively from 1,103 trauma patients presented to the ED. The score was then prospectively validated by including 481 patients admitted to a level 1 trauma center in Germany over 4 years ⁽²⁵⁾. Nine parameters collected after first 10 min of arrival to trauma were used to predict the need of MT in patients. Logistic regression was performed to determine the predictive value for the need of blood transfusion. ETS totals up to 9.5 points including variables such as SBP < 90 (2.5 points), 90 to 120 (1.5 points); FAST positive (2 points); clinically unstable pelvic ring fracture (1.5 points); age 20 to 60 years (0.5 points), > 60 years (1.5 points); admission from scene (1.0 points); traffic accident (1.0 points); fall from >3 m (1.0 points). The transfusion rate was 0.7% at one point; 5% at three points; 97% at 9.5 points which shows an exponential increase in probability. The cut-off point (probability/risk of transfusion ≤5%) was ETS ≤3, a low risk group with PPV 0.222; NPV 0.998; sensitivity 97.5% and specificity 68%.

Rapid thrombelastography (r-TEG)

Rapid thrombelastography (r-TEG) offers a complete assessment of the coagulation system. Cotton et al. ⁽¹⁰⁾ prospectively studied 272 trauma patients and found that early r-TEG values including activated clotting time (ACT), k-time, and r-value were correlated with prothrombin time, INR, and partial thromboplastin time significantly. ACT >128s was predictive for MT in patients during the first 6 h (OR 5.15; 95% CI, 1.36–19.49). On the other hand, ACT <105s predicted patients who did not receive any transfusions in the first 24 h (OR 2.80; CI, 1.02–7.07).

Rotational thromboelastometry

Rotational thromboelastometry is referred to graphic representation of viscoelastic changes during the process of fibrin polymerization. The variables include time, size, dynamics, and firmness of the clot. Davenport et al. ⁽²⁶⁾ prospectively studied 300 trauma patients to predict transfusion requirements based on rotational thromboelastometry variables. The study found that clot amplitude at 5 min predicted the need for MT with a significantly higher detection rate when compared to prothrombin time ratio >1.2 (71% vs. 43%, P<0.001).

Trauma bleeding severity score (TBSS)

It is a Japanese scoring system specially developed to compensate for the elderly as the mean age of other scores especially ABC and TASH was around 40 years whereas in Japan elderly (>65 years) constitute 22% of their population ⁽²⁷⁾. It was derived from age, SBP, number of region points on FAST scan, pelvic fracture, and serum lactate, with scoring from 0 to 57 developed and it was validated by same authors ⁽²⁷⁾. The cutoff point of 15 points provides sensitivity of 97.5% and specificity of 96.2%.

Shock index (SI), modified SI, age-adjusted SI

SI is used to predict the severity of hypovolemic shock in trauma patients. It is defined as the ratio of heart rate to SBP and modified SI (MSI) is the ratio of HR to mean arterial pressure (MAP). SI could be also adjusted for age (SI × age). Relying solely on the separated vital signs (HR alone or SBP alone) might underestimate and mask the injury severity and level of shock in patients who are in urgent need for intervention. In addition, SI was found to predict the injured patient who would benefit from MTP, however, currently there is no consensus for the optimal SI cutoff for prediction of MT (SI = 0.70, 0.8, 0.90, or 1.0).

Prior work in 2018 demonstrated that the cutoff value 0.8 has excellent negative predictive value and sensitivity ⁽²⁸⁾. SI was well correlated with other physiological, laboratory and anatomical scores and predictors such as pulse pressure, injury severity scores (GCS, AIS, and ISS), base deficit and the amount of blood units transfused. Furthermore, it correlates with the probability of survival (TRISS) and hospital length of stay. Multivariable regression analysis confirmed the utility of SI as an independent predictor for transfusion (OR 3.6) and mortality (OR 2.5) after adjusting for age, sex, GCS and ISS. Even in patients with head injuries, the presence of high SI was associated with higher rate of blood transfusion and worse outcomes in comparison with those who had SI < 0.8 ⁽²⁸⁾.

