Fig 1 - uploaded by Eric Elster
Content may be subject to copyright.
ROC curves. a RF with variables ISS, AIS chest, and cryoprecipitate (received prior to WRNMMC). b LR with same variables as in a. c RF with variables, FGF-basic, IL-2R, and IL-6. d LR with same variables as c
Source publication
Background
Tools to assist clinicians in predicting pneumonia could lead to a significant decline in morbidity. Therefore, we sought to develop a model in combat trauma patients for identifying those at highest risk of pneumonia.
Methods
This was a retrospective study of 73 primarily blast-injured casualties with combat extremity wounds. Binary cl...
Contexts in source publication
Context 1
... for predicting pneumonia was the RF algorithm using the variables ISS, AIS chest, and cryoprecipitate given within the first 24 h, which were selected from BE performed on all included variables in the dataset. This RF algorithm produced a sensitivity of 1.0, specificity of 0.89, and AUC of 0.97. ROC curves for the four algorithms are shown in Fig. 1. The solitary contributions of each variable can be seen from the plots (Fig. 3 ...
Context 2
... for predicting pneumonia was the RF algorithm using the variables ISS, AIS chest, and cryoprecipitate given within the first 24 h, which were selected from BE performed on all included variables in the dataset. This RF algorithm produced a sensitivity of 1.0, specificity of 0.89, and AUC of 0.97. ROC curves for the four algorithms are shown in Fig. 1. The solitary contributions of each variable can be seen from the plots (Fig. 3 ...
Citations
... P < 0.05 was considered statistically significant. their performance: logistic regression (LR) (22), random forest (RF) (22), support vector machine (SVM) (23), gradient boosting decision tree (GBDT) (24), backpropagation artificial neural network (ANN) (23), extreme gradient boosting (XGBoost) (25), and naive Bayes (NB) (26). These algorithms have previously been shown to be stable and suitable for clinical datasets (27). ...
... P < 0.05 was considered statistically significant. their performance: logistic regression (LR) (22), random forest (RF) (22), support vector machine (SVM) (23), gradient boosting decision tree (GBDT) (24), backpropagation artificial neural network (ANN) (23), extreme gradient boosting (XGBoost) (25), and naive Bayes (NB) (26). These algorithms have previously been shown to be stable and suitable for clinical datasets (27). ...
... The CPIS tool has shown moderate reliability and accuracy in diagnosing nosocomial pneumonia (33, 34). The CPIS, however, is based on radiographic and laboratory values following symptom onset, as with other tools (22). Thus, CPIS may assist clinicians in narrowing down the identification of patients with pneumonia, but treatment becomes delayed rather than early prevention. ...
Objective
Pneumonia is a common pulmonary complication of flail chest, causing high morbidity and mortality rates in affected patients. The existing methods for identifying pneumonia have low accuracy, and their use may delay antimicrobial therapy. However, machine learning can be combined with electronic medical record systems to identify information and assist in quick clinical decision-making. Our study aimed to develop a novel machine-learning model to predict pneumonia risk in flail chest patients.
Methods
From January 2011 to December 2021, the electronic medical records of 169 adult patients with flail chest at a tertiary teaching hospital in an urban level I Trauma Centre in Chongqing were retrospectively analysed. Then, the patients were randomly divided into training and test sets at a ratio of 7:3. Using the Fisher score, the best subset of variables was chosen. The performance of the seven models was evaluated by computing the area under the receiver operating characteristic curve (AUC). The output of the XGBoost model was shown using the Shapley Additive exPlanation (SHAP) method.
Results
Of 802 multiple rib fracture patients, 169 flail chest patients were eventually included, and 86 (50.80%) were diagnosed with pneumonia. The XGBoost model performed the best among all seven machine-learning models. The AUC of the XGBoost model was 0.895 (sensitivity: 84.3%; specificity: 80.0%).
Pneumonia in flail chest patients was associated with several features: systolic blood pressure, pH value, blood transfusion, and ISS.
