Severely injured patients often progress from early hypocoagulable to normal and eventually hypercoagulable states, developing increased risk for venous thromboembolism (VTE). Prophylactic ...anticoagulation can decrease this risk, but its initiation is frequently delayed for extended periods due to concerns for bleeding. To facilitate timely introduction of VTE chemoprophylaxis, we characterized the transition from hypo- to hypercoagulability and hypothesized that trauma-induced coagulopathy resolves within 24 hours after injury.
Serial blood samples were collected prospectively from critically injured patients for 120 hours after arrival at an urban Level I trauma center. Extrinsic thromboelastometry maximum clot firmness was used to classify patients as hypocoagulable (HYPO, <49 mm), normocoagulable (NORM, 49-71 mm), or hypercoagulable (HYPER, >71 mm) at each time point. Changes in coagulability over hospital course, VTE occurrence, and timing of prophylaxis initiation were analyzed.
898 patients (median Injury Severity Score, 13; mortality, 12%; VTE, 8%) were enrolled. Upon arrival, 3% were HYPO (90% NORM, 7% HYPER), which increased to 9% at 6 hours before down-trending. Ninety-seven percent were NORM by 24 hours, and 53% were HYPER at 120 hours. Median maximum clot firmness began in the NORM range, up-trended gradually, and entered the HYPER range at 120 hours. Patients with traumatic brain injury (TBI) followed a similar course and were not more HYPO at any time point than those without TBI. Failure to initiate prophylaxis by 72 hours was predicted by TBI and associated with VTE development (27% vs 16%, p < 0.05).
Regardless of injury pattern, trauma-induced coagulopathy largely resolves within 24 hours, after which hypercoagulability becomes increasingly more prevalent. Deferring initiation of chemoprophylaxis, which is often biased toward patients with intracranial injuries, is associated with VTE development.
Prognostic study, level III; Therapeutic, level IV.
Machine learning techniques have demonstrated superior discrimination compared to conventional statistical approaches in predicting trauma death. The objective of this study is to evaluate whether ...machine learning algorithms can be used to assess risk and dynamically identify patient-specific modifiable factors critical to patient trajectory for multiple key outcomes after severe injury.
SuperLearner, an ensemble machine-learning algorithm, was applied to prospective observational cohort data from 1494 critically-injured patients. Over 1000 agnostic predictors were used to generate prediction models from multiple candidate learners for outcomes of interest at serial time points post-injury. Model accuracy was estimated using cross-validation and area under the curve was compared to select among predictors. Clinical variables responsible for driving outcomes were estimated at each time point.
SuperLearner fits demonstrated excellent cross-validated prediction of death (overall AUC 0.94-0.97), multi-organ failure (overall AUC 0.84-0.90), and transfusion (overall AUC 0.87-0.9) across multiple post-injury time points, and good prediction of Acute Respiratory Distress Syndrome (overall AUC 0.84-0.89) and venous thromboembolism (overall AUC 0.73-0.83). Outcomes with inferior data quality included coagulopathic trajectory (AUC 0.48-0.88). Key clinical predictors evolved over the post-injury timecourse and included both anticipated and unexpected variables. Non-random missingness of data was identified as a predictor of multiple outcomes over time.
Machine learning algorithms can be used to generate dynamic prediction after injury while avoiding the risk of over- and under-fitting inherent in ad hoc statistical approaches. SuperLearner prediction after injury demonstrates promise as an adaptable means of helping clinicians integrate voluminous, evolving data on severely-injured patients into real-time, dynamic decision-making support.
Full text
Available for:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
Historically, hemorrhage has been attributed as the leading cause (40%) of early death. However, a rigorous, real-time classification of the cause of death (COD) has not been performed. This study ...sought to prospectively adjudicate and classify COD to determine the epidemiology of trauma mortality.
Eighteen trauma centers prospectively enrolled all adult trauma patients at the time of death during December 2015 to August 2017. Immediately following death, attending providers adjudicated the primary and contributing secondary COD using standardized definitions. Data were confirmed by autopsies, if performed.
One thousand five hundred thirty-six patients were enrolled with a median age of 55 years (interquartile range, 32-75 years), 74.5% were male. Penetrating mechanism (n = 412) patients were younger (32 vs. 64, p < 0.0001) and more likely to be male (86.7% vs. 69.9%, p < 0.0001). Falls were the most common mechanism of injury (26.6%), with gunshot wounds second (24.3%). The most common overall primary COD was traumatic brain injury (TBI) (45%), followed by exsanguination (23%). Traumatic brain injury was nonsurvivable in 82.2% of cases. Blunt patients were more likely to have TBI (47.8% vs. 37.4%, p < 0.0001) and penetrating patients exsanguination (51.7% vs. 12.5%, p < 0.0001) as the primary COD. Exsanguination was the predominant prehospital (44.7%) and early COD (39.1%) with TBI as the most common later. Penetrating mechanism patients died earlier with 80.1% on day 0 (vs. 38.5%, p < 0.0001). Most deaths were deemed disease-related (69.3%), rather than by limitation of further aggressive care (30.7%). Hemorrhage was a contributing cause to 38.8% of deaths that occurred due to withdrawal of care.
