Optimizing damage control resuscitation through early patient identification and real‐time performance improvement

Uncontrolled hemorrhage continues to be the most common cause of preventable deaths both for combat1, 2 and civilian3-5 injuries. Injury outcomes have improved with the establishment of trauma systems6 and damage control resuscitation (DCR) clinical practice guidelines (CPG).7-9 Additionally, implementing massive transfusion protocols (MTP) and maintaining a high ratio of plasma and platelets to red blood cells is associated with improved survival for critically injured patients.7, 10 However, early identification of patients with occult massive hemorrhage in time to prevent progressing hemorrhage, repay the oxygen debt, and halt the "bloody vicious cycle" of death during a team-based resuscitation across multiple phases of care proves difficult even under the best circumstances.11-14 Though challenges in DCR exist, several promising solutions may improve hemorrhagic shock recognition, optimize MTP triggers for early blood product mobilization, refine metrics used for optimal DCR practice, and improve adherence to guidelines. Clinical decision support systems (CDSS) represent a range of tools for handling and presenting information to clinical care teams in a way that improves diagnosis and treatment.15 For example, real-time DCR decision support systems can help trauma teams track blood products and adjuncts across multiple phases of patient care.16 Nudges change the architecture of critical choices leading to altered behavior in a predictable way without forbidding options or altering incentives.17 Today, the use of behavioral insights and nudge theory is common practice and used in fields like economics, politics, marketing, and even health care.18-20 With proper optimization for clinical workflow, nudge interventions have significant potential to improve both patient outcomes and the delivery of health care.18-21 Lastly, machine learning algorithms excel in real-time responsiveness and the iterative process helps the system adapt to new patterns often missed by the human eye. Machine learning has the potential to help us identify patients with occult bleeding early and navigate the intricacies and challenges of treating patients with traumatic injuries.22 This manuscript outlines the importance of DCR, identifies barriers in early identification of patients requiring massive transfusion and adherence to DCR CPGs, and highlights promising solutions for addressing these barriers so care teams can consistently execute optimized resuscitations. As with many other concepts in medicine, DCR emerged from innovations in military resuscitation practices and describes a strategic approach to the resuscitation of critically ill trauma patients.23, 24 DCR is based on the understanding that early and aggressive correction and prevention of metabolic derangements can mitigate against further hemorrhage and mortality (Table 1). Initial DCR protocols focused on permissive hypotension and early transfusion of blood products.24 However, the DCR concept evolved to include a more balanced approach to trauma resuscitation with a focus on rapid hemorrhage control, early and balanced blood product transfusion, and prevention of acidosis and hypothermia. Minimize blood loss with early hemorrhage control measures Transfuse blood components that optimize hemostasis Activate MTP Obtain functional laboratory measures of coagulation to refine ongoing resuscitation Give pharmacologic adjuncts to safely promote hemostasis The initial stages of resuscitation may consist of limited volumes of crystalloid fluid to expand plasma volume while awaiting blood products; however, large amounts of crystalloid fluid can lead to acidosis, hemodilution, and decreased oxygen delivery. Thus crystalloid resuscitation ultimately impedes hemostasis and fails to resolve the accumulating oxygen debt.25 Whole blood represents the optimal resuscitation fluid, providing the patient with physiologic ratios of coagulation factors, functioning platelets, and oxygen-carrying hemoglobin.26 In the absence of whole blood, a balanced, 1:1:1 ratio of packed red blood cells (PRBCs), plasma, and platelets should be started early, a practice supported by a large prospective cohort study and a randomized clinical trial.27, 28 However, this ratio-based treatment is not equivalent to whole blood for many reasons; for example, whole blood contains increased concentration of cellular components, lower amounts of anticoagulants and additives, increased oxygen carrying capacity, single-donor exposure, preserved platelet function, and other potential unmeasured benefits.29-31 Additionally, whole blood carries logistical advantages over administering multiple component units. At the same time, avoidance of over-resuscitation is essential to DCR success.32, 33 The "Lethal Diamond of Trauma" consists of