Conclusion
Our study demonstrated that the XGBoost model with 32 variables had high reliability in assessing risk indicators of pneumonia in flail chest patients. The SHAP method can identify vital pneumonia risk factors, making the XGBoost model's output clinically meaningful.
... 4 Regardless, emerging models can theoretically become important tools, establishing anticipated clinical trajectories, identifying deviation from those trajectories, suggesting treatment course corrections, and redirecting resourcesdall while providing a deeper understanding of disease processes. 5 Machine learning algorithms and advanced modeling tools have been successfully used to predict post-traumatic complications, such as multiorgan failure, 6 limb loss, 7 acute kidney injury (AKI), 8 pneumonia, 3 and infection. 2,9,10 The response to traumatic injury is complex, requiring hormonal 11 and immunological 12 adaptations in addition to more intuitive social, demographic, mechanical, and hemostatic elements. ...
Background:
An emerging body of literature supports the role of individualized prognostic tools to guide the management of patients after trauma. The aim of this study was to develop advanced modeling tools from multidimensional data sources, including immunological analytes and clinical and administrative data, to predict outcomes in trauma patients.
Methods:
This was a prospective study of trauma patients at Level 1 centers from 2015 to 2019. Clinical, flow cytometry, and serum cytokine data were collected within 48 hours of admission. Sparse logistic regression models were developed, jointly selecting predictors and estimating the risk of ventilator-associated pneumonia, acute kidney injury, complicated disposition (death, rehabilitation, or nursing facility), and return to the operating room. Model parameters (regularization controlling model sparsity) and performance estimation were obtained via nested leave-one-out cross-validation.
Results:
A total of 179 patients were included. The incidences of ventilator-associated pneumonia, acute kidney injury, complicated disposition, and return to the operating room were 17.7%, 28.8%, 22.5%, and 12.3%, respectively. Regarding extensive resource use, 30.7% of patients had prolonged intensive care unit stay, 73.2% had prolonged length of stay, and 23.5% had need for prolonged ventilatory support. The models were developed and cross-validated for ventilator-associated pneumonia, acute kidney injury, complicated dispositions, and return to the operating room, yielding predictive areas under the curve from 0.70 to 0.91. Each model derived its optimal predictive value by combining clinical, administrative, and immunological analyte data.
Conclusion:
Clinical, immunological, and administrative data can be combined to predict post-traumatic outcomes and resource use. Multidimensional machine learning modeling can identify trauma patients with complicated clinical trajectories and high resource needs.
... A further challenge is the interpretation of microbial metagenomic data and assessment of its predictive value. Previous studies have demonstrated the promise of employing machine learning and other biostatistical techniques for predicting the outcome of trauma from biomolecular features [18][19][20] . Machine learning methods have been used to process multi-omic datasets for identification of features predictive of trauma 21 and have associated viral micro-RNA detection with injury outcome 22 . ...
... Machine learning classifiers have been applied to combat injuries for prediction of pneumonia 19 , infection 48 , closure timing 29 , venous thromboembolism 18 , and heterotopic ossification 49 . The utility of clinical variables for predicting outcomes was demonstrated in these studies, thus the current study employed a microbe-centric focus. ...
Battlefield injury management requires specialized care, and wound infection is a frequent complication. Challenges related to characterizing relevant pathogens further complicates treatment. Applying metagenomics to wounds offers a comprehensive path toward assessing microbial genomic fingerprints and could indicate prognostic variables for future decision support tools. Wound specimens from combat-injured U.S. service members, obtained during surgical debridements before delayed wound closure, were subjected to whole metagenome analysis and targeted enrichment of antimicrobial resistance genes. Results did not indicate a singular, common microbial metagenomic profile for wound failure, instead reflecting a complex microenvironment with varying bioburden diversity across outcomes. Genus-level Pseudomonas detection was associated with wound failure at all surgeries. A logistic regression model was fit to the presence and absence of antimicrobial resistance classes to assess associations with nosocomial pathogens. A. baumannii detection was associated with detection of genomic signatures for resistance to trimethoprim, aminoglycosides, bacitracin, and polymyxin. Machine learning classifiers were applied to identify wound and microbial variables associated with outcome. Feature importance rankings averaged across models indicated the variables with the largest effects on predicting wound outcome, including an increase in P. putida sequence reads. These results describe the microbial genomic determinants in combat wound bioburden and demonstrate metagenomic investigation as a comprehensive tool for providing information toward aiding treatment of combat-related injuries.
