The Battle of the Titans-Comparing Resuscitation Between Five Major Burn Centers Using the Burn Navigator.

The goal of burn resuscitation is to provide the optimal amount of fluid necessary to maintain end-organ perfusion and prevent burn shock. The objective of this analysis was to examine how the Burn Navigator (BN), a clinical decision support tool in burn resuscitation, was utilized across five major burn centers in the United States, using an observational trial of 300 adult patients. Subject demographics, burn characteristics, fluid volumes, urine output, and resuscitation-related complications were examined. Two hundred eighty-five patients were eligible for analysis. There was no difference among the centers on mean age (45.5 ± 16.8 years), body mass index (29.2 ± 6.9), median injury severity score (18 [interquartile range: 9-25]), or total body surface area (TBSA) (34 [25.8-47]). Primary crystalloid infusion volumes at 24 h differed significantly in ml/kg/TBSA (range: 3.1 ± 1.2 to 4.5 ± 1.7). Total fluids, including colloid, drip medications, and enteral fluids, differed among centers in both ml/kg (range: 132.5 ± 61.4 to 201.9 ± 109.9) and ml/kg/TBSA (3.5 ± 1.0 to 5.3 ± 2.0) at 24 h. Post-hoc adjustment using pairwise comparisons resulted in a loss of significance between most of the sites. There was a total of 156 resuscitation-related complications in 92 patients. Experienced burn centers using the BN successfully titrated resuscitation to adhere to 24 h goals. With fluid volumes near the Parkland formula prediction and a low prevalence of complications, the device can be utilized effectively in experienced centers. Further study should examine device utility in other facilities and on the battlefield.

Initial Results of the American Burn Association Observational Multicenter Evaluation on the Effectiveness of the Burn Navigator.

The objective of this multicenter observational study was to evaluate resuscitation volumes and outcomes of patients who underwent fluid resuscitation utilizing the Burn Navigator (BN), a resuscitation clinical decision support tool. Two analyses were performed: examination of the first 24 hours of resuscitation and the first 24 hours postburn regardless of when the resuscitation began, to account for patients who presented in a delayed fashion. Patients were classified as having followed the BN (FBN) if all hourly fluid rates were within ±20 ml of BN recommendations for that hour at least 83% of the time; otherwise, they were classified as not having followed BN (NFBN). Analysis of resuscitation volumes for FBN patients in the first 24 hours resulted in average volumes for primary crystalloid and total fluids administered of 4.07 ± 1.76 ml/kg/TBSA (151.48 ± 77.46 ml/kg) and 4.68 ± 2.06 ml/kg/TBSA (175.01 ± 92.22 ml/kg), respectively. Patients who presented in a delayed fashion revealed average volumes for primary and total fluids of 5.28 ± 2.54 ml/kg/TBSA (201.11 ± 106.53 ml/kg) and 6.35 ± 2.95 ml/kg/TBSA (244.08 ± 133.5 ml/kg), respectively. There was a significant decrease in the incidence of burn shock in the FBN group (P < .05). This study shows that the BN provides comparable resuscitation volumes of primary crystalloid fluid to the Parkland formula, recommends total fluid infusion less than the Ivy index, and was associated with a decreased incidence of burn shock. Early initiation of the BN device resulted in lower overall fluid volumes.

Quantifying the effects of wound healing risk and potential on clinical measurements and outcomes of severely burned patients: A data-driven approach.

INTRODUCTION: Given recent advances in computational power, the goal of this study was to quantify the effects of wound healing risk and potential on clinical measurements and outcomes of severely burned patients, with the hope of providing more insight on factors that affect wound healing. METHODS: This retrospective study involved patients who had at least 10% TBSA% "burned" and three burn mappings each. To model risk to wounds, we defined the variable θ, a hypothetical threshold for TBSA% "open wound" used to demarcate "low-risk" from "high-risk" patients. Low-risk patients denoted those patients whose actual TBSA% "open wound" ≤θ, whereas high-risk patients denoted those patients whose actual TBSA% "open wound" >θ. To consider all possibilities of risk, 100 sub analyses were performed by (1) varying θ from 100% to 1% in decrements of 1%, (2) grouping all patients as either "low-risk" or "high-risk" for each θ, and (3) comparing all means and deviations of variables and outcomes between the two groups for each θ. Hence, this study employed a data-driven approach to capture trends in clinical measurements and outcomes. Plots and tables were also obtained. RESULTS: For 303 patients, median age and weight were 43 [29-59] years and 85 [72-99]kg, respectively. Mean TBSA% "burned" was 25 [17-39] %, with a full-thickness burn of 4 [0-15] %. Average crystalloid volumes were 4.25±2.27mL/kg/TBSA% "burned" in the first 24h. Importantly, for high-risk patients, decreasing θ was matched by significant increases in PaO2-FiO2 ratio, platelet count, Glasgow coma score (GCS), and MAP. On the other hand, increasing their risk θ was also matched by significant increases in creatinine, bilirubin, lactate, blood, estimated blood loss, and 24-h and total fluid volumes. As expected, for low-risk patients, clinical measurements were more stable, despite decreasing or increasing θ. At a θ of 80%, statistical tests indicated much disparity between high-risk and low-risk patients for TBSA% "burned", full thickness burn, bilirubin (1.66±1.16mg/dL versus 0.83±0.65mg/dL, p=0.005), GCS (7±2 versus 12±3, p<0.001), MAP (42±22mm Hg versus 59±22mm Hg, p=0.004), 24-h blood, estimated blood loss, 24-h fluid, total fluid, and ICU length of stay (81±113 days versus 24±27 days, p=0.002). These differences were all statistically significant and remained significant down to θ=10%. CONCLUSION: Wound healing risk and potential may be forecasted by many different clinical measurements and outcomes and has many implications on multi-organ function. Future work will be needed to further explain and understand these effects, in order to facilitate development of new predictive models for wound healing.