Code red

Weaver et al. in 2016 established a prehospital transfusion request policy based on three simple criteria such as evidence or suspicion of active hemorrhage; SBP < 90 mm Hg; and blood pressure failure to respond to bolus intravenous fluid ⁽²⁹⁾. Code Red was shown as an important tool to activate MT at the receiving hospital which facilitated the availability of blood products and earlier transfusion. The study demonstrated that 91% of patients with Cod Red activation received MT in trauma center. The majority of patients were injured due to motor vehicle collisions (45%) whereas penetrating trauma occurred in 30% with a mean ISS of 29 ⁽²⁹⁾.

Coagulopathy of severe trauma score (COAST)

Mitra et al. in 2011 developed the COAST score to predict acute traumatic coagulopathy using prehospital variables associated with coagulopathy in major trauma patients ⁽³⁰⁾. The study was based on retrospective data of 1,680 major trauma patients in whom 151 patients were coagulopathic. Acute traumatic coagulopathy was found dependent on prehospital variables such as entrapment (1 point), body temperature (< 35°C: 1 point, < 32°C: 2 points), SBP (< 100 mm Hg: 1 point, < 90 mm Hg: 2 points), pelvic content or abdominal injury (1 point), and chest decompression (1 point). The score ranges from 0 to 7 points. The prospective validation of 1,224 patients revealed that COAST ≥ 3 had a specificity of 96% with a sensitivity of 60% for acute traumatic coagulopathy with an AUC of 0.83 ⁽³⁰⁾.

Wade model

Wade et al. developed a model to predict the need for MT based on a retrospective data obtained from 1,574 patients received transfusion during a 1-year period from 17 level I trauma centers across the USA ^{(31, 32)}. Multivariate logistic regression was performed by incorporating covariates such as SBP, HR, pH, and hematocrit. The model found that a patient was predicted to require MT if π >0.5 (see Table 1 ). This model yielded PPV: 75%; NPV: 72%; sensitivity 87% and specificity of 53% ^{(31, 32)}.

Models or scores from military trauma

McLaughlin score

McLaughlin et al. developed a model to predict the requirement of MT in trauma patients in the military setting based on the retrospective data obtained from 302 patients ⁽¹⁷⁾. Independent risk factors for MT identified were: HR>105, SBP <110, pH <7.25, and hematocrit <32%. These components were non-weighted but identified dichotomously as “yes” or “no.” MT rate increased from 20% where the patient had one of these variables positive on admission to 80% when all four were present. The AUC was 0.839. In an independent internal dataset comprised of 396 patients, the ability to identify patients with combat-related injuries in need for MT accurately was 66% (AUC = 0.747).

Larson model

Larson et al. developed a model based on a retrospective review of data on combat casualties obtained from Joint Theatre Trauma Registry (JTTR) and transfusion data from US Department of Defense database ⁽³³⁾. The model included 1,124 military casualties that occurred between 2003 and 2008 and had received at least 1 unit of blood during initial resuscitation. Transfusion of PRBCs and/or fresh whole blood was included. MT was defined as ≥10 units of PRBCs within 24 h after admission. Nearly 37% had received MT.

Four variables included for determining the likelihood of needing an MT were HR > 110, SBP <110, Hb ≤11, and BD ≤ 6. The presence of at least 2 data points showed the MT incidence 54% with PPV 54%, NPV 78%, sensitivity 69%, and specificity 65%. Patients predicted for MT, but not actually had MT, died earlier and majority of them head injuries compared to those with predicted and observed to have had MT. On the other hand, patients not predicted for MT, but actually had MT, had increased chest, abdominal, and extremity injuries than those neither predicted nor observed to have had an MT.

Schreiber model

Schreiber et al. ⁽¹⁸⁾ conducted a retrospective study among 558 combat victims at two level III combat support hospitals in Iraq. Nearly 44% patients required MT defined as transfusion of ≥ 10 units of a combination of stored PRBCs and Fresh whole blood (FWB) in the first 24 h postinjury. Eight potentially predictive variables were analyzed for association with the MT requirement and stepwise logistic regression was performed. Following variables independently predicted the requirement for MT: Hb ≤11, INR >1.5, and the penetrating trauma; with most predictive potential for Hb ≤11 g/dL (OR 7.7). AUROC of the predictive model with these three variables was 0.804 and the HL goodness-of-fit test was 0.98.