... Cytokine and chemokine expression were evaluated in serum samples using a multiplex assay approach as previously described. 5 Serum aliquots were thawed and filtered (0.65 μm; Millipore, Billerica, MS). Human Cytokine Magnetic 35-Plex Panel assays (cat. ...
Introduction:
The pathophysiology of the inflammatory response after major trauma is complex and the magnitude correlates with severity of tissue injury and outcomes. Study of infection-mediated immune pathways have demonstrated that cellular microRNAs may modulate the inflammatory response. The authors hypothesize that the expression of microRNAs would correlate to Complicated Recoveries in polytrauma patients (PtPs).
Methods:
PtPs enrolled in the prospective observational Tissue and Data Acquisition Protocol with Injury Severity Score (ISS) >15 were selected for this study. PtPs were divided into Complicated Recoveries and uncomplicated recovery groups. PtPs blood samples were obtained at the time of admission (T0). Established biomarkers of systemic inflammation, including cytokines and chemokines were measured using multiplexed Luminex-based methods and novel microRNAs were measured in plasma samples using multiplex RNA hybridization.
Results:
PtPs (n = 180) had high ISS (26 [20-34]) and Complicated Recovery rate of 33%. microRNAs were lower in PtPs at T0 compared to healthy controls, and bivariate analysis demonstrated that variations of microRNAs correlated with age, race, comorbidities, venous thromboembolism, pulmonary complications, Complicated Recovery, and mortality. Positive correlations were noted between microRNAs and IL-10, VEGF, APACHE, and SOFA scores. Multivariable LASSO analysis of predictors of Complicated Recovery based on microRNAs, cytokines and chemokines, revealed that miR-21-3p and MCP-1 were predictive of Complicated Recovery with an AUC of 0.78.
Conclusion:
Systemic microRNAs were associated with poor outcomes in PtPs, and results are consistent with previously described trends in critically ill patients. These early biomarkers of inflammation might provide predictive utility in early Complicated Recovery diagnosis and prognosis. Due to their potential to regulate immune responses microRNAs may provide therapeutic targets for immunomodulation.
Level of evidence:
II, Diagnostic Tests.
... Plasma levels of inflammatory cytokines and chemokines were determined in using a multiplexing immune assay approach, as previously described (21). Plasma aliquots were thawed and filtered (0.65 µm; Millipore, Billerica, MS). ...
Objectives:
To evaluate early activation of latent viruses in polytrauma patients and consider prognostic value of viral micro-RNAs in these patients.
Design:
This was a subset analysis from a prospectively collected multicenter trauma database. Blood samples were obtained upon admission to the trauma bay (T0), and trauma metrics and recovery data were collected.
Setting:
Two civilian Level 1 Trauma Centers and one Military Treatment Facility.
Patients:
Adult polytrauma patients with Injury Severity Scores greater than or equal to 16 and available T0 plasma samples were included in this study. Patients with ICU admission greater than 14 days, mechanical ventilation greater than 7 days, or mortality within 28 days were considered to have a complicated recovery.
Interventions:
None.