Relationship Between Burn Wound Location and Outcomes in Severely Burned Patients: More Than Meets the Size.

We hypothesized that burn location plays an important role in wound healing, mortality, and other outcomes and conducted the following study to test this multifold hypothesis. We conducted a study to retrospectively look at patients with burns ≥10% TBSA. Demographics, TBSA, partial/full thickness burns (PT/FT) in various wound locations, fluids, inhalation injury, mortality, ICU duration, and hospital duration were considered. Initial wound healing rates (%/d) were also calculated as a slope from the time of the first mapping of open wound size to the time of the third mapping of open wound size. Multivariate logistic regression and operating curves were used to measure mortality prediction performance. All values were expressed as median [interquartile range]. The mortality rate for 318 patients was 17% (54/318). In general, patients were 43 years [29, 58 years] old and had a TBSA of 25% [17, 39%], PT of 16% [10, 25%], and FT of 4% [0, 15%]. Between patients who lived and did not, age, TBSA, FT, 24-hour fluid, and ICU duration were statistically different (P < .001). Furthermore, there were statistically significant differences in FT head (0% [0, 0%] vs 0% [0, 1%], P = .048); FT anterior torso (0% [0, 1%] vs 1% [0, 4%], P < .001); FT posterior torso (0% [0, 0%] vs 0% [0, 4%], P < 0.001); FT upper extremities (0% [0, 3%] vs 2% [0, 11%], P < .001); FT lower extremities (0% [0, 2%] vs 6% [0, 17%], P < .001); and FT genitalia (0% [0, 0%] vs 0% [0, 2%], P < .001). Age, presence of inhalation injury, PT/FT upper extremities, and FT lower extremities were independent mortality predictors and per unit increases of these variables were associated with an increased risk for mortality (P < .05): odds ratio of 1.09 (95% confidence interval [CI] = 1.61-1.13; P < .001) for mean age; 2.69 (95% CI = 1.04-6.93; P = .041) for inhalation injury; 1.14 (95% CI = 1.01-1.27; P = .031) for mean PT upper extremities; 1.26 (95% CI = 1.11-1.42; P < .001) for mean FT upper extremities; and 1.07 (95% CI = 1.01-1.12; P = .012) for mean FT lower extremities. Prediction of mortality was better using specific wound locations (area under the curve [AUC], AUC of 0.896) rather than using TBSA and FT (AUC of 0.873). Graphs revealed that initial healing rates were statistically lower and 24-hour fluids and ICU length of stay were statistically higher in patients with FT upper extremities than in patients without FT extremities (P < .001). Burn wound location affects wound healing and helps predict mortality and ICU length of stay and should be incorporated into burn triage strategies to enhance resource allocation or stratify wound care.

Predicting the Ability of Wounds to Heal Given Any Burn Size and Fluid Volume: An Analytical Approach.

