Shock Volume

A Patient-Specific Index that Predicts Transfusion Requirements and Organ Dysfunction in Multiply Injured Patients

Abstract

Multiply injured patients (MIPs) in hemorrhagic shock develop oxygen debt which causes organ dysfunction and can lead to death. We developed a noninvasive patient-specific index, Shock Volume (SV), to quantify the magnitude of hypoperfusion. SV integrates the magnitude and duration that incremental shock index values are elevated above known thresholds of hypoperfusion using serial individual vital sign data. SV can be monitored in real time to assess ongoing hypoperfusion. The goal of this study was to determine how SV corresponded to transfusion requirements and organ dysfunction in a retrospective cohort of 74 MIPs. We measured SV in 6-h increments for 48 h after injury in multiply injured adults (18–65; Injury Severity Score ≥18). Patients who had accumulated 40 units of SV within 6 h of injury and 100 units of SV within 12 h of injury were at high risk for requiring massive transfusion or multiple critical administration transfusions. SV measurements were equally sensitive and specific as compared with base deficit values in predicting transfusions. SV measurements at 6 h after injury stratified patients at risk for multiple organ failure determined by Denver scores. In addition, SV values corresponded to the magnitude of organ failure determined by Sequential Organ Failure Assessment scores. SV is a patient-specific index that can be quantified in real time in critically injured patients. It is a surrogate for cumulative hypoperfusion and it predicts high-volume transfusions and organ dysfunction.

INTRODUCTION

Hemorrhagic shock (HS) causes tissue oxygen deficits leading to cellular dysfunction secondary to inadequate combustible energy substrate and oxygen ^{(1, 2)}. Restoration of perfusion rapidly generates regional oxidants that damage a variety of cellular structures ^(3–6). Collectively, these processes result in cellular-level pathologic changes ranging from mild cellular dysfunction to acute cellular necrosis. Corresponding clinical phenotypic changes range from minimal hemodynamic decompensation to organ dysfunction and death. The pathophysiologic common denominator of cellular and organ dysfunction secondary to HS is the magnitude of cellular oxygen debt resulting from cumulative tissue hypoperfusion ^(7–9).

Clinicians typically monitor hypoperfusion by interpreting progression of traditional hemodynamic indices including heart rate (HR) and systolic blood pressure (SBP), along with serum markers of hypoperfusion including lactate, pH, and corresponding base deficit (BD) calculations. Hemodynamic and serum indices best reflect the current hemodynamic and corresponding metabolic status of the patient, and yield some information on response to resuscitation. However, these indices are sampled at discrete time points and poorly reflect cumulative hypoperfusion secondary to HS.

In this study, we explored a novel technique developed to quantify accumulating hypoperfusion on a patient-specific basis, and tested the hypothesis that this patient-specific cumulative hypoperfusion index would correspond to transfusion requirements and clinical trajectories in multiply injured patients (MIPs). While HR and SBP are neither sensitive nor specific in predicting hypoperfusion, the ratio of HR/SBP, the Shock Index (SI) has been shown to correspond closely to blood loss in trauma patients ⁽¹⁰⁾. Incremental SI values are readily available on a relatively continuous basis in MIPs during the acute injury and resuscitation time frame. In this study, we integrated serial SI values, above known hypoperfusion thresholds, in a retrospective cohort of MIPs to yield a patient-specific cumulative index of hypoperfusion termed Shock Volume (SV). Our primary hypothesis was that SV would correspond to blood transfusion requirements. In addition, we hypothesized that increasing SV would correspond to phenotypic changes associated with hypoperfusion including organ dysfunction.

MATERIALS AND METHODS

Patient screening

We obtained approval from our Institutional Review Board to perform a retrospective review on MIPs admitted to a single Level-I trauma center. Multiply injured (Injury Severity Score ≥18) adults (18–65 inclusive) were identified from our trauma data registry (467 sequential patients from November 1, 2011 through December 31, 2012). Patients had to be admitted to an intensive care unit (ICU) for a minimum of 1 week after injury. We excluded patients with known immune disorders, hematologic diseases, or patients taking immunomodulating drugs. Patients had to have unambiguous vital sign and laboratory data to do all SV and clinical index calculations. This yielded a cohort of 74 patients.

Shock Volume calculations

Shock Volume (SV) quantifies the magnitude and duration that SI values exceed 0.9, which is a threshold shown to predict hypoperfusion ^{(11, 12)}. Sequential vital sign data were recorded for 48 h beginning from the initial encounter in the emergency department, and incremental SI values (HR/SBP; SI_(i)) were calculated at each vital sign sampling time point. Subsequently, 0.9 was subtracted from each SI_(i) value at each time interval to yield an SI threshold value (SI_th(i)):

SI_th(i) values less than zero were automatically set to zero. For example, for an SI_i value of 1.1, SI_th(i) = 0.2. Similarly, for an SI_i of 0.8, SI_th(i) = −0.1 = 0. SI_th(i) values represented the magnitude of hypoperfusion at each discrete sampling time point.

