Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis

Machine learning in infection management using routine electronic health records: tools, techniques, and reporting of future technologies

BACKGROUND

Machine learning (ML) is increasingly being used in many areas of health care. Its use in infection management is catching up as identified in a recent review in this journal. We present here a complementary review to this work.

OBJECTIVES

To support clinicians and researchers in navigating through the methodological aspects of ML approaches in the field of infection management.

SOURCES

A Medline search was performed with the keywords artificial intelligence, machine learning, infection∗, and infectious disease∗ for the years 2014-2019. Studies using routinely available electronic hospital record data from an inpatient setting with a focus on bacterial and fungal infections were included.

CONTENT

Fifty-two studies were included and divided into six groups based on their focus. These studies covered detection/prediction of sepsis (n = 19), hospital-acquired infections (n = 11), surgical site infections and other postoperative infections (n = 11), microbiological test results (n = 4), infections in general (n = 2), musculoskeletal infections (n = 2), and other topics (urinary tract infections, deep fungal infections, antimicrobial prescriptions; n = 1 each). In total, 35 different ML techniques were used. Logistic regression was applied in 18 studies followed by random forest, support vector machines, and artificial neural networks in 18, 12, and seven studies, respectively. Overall, the studies were very heterogeneous in their approach and their reporting. Detailed information on data handling and software code was often missing. Validation on new datasets and/or in other institutions was rarely done. Clinical studies on the impact of ML in infection management were lacking.

IMPLICATIONS

Promising approaches for ML use in infectious diseases were identified. But building trust in these new technologies will require improved reporting. Explainability and interpretability of the models used were rarely addressed and should be further explored. Independent model validation and clinical studies evaluating the added value of ML approaches are needed.

Medicina de precisión en sepsis: utilidad de los biomarcadores en pacientes biomarcadores en pacientes críticamente enfermos

Durante años la evolución del cuidado intensivo ha intentado ofrecer una atención basada en protocolos y paquetes de manejo agrupados por patologías y cuadro sindromáticos. Aunque se logró disminuir la mortalidad en diferentes patologías (sepsis y síndromes coronario agudo y de distrés respiratorio agudo), no se han resuelto por completo los problemas clínicos, en especial el diagnóstico y el manejo. Una nueva opción ha surgido en el horizonte denominada “medicina de precisión”, entendida como estrategia de prevención y tratamiento que tiene en cuenta la variabilidad individual. La sepsis es un síndrome con múltiples aristas en cuanto al fenotipo y genotipo, cuyo diagnóstico temprano es relevante para los desenlaces clínicos. Hasta el momento el enfoque principal ha sido la identificación de un germen etiológico para diferenciarla del síndrome de respuesta inflamatoria sistémica (SIRS). En los últimos años el paradigma en enfermedades infecciosas ha cambiado debido a estudios que demuestran como la respuesta inmunitaria del paciente séptico tiene un papel clave en el desarrollo de la enfermedad, con implicaciones en el diagnóstico, pronóstico y tratamiento, que podrían ayudar a cambiar el abordaje en los próximos años gracias a una estrategia basada en medicina de precisión. Hoy los aislamientos microbiológicos y los cultivos siguen siendo el estándar de referencia con varias desventajas como el tiempo para obtener resultados, sobre todo en infecciones por gérmenes resistentes u hongos, que pueden retrasar el inicio de la terapia antimicrobiana. Como alternativa se ha planteado el uso de biomarcadores en sepsis que siendo productos de la respuesta inflamatoria del individuo ante la infección, son útiles para el diagnóstico y pronóstico primordialmente en los críticamente enfermos. Decidimos realizar esta revisión narrativa acerca de la utilidad de los biomarcadores en pacientes con sepsis críticamente enfermos, para enfocarlos en un modelo de medicina personalizada.