hypothermia, acidosis, coagulopathy, and hypocalcemia.34, 35 The interplay among these four factors is complex and multifactorial, and each should be rapidly identified and corrected. For example, acidosis prevents adequate regulation of vascular tone and cardiac function leading to impaired circulatory function and hypothermia. Both acidosis and hypothermia contribute to coagulopathy by numerous mechanisms, such as impaired thrombin production and accelerated fibrinogen consumption.36-38 Acute hypocalcemia after injury, mediated by direct tissue damage and the chelating effect of blood product preservatives, is associated with prehospital hypotension, increased blood product administration, and acute traumatic coagulopathy (INR >1.5) and is more predictive of mortality than base deficit.39-42 Several studies identified ionized calcium (iCa) as an independent predictor for mortality.34 As such, iCa levels could serve as a substantial predictor for mortality. Current approaches to mitigating this lethal diamond include the empiric use of adjuncts such as calcium and tranexamic acid (TXA).43, 44 The optimal dosing strategy for calcium is not fully elucidated; however, the Joint Trauma Systems Damage Control Clinical Practice Guideline recommends first administering 1 g of calcium chloride or 3 g of calcium gluconate empirically before or with the first unit of blood and then repeating administration every four blood products.9 CRASH-2 demonstrated that early TXA (1 g over 10 min, followed by 1 g over 8 h), administered within 3 h of injury, was associated with a lower rate of early mortality.44 Recently, administering a single 2 g bolus of TXA has gained increased attention, and a systematic review and meta-analysis suggest that a single high-dose bolus may be associated with decreased transfusion requirements without increased complications.45 The care of trauma patients begins at the scene of the injury and prehospital providers play a crucial role in thwarting the ensuing lethal diamond.46 While prehospital blood product resuscitation has been associated with improved survival, the STAAMP trial found that prehospital TXA administration did not improve mortality at 30 days.47-50 Additionally, a recent trial of prehospital TXA showed no improvement in favorable functional outcomes at 6 months.51 In the CRYOSTAT-2 trial patients received three pools of cryoprecipitate within 3 hours of injury in addition to standard care; the investigators found no difference in all-cause mortality but also no difference in adverse events.52 Airway management in patients with hemorrhagic shock also requires a thoughtful and measured approach. While many of these patients will have compromised airways, due to either altered mental status or a contaminated airway, rapid sequence induction and positive pressure ventilation (PPV) can be perilous in a hypovolemic patient. Post-intubation, hypotension can result from several mechanisms. PPV reduces venous return due to increased intrathoracid pressure while loss of vasomotor tone results in vasodilation with associated decreased afterload and venous return.53, 54 Induction medications further suppress the release of catecholamines needed to maintain adequate perfusion.55 Definitive airway management with intubation should be delayed until after resuscitation has been initiated to prevent hemodynamic collapse.56, 57 Recent research highlights the importance of addressing hypovolemia prior to performing rapid sequence intubation (RSI) and will likely be reflected in future editions of ATLS.57, 58 To ensure ultimate patient survival, damage control surgery (DCS) must accompany DCR. In this synchronized approach, surgical hemostasis staunches ongoing hemorrhage while DCR mitigates or reverses trauma-induced coagulopathy. Definitive surgical repair or reconstruction then commences after all metabolic derangements of hemorrhage are corrected.59, 60 Bogert and colleagues astutely noted that DCS is truly a component of DCR, despite its origins being decades before the advent of DCR.23 The entire process of DCS occurs in conjunction with resuscitation practices guided by DCR concepts. Both systems are complex and require coordination among a team of healthcare providers along the entire continuum of care for the injured trauma patient. Modern-day DCR CPGs, including the American College of Surgeons (ACS) Committee on Trauma (COT) Massive Transfusion in Trauma Guidelines,7-9 include MTP with indications for activation, transfusion service processes, blood product ratios, adjunctive medications, termination of MTPs, and performance improvement (PI) metrics while considering multi-disciplinary viewpoints. The most commonly accepted DCR CPGs outside of the ACS include DCR in patients with severe traumatic hemorrhage: A practice management guideline from the Eastern Association for the Surgery of Trauma7 and the Joint Trauma System Damage Control