Revised trauma score (RTS) and field triage score 07 (FTS07)

RTS and the modified FTS07 predict the need for MT (i.e., ≥10 units of PRBCs or FWB) in combat casualties. Cancio et al. ⁽¹⁹⁾ compared the potential predictive power of these scores in 536 combat casualties admitted to level III US hospitals in Iraq. The RTS was based mostly on physiology including three weighted parameters such as GCS, SBP, and respiratory rate (RR) ⁽³⁴⁾. Eastridge et al. ⁽³⁵⁾ analyzed the data on 4,988 military casualties from Iraq and Afghanistan from the JTTR in the duration 2002 to 2008 and found that SBP and motor component of GCS was helpful in predicting the outcomes including mortality. The authors validated the modified FTS07 (GCS _total <8 and SBP < 100 as cut-points) with a range of 0 to 2. In addition, the RR was used for score calculation in artificially ventilated and spontaneously breathing patients. The accuracy in predicting the need for MT associated with RTS and FTS07 scores were relatively low; AUROCs of 0.638 and 0.618 respectively. FTS07 is advantageous over RTS because of its easiness in computation.

Clinical gestalt

Clinical gestalt is a shortcut of thinking for decision making to quickly forming a diagnosis and/ or treatment plan, it is often intuitively recalled within seconds of data observation and collection. Pommerening et al. ⁽³⁶⁾ assessed the ability of clinical gestalt to predict the need for MT in patients with bleeding based on the PROMMTT study. Each trauma attending physician was asked to answer “yes” or “no” to the gestalt question “is this patient likely to be massively transfused” after 10 min of the patient arrival. The investigators used three previously validated MT scoring systems (TASH, ABC, and McLaughlin scores) to be correlated with the clinical gestalt. They found that clinical gestalt is an unreliable predictor of MT with a sensitivity of only 66%; it worked poorly as a screening test for MT and missed over one-third of patients who ultimately required MT. This finding indicates that trauma surgeons’ threshold for MT activation is still questionable as they missed a substantial number of cases that were potentially under resuscitated. The inclusion criteria of the PROMMTT study could be the source of selection bias. The study included patients who received at least one unit of blood in the first 6 h and excluded patients who were ill enough to die early post-arrival. In this case, clinical gestalt worked well in the two extremes of patients (i.e., those who massively bleed and those who were not bleed). Among the gestalt negative group, false negatives (missed MT) were three times as likely to have bleeding in the pelvis with higher degrees of physiologic derangements. The TASH score correctly classified more patients than gestalt (TASH includes areas such as pelvis with relative blindness for the clinician observation and correct estimation). However, no statistically significant difference in predictive ability was detected between clinical gestalt and the ABC or McLaughlin scores ⁽³⁶⁾. The authors also concluded that “over-triaging is an acceptable consequence, as it is much easier to return unused blood than manage the patients awaiting product delivery.”