The intrinsic relationship between fluid volume and open wound size (%) has not been previously examined. Therefore, we conducted this study to investigate whether open wound size can be predicted from fluid volume plus other significant factors over time and to evaluate how machine learning may perform in predicting open wound size. This retrospective study involved patients with at least 20% TBSA burned. Various predictive models were developed and compared using goodness-of-fit statistics (R2, error [mean absolute error (MAE), root mean squared error (RMSE)]). Bland-Altman analysis was also performed to determine bias. A total of 121 patients were included in the analysis. Median TBSA burned was 31% (interquartile range: 26-46%). Average crystalloid volumes were 4.0 ± 2.7 ml/kg/TBSA in the first 24 hours. There were 24 (20%) patients who died. Importantly, multivariate analysis identified seven independent predictors of open wound size. Also, machine learning analysis was able to stratify patients based on the 20th day after admission, ~40% TBSA burned, and fluid volumes. Models for predicting open wound size varied in performance (R2 = .79-.90, MAE = 3.97-7.52, RMSE = 7.11-10.69). Notably, a combined machine learning model using only four features (fluid volume, days since admission, TBSA burned, age) performed the best and was sufficient to predict open wound size, with >90% goodness of fit and <4% absolute error. Bland-Altman analysis showed that there were no biases in the models. Open wound size can be predicted reliably using machine learning and fluid volume, days since admission, TBSA burned, and age. Future work will be needed to validate the utility of this study's models in a clinical environment.

Trend Analysis of Current Modalities for Monitoring Fluid Therapy in Patients With Large Burns: Echoing the Call for Better Resuscitation Indices.

The aim of this study was to investigate whether current standards at our institution have effectively monitored resuscitations of severely burned patients during the first 48 hours postburn. Demographics, injuries assessed by TBSA and full thickness (FT), and resuscitation volumes (lactated Ringer's [LR]) were compared for all patients and those who died or survived. Means and standard deviations of hourly indices (urinary output [UOP], lactate [LAC], base excess [BE]) vs LR were analyzed. Waveforms, four-quadrant concordance, and correlation were also employed to compare the trending abilities (hourly changes [∆]) of aforementioned variables vs LR. A total of 203 patients were included in the analysis. Of these patients, 71 (35%) died, and 50 (25%) had inhalation injuries. Mean age and weight were 47 ± 19 years and 87 ± 18 kg, respectively. Mean TBSA burned was 41 ± 20%, with a mean FT of 18 ± 24%. Importantly, normalized waveform plots demonstrated the inability of UO, LAC, and BE to follow hourly changes in LR. Correlation of these variables was weak (r>>-1). This was confirmed by concordance plots. Slopes in all groups demonstrated that UOP was a better resuscitative monitor than LAC or BE. ∆UOP responded to ∆LR better in patients who survived than died. Reliance on hourly UOP as the sole index of optimal resuscitation is not supported. This study echoed the call for better resuscitation indices.

Machine Learning for Predicting Outcomes in Trauma.

To date, there are no reviews on machine learning (ML) for predicting outcomes in trauma. Consequently, it remains unclear as to how ML-based prediction models compare in the triage and assessment of trauma patients. The objective of this review was to survey and identify studies involving ML for predicting outcomes in trauma, with the hypothesis that models predicting similar outcomes may share common features but the performance of ML in these studies will differ greatly. MEDLINE and other databases were searched for studies involving trauma and ML. Sixty-five observational studies involving ML for the prediction of trauma outcomes met inclusion criteria. In total 2,433,180 patients were included in the studies. The studies focused on prediction of the following outcome measures: survival/mortality (n = 34), morbidity/shock/hemorrhage (n = 12), hospital length of stay (n = 7), hospital admission/triage (n = 6), traumatic brain injury (n = 4), life-saving interventions (n = 5), post-traumatic stress disorder (n = 4), and transfusion (n = 1). Six studies were prospective observational studies. Of the 65 studies, 33 used artificial neural networks for prediction. Importantly, most studies demonstrated the benefits of ML models. However, algorithm performance was assessed differently by different authors. Sensitivity-specificity gap values varied greatly from 0.035 to 0.927. Notably, studies shared many features for model development. A common ML feature base may be determined for predicting outcomes in trauma. However, the impact of ML will require further validation in prospective observational studies and randomized clinical trials, establishment of common performance criteria, and high-quality evidence about clinical and economic impacts before ML can be widely accepted in practice.

Inefficacy of standard vital signs for predicting mortality and the need for prehospital life-saving interventions in blunt trauma patients transported via helicopter: A repeated call for new measures.