Resuscitation Clinical Practice Guideline.9 The 10th edition of the Advanced Trauma Life Support guidelines acknowledges that massive infusion of crystalloid is associated with higher mortality rates and advocates for earlier blood product transfusion.61 To best approximate the blood lost by an exsanguinating trauma victim, blood product ratios should target a high ratio of plasma and platelets to PRBCs.7 AB and A plasma are considered universal donors for plasma, while type O is universal for PRBCs. An alternative transfusion strategy that delivers a maximally balanced resuscitation involves the transfusion of Low Titer O Whole Blood (LTOWB) with platelet functionality.9 Preparing coolers with a predetermined 1:1:1 ratio27, 28 of Plasma: Platelets: PRBCS (6 units of plasma, 1 unit of apheresis platelets, and 6 units of PRBCs) can mitigate significant ratio imbalances and time spent within ratio imbalances.7, 10 DCR CPGs recommend protocolized ratios and adjunctive medications to minimize the lethal diamond.9 Empiric TXA should be infused within 3 hours after traumatic insult, while calcium infusion should occur after the first unit of blood and for every four units of blood product after that with a repletion goal of normocalcemia (1.2 mmol/L as mentioned in the Joint Trauma System (JTS) CPG).7 The infusion of recombinant human-activated factor VIIa and the use of hydroxyethyl starch (Hextend, Hespan) as a resuscitative fluid are not endorsed by current DCR CPGs.7-9 Prothrombin complex concentrate (PCC) effectively reverses anticoagulant medications as per its Food and Drug Administration cleared use and is also occasionally used off-label to manage trauma-induced coagulopathy that proves refractory to platelets, cryoprecipitate, or fibrinogen and TXA. Early use of PCC for acutely bleeding trauma patients is the subject of a recently completed study and an ongoing large randomized controlled trial (ClinicalTrials.gov identifier NCT05568888).62, 63 In the PROCOAG trial, a double-blind, randomized, placebo-controlled multi-center superiority trial, administration of 4F-PCC showed no significant reduction in 24-hour blood product consumption for patients with traumatic injuries at risk for massive transfusion.63 The target systolic blood pressure during DCR is 100 mmHg (110 mmHg for traumatic brain injuries) to balance resuscitation while avoiding the disruption of unstable, early hemostatic clots by having higher blood pressure.9 The use of mechanical hemostatic adjuncts and DCR tools (tourniquets, resuscitative endovascular occlusion of the aorta, direct peritoneal resuscitation) used during DCR lies outside the scope of this review. Early identification of the high-risk patient has several benefits for the team and the patient.64 First, the patient can be more quickly and accurately triaged to the appropriate level of care. The team can better prepare with a focus on clear communication that highlights the patient's circumstances. For example, anticipating the need for a massive transfusion provides a clinically valuable advantage by enabling early and effective communication with the blood bank. Finally, the team can initiate treatment with the optimal level of aggressiveness, without over- or underutilizing valuable resources. By accurately predicting the requirement for a significant volume of blood products, healthcare providers can allocate resources more efficiently, ensuring an adequate supply is readily available. Taken together, these improvements will translate into better patient outcomes. Identification of the critically ill and coagulopathic trauma patient is beneficial, but challenging with clinicians often relying on gestalt and experience to recognize patients at risk for decompensation and death.65 Rapid detection of hemorrhagic shock is a crucial step in the early initiation of DCR. Physical exams and vital signs are typically the only data available immediately upon arrival to the Emergency Department (ED). Compensatory mechanisms, however, can hide signs of severe shock until they tire and fail. Shock Index and the Assessment of Blood Consumption (ABC) score attempt to quantify the degree of hemorrhagic shock and the possible need for massive transfusion. Conventional lab tests generally take too much time to guide immediate treatment and individual tests have specific inherent flaws. Viscoelastic hemostatic assays (VHAs) have shown potential to guide resuscitation and transfusion of blood products and adjuncts66; additionally, VHA-guided resuscitation generally utilizes fewer blood products than empiric transfusion.67 Real-time PI is the process of monitoring, analyzing, and optimizing processes in real-time.68, 69 This process is ongoing and iterative, with the goal of improving patient care and outcomes. Traditional PI approaches such as "Plan-Do-Check-Act" or the JTS PI pillars of conduct, support, inform, and consult can be used as a