BackgroundSepsis is one of the leading causes of mortality in hospitalized patients. Despite this fact, a reliable means of predicting sepsis onset remains elusive. Early and accurate sepsis onset predictions could allow more aggressive and targeted therapy while maintaining antimicrobial stewardship. Existing detection methods suffer from low performance and often require time-consuming laboratory test results.ObjectiveTo study and validate a sepsis prediction method, InSight, for the new Sepsis-3 definitions in retrospective data, make predictions using a minimal set of variables from within the electronic health record data, compare the performance of this approach with existing scoring systems, and investigate the effects of data sparsity on InSight performance.MethodsWe apply InSight, a machine learning classification system that uses multivariable combinations of easily obtained patient data (vitals, peripheral capillary oxygen saturation, Glasgow Coma Score, and age), to predict sepsis using the retrospective Multiparameter Intelligent Monitoring in Intensive Care (MIMIC)-III dataset, restricted to intensive care unit (ICU) patients aged 15 years or more. Following the Sepsis-3 definitions of the sepsis syndrome, we compare the classification performance of InSight versus quick sequential organ failure assessment (qSOFA), modified early warning score (MEWS), systemic inflammatory response syndrome (SIRS), simplified acute physiology score (SAPS) II, and sequential organ failure assessment (SOFA) to determine whether or not patients will become septic at a fixed period of time before onset. We also test the robustness of the InSight system to random deletion of individual input observations.ResultsIn a test dataset with 11.3% sepsis prevalence, InSight produced superior classification performance compared with the alternative scores as measured by area under the receiver operating characteristic curves (AUROC) and area under precision-recall curves (APR). In detection of sepsis onset, InSight attains AUROC = 0.880 (SD 0.006) at onset time and APR = 0.595 (SD 0.016), both of which are superior to the performance attained by SIRS (AUROC: 0.609; APR: 0.160), qSOFA (AUROC: 0.772; APR: 0.277), and MEWS (AUROC: 0.803; APR: 0.327) computed concurrently, as well as SAPS II (AUROC: 0.700; APR: 0.225) and SOFA (AUROC: 0.725; APR: 0.284) computed at admission (P<.001 for all comparisons). Similar results are observed for 1-4 hours preceding sepsis onset. In experiments where approximately 60% of input data are deleted at random, InSight attains an AUROC of 0.781 (SD 0.013) and APR of 0.401 (SD 0.015) at sepsis onset time. Even with 60% of data missing, InSight remains superior to the corresponding SIRS scores (AUROC and APR, P<.001), qSOFA scores (P=.0095; P<.001) and superior to SOFA and SAPS II computed at admission (AUROC and APR, P<.001), where all of these comparison scores (except InSight) are computed without data deletion.ConclusionsDespite using little more than vitals, InSight is an effective tool for predic...

IntroductionSeveral methods have been developed to electronically monitor patients for severe sepsis, but few provide predictive capabilities to enable early intervention; furthermore, no severe sepsis prediction systems have been previously validated in a randomised study. We tested the use of a machine learning-based severe sepsis prediction system for reductions in average length of stay and in-hospital mortality rate.MethodsWe conducted a randomised controlled clinical trial at two medical-surgical intensive care units at the University of California, San Francisco Medical Center, evaluating the primary outcome of average length of stay, and secondary outcome of in-hospital mortality rate from December 2016 to February 2017. Adult patients (18+) admitted to participating units were eligible for this factorial, open-label study. Enrolled patients were assigned to a trial arm by a random allocation sequence. In the control group, only the current severe sepsis detector was used; in the experimental group, the machine learning algorithm (MLA) was also used. On receiving an alert, the care team evaluated the patient and initiated the severe sepsis bundle, if appropriate. Although participants were randomly assigned to a trial arm, group assignments were automatically revealed for any patients who received MLA alerts.ResultsOutcomes from 75 patients in the control and 67 patients in the experimental group were analysed. Average length of stay decreased from 13.0 days in the control to 10.3 days in the experimental group (p=0.042). In-hospital mortality decreased by 12.4 percentage points when using the MLA (p=0.018), a relative reduction of 58.0%. No adverse events were reported during this trial.ConclusionThe MLA was associated with improved patient outcomes. This is the first randomised controlled trial of a sepsis surveillance system to demonstrate statistically significant differences in length of stay and in-hospital mortality.Trial registrationNCT03015454.