Incremental SI_th(i) values were integrated over time to quantify Shock Volume (SV). Initially incremental SV measurements (SV_i) were made by calculating average SI_th(i) values between two adjacent time points, and by multiplying the incrementally averaged SI_th(i) by the duration of the corresponding time interval.

The units of SV are (mmHg)⁻¹ which carry little physical significance. Therefore, SV is presented as dimensionless.

Finally, SV measurements for any designated duration, T, were made by summing SV_i values over the desired duration:

It should be noted that SV_i is dependent on the time interval duration, [(t+1) – (t)], which is affected by the sampling frequency. Sampling frequency in the first 24 h typically ranged between 5 and 20 min, in contrast to the second 24 h in which sampling frequency typically ranged between 30 and 60 min. SV measurements are inherently more accurate with increased sampling frequency.

Metabolic indices

SV values were compared to base deficit (BD) values for correspondence to transfusion requirements. We determined BD values at presentation to the emergency department (initial), and maximum values measured 6 and 12 h after injury to correspond to timing of SV calculations.

Transfusion calculations

Blood product transfusions were recorded directly from the medical records. Transfusions included PRBCs given during surgery. We recorded the number of units of packed red blood cells (PRBC) transfused during the initial 24 h after injury. Mass transfusion (MT) was defined as patients receiving 10 or more units of PRBCs in the first 24 h after injury ^{(13, 14)}. In addition, we determined the number of critical administration transfusions (CATs) defined as receiving three or more units of blood within a 60-min time period ^{(15, 16)}. We recorded the number of patients who received ≥2 CATs in the first 24 h after injury.

Organ dysfunction calculations

The presence of multiple organ failure was determined using the Denver Organ Failure assessment score, and defined as a Denver score >3 after 48 h postinjury ⁽¹⁷⁾. The Denver score assesses pulmonary, hepatic, renal, and cardiac function. The magnitude of organ dysfunction was estimated by averaging daily Sequential Organ Failure Assessment (SOFA) scores over the entire duration of ICU admission. SOFA scores assess six organ systems (Pulmonary, Cardiac, Renal, Hepatic, Hematologic, Central Nervous (CNS)), assigning an integer score from 0 to 4 daily to each system ⁽¹⁸⁾. The highest component score from each organ system was used to calculate a daily SOFA score. The study cohort had a significant number of patients who had prolonged intubation for non-CNS reasons which has been shown to inadvertently increase SOFA scores ⁽¹⁹⁾. Therefore, we removed the CNS component from daily SOFA scores to calculate a daily Modified SOFA score. Daily Modified SOFA scores were averaged over the entire ICU admission to calculate the Mean Modified SOFA score ⁽²⁰⁾.

Statistical analyses

Our primary hypothesis was that SV would correspond to transfusion requirements. We hypothesized that SV would be as sensitive and specific as BD in predicting MT and multiple (≥2) CATs. Initially, scatter plots comparing SV and BD to both MT and CATs were created to explore correspondence between clinical indices and transfusion metrics. Further analysis involved the use of Receiver Operating Characteristic (ROC) curves. Individual ROC curves were created for SV (6 and 12 h) and BD (initial, 6 h, and 12 h) predicting both MT and ≥2 CATs. For each ROC curve the Area Under the Curve (AUC) was calculated. Additionally, using bootstrapping, 90% confidence intervals were found for each AUC. All local maximas were considered for potential cutoff thresholds. Optimal thresholds with both high sensitivity and high specificity were sought, with priority being high sensitivity.

The clinical utility of SV predicting organ dysfunction was examined using regression models. Since the distribution of SV was highly skewed, the natural logarithm of SV was used in the regression models. The relationship between SV and Mean Modified SOFA scores was found to be curvilinear, especially at later time points; therefore, polynomial regression was used to model these relationships. For the models using the 6 and 12 h SV scores, the quadratic terms did not contribute significantly and were removed, resulting in final regression models that were linear.

In addition, we quantified mean and median SV values at each time point in patients who developed multiple organ failure (MOF) (Denver score >3) with patients who did not develop MOF. Mean SV values at each time point in patients who did or did not develop MOF were compared using t tests.