Exsanguination remains the predominant early primary COD with TBI accounting for most deaths at later time points. Timing and primary COD vary significantly by mechanism. Contemporaneous adjudication of COD is essential to elucidate the true understanding of patient outcome, center performance, and future research.
Epidemiologic, level II.
It has been observed that trauma patients often display elevated procoagulant activity that could be caused, in part, by tissue factor (TF). We previously observed that trauma patients with thermal, ...blunt, and penetrating injuries have active FIXa and FXIa in their plasma. In the current study, we evaluated the effect of injury severity, with or without accompanying shock, on the frequency and concentration of TF, FIXa, and FXIa in plasma from trauma patients.
Eighty trauma patients were enrolled and divided equally into four groups based on their Injury Severity Score and base deficit:Blood was collected at a 0 time-point (first blood draw upon arrival at hospital) and citrate plasma was prepared, frozen, and stored at -80 °C. FXIa, FIXa, and TF activity assays were based on a response of thrombin generation to corresponding monoclonal inhibitory antibodies.
The frequency and median concentrations of TF were relatively low in non-severe injury groups (17.5% and 0 pM, respectively) but were higher in those with severe injury (65% and 0.5 pM, respectively). Although FXIa was observed in 91% of samples and was high across all four groups, median concentrations were highest (by approximately fourfold) in groups with shock. FIXa was observed in 80% of plasma samples and concentrations varied in a relatively narrow range between all four groups. No endogenous activity was observed in plasma from healthy individuals.
(1) Frequency and concentration of TF is higher in patients with a higher trauma severity. (2) Concentration of FXIa is higher in patients with shock. (3) For the first time reported, the vast majority of plasma samples from trauma patients contain active FIXa and FXIa.
Prognostic/epidemiological study, level II.
Payers have approached select complications as never events, yet there is rationale that achieving a zero incidence of these events is impractical. Prior 2005 National Trauma Data Bank (NTDB) ...analysis showed high rates (37%) of centers reporting no complications data making national estimates for determining standardized complication rates difficult to ascertain.
The 2008-2012 NTDB National Sample Program nationally weighted files were used to calculate yearly national estimates. Rates were compared in all centers and those reporting complications data. Hospital characteristics were compared using Student t test. In 2011, an other complication category was introduced; overall rates were calculated with and without this category. Yearly estimates were reported for patients receiving care within centers reporting complications data.
From 2008-2012 NTDB, there were raw data on 3,657,884 patients. A total of 594,894 patients (16.3%) experienced one or more complications (82.7% one complication; 17.3% two or more complications). Excluding the other complication category, the overall weighted rate was 8.4% to 9.2%. Pneumonia was the most common complication (2.7-3.0%), occurring at twice the 2005 rate. The number of centers reporting no complications data dropped to 8.1% in 2011 (2008, 14.5%; 2009, 18.2%; 2010, 15.9%; 2012, 8.9%). By 2012, nearly all level I centers reported complications, whereas 46.4% of level IVs reported none (I 0.5%, II 2.7%, III 8.5%, p = 0.04). Data were reported the least frequently in nonteaching hospitals (15.8%, p = 0.007), those in the South (19.6%, p = 0.007), and those with less than 200 beds (23.6%, p = 0.005).
Overall rates of complications from 2008 to 2012 were nearly twofold higher than 2005 data. Reporting has increased, and NTDB may provide a valuable platform for establishing rational and achievable measures for specific complications.
Prognostic and epidemiological, level IV.
Injury to the blood-brain barrier exposes endothelium rich in von Willebrand factor (vWF), which may play a role in altered platelet aggregation following traumatic brain injury (TBI). Ristocetin is ...an antimicrobial substance that induces vWF-mediated aggregation of platelets. We examined these mechanisms in injured patients by measuring the aggregation response of platelets to stimulating agonists (including ristocetin) via whole-blood multiple-electrode platelet aggregometry. We hypothesized that patients with TBI have an altered platelet aggregation response to ristocetin stimulation compared with patients without TBI.