model for how to review, analyze, and implement PI actions as well as future standards of practice.69 However, these approaches use an asynchronous approach to PI that takes weeks, months, or even years to appreciate any positive impact. We propose that for time-sensitive, complex clinical situations like hemorrhagic shock, a real-time feedback loop represents a more logical and ultimately more effective approach. Though we know that early identification of patients with severe hemorrhage is a critically important step in the initiation of DCR, robust compensatory physiologic mechanisms make this exceedingly difficult to achieve consistently. Keeping track of and administering blood products and adjuncts appropriately across multiple phases of care while coordinating a team-based resuscitation proves difficult.16 As such, adherence to DCR principles remains surprisingly low.10, 27, 28 In light of these myriad challenges, using CDSS, nudges, and machine learning as part of a real-time DCR PI process may improve adherence to DCR best practices leading to optimal outcomes from team-based resuscitations. One key aspect of real-time PI in healthcare is decision support. Broadly, the advantages of CDSS include reduction of clinician errors, increased adherence to clinical management guidelines, and diagnostic support.11 They have been developed for diverse applications,15 and work by leveraging technology, data analytics, and monitoring to provide clinicians with real-time evidence-based recommendations at the point of care.70, 71 CDSS also enables clinical teams to identify and address deviations from guidelines, either in real-time or in the form of an after-action review. In clinical settings where decisions are often made under constraints of time and uncertainty, such as the Intensive Care Unit (ICU) and ED, CDSS have supported clinicians' management of acutely ill and critically injured patients.11, 72 In the context of DCR, an iterative development and human factors testing approach resulted in a clinically usable CDSS capable of prompting activation of MTP, tracking and prompting blood product and adjunct administration, viewing viscoelastic testing, and after-action review (Figure 1).16 Device-level feedback highlighted the device's ease of use, and users overall had a positive impression. In a prospective pilot study, use of a DCR CDSS with a real-time PI process for massively transfused patients resulted in more time spent in target ratios of plasma and platelets to PRBCs compared to both controls at the same institution and PROMMTT patients. This initial pilot study highlights the application of CDSS for DCR; however, multi-center validation is warranted.73, 74 Nudge theory was initially applied in the field of behavioral economics, but has since been used in many other fields, health care included.18-20 For example, in health care, nudges have been described in relation to both behavioral intervention and ordering practices.75-77 Some academic and medical instititutions have created Nudge Units or embedded behavioral design teams within the health care system.19, 20 Such Nudge Units have helped clinicians and researchers utilize nudge theory to improve the delivery of care. Nudges can take many forms and produce varying degrees of behavioral impact (Figure 2), with information framing exerting lighter influence and guiding choices through defaults exerting stronger influence.18-20 While nudges at the bottom of the ladder passively influence members of the care team (i.e., an email reminding clinicians of existing guidelines or a poster with information placed strategically in a Trauma Bay or an Operating Room (OR)), nudges at the top of the ladder more directly influence clinicians at the time of decision making (i.e., enabling active choice between two options of medications or changing the default selection for a medication in the electronic medical record). The most effective nudges tend to be the more assertive ones, often limiting the set of choices or changing default options.18 That being said, the more assertive nudges are not always feasible given the existing workflows in clinical spaces. For that reason, it is important to evaluate how best to optimize the nudge for the environment.20 In a pre-post study of severely injured patients in an urban Level I trauma center, a nudge providing information on calcium-specific guidelines during blood product resuscitation was a simple solution and barriers to implementation were minimal.78 In this study, a sign indicating that 1 g calcium chloride was to be administered empirically following the administration of the fourth blood product was posted in the trauma bay, OR, and ICU. While the simplest solution in this case did not significantly improve adherence to the institutional guidelines, it was the most feasible to implement. Given competing priorities during massive transfusion, a more aggressive