ObjectivesWe validate a machine learning-based sepsis-prediction algorithm (InSight) for the detection and prediction of three sepsis-related gold standards, using only six vital signs. We evaluate robustness to missing data, customisation to site-specific data using transfer learning and generalisability to new settings.DesignA machine-learning algorithm with gradient tree boosting. Features for prediction were created from combinations of six vital sign measurements and their changes over time.SettingA mixed-ward retrospective dataset from the University of California, San Francisco (UCSF) Medical Center (San Francisco, California, USA) as the primary source, an intensive care unit dataset from the Beth Israel Deaconess Medical Center (Boston, Massachusetts, USA) as a transfer-learning source and four additional institutions’ datasets to evaluate generalisability.Participants684 443 total encounters, with 90 353 encounters from June 2011 to March 2016 at UCSF.InterventionsNone.Primary and secondary outcome measuresArea under the receiver operating characteristic (AUROC) curve for detection and prediction of sepsis, severe sepsis and septic shock.ResultsFor detection of sepsis and severe sepsis, InSight achieves an AUROC curve of 0.92 (95% CI 0.90 to 0.93) and 0.87 (95% CI 0.86 to 0.88), respectively. Four hours before onset, InSight predicts septic shock with an AUROC of 0.96 (95% CI 0.94 to 0.98) and severe sepsis with an AUROC of 0.85 (95% CI 0.79 to 0.91).ConclusionsInSight outperforms existing sepsis scoring systems in identifying and predicting sepsis, severe sepsis and septic shock. This is the first sepsis screening system to exceed an AUROC of 0.90 using only vital sign inputs. InSight is robust to missing data, can be customised to novel hospital data using a small fraction of site data and retains strong discrimination across all institutions.

Acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) are common causes of morbidity and mortality in the intensive care unit. ALI/ARDS occurs as a result of systemic inflammation, usually triggered by a microorganism. Activation of leukocytes and release of proinflammatory mediators from multiple cellular sources result in both local and distant tissue injury. Tumor necrosis factor-alpha and interleukin-1 beta are the best characterized of the proinflammatory cytokines contributing to ALI/ARDS and subsequent fibrosis. The ultimate clinical course of ALI/ARDS often is determined by the ability of the injured lung to repopulate the alveolar epithelium with functional cells. Death may occur when fibrosis predominates the healing response, as it results in worsening lung compliance and oxygenation. The rodent bleomycin model of lung fibrosis allows the use of molecular tools to dissect the cellular and subcellular processes leading to fibrosis. The elements of this response may provide therapeutic targets for the prevention of this devastating complication of ALI/ARDS.

The association of O-antigen serotypes with type III secretory toxins was analyzed in 99 clinical isolates of Pseudomonas aeruginosa. Isolates secreting ExoU were frequently serotyped as O11, but none were serotype O1. Most of the isolates that were nontypeable for O antigen did not secrete type III secretory toxins.Lung infections caused by Pseudomonas aeruginosa are frequently associated with a high rate of mortality, particularly in immunocompromised patients (1,4). In addition to an increase in the prevalence of drug-resistant organisms, these poor outcomes of P. aeruginosa pneumonia appear to be due to the development of acute lung injury and septic shock (1,3,4,28). Among the various virulence factors of P. aeruginosa, lung injury and sepsis in infected hosts depend largely on the expression of exoenzyme S and its coregulated toxins secreted by the type III secretion system (TTSS) (10,15,29,30,34). The TTSS, which delivers toxins directly into the cytosol of cells, is utilized by most pathogenic gram-negative bacteria (11,14).The TTSS, including secretion, translocation, and regulation apparatuses, is encoded by the exoenzyme S regulon in P. aeruginosa (10, 33). However, the genes for the type III secretory toxins (TTS toxins) are distributed in various regions of the P. aeruginosa chromosomal DNA separate from the exoenzyme S regulon (8, 26). To date, four TTS toxins have been identified in P. aeruginosa (10,33). ExoS (exoenzyme S) and ExoT (exoenzyme T), having ADP-ribosyltransferase activities, diminish macrophage motility and phagocytosis (12) and are associated with mortality in animal models (2,(20)(21)(22). ExoY possesses adenylate cyclase activity and affects cell morphology (32). ExoU, a cytotoxin, contributes to epithelial cell toxicity, lung injury, and sepsis in infected animals, but the mechanism of its action remains unknown (8,16).While almost all strains of P. aeruginosa appear to possess a set of genes for the TTSS itself (7, 13), not all strains carry genes for all of the four TTS toxins. For instance, strain PAO1 has a negative genotype for exoU and strain PA103 has a negative genotype for exoS (8,9,26). In addition, some chronic isolates suppress the expression of the TTSS (24). It has been reported that patients infected with P. aeruginosa expressing the TTSS had a sixfold higher rate of mortality and an increased incidence of bacteremia than patients infected with P. aeruginosa not expressing the TTSS (24). A poor prognosis for patients with ventilator-associated pneumonia due to P. aeruginosa is associated with strains expressing the TTSS (13). Therefore, characterizing the phenotypes of TTS toxins in P. aeruginosa isolates could help in the identification of virulent strains.The lipopolysaccharide (LPS) O antigen has been used for the classification of P. aeruginosa isolates. There are 20 different International Antigenic Typing Scheme serotypes of P. aeruginosa based on differences of the B-band LPS. Our group previously obtained 108 clinical isolates of P. aeruginosa, and 99 of th...