Measurements and main results:
Polytrauma patients (n = 180) were identified, and complicated recovery was noted in 33%. Plasma samples from T0 underwent reverse transcriptase-quantitative polymerase chain reaction analysis for Kaposi's sarcoma-associated herpesvirus micro-RNAs (miR-K12_10b and miRK-12-12) and Epstein-Barr virus-associated micro-RNA (miR-BHRF-1), as well as Luminex multiplex array analysis for established mediators of inflammation. Ninety-eight percent of polytrauma patients were found to have detectable Kaposi's sarcoma-associated herpesvirus and Epstein-Barr virus micro-RNAs at T0, whereas healthy controls demonstrated 0% and 100% detection rate for Kaposi's sarcoma-associated herpesvirus and Epstein-Barr virus, respectively. Univariate analysis revealed associations between viral micro-RNAs and polytrauma patients' age, race, and postinjury complications. Multivariate least absolute shrinkage and selection operator analysis of clinical variables and systemic biomarkers at T0 revealed that interleukin-10 was the strongest predictor of all viral micro-RNAs. Multivariate least absolute shrinkage and selection operator analysis of systemic biomarkers as predictors of complicated recovery at T0 demonstrated that miR-BHRF-1, miR-K12-12, monocyte chemoattractant protein-1, and hepatocyte growth factor were independent predictors of complicated recovery with a model complicated recovery prediction area under the curve of 0.81.
Conclusions:
Viral micro-RNAs were detected within hours of injury and correlated with poor outcomes in polytrauma patients. Our findings suggest that transcription of viral micro-RNAs occurs early in the response to trauma and may be associated with the biological processes involved in polytrauma-induced complicated recovery.
... Clinical features alone have not shown adequate prognostic performance in SBE, despite having some predictive value (37,38,40,55,56). Combining clinical features with immunologic markers has been used to prognosticate outcomes in other disease states (57)(58)(59)(60). An improved understanding of the immunologic response combined with available clinical features in SBE could inform a prognostic model to predict severity and recovery in SBE. ...
... In 2019, Gelbard et al. developed models using clinical data, cytokines, chemokines and growth factors that predicted severe sepsis and organ space infections following laparotomy for abdominal trauma (58). This modeling approach has expanded to combat trauma, where a predictive model composed of clinical features and immunologic biomarkers can accurately predict pneumonia in a predominantly blast-injured cohort of patients (59). By expanding this technique to SBE, we have built on the prior work evaluating both the clinical features predicting severity and studies of the association of soluble biomarkers with severity (47,48,65,66). ...
Background
The immunologic pathways activated during snakebite envenoming (SBE) are poorly described, and their association with recovery is unclear. The immunologic response in SBE could inform a prognostic model to predict recovery. The purpose of this study was to develop pre- and post-antivenom prognostic models comprised of clinical features and immunologic cytokine data that are associated with recovery from SBE.
Materials and Methods
We performed a prospective cohort study in an academic medical center emergency department. We enrolled consecutive patients with Crotalinae SBE and obtained serum samples based on previously described criteria for the Surgical Critical Care Initiative (SC2i)(ClinicalTrials.gov Identifier: NCT02182180). We assessed a standard set of clinical variables and measured 35 unique cytokines using Luminex Cytokine 35-Plex Human Panel pre- and post-antivenom administration. The Patient-Specific Functional Scale (PSFS), a well-validated patient-reported outcome of functional recovery, was assessed at 0, 7, 14, 21 and 28 days and the area under the patient curve (PSFS AUPC) determined. We performed Bayesian Belief Network (BBN) modeling to represent relationships with a diagram composed of nodes and arcs. Each node represents a cytokine or clinical feature and each arc represents a joint-probability distribution (JPD).
Results
Twenty-eight SBE patients were enrolled. Preliminary results from 24 patients with clinical data, 9 patients with pre-antivenom and 11 patients with post-antivenom cytokine data are presented. The group was mostly female (82%) with a mean age of 38.1 (SD ± 9.8) years. In the pre-antivenom model, the variables most closely associated with the PSFS AUPC are predominantly clinical features. In the post-antivenom model, cytokines are more fully incorporated into the model. The variables most closely associated with the PSFS AUPC are age, antihistamines, white blood cell count (WBC), HGF, CCL5 and VEGF. The most influential variables are age, antihistamines and EGF. Both the pre- and post-antivenom models perform well with AUCs of 0.87 and 0.90 respectively.
Discussion
Pre- and post-antivenom networks of cytokines and clinical features were associated with functional recovery measured by the PSFS AUPC over 28 days. With additional data, we can identify prognostic models using immunologic and clinical variables to predict recovery from SBE.