Blood was collected from 233 trauma patients without thrombocytopenia. Platelet aggregation was assessed using multiple-electrode platelet aggregometry (Multiplate). Platelet aggregation response to stimulating agonists collagen, thrombin receptor-activating peptide 6, adenosine diphosphate, arachidonic acid, and ristocetin was measured. Factor activity was measured.
Of the 233 patients, 23% had TBI. There were no differences in platelet aggregation responses to any agonists between TBI and non-TBI patients except ristocetin. Platelet aggregation response to ristocetin stimulation was significantly lower in TBI patients (p = 0.03). Patients with TBI also had higher factor VIII activity (215% vs. 156%, p = 0.01). In multivariate analysis, there was a significant independent association of impaired platelet aggregation response to ristocetin stimulation with TBI (odds ratio, 3.05; p = 0.04).
Given the importance of platelets in hemostasis, understanding the mechanisms of impaired platelet aggregation following injury is critical. The impaired platelet aggregation response to ristocetin stimulation and corresponding increase in factor VIII activity in TBI patients may be secondary to a TBI-induced effect on vWF quantity (due to injury-driven consumption of vWF) or vWF function with resultant increase in circulating factor VIII activity (due to impaired carrying capacity of vWF). Given there are multiple known therapies for vWF deficits including desmopressin, purified and recombinant vWF, and estrogens, these lines of investigation are particularly compelling in patients with TBI and hemorrhage.
Prognostic study, level II.
Previous work proposed a Massive Transfusion Score (MTS) calculated from values obtained in the emergency department to predict likelihood of massive transfusion (MT). We hypothesized the MTS could ...be used at Hour 6 to differentiate who continues to require balanced resuscitation in Hours 7 to 24 and to predict death at 28 days.
We prospectively enrolled patients in whom the MT protocol was initiated from 2005 to 2011. Data including timing of blood products were determined at Hours 0, 6, 12, and 24. For each patient, transfusion needs were defined based on either an inappropriately low hemoglobin response to transfusion or a hemoglobin decrease of greater than 1 g/dL if no transfusion. Timing and cause of death were used to account for survivor bias. Multivariate logistic regression was used to determine independent predictors of outcome.
A total of 190 MT protocol activations were included, and by Hour 6, 61% required 10 U or greater packed red blood cells. Calculated at initial presentation, a revised MTS (systolic blood pressure < 90 mm Hg, base deficit ≥ 6, temperature < 35.5°C, international normalized ratio > 1.5, hemoglobin < 11 g/dL) was superior to the original MTS (including heart rate ≥ 120 beats per minute, Focused Assessment With Sonography in Trauma FAST status, mechanism) or the Assessment of Blood Consumption (ABC) score for predicting MT (area under the curve AUC MT at 6 hours, 0.68; 95% confidence interval CI, 0.57-0.79; at 24 hours, 0.72; 0.61-0.83; p < 0.05). For those alive at Hour 6, the revised MTS was predictive of future packed red blood cell need (AUC, 0.87) in Hours 7 to 12, 24-hour mortality (AUC, 0.95), and 28-day mortality (AUC, 0.77). For each additional positive trigger of the MTS at Hour 6, the odds of death at 24 hours and 28 days were substantially increased (24-hour odds ratio, 4.6; 95% CI, 2.3-9.3; 28-day odds ratio, 2.2; 95% CI, 1.5-3.2; p < 0.0001).
Early end points of resuscitation adopted from the components of the revised MTS are predictive of ongoing transfusion. Failure to normalize these components by Hour 6 portends a particularly poor prognosis.
Prognostic study, level 3.
Goal-directed hemostatic resuscitation based on thrombelastography (TEG) has a survival benefit compared with conventional coagulation assays such as international normalized ratio, activated partial ...thromboplastin time, fibrinogen level, and platelet count. While TEG-based transfusion thresholds for patients at risk for massive transfusion (MT) have been defined using rapid TEG, cutoffs have not been defined for TEG using other activators such as kaolin. The purpose of this study was to develop thresholds for blood product transfusion using citrated kaolin TEG (CK-TEG) in patients at risk for MT.
CK-TEG was assessed in trauma activation patients at two Level 1 trauma centers admitted between 2010 and 2017. Receiver operating characteristic (ROC) curve analyses were performed to test the predictive performance of CK-TEG measurements in patients requiring MT, defined as >10 units of red blood cells or death within the first 6 hours. The Youden Index defined optimal thresholds for CK-TEG-based resuscitation.