RESULTS

Patient cohort

We identified 467 adults (18–65) with an ISS of ≥18 admitted to ICU (n = 353) for 1 week (n = 81). Patients with hematologic or immune disorders (n = 4) and incomplete medical records (n = 3) were excluded resulting in 74 patients. The overall Injury Severity Score was 31.6 mortality rate in the cohort was 9.5% (7/74 deaths). Only one patient who died received an MT and multiple CATs.

SV and BD versus blood transfusion requirements

Thirteen (18%) patients received MT and 17 patients (23%) received ≥2 CATs in the first 24 h after injury. Scatter plots depicting MT and CATs versus SV (Fig. 1) and BD (Fig. 2) demonstrate distinct threshold values of both indices, above which the incidence of MT and CATs increases rapidly. ROC curves for 6 and 12 h SV (Fig. 3), and for initial, 6, and 12 h BD (Fig. 4) show both indices effectively predict MT and ≥2 CATs. The area under the curve (AUC) for all ROC curves and sensitivities and specificities for select thresholds are presented in Table 1. The AUC for all SV and BD measures were very similar for MT prediction (all approximately 90%). The cutoff threshold identified for SV_6h was 40.2, and prediction using that threshold had a sensitivity of 0.85 and a specificity of 0.82. The prediction threshold identified for SV_12h was 80.2 and its corresponding sensitivity and specificity were 0.92 and 0.84, respectively. The prediction thresholds for BD were all between 6.5 and 8.5 with sensitivities of 0.85 and specificities ranging from 0.69 to 0.85. Values for the prediction of ≥2 CATs were generally lower than those for MT prediction.

Fig. 1: At 6 h after injury, patients with SV_6h of ≥40 had rapid increase in receiving MT (A) and ≥2 CATs (B).Likewise, at 12-h postinjury, patients with SV_12h ≥100 had rapid increases in receiving MT (C) or ≥2 CATs (D).

Fig. 2: At 6 h after injury, patients with BD_6h of ≥8 had rapid increase in receiving MT (A) and ≥2 CATs (B).Likewise, at 12-h postinjury, patients with BD_12h ≥8 had rapid increases in receiving MT (C) or ≥2 CATs (D).

Fig. 3: ROC curves depicting SV vs. MT (A) demonstrate that SV_12h is modestly more sensitive and specific compared with SV_6h, but the differences are minimal.ROC curves of SV vs. ≥2 CATs (B) demonstrate that both SV_6h and SV_12h demonstrate excellent predictive value in identifying patients who will receive multiple CATs.

Fig. 4: ROC curves depicting initial and maximum BD vs. MT (A) demonstrate that BD_6h and BD_12h are functionally identical in predicting patients who will receive MT and are both modestly more sensitive and specific in predicting MT compared with initial BD values.ROC curves depicting maximum BD vs. ≥2 CATs (B) show that BD_initial, BD_6h, and BD_12h are very similar in predicting patients who will receive multiple CATs.

Table 1: Area under the curve values are functionally similar between SV and BD values

Organ dysfunction

Twenty-seven patients (36%) developed MOF by Denver Organ Failure criteria. Mean SV_6h measured 48 units in MOF patients compared with SV_6h of 23 units in patients who did not develop MOF (P = 0.033) (Fig. 5). Corresponding median values of SV_6h were 31 units and 6 units respectively. At the other time points, there were trends of increased mean SV values in patients who developed MOF, but these increases were not statistically significant (P values between 0.071 and 0.086). The magnitude of organ dysfunction showed modest correspondence with SV, with better correspondence measured between Mean Modified SOFA scores and SV at later time points including SV_24h (R2 = 29.6%) and SV_48h (28.9%) (Fig. 6).

Fig. 5: Mean SV values were higher at each time increment in patients who developed MOF compared with patients who did not develop MOF.Increases in SV were statistically significant at the 6-h time point in MOF patients, and trended toward statistical significance at the other time points measured.

Fig. 6: The magnitude of organ dysfunction showed increasing correspondence to SV measurements measured at later time points up to 24 h.SV_24h values and SV_48h values showed equivalent correspondence with R ² values of approximately 0.29.