BACKGROUND: The aim of this study was to investigate the efficacy of traditional vital signs for predicting mortality and the need for prehospital lifesaving interventions (LSIs) in blunt trauma patients requiring helicopter transport to a Level I trauma center. Our hypothesis was that standard vital signs are not sufficient for identifying or determining treatment for those patients most at risk. METHODS: This study involved prehospital trauma patients suffering from blunt trauma (motor vehicle/cycle collision) and transported from the point of injury via helicopter. Means and standard deviations for vital signs and Glasgow Coma Scale (GCS) scores were obtained for non-LSI versus LSI and survivor versus nonsurvivor patient groups and then compared using Wilcoxon statistical tests. Variables with statistically significant differences between patient groups were then used to develop multivariate logistic regression models for predicting mortality and/or the need for prehospital LSIs. Receiver-operating characteristic (ROC) curves were also obtained to compare these models. RESULTS: A final cohort of 195 patients was included in the analysis. Thirty (15%) patients received a total of 39 prehospital LSIs. Of these, 12 (40%) died. In total, 33 (17%) patients died. Of these, 21 (74%) did not receive prehospital LSIs. Model variables were field heart rate, lowest systolic blood pressure, shock index, pulse pressure, and GCS components. Using vital signs alone, ROC curves demonstrated poor prediction of LSI needs, mortality, and nonsurvivors who did not receive LSIs (area under the curve [AUC], AUCs: 0.72, 0.65, and 0.61). When using both vital signs and GCS, ROC curves still demonstrated poor prediction of nonsurvivors overall and nonsurvivors who did not receive LSIs (AUCs: 0.67, 0.74). CONCLUSION: The major implication of this study was that traditional vital signs cannot identify or determine treatment for many prehospital blunt trauma patients who are at great risk. This study reiterated the need for new measures to improve blunt trauma triage and prehospital care. LEVEL OF EVIDENCE: Therapeutic/care management, level IV.

The impact of patient weight on burn resuscitation.

BACKGROUND: Optimal fluid resuscitation of burn patients with burns greater than 20% total body surface area is critical to prevent burn shock during the initial 24 hours to 48 hours postburn. Currently, most resuscitation formulas incorporate the patient's weight when estimating 24-hour fluid requirements. The objective of this study was to determine the impact of weight on fluid resuscitation requirements and outcomes during the initial 24 hours after admission. METHODS: We performed a retrospective review of patients admitted to our burn intensive care unit from December 2007 to April 2013, resuscitated with a computerized decision support system. We classified patients into body mass index (BMI) categories of underweight (BMI: <18.5), normal (BMI: 18.5-24.9), overweight (BMI: 25.0-29.9), or obese (BMI: >30.0). We also calculated the percent difference from ideal body weight (IBW) and compared 24-hour fluid volumes received. RESULTS: Patients with missing weight and/or height values were excluded from the study, resulting in a final cohort of 161 patients for analysis. Mean total body surface area was 42 ± 20% with a full thickness burn of 18 ± 23%. Mean age, weight, and height were 47 ± 19 years, 83 ± 19 kg, and 68 ± 4 inches, respectively. IBW for this cohort was 68 ± 11 kg with a BMI of 28 ± 6. Univariate analysis showed significant differences in 24-hour resuscitation volumes (mL/kg) between normal and obese patients (p < 0.05). Further analysis revealed that increasing percent difference from IBW was associated with lower fluid volumes. Although obesity was not associated with inhalation injury or renal replacement therapy, it was correlated to an increased risk for mortality (p < 0.05). CONCLUSION: This analysis showed that increasing weight was associated with lower fluid resuscitation volume requirements and a higher mortality rate, despite the low incidence of inhalation injury and renal replacement therapy in our obese patients. The use of actual body weight to drive resuscitation volumes may result in overresuscitation of obese patients, depending on the resuscitation formula. Further studies are needed to better explain the relationship between mortality and obesity in burn patients. LEVEL OF EVIDENCE: Therapeutic/care management, level IV.

Closed-Loop Control of FiO2 Rapidly Identifies Need For Rescue Ventilation and Reduces ARDS Severity in a Conscious Sheep Model of Burn and Smoke Inhalation Injury.

Pulmonary injury can be characterized by an increased need for fraction of inspired oxygen or inspired oxygen percentage (FiO2) to maintain arterial blood saturation of oxygenation (SaO2). We tested a smart oxygenation system (SOS) that uses the activity of a closed-loop control FiO2 algorithm (CLC-FiO2) to rapidly assess acute respiratory distress syndrome (ARDS) severity so that rescue ventilation (RscVent) can be initiated earlier. After baseline data, a pulse-oximeter (noninvasive saturation of peripheral oxygenation [SpO2]) was placed. Sheep were then subjected to burn and smoke inhalation injury and followed for 48 h. Initially, sheep were spontaneously ventilating and then randomized to standard of care (SOC) (n = 6), in which RscVent began when partial pressure of oxygen (PaO2) < 90 mmHg or FiO2 < 0.6, versus SOS (n = 7), software that incorporates and displays SpO2, CLC-FiO2, and SpO2/CLC-FiO2 ratio, at which RscVent was initiated when ratio threshold < 250. RscVent was achieved using a G5 Hamilton ventilator (Bonaduz, Switzerland) with adaptive pressure ventilation and adaptive support ventilation modes for SOC and SOS, respectively. OUTCOMES: the time difference from when SpO2/FiO2 < 250 to RscVent initiation was 4.7 ±â€Š0.6 h and 0.2 ±â€Š0.1 h, SOC and SOS, respectively (P < 0.001). Oxygen responsiveness after RscVent, defined as SpO2/FiO2 > 250 occurred in 4/7, SOS and 0/7, SOC. At 48 h the SpO2/FiO2 ratio was 104 ±â€Š5 in SOC versus 228 ±â€Š59 in SOS (P = 0.036). Ventilatory compliance and peak airway pressures were significantly improved with SOS versus SOC (P < 0.001). Data suggest that SOS software, e.g. SpO2/CLC-FiO2 ratio, after experimental ARDS can provide a novel continuous index of pulmonary function that is apparent before other clinical symptoms. Earlier initiation of RscVent translates into improved oxygenation (reduces ARDS severity) and ventilation.