TIME AND COMPLIANCE TO MT ACTIVATION AND IMPACT ON SURVIVAL

The recommendation of the American College of Surgeons Trauma Quality Improvement guidelines for MT included the delivery of the first blood product cooler within 15 min of activation, and the delivery of each subsequent cooler within 10 min of the request ⁽³⁷⁾. However, one of the challenges to improve the timing of blood product delivery is to decrease the time to activation of MTP by decreasing the time to physician recognition of the need for blood product use. A subanalysis of the Pragmatic, Randomized Optimal Platelets and Plasma Ratios (PROPPR) study evaluated the impact of time to activation of MTPs and then time from activation to delivery of the first MTP cooler ⁽³⁸⁾. Time of arrival, time to MTP activation, and time to arrival of initial and each subsequent cooler were identified. The content of each MTP cooler was identical between the study centers, varying only by randomization group (1:1:1 vs. 1:1:2 group). Moreover, to attain rapid and balanced transfusion strategy, the sequence of transfusion was the same among centers (12 sites with 14,313 highest-level trauma activations and 680 randomized patients). The median time from arrival to MTP activation was 9 min with a median time from MTP activation call to delivery of first cooler of 8 min. The investigators found that increased time to MT activation and time to arrival of first cooler were associated with increased mortality. Furthermore, adjusting for injury severity, physiology, resuscitation intensity, and treatment arm (1:1:1 vs. 1:1:2), increased time to arrival of first cooler was significantly associated with an increased mortality at 24 h and 30 days. Therefore, independent of products ratios, every minute from time of MT protocol activation to time of initial cooler arrival increased odds of mortality by 5% ⁽³⁸⁾. In one study, Bawazeer et al. ⁽³⁹⁾ assessed the compliance in 72 consecutive MTP activations utilizing 13 compliance criteria. The authors found 66% as an average compliance rate. The mortality rates in the compliance groups were 62%, 50%, and 10% in the low (A; <60%), moderate (B; 60–80%), and high (C; >80%) compliance groups, respectively. The adjusted OR for mortality between groups A and B was 1.1 (P = 0.89) while the OR for mortality between groups C and B was 0.02 (P = 0.04) ⁽³⁹⁾.

SUMMARY

This review summarizes the current scoring models in predicting the need for MT in civilian and military trauma. Several questions remain unanswered. These include whether we need to develop new scores, merge, modify, or adopt exist ones and apply them to specific trauma settings. Different models and scores for identifying patients requiring MT in military and civilian settings have been described. These systems or models are based on hemodynamic, physiologic, laboratory, injury pattern and severity, mechanism and demographic triggers. Many of the scores are complex requiring difficult calculation and time consuming lab tests whereas some were simple and easy to remember using physiologic parameters, injury characteristics, and/or simple procedures done as point of care (Fig. 2). Of note, predicting scores do not account for survivorship bias (patients have to live long enough to be massively transfused and those dying before MT are not accounted for). Predicting scores are based on potentially incorrect—assumption that all transfusions are appropriately done (at least 30% of plasma transfusions are inappropriate). Finally, some scores do not account for head injury, which is the most common cause of deaths and many patients presumed to die from bleeding in fact died of head injury. The retrospective nature of some studies and lack of prospective validation may be an issue of concern for some scores.

Fig. 2: Classification and categorization of massive transfusion scores.

The most frequently used variables in these models include SBP, HR, and Hb. Injury mechanisms of concerns were MVC, penetrating trauma, pelvic fracture, chest trauma, or abdominal trauma. Base deficit, serum lactate, INR, and FAST results were also used as covariates in some models. Furthermore, gender, age, respiratory rate, GCS, ISS, AIS, hematocrit, lactate, pH, CT scan, and temperature are less frequently used variables in assessing the MT need. Utility of these models in terms of sensitivity, specificity, PPV, and NPV varied among studies (Table 1 ). The overall accuracy among the scores discussed here ranged between AUC 0.618 and AUC 0.905 with both sensitivity and specificity ranges 53% to 98%.

Some of the scores such as TASH, PWH, ETS and FTS used weighted variables whereas ABC, MTS, Vandromme, Baker, McLaughlin, Schreiber and Larson scores were non-weighted. Wade model, Moore model and RTS were based on prediction formula. CITT used individual triggers. Non-weighted scoring systems are commonly based on sum of predictors in the model; however, Schreiber model was based on individual predictive variables with Hb as the most predictive variable. Rates for correct classification are frequently higher for models with weighted variables, for example, 90% of the patients who had MT were correctly classified by TASH>16. Furthermore, correct classification rate for PWH>6 was 97%. Models with non-weighted variables were dependent on number of predictive triggers present.