... They suggest the use of advanced modeling tools, combining immunologic and clinical data, to guide and support clinical management and decision making strategies in an individualized and cost-effective manner. [7][8][9] The purpose of this study was to describe the pattern of nosocomial infection among severely injured trauma patients undergoing exploratory laparotomy and to develop multivariable models to address organ space surgical site infection (OSI), severe sepsis (SS) and ventilator-associated pneumonia (VAP) risk after major trauma. We hypothesized that a model, which integrated both clinical and immunologic parameters, could be created to accurately anticipate these complications and form the basis of a clinical decision support tool to facilitate bedside decision making. ...
Background:
Flow cytometry (FCM) is a rapid diagnostic tool for monitoring immune cell function. We sought to determine if assessment of cell phenotypes using standardized FCM could be used to identify nosocomial infection after trauma.
Methods:
Prospective study of trauma patients at a Level 1 center from 2014-2018. Clinical and FCM data were collected within 24h of admission. Random forest (RF) models were developed to estimate the risk of severe sepsis (SS), organ space infection (OSI) and ventilator-associated pneumonia (VAP). Variables were selected using backwards elimination and models were validated with leave-one-out.
Results:
138 patients were included (median age 30y (23-44), median ISS 20 (14-29), 76% (105/138) black, 60% (83/138) gunshots). The incidence of SS was 8.7% (12/138), OSI 16.7% (23/138), and VAP 18% (25/138). The final RF SS model resulted in 5 variables [RBCs transfused in first 24h; absolute counts of CD56- CD16+ lymphocytes, CD4+ T cells, and CD56 bright natural killer (NK) cells; percentage of CD16+ CD56+ NK cells] that identified SS with AUC 0.89, sensitivity 0.98, and specificity 0.78. The final RF OSI model resulted in 4 variables (RBC in first 24h, shock index, absolute CD16+ CD56+ NK cell counts, percentage of CD56 bright NK cells) that identified OSI with AUC 0.76, sensitivity 0.68, and specificity 0.82. The RF VAP model resulted in 6 variables (SOFA score; ISS; CD4- CD8- T cell counts; percentages of CD16- CD56- NK cells, CD16- CD56+ NK cells, and CD19+ B lymphocytes) that identified VAP with AUC 0.86, sensitivity 0.86, and specificity 0.83.
Conclusion:
Combined clinical and FCM data can assist with early identification of posttraumatic infections. The presence of NK cells supports the innate immune response that occurs during acute inflammation. Further research is needed to determine the functional role of these innate cell phenotypes and their value in predictive models immediately after injury.
Level of evidence:
Level III, prognostic.
... Nonetheless, no successful univariate metric emerged from our analysis of 31 cytokines. Multivariate and/or machine learning methods have been successful categorizing patient outcomes in sepsis 44 and trauma, 45 but the variables used are often physiologic rather than molecular. Also, combinations of physiologic measurements with molecular markers made over time have had some success in predicting outcomes in sepsis. ...
Objectives:
Sepsis and sterile both release "danger signals' that induce the systemic inflammatory response syndrome (SIRS). So differentiating infection from SIRS can be challenging. Precision diagnostic assays could limit unnecessary antibiotic use, improving outcomes.
Methods:
After surveying human leukocyte cytokine production responses to sterile damage-associated molecular patterns (DAMPs), bacterial pathogen-associated molecular patterns, and bacteria we created a multiplex assay for 31 cytokines. We then studied plasma from patients with bacteremia, septic shock, "severe sepsis," or trauma (ISS ≥15 with circulating DAMPs) as well as controls. Infections were adjudicated based on post-hospitalization review. Plasma was studied in infection and injury using univariate and multivariate means to determine how such multiplex assays could best distinguish infective from noninfective SIRS.