Predicting the proportion of full-thickness involvement for any given burn size based on burn resuscitation volumes.

INTRODUCTION: The depth of burn has been an important factor often overlooked when estimating the total resuscitation fluid needed for early burn care. The goal of this study was to determine the degree to which full-thickness (FT) involvement affected overall 24-hour burn resuscitation volumes. METHODS: We performed a retrospective review of patients admitted to our burn intensive care unit from December 2007 to April 2013, with significant burns that required resuscitation using our computerized decision support system for burn fluid resuscitation. We defined the degree of FT involvement as FT Index (FTI; percentage of FT injury/percentage of total body surface area (TBSA) burned [%FT / %TBSA]) and compared variables on actual 24-hour fluid resuscitation volumes overall as well as for any given burn size. RESULTS: A total of 203 patients admitted to our burn center during the study period were included in the analysis. Mean age and weight were 47 ± 19 years and 87 ± 18 kg, respectively. Mean %TBSA was 41 ± 20 with a mean %FT of 18 ± 24. As %TBSA, %FT, and FTI increased, so did actual 24-hour fluid resuscitation volumes (mL/kg). However, increase in FTI did not result in increased volume indexed to burn size (mL/kg per %TBSA). This was true even when patients with inhalation injury were excluded. Further investigation revealed that as %TBSA increased, %FT increased nonlinearly (quadratic polynomial) (R = 0.994). CONCLUSION: Total burn size and FT burn size were both highly correlated with increased 24-hour fluid resuscitation volumes. However, FTI did not correlate with a corresponding increase in resuscitation volumes for any given burn size, even when patients with inhalation injury were excluded. Thus, there are insufficient data to presume that those who receive more volume at any given burn size are likely to be mostly full thickness or vice versa. This was influenced by a relatively low sample size at each 10%TBSA increment and larger burn sizes disproportionately having more FT burns. A more robust sample size may elucidate this relationship better. LEVEL OF EVIDENCE: Therapeutic/care management study, level IV.

Acute Respiratory Distress Syndrome in Burn Patients: A Comparison of the Berlin and American-European Definitions.

The purpose of this study was to compare the Berlin definition to the American-European Consensus Conference (AECC) definition in determining the prevalence of acute respiratory distress syndrome (ARDS) and associated mortality in the critically ill burn population. Consecutive patients admitted to our institution with burn injury that required mechanical ventilation for more than 24 hours were included for analysis. Included patients (N = 891) were classified by both definitions. The median age, % TBSA burn, and injury severity score (interquartile ranges) were 35 (24-51), 25 (11-45), and 18 (9-26), respectively. Inhalation injury was present in 35.5%. The prevalence of ARDS was 34% using the Berlin definition and 30.5% using the AECC definition (combined acute lung injury and ARDS), with associated mortality rates of 40.9 and 42.9%, respectively. Under the Berlin definition, mortality rose with increased ARDS severity (14.6% no ARDS; 16.7% mild; 44% moderate; and 59.7% severe, P < 0.001). By contrast, under the AECC definition increased mortality was seen only for ARDS category (14.7% no ARDS; 15.1% acute lung injury; and 46.0% ARDS, P < 0.001). The mortality of the 22 subjects meeting the AECC, but not the Berlin definition was not different from patients without ARDS (P = .91). The Berlin definition better stratifies ARDS in terms of severity and correctly excludes those with minimal disease previously captured by the AECC.

Machine learning and new vital signs monitoring in civilian en route care: A systematic review of the literature and future implications for the military.