nudge like an automated default order for calcium in accordance with the institutional guidelines might have been more impactful.78 For example, modeling after the INPUT trial, additional strategies might be explored for improving adherence using the electronic health record.79 First, a default order for calcium can be programed following the scanning and administration of the fourth blood product during the resuscitation. Alternatively, the accountable justification strategy might be implemented. This would require the clinician to provide a justification for not administering calcium after the fourth product before being able to order or administer any other products (i.e., blood products, adjuncts). Researchers and clinicians have employed the concept of machine learning to recognize patterns among often heterogeneous and noisy datasets in the setting of both pediatric and adult critical care.80-82 Models have been used to predict the onset of sepsis, readmission to the ICU, and volume responsiveness.81, 83-86 Machine learning algorithms have even predicted mortality and length of stay in ICU patients with high accuracy.87 Patients at risk for hemorrhagic shock often present with a heterogeneous range of compensatory mechanisms leading to varying degrees of physiologic compensation. Traditional scoring systems like the ABC and Trauma-Associated Severe Hemorrhage (TASH) serve as effective tools for predicting the need for MTP.88, 89 However, these systems simplify variables for ease of clinical use—a practical yet limiting approach. This simplification can miss nuanced relationships between critical variables for accurate prediction. Machine learning offers a transformative approach to address these exact limitations. Machine learning, a subset of artificial intelligence, employs algorithms to autonomously analyze data and make predictions. Leveraging the power of granular data, machine learning algorithms can identify complex relationships often overlooked by traditional scoring systems. These algorithms have been shown to outperform established methods like the ABC score, even when operating on a very limited set of variables.22, 90 Beyond their predictive accuracy, machine learning algorithms excel in adaptability and real-time responsiveness. These models can seamlessly integrate into existing clinical workflows, providing continuously updated predictions as new data and lab results become available.91, 92 Furthermore, such analysis can reveal previously unexpected relationships between input variables and clinical outcomes. For example, a recent machine learning algorithm designed to predict MT demonstrated a surprising relationship between small perturbations in serum glucose in admission labs and the need for MT (Figure 3).22 The ability to exploit these subtle relationships and the dynamic nature of machine learning algorithms serve to equip trauma providers with the most current and accurate predictive data, facilitating timely and appropriate interventions while limiting unnecessary usage of limited resources. In addition, as more data becomes available for training, the algorithms can refine their predictive models to enhance precision and accuracy. This iterative learning process allows the system to adapt to new patterns and presentations, making it an optimal tool for navigating the intricate, time-critical challenges inherent to treating the critically injured trauma patients. Early identification of individuals requiring massive transfusion remains challenging despite existing guidelines and best practices. However, emerging technologies such as decision support systems, nudges, and machine learning may prove useful. Real-time PI decision support systems have demonstrated potential benefits with favorable reviews by end users.16 The ability to track blood product ratios, time spent in high ratio targets during the resuscitation, time to MTP, and time to adjuncts is a way to combat challenges preventing adherence to DCR best practice guidelines and optimized resuscitations. The use of nudges in acute resuscitations warrants further exploration as well. The implementation of a high-impact nudge, like a default order in the electronic medical record, may also improve adherence to guidelines by clinical care teams. Lastly, machine learning models should be further explored to both help identify patients with severe hemorrhage and prompt MTP.22 Though challenges in the early identification of patients and optimal DCR practice have persisted over the years, there are promising solutions on the rise to help improve patient outcomes. This work was supported by a Measey Fund Scholarship from the Department of Surgery, Perelman School of Medicine at the University of Pennsylvania (DS), and by the US Army Medical Research and Materiel Command under contract number W81XWH-18-C-0163 (DS, JWC). The authors have no disclosures related to this work.