Despite the advancements in surgical and interventional techniques and overall trauma centers expertise, exsanguinating hemorrhage remains a leading cause of preventable death in trauma settings, and the majority of these deaths occur within the first 6 to 12 h post injury ^(40–44). The situation is worse in patients with coagulopathy, notably, a quarter of trauma patients presented to the ED had coagulopathy on admission ^(42–48). On the other hand, the incidence rate of coagulopathy was reported in up to 60% according to the International Society on Thrombosis and Haemostasis (ISTH) definition ⁽⁴⁹⁾. Therefore, early and accurate identification of patients in need for transfusion and particularly MT remain crucial in improving the outcomes, particularly in patients who present in the hyperfibrinolytic phase post-trauma. Scores for prehospital evaluation and transfusion (TICCS, CODE red, and COAST) used, in addition to SBP, simple parameters including patients disposition (resuscitation room vs regular ED room) and injured body region (TICCS), response to fluid challenge (CODE Red), and body temperature (COAST). The latter score aimed to predict acute traumatic coagulopathy (fibrinolytic dysregulation). Fibrinolysis shutdown is more common response to injury than hyperfibrinolysis. The first response may cause delayed death with organ failure; whereas the latter response may cause early death with exsanguinations ^{(50, 51)}. Therefore, early scores need to consider the two different posttraumatic responses for better management. However, this is not easy task using only physiological and clinical parameters and may require the integration with other parameters such as TEG.

Significant improvements in the knowledge regarding resuscitation of exsanguinating patients were attained in recent years. It was demonstrated that mortality has reduced when patients were resuscitated with more balanced ratios of PRBCs and FFP ^{(52, 53)}. Many centers have altered their MTPs to reflect the concept of “DCR” ^(54–56). The ACS guidelines for MTP suggested that these protocols should be based on the “DCR principles” ⁽²⁾. The ACS/COT guideline further stressed about the criteria that triggers MTP ⁽²⁾. Infusion of universal blood product must be employed when criteria are met in such a way that achieve high plasma to RBC ratio; 1:1 and 1:2 ⁽²⁾.

Notably, while developing the ABC score, Nunez et al. ⁽⁹⁾ applied both TASH and McLaughlin Score in their database and found that all scores were equally good predictors for MT with AUROC 0.842 for both TASH and ABC and AUROC was 0.846 for McLaughlin. Cotton et al. studied the predictive ability of ABC score and found that the sensitivity and specificity were in a range from 75 to 90% and 67 to 88% respectively. However, the correct classification rate was 84 to 87% and the AUROCs were 0.83 to 0.90 ⁽⁵⁷⁾. Mitra et al. compared PWH Score with ABC and TASH, in which the performance of TASH was higher with AUROC of 0.8986, followed by PWH Score with AUROC = 0.8419 and ABC ⁽⁵⁸⁾. Maegele et al. demonstrated a comparison of performances of ABC, Vandromme, PWH, Baker, Larson, Schreiber and TASH scores based on analysis of 7,042 adult trauma patients with ISS > 16 and found that weighted scores such as TASH and PWH were more accurate with AUROCs 0.890 and 0.858 respectively ⁽³²⁾. It was followed by non-weighted scores or algorithms, such as Vandromme, Larson, and Schreiber. The least performance was reported for Baker and ABC scores with AUROCs 0.768 and 0.779 respectively ⁽³²⁾. In short, weighted and more advanced models that include the most relevant variables are better predictors for MT than simple non-weighted models. However, the retrospective nature of commonly used models remains as the major limitation and prospective validations are warranted to define the best score or the most precise criterion.

Beside these clinical scoring systems, there are objective laboratory indicators that predict MT ⁽⁵⁹⁾. Tissue oxygenation (tissue spectrometer) was found in a prior study as the only parameter that could provide early (first 1–3 h) prediction of worse outcomes in patients requiring MT reflecting the clinically significant hypoperfusion status ^{(9, 59, 60)}. However, this tissue oxygen saturation monitoring is a relatively new technology based on near-infrared spectroscopy which may not be available in many of the EDs ⁽⁶¹⁾. Furthermore, blood lactate is another surrogate for assessment of prehospital hypoperfusion; however, a prior study showed that lactate accumulation usually lags behind the initial insult which called occult shock (persistent occult hypoperfusion in the setting of normal vital signs and lactate levels) ⁽⁶²⁾.