BACKGROUND: Although air transport medical services are today an integral part of trauma systems in most developed countries, to date, there are no reviews on recent innovations in civilian en route care. The purpose of this systematic review was to identify potential machine learning and new vital signs monitoring technologies in civilian en route care that could help close civilian and military capability gaps in monitoring and the early detection and treatment of various trauma injuries. METHODS: MEDLINE, the Cochrane Database of Systematic Reviews, and citation review of relevant primary and review articles were searched for studies involving civilian en route care, air medical transport, and technologies from January 2005 to November 2015. Data were abstracted on study design, population, year, sponsors, innovation category, details of technologies, and outcomes. RESULTS: Thirteen observational studies involving civilian medical transport met inclusion criteria. Studies either focused on machine learning and software algorithms (n = 5), new vital signs monitoring (n = 6), or both (n = 2). Innovations involved continuous digital acquisition of physiologic data and parameter extraction. Importantly, all studies (n = 13) demonstrated improved outcomes where applicable and potential use during civilian and military en route care. However, almost all studies required further validation in prospective and/or randomized controlled trials. CONCLUSION: Potential machine learning technologies and monitoring of novel vital signs such as heart rate variability and complexity in civilian en route care could help enhance en route care for our nation's war fighters. In a complex global environment, they could potentially fill capability gaps such as monitoring and the early detection and treatment of various trauma injuries. However, the impact of these innovations and technologies will require further validation before widespread acceptance and prehospital use. LEVEL OF EVIDENCE: Systematic review, level V.

Blood pressure and heart rate from the arterial blood pressure waveform can reliably estimate cardiac output in a conscious sheep model of multiple hemorrhages and resuscitation using computer machine learning approaches.

BACKGROUND: This study was a first step to facilitate the development of automated decision support systems using cardiac output (CO) for combat casualty care. Such systems remain a practical challenge in battlefield and prehospital settings. In these environments, reliable CO estimation using blood pressure (BP) and heart rate (HR) may provide additional capabilities for diagnosis and treatment of trauma patients. The aim of this study was to demonstrate that continuous BP and HR from the arterial BP waveform coupled with machine learning (ML) can reliably estimate CO in a conscious sheep model of multiple hemorrhages and resuscitation. METHODS: Hemodynamic parameters (BPs, HR) were derived from 100-Hz arterial BP waveforms of 10 sheep records, 3 hours to 4 hours long. Two models (mean arterial pressure, Windkessel) were then applied and merged to estimate COVS. ML was used to develop a rule for identifying when models required calibration. All records contained 100-Hz recording of pulmonary arterial blood flow using Doppler transit time (COFP). COFP and COVS were analyzed using equivalence tests and Bland-Altman analysis, as well as waveform and concordance plots. RESULTS: Baseline COFP varied from 3.0 L/min to 5.4 L/min, while posthemorrhage COFP varied from 1.0 L/min to 1.8 L/min. A total of 315,196 pairs of data were obtained. Equivalence tests for individual records showed that COVS was statistically equivalent to COFP (p < 0.05). Smaller equivalence thresholds (<0.3 L/min) indicated an overall high COFP accuracy. The agreement between COFP and COVS was -0.13 (0.69) L/min (Bland-Altman). In an exclusion zone of 12%, trending analysis found a 92% concordance between 5-minute changes in COFP and COVS. CONCLUSION: This study showed that CO can be reliably estimated using BPs and HR from the arterial BP waveform in combination with ML. A next step will be to test this approach using noninvasive BPs and HR.

Machine learning in burn care and research: A systematic review of the literature.

BACKGROUND: To date, there are no reviews on machine learning (ML) in burn care. Considering the growth of ML in medicine and the complexities and challenges of burn care, this review specializes on ML applications in burn care. The objective was to examine the features and impact of applications in targeting various aspects of burn care and research. METHODS: MEDLINE, the Cochrane Database of Systematic Reviews, ScienceDirect, and citation review of relevant primary and review articles were searched for studies involving burn care/research and machine learning. Data were abstracted on study design, study size, year, population, application of burn care/research, ML technique(s), and algorithm performance. RESULTS: 15 retrospective observational studies involving burn patients met inclusion criteria. In total 5105 patients with acute thermal injury, 171 clinical burn wounds, 180 9-mer peptides, and 424 12-mer peptides were included in the studies. Studies focused on burn diagnosis (n=5), aminoglycoside response (n=3), hospital length of stay (n=2), survival/mortality (n=4), burn healing time (n=1), and antimicrobial peptides in burn patients (n=1). Of these 15 studies, 11 used artificial neural networks. Importantly, all studies demonstrated the benefits of ML in burn care/research and superior performance over traditional statistical methods. However, algorithm performance was assessed differently by different authors. Feature selection varied among studies, but studies with similar applications shared specific features including age, gender, presence of inhalation injury, total body surface area burned, and when available, various degrees of burns, infections, and previous histories/conditions of burn patients. CONCLUSION: A common feature base may be determined for ML in burn care/research, but the impact of ML will require further validation in prospective observational studies and randomized clinical trials, establishment of common performance metrics, and high quality evidence about clinical and economic impacts. Only then can ML applications be advanced and accepted widely in burn care/research.