Of the 825 trauma activations, 671 (81.3%) were men, 419 (50.8%) suffered a blunt injury, and 62 (7.5%) received a MT. Patients who had a MT were more severely injured, had signs of more pronounced shock, and more abnormal coagulation assays. CK-TEG R-time was longer (4.9 vs. 4.4 min, p = 0.0084), angle was lower (66.2 vs. 70.3 degrees, p < 0.0001), maximum amplitude was lower in MT (57 vs. 65.5 mm, p < 0.0001), and LY30 was greater (1.8% vs. 1.2%, p = 0.0012) in patients with MT compared with non-MT. To predict MT, R-time yielded an area under the ROC curve (AUROC) = 0.6002 and a cut point of >4.45 min. Angle had an AUROC = 0.6931 and a cut point of <67 degrees. CMA had an AUROC = 0.7425, and a cut point of <60 mm. LY30 had an AUROC = 0.623 with a cut point of >4.55%.
We have identified CK-TEG thresholds that can guide MT in trauma. We propose plasma transfusion for R-time >4.45 min, fibrinogen products for an angle <67 degrees, platelet transfusion for MA <60 mm, and antifibrinolytics for LY30 >4.55%.
Therapeutic study, level V.
Impaired postinjury platelet aggregation is common, but the effect of transfusion on this remains unclear. Data suggest that following injury platelet transfusion may not correct impaired platelet ...aggregation, and impaired platelet aggregation may not predict the need for platelet transfusion. We sought to further investigate platelet aggregation responses to transfusions, using regression statistics to isolate the independent effects of transfusions given in discrete time intervals from injury on both immediate and longitudinal platelet aggregation. We hypothesized that platelet aggregation response to platelet transfusion increases over time from injury.
Serial (0-96 hours) blood samples were collected from 248 trauma patients. Platelet aggregation was assessed in vitro with impedance aggregometry stimulated by adenosine diphosphate, collagen, and thrombin receptor-activating peptide-6. Using regression, transfusion exposure was modeled against platelet aggregation at each subsequent timepoint and adjusted for confounders (Injury Severity Score, international normalized ratio (INR), base deficit, platelet count, and interval transfusions). The expected change in platelet aggregation at each timepoint under the intervention of transfusion exposure was calculated and compared with the observed platelet aggregation.
The 248 patients analyzed were severely injured (Injury Severity Score, 21 ± 19), with normal platelet counts (mean, 268 × 10/L ± 90), and 62% were transfused in 24 hours. The independent effect of transfusions on subsequent platelet aggregation over time was modeled with observed platelet aggregation under hypothetical treatment of one unit transfusion of blood, plasma, or platelets. Platelet transfusions had increasing expected effects on subsequent platelet aggregation over time, with the maximal expected effect occurring late (4-5 days from injury).
Controversy exists on whether transfusions improve impaired postinjury platelet aggregation. Using regression modeling, we identified that expected transfusion effects on subsequent platelet aggregation are maximal with platelet transfusion given late after injury. This is critical for tailored resuscitation, identifying a potential early period of resistance to platelet transfusion that resolves by 96 hours.
Therapeutic, level V.
Acute respiratory distress syndrome (ARDS) is common after traumatic brain injury (TBI) and is associated with worse neurologic outcomes and longer hospitalization. However, the incidence and ...associated causes of ARDS in isolated TBI have not been well studied.
We performed a subgroup analysis of 210 consecutive patients with isolated severe TBI enrolled in a prospective observational cohort at a Level 1 trauma center between 2005 and 2014. Subjects required endotracheal intubation and had isolated severe TBI defined by a head Abbreviated Injury Scale (AIS) score of 3 or greater and AIS score lower than 3 in all other categories. ARDS within the first 8 days of admission was rigorously adjudicated using Berlin criteria. Regression analyses were used to test the association between predictors of interest and ARDS.
The incidence of ARDS in the first 8 days after severe isolated TBI was 30%. Patients who developed ARDS were administered more crystalloids (4.3 L vs. 3.5 L, p = 0.005) and blood products in the first 12 hours of admission. Patients with ARDS had significantly worse clinical outcomes measured at 28 days, including longer median intensive care unit and hospital stays (4 days vs. 13 days, p < 0.001, and 7.5 days vs. 14.5 days, p < 0.001, respectively). In unadjusted logistic regression analyses, the odds of developing ARDS were significantly associated with head AIS score (odds ratio OR, 1.8; p = 0.018), male sex (OR, 2.9; p = 0.012), and early transfusion of platelets (OR, 2.8; p = 0.003). These associations were similar in a multivariate logistic regression model.
In the era of balanced hemostatic resuscitation practices, severity of head injury, male sex, early crystalloids, and early transfusion of platelets are associated with a higher risk of ARDS after severe isolated TBI. Early transfusion of platelets after severe TBI may be a modifiable risk factor for ARDS, and these findings invite further investigation into causal mechanisms driving this observed association.
Prognostic/epidemiologic study, level III.