Results:
Infected patients had high plasma interleukin (IL)-6, IL-1α, and triggering receptor expressed on myeloid cells-1 (TREM-1) compared to controls [false discovery rates (FDR) <0.01, <0.01, <0.0001]. Conversely, injury suppressed many mediators including MDC (FDR <0.0001), TREM-1 (FDR <0.001), IP-10 (FDR <0.01), MCP-3 (FDR <0.05), FLT3L (FDR <0.05), Tweak, (FDR <0.05), GRO-α (FDR <0.05), and ENA-78 (FDR <0.05). In univariate studies, analyte overlap between clinical groups prevented clinical relevance. Multivariate models discriminated injury and infection much better, with the 2-group random-forest model classifying 11/11 injury and 28/29 infection patients correctly in out-of-bag validation.
Conclusions:
Circulating cytokines in traumatic SIRS differ markedly from those in health or sepsis. Variability limits the accuracy of single-mediator assays but machine learning based on multiplexed plasma assays revealed distinct patterns in sepsis- and injury-related SIRS. Defining biomarker release patterns that distinguish specific SIRS populations might allow decreased antibiotic use in those clinical situations. Large prospective studies are needed to validate and operationalize this approach.
... Clinical features alone have not shown adequate prognostic performance in SBE, despite having some predictive value (37,38,40,55,56). Combining clinical features with immunologic markers has been used to prognosticate outcomes in other disease states (57)(58)(59)(60). An improved understanding of the immunologic response combined with available clinical features in SBE could inform a prognostic model to predict severity and recovery in SBE. ...
... In 2019, Gelbard et al. developed models using clinical data, cytokines, chemokines and growth factors that predicted severe sepsis and organ space infections following laparotomy for abdominal trauma (58). This modeling approach has expanded to combat trauma, where a predictive model composed of clinical features and immunologic biomarkers can accurately predict pneumonia in a predominantly blast-injured cohort of patients (59). By expanding this technique to SBE, we have built on the prior work evaluating both the clinical features predicting severity and studies of the association of soluble biomarkers with severity (47,48,65,66). ...
Background: Pneumonia is associated with increased morbidity and costs in the intensive care unit (ICU). Its early identification is key for optimal outcomes, but early biomarkers are lacking. Studies suggest that fibrinolysis resistance (FR) after major abdominal surgery is linked to an increased risk of infection. Patients and Methods: Patients in a randomized controlled trial for hemorrhagic shock were evaluated for FR. Fibrinolysis resistance was quantified by thrombelastography with exogenous tissue plasminogen activator (tPA-TEG) at 24- and 48-hours post-injury and measuring LY30 (%). A receiver-operating characteristics (ROC) curve analysis was used to identify a cutoff for increased risk of pneumonia, which was then validated in ICU patients at risk for venous thromboembolism (VTE). Multivariable logistic regression was used to control for confounders. Results: Forty-nine patients in the hemorrhagic shock cohort had tPA-TEGs at 24- and 48-hours (median ISS, 27; 7% pneumonia). A composite tPA-TEG LY30 of less than 4% at 24 and 48 hours was found to be the optimal cutoff for increased risk of pneumonia. This cohort had a seven-fold increased rate of pneumonia (4% vs. 28%; p = 0.048). Eighty-eight patients in the VTE cohort had tPA-TEGs at 24 and 48 hours post-ICU admission (median ISS, 28; 6% pneumonia). The tPA-TEG LY30 of less than 4% was associated with a 10-fold increased rate of pneumonia (19% vs. 1.5%; p = 0.002). In patients with traumatic brain injury, the same association was found (33% vs. 3.2%; p = 0.006). Adjusting for confounders, the tPA-TEG persisted as a substantial risk factor for pneumonia (adjusted odds ratio [OR], 35.7; 95% confidence interval [CI], 1.9-682; p = 0.018). Conclusions: Fibrinolysis resistance quantified by tPA-TEG within 48 hours of ICU admission is associated with an increased risk of pneumonia in patients in hemorrhagic shock and those at risk for VTE. Prospective validation of the tPA-TEG LY30 optimal cutoff for pneumonia and further investigation into whether endogenous FR is a cause of an altered immunity is warranted.