Is heart-rate complexity a surrogate measure of cardiac output before, during, and after hemorrhage in a conscious sheep model of multiple hemorrhages and resuscitation?

BACKGROUND: Despite its medical utility, continuous cardiac output (CO) monitoring remains a practical challenge on the battlefield and in the prehospital environment. Measuring a CO surrogate, perhaps heart-rate complexity (HRC), might be a viable solution when no direct monitoring of CO is available. Changes in HRC observed before and during hemorrhagic shock may be able to track the simultaneous changes in CO. The goal of this study was to test whether HRC is a surrogate measure of CO before, during, and after hemorrhage in a conscious sheep model of multiple hemorrhages and resuscitation. METHODS: HRC was derived from 100-Hz electrocardiograms of 10 sheep records, 3 hours to 4 hours long, using the method of sample entropy. A real-time detection algorithm was used to detect the R-R interval sequences for HRC calculations. All records contained 100-Hz recordings of pulmonary arterial blood flow using Doppler transit time (criterion standard CO). Gold CO and estimated HRC values were analyzed using overlaid time-synchronized waveform plots as well as Bland-Altman, regression, and four-quadrant analysis. RESULTS: Baseline CO varied from 3.0 L/min to 5.4 L/min, while posthemorrhage CO varied from 1.0 L/min to 1.8 L/min. Importantly, overlaid plots demonstrated an overall high similarity between CO and HRC waveforms before and during hemorrhage, but not during resuscitation. When the electrocardiogram quality was high, the correlation between CO and HRC within the first 45 minutes was greater than 0.75 (p < 0.0001; maximum r, 0.972). Scatter plots also depicted high linearity before and during hemorrhage. Four-quadrant analysis showed that instantaneous changes between consecutive beat-to-beat HRC measurements followed CO measurements (100% concordance rate), while 5-minute time points yielded a 76.19% concordance rate. CONCLUSION: HRC has potential utility as a noninvasive tool for assessing the response of CO to life-threatening injuries such as hemorrhagic shock. However, further investigation and other animal models or human studies are needed.

Data quality of a wearable vital signs monitor in the pre-hospital and emergency departments for enhancing prediction of needs for life-saving interventions in trauma patients.

This study was designed to investigate the quality of data in the pre-hospital and emergency departments when using a wearable vital signs monitor and examine the efficacy of a combined model of standard vital signs and respective data quality indices (DQIs) for predicting the need for life-saving interventions (LSIs) in trauma patients. It was hypothesised that prediction of needs for LSIs in trauma patients is associated with data quality. Also, a model utilizing vital signs and DQIs to predict the needs for LSIs would be able to outperform models using vital signs alone. Data from 104 pre-hospital trauma patients transported by helicopter were analysed, including means and standard deviations of continuous vital signs, related DQIs and Glasgow coma scale (GCS) scores for LSI and non-LSI patient groups. DQIs involved percentages of valid measurements and mean deviation ratios. Various multivariate logistic regression models for predicting LSI needs were also obtained and compared through receiver-operating characteristic (ROC) curves. Demographics of patients were not statistically different between LSI and non-LSI patient groups. In addition, ROC curves demonstrated better prediction of LSI needs in patients using heart rate and DQIs (area under the curve [AUC] of 0.86) than using heart rate alone (AUC of 0.73). Likewise, ROC curves demonstrated better prediction using heart rate, total GCS score and DQIs (AUC of 0.99) than using heart rate and total GCS score (AUC of 0.92). AUCs were statistically different (p < 0.05). This study showed that data quality could be used in addition to continuous vital signs for predicting the need for LSIs in trauma patients. Importantly, trauma systems should incorporate processes to regulate data quality of physiologic data in the pre-hospital and emergency departments. By doing so, data quality could be improved and lead to better prediction of needs for LSIs in trauma patients.

Improving the prediction of mortality and the need for life-saving interventions in trauma patients using standard vital signs with heart-rate variability and complexity.