DISCUSSION

Clinical indices (HR, SBP) and metabolic indices (pH, BD, lactate) of hypoperfusion reflect current patient conditions, but do not quantify cumulative hypoperfusion. SV was developed to quantify patient-specific accumulation of hypoperfusion. SV is calculated from SI values above hypoperfusion thresholds. SI has been shown to predict mortality, transfusion requirements, and organ dysfunction, and has been shown to be equally as accurate as BD in predicting transfusion requirements ^{(11, 22)}. A recent review and metanalysis have concluded that an SI of 0.9 was the best threshold for hypoperfusion ^{(11, 12, 22)}. Our data demonstrated that well-delineated thresholds of SV corresponded closely to patients requiring MT and multiple CATs. By 6-h postinjury, patients who had accumulated 40 units of SV were at risk for MT and multiple CATs. By 12-h postinjury and at all subsequent time points, patients who had accumulated approximately 100 units of SV were at risk for MT and multiple CATs. These observations are clinically meaningful as SV progression is a noninvasive calculation that can easily be automated and monitored in real time. Ongoing hypoperfusion and response to resuscitative interventions would be reflected in progression of SV. SV measurements were equally predictive of MT and multiple CATs compared with the standard metabolic marker of BD.

Decisions for transfusions in the patient cohort were not made from standing protocols in the emergency department or intensive care unit, but were made from composite clinical and laboratory information specific to the patient. Clinical parameters included SBP, heart rate, and physical signs of shock including mental status, urine output, and ongoing external hemorrhage. Laboratory measures included circulating concentration of hemoglobin and pH (lactates were infrequently measured at our institution), and calculated base deficit. Response of clinical and laboratory indices to resuscitation were the primary factors that determined transfusion requirements. Patients who received MT or multiple CATs were commonly found to have circulating hemoglobin <6.0 gm/dL, decreases in hemoglobin of ≥3 gm/dL in less than 1 to 2 h, and a BD >6 or pH <7.20 that was not improving with resuscitation. However, no single index was used as a threshold for transfusion.

By 6 h after injury, SV measurements stratified patients at risk for MOF (Fig. 5). The magnitude of organ dysfunction corresponded to SV measurements, especially at later time points (SV_24h and SV_48h; Fig. 6). Current pathomechanical models of trauma-associated organ dysfunction have focused on ischemia-associated changes leading to localized energy depletion causing cessation of tissue and organ function ^{(23, 24)}. Oxygen debt plays a central role in organ dysfunction and mortality in trauma patients. Oxygen debt correlated with mortality in several large animal models ^{(1, 8, 9, 21)}, with notably similar magnitudes of oxygen debt corresponding to exponential increases in mortality in different species. Higher SV values possibly reflect accumulation of ischemia-mediated oxygen debt leading to organ dysfunction. However, the correspondence between SV and oxygen debt is unknown.

Our patient cohort was severely injured with a mortality rate of 9.5%, mean ISS of 31, and minimum of 1-week admission to intensive care. Presumably our patient population was more severely injured than trauma patients who did not require 1 week of intensive care. In contrast, we did not investigate patients who died within 1 week of injury. This population of trauma patients would likely be more injured than the cohort in this study. In addition, our patient cohort was a single-institution consecutive population of trauma patients selected over a 14-month period. Taken together, it is unknown how our findings would extrapolate on a broader population of trauma patients from multiple institutions and with broader ranges of injury severity. However, this study is an initial exploration of a new index. We conducted a rigorous statistical analysis, comparing SV to the established clinical index BD in predicting MT or multiple CATs. Subsequent prospective investigation is warranted to determine the clinical utility of measuring SV.

The accuracy of SV inherently increases with greater sampling frequency. As vital sign sampling frequency decreases, abnormal influence of a high or low SI value on the corresponding incremental SV values would be magnified. For example, a spuriously elevated SI value recorded while vital signs were collected once per hour (typical vital monitoring frequency once a patient is stabilized) would inadvertently increase the preceding and subsequent incremental SI_i values, both of which would be multiplied by 60 causing an inaccurate increase in SV. Therefore, SV calculations need to be scrutinized closely, especially in patients with less frequent vital sign recordings. In this study, the SV values measured after 24 likely became increasingly inaccurate. The presence of organ dysfunction was determined by Denver scores and the magnitude of organ dysfunction was quantified using Mean Modified SOFA scores. There are multiple organ dysfunction scoring indices available, and none of the scores has demonstrated a clearly defined superiority in quantifying organ dysfunction in trauma patients ^(25–27). We had a significant number of patients who had daily CNS component scores of ≥3 secondary to prolonged intubation without any head trauma. Therefore, we removed CNS component scores from SOFA calculations, and we chose Mean Modified SOFA scores as a more accurate index of the magnitude of organ dysfunction ⁽¹⁹⁾.

SV is a novel, easily measured, noninvasive, patient-specific index of cumulative hypoperfusion in trauma patients. From this study we conclude that SV measured at both 6 and 12 h after injury is a strong indicator of patients who need MT or multiple CATs. It is as effective as base deficit measurements in predicting MT and multiple CATs. In addition, SV also predicted patients at risk for multiple organ failure and corresponded to the magnitude of organ dysfunction.