The current prevailing practice that is best described as institutional or provider centered should be supplemented with score based protocol with auditing and monitoring tools to refine it. Furthermore, better understanding and advances in resuscitating patient with exsanguinating bleeding should be incorporated in a comprehensive approach. Such approach that could match the complexities and unique needs of patients are best referred to as DCR. The DCR approach involves multidisciplinary teams to assure permissive hypotension, high ratio based transfusion, limited crystalloid, prompt interventions to stop bleeding, hemostatic resuscitation, point of care testing, use of viscoelastic clot properties testing to guide the resuscitation, and direct peritoneal resuscitation. One of the limitations in this regard, is that predicting scores may identify patients at risk of requiring MT, but then, if early appropriate resuscitation strategies are used, they may halt the bleeding and prevent these patients from requiring massive transfusion. In this respect, the goal of the score is fulfilled; while at the same time the score is disproved as identifying patients receiving MT.

“WHERE DO WE STAND?” AND FUTURE DIRECTIONS

The present review identified 24 different massive transfusion prediction scores, most created and reported over the last decade. The existence of so many scores reflects the contrasting approaches and lack of agreement in resuscitation, and the common objective of identifying the injured patients requiring massive transfusion, which remains an unmet need. The scores reviewed were not created equal but share many similarities. One similarity is in sharing limitations. Most scores were created from single institutions or repositories, in addition to the lack of adequate validation and complexity of the used formula. Therefore, many scores are not widely adopted. The timing of gathering and interpreting the information and how the formula is intended to be used limit the data that can be used. For example in the Moore formula they utilized ISS which requires an extensive data set, including CT reading and operative findings, and is usually score late or even peri-discharge. This severely limits the utility of the formula. The same applies to many of the laboratory findings used in different scoring models. Furthermore, the availability of adequate amounts of blood products at the bed side in the trauma bay prior to the patient's arrival could improve the utility of ABC at the prehospital setting, for example. Notably, the time and compliance to MTP activation should be revisited to assure better quality of the care in trauma patients.

One of the important questions that needs further elaboration is “Do different scoring systems decrease resource utilization or blood product waste?” Because blood products have a limited half-life, accurate strategies should have been enforced for blood reserves in order to prevent loss and reduce wastes as much as possible. One study showed that the dropping pattern of blood wastage was noted after the hemovigilance performance and training of healthcare center team ⁽⁶³⁾. Besides the fact some scores are complex (and thus difficult to apply during resuscitation) and made for specific populations (civilian vs. military), no scoring method is perfect and may lead to inappropriate (iatrogenic) interventions (for example transfusing blood to a patient with tension pneumothorax just because the ABC score is positive). Scores must be used within a clinical context and as part of the clinical armamentarium.

The desirable predictive power and accuracy remain low, as they have mostly been evaluated in retrospective studies ^(64–66). Despite their limitations predicting scores perform better than clinical “gestalt” alone suggesting that they are relevant and have a role in supporting clinical decision-making. MT is defined by the number of blood transfusions without consideration to whether the transfusions were appropriate or not. If, for example, a patient with a negative score is inappropriately transfused 10U RBC, the predictive score is considered wrong when in fact the patient did not need 10U. Furthermore, the reviewed scores also share similar accuracy (sensitivity, specificity and AUROC) for predicting massive transfusion. Considering the similar limitations and accuracy, it is not surprising to note that the most widely used scores are those requiring the simplest and least number of variables, such as the ABC score and Shock Index.

According to the fifth edition of The European guideline on management of major bleeding following trauma in 2019 ⁽⁶⁷⁾, there are a large number of scoring systems predicting the risk of ongoing bleeding and transfusion needs; however, all of these lack prospective validation and each scoring system has its unique advantages, disadvantages, and utility that may affect its widespread applicability and statistical performance ⁽⁶⁷⁾.

Thus, the present review indicates that massive transfusion predicting scores are valid and clinically relevant despite their limitations. Arguably, the ideal score should be constructed over resuscitation protocols that are continuously audited and monitored, allowing improvement of the score itself. This narrative review could be a platform for seeking a consensus for a specific score in a specific trauma setting; for instance, score should consider certain parameters such as the mechanism of injury (blunt vs. penetrating), location (civilian or military) injured body region (pelvic, head,), prehospital staff and resources, hospital ED settings (level I, II, or III), and trauma center facilities (point of care lab and resources).