Background Transfusion can cause severe acute lung injury, although most transfusions do not appear to induce complications. We tested the hypothesis that transfusion can cause mild pulmonary dysfunction that has not been noticed clinically and is not sufficiently severe to fit the definition of transfusion-related acute lung injury. Methods We studied 35 healthy normal volunteers who donated one unit of blood 4 weeks and another 3 weeks before two study days separated by one week. On study days two units of blood were withdrawn while maintaining isovolemia, followed by transfusion with either the volunteer’s autologous fresh red cells (RBCs) removed 2 hours earlier or their autologous stored RBCs (random order). The following week each volunteer was studied again, transfused with the RBCs of the other storage duration. The primary outcome variable was the change in alveolar to arterial difference in oxygen partial pressure (AaDO2) from before to 60 min after transfusion with fresh or older RBCs. Results Fresh RBCs and RBCs stored for 24.5 days equally (P=0.85) caused an increase of AaDO2 (fresh: 2.8 mmHg [95% CI: 0.8 - 4.8; (P=0.007)]; stored 3.0 mmHg [1.4 - 4.7; (P=0.0006)]). Concentrations of all measured cytokines, except for interleukin-10 (P=0.15), were less in stored leukoreduced (LR) than stored non-LR packed RBCs; however, vascular endothelial growth factor was the only measured in vivo cytokine that increased more after transfusion with LR than non-LR stored packed RBCs. Vascular endothelial growth factor was the only cytokine tested with in vivo concentrations that correlated with AaDO2. Conclusion RBC transfusion causes subtle pulmonary dysfunction, as evidenced by impaired gas exchange for oxygen, supporting our hypothesis that lung impairment after transfusion includes a wide spectrum of physiologic derangements and may not require an existing state of altered physiology. These data do not support the hypothesis that transfusion of RBCs stored for >21 days is more injurious than that of fresh RBCs.

The recent H1N1 epidemic has resulted in a large number of deaths, primarily from acute hypoxemic respiratory failure. We reviewed the current strategies to rescue patients with severe hypoxemia. Included in these strategies are high-frequency oscillatory ventilation, airway pressure release ventilation, inhaled vasodilators, and the use of extracorporeal life support. All of these strategies are targeted at improving oxygenation, but improved oxygenation alone has yet to be demonstrated to correlate with improved survival. The risks and benefits of these strategies, including cost-effectiveness data, are discussed.

In order to evaluate mortality predictions based on boosted trees, this retrospective study uses electronic medical record data from three academic health centers for inpatients 18 years or older with at least one observation of each vital sign. Predictions were made 12, 24, and 48 hours before death. Models fit to training data from each institution were evaluated using hold-out test data from the same institution, and from the other institutions. Gradient-boosted trees (GBT) were compared to regularized logistic regression (LR) predictions, support vector machine (SVM) predictions, quick Sepsis-Related Organ Failure Assessment (qSOFA), and Modified Early Warning Score (MEWS) using area under the receiver operating characteristic curve (AUROC). For training and testing GBT on data from the same institution, the average AUROCs were 0.96, 0.95, and 0.94 across institutional test sets for 12-, 24-, and 48-hour predictions, respectively. When trained and tested on data from different hospitals, GBT AUROCs achieved up to 0.98, 0.96, and 0.96, for 12-, 24-, and 48-hour predictions, respectively. Average AUROC for 48-hour predictions for LR, SVM, MEWS, and qSOFA were 0.85, 0.79, 0.86 and 0.82, respectively. GBT predictions may help identify patients who would benefit from increased clinical care.