The goal of this study was to determine the effectiveness of using traditional and new vital signs (heart rate variability and complexity [HRV, HRC]) for predicting mortality and the need for life-saving interventions (LSIs) in prehospital trauma patients. Our hypothesis was that statistical regression models using traditional and new vital signs would be superior in predictive performance over models using standard vital signs alone. This study involved 108 prehospital trauma patients transported from the point of injury via helicopter. Heart rate variability and HRC were calculated using criterion standard R-R interval sequences manually verified from the patients' electrocardiograms. Means and standard deviations for vital signs, HRV, HRC, and Glasgow coma scale (GCS) scores were obtained for nonsurvivors versus survivors and LSI versus non-LSI patient groups and then compared using Wilcoxon statistical tests. Receiver-operating characteristic curves were also obtained to compare different regression models for predicting mortality and the need for LSIs. Seventeen patients (16%) died. Eighty-two patients (76%) received a total of 142 LSIs. Receiver-operating characteristic curves demonstrated better prediction of mortality and LSI needs using heart rate and HRC (area under the curve [AUC]; AUCs, 0.86 and 0.86) than using heart rate alone (AUCs, 0.79 and 0.57). Likewise, receiver-operating characteristic curves demonstrated better prediction using total GCS score and HRC (AUCs, 0.82 and 0.97) than using total GCS score (AUCs, 0.81 and 0.91). Similar results were obtained for heart rate and HRV (AUCs, 0.86 and 0.73). The major implication of this study was that traditional and new vital signs (HRV and HRC) should be used simultaneously to improve prediction of mortality and the need for LSIs in prehospital trauma patients during all echelons of trauma care. Improvements in the timely use and diagnostic accuracy of transportable vital signs monitors will require use of traditional and new vital signs from the trauma patient cohort.

Evaluation of standard versus nonstandard vital signs monitors in the prehospital and emergency departments: results and lessons learned from a trauma patient care protocol.

BACKGROUND: This study aimed to determine the effectiveness of using a wireless, portable vital signs monitor (WVSM) for predicting the need for lifesaving interventions (LSIs) in the emergency department (ED) and use a multivariate logistic regression model to determine whether the WVSM was an improved predictor of LSIs in the ED over the standard of care monitor currently being used. METHODS: This study analyzed 305 consecutive patients transported from the scene via helicopter to a Level I trauma center. For 104 patients in the study, a WVSM was also attached to the patient's arm and used to record and display prehospital and hospital physiologic data in real time on a handheld computer and in the trauma bay. Multivariate logistic regression analyses were performed for accuracy in predicting needs for LSIs in control and WVSM subjects. In addition, receiver operating characteristic curves were obtained to examine the discriminating power of the models for the outcome of one or more LSIs in the ED. RESULTS: Of the 305 patients, 73 underwent 109 LSIs in the ED. Of these, 21 patients wore the WVSM during transport in addition to the standard monitor. Logistic regression analysis revealed that heart rate, respiratory rate, and systolic blood pressure were significantly associated with an increased risk for LSIs in the ED (p < 0.05). Receiver operating characteristic curve analysis also demonstrated better prediction for LSIs performed in the ED in WVSM subjects than in control subjects (area under the curve, 0.86 vs. 0.81, respectively). CONCLUSION: The WVSM system leads to improved LSI accuracy in the ED. In addition, many important lessons have been learned in preparation for this study. Adoption of nonstandard vital signs monitors into critical care/trauma medicine may require a new paradigm of personnel education, training, and practice. LEVEL OF EVIDENCE: Therapeutic/care management, level IV.

Utility of vital signs, heart rate variability and complexity, and machine learning for identifying the need for lifesaving interventions in trauma patients.

To date, no studies have attempted to utilize data from a combination of vital signs, heart rate variability and complexity (HRV, HRC), as well as machine learning (ML), for identifying the need for lifesaving interventions (LSIs) in trauma patients. The objectives of this study were to examine the utility of the above for identifying LSI needs and compare different LSI-associated models, with the hypothesis that an ML model would be superior in performance over multivariate logistic regression models. One hundred four patients transported from the injury scene via helicopter were selected for the study. A wireless vital signs monitor was attached to the patient's arm and used to capture physiologic data, including HRV and HRC. The power of vital sign measurements, HRV, HRC, and Glasgow Coma Scale score (GCS) to identify patients requiring LSIs was estimated using multivariate logistic regression and ML. Receiver operating characteristic (ROC) curves were also obtained. Thirty-two patients underwent 75 LSIs. After logistic regression, ROC curves demonstrated better identification for LSIs using heart rate (HR) and HRC (area under the curve [AUC] of 0.81) than using HR alone (AUC of 0.73). Likewise, ROC curves demonstrated better identification for LSIs using GCS and HRC (AUC of 0.94) than using GCS and HR (AUC of 0.92). Importantly, ROC curves demonstrated that an ML model using HR, GCS, and HRC (AUC of 0.99) had superior performance over multivariate logistic regression models for identifying the need for LSIs in trauma patients. Development of computer decision support systems should utilize vital signs, HRC, and ML in order to achieve more accurate diagnostic capabilities, such as identification of needs for LSIs in trauma patients.