Phenotyping Multiple Organ Dysfunction Syndrome Using Temporal Trends in Critically Ill Children

Multiple organ dysfunction syndrome (MODS) is one of the most common causes of death in critically ill children. However, despite decades of clinical trials, there are no comprehensive approaches to the management of MODS or effective targeted therapies that have consistently improved outcomes. Better understanding the heterogeneity of MODS and characterizing subgroups of MODS patients could improve our understanding of the syndrome and help us develop new management strategies. We analyzed a cohort of 5,297 children with MODS from two children's hospitals and used subgraph-augmented non-negative matrix factorization (SANMF) to identify unique temporal patterns in organ dysfunction across four novel subgroups. We demonstrate that these subgroups are composed of patients with distinct clinical characteristics and are independently predictive of clinical outcomes. Our work suggests that these subgroups represent four relevant phenotypes of pediatric MODS that could be used to identify novel management strategies.

Machine Learning for Pulmonary and Critical Care Medicine: A Narrative Review

Machine learning (ML) is a discipline of computer science in which statistical methods are applied to data in order to classify, predict, or optimize, based on previously observed data. Pulmonary and critical care medicine have seen a surge in the application of this methodology, potentially delivering improvements in our ability to diagnose, treat, and better understand a multitude of disease states. Here we review the literature and provide a detailed overview of the recent advances in ML as applied to these areas of medicine. In addition, we discuss both the significant benefits of this work as well as the challenges in the implementation and acceptance of this non-traditional methodology for clinical purposes.

Time to Recognition of Sepsis in the Emergency Department Using Electronic Health Record Data: A Comparative Analysis of Systemic Inflammatory Response Syndrome, Sequential Organ Failure Assessment, and Quick Sequential Organ Failure Assessment

OBJECTIVES

Early identification of sepsis is critical to improving patient outcomes. Impact of the new sepsis definition (Sepsis-3) on timing of recognition in the emergency department has not been evaluated. Our study objective was to compare time to meeting systemic inflammatory response syndrome (Sepsis-2) criteria, Sequential Organ Failure Assessment (Sepsis-3) criteria, and quick Sequential Organ Failure Assessment criteria using electronic health record data.

DESIGN

Retrospective, observational study.

SETTING

The emergency department at the University of California, San Francisco.

PATIENTS

Emergency department encounters between June 2012 and December 2016 for patients greater than or equal to 18 years old with blood cultures ordered, IV antibiotic receipt, and identification with sepsis via systemic inflammatory response syndrome or Sequential Organ Failure Assessment within 72 hours of emergency department presentation.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

We analyzed timestamped electronic health record data from 16,612 encounters identified as sepsis by greater than or equal to 2 systemic inflammatory response syndrome criteria or a Sequential Organ Failure Assessment score greater than or equal to 2. The primary outcome was time from emergency department presentation to meeting greater than or equal to 2 systemic inflammatory response syndrome criteria, Sequential Organ Failure Assessment greater than or equal to 2, and/or greater than or equal to 2 quick Sequential Organ Failure Assessment criteria. There were 9,087 patients (54.7%) that met systemic inflammatory response syndrome-first a median of 26 minutes post-emergency department presentation (interquartile range, 0-109 min), with 83.1% meeting Sequential Organ Failure Assessment criteria a median of 118 minutes later (interquartile range, 44-401 min). There were 7,037 patients (42.3%) that met Sequential Organ Failure Assessment-first, a median of 113 minutes post-emergency department presentation (interquartile range, 60-251 min). Quick Sequential Organ Failure Assessment was met in 46.4% of patients a median of 351 minutes post-emergency department presentation (interquartile range, 67-1,165 min). Adjusted odds of in-hospital mortality were 39% greater in patients who met systemic inflammatory response syndrome-first compared with those who met Sequential Organ Failure Assessment-first (odds ratio, 1.39; 95% CI, 1.20-1.61).

CONCLUSIONS

Systemic inflammatory response syndrome and Sequential Organ Failure Assessment initially identified distinct populations. Using systemic inflammatory response syndrome resulted in earlier electronic health record sepsis identification in greater than 50% of patients. Using Sequential Organ Failure Assessment alone may delay identification. Using systemic inflammatory response syndrome alone may lead to missed sepsis presenting as acute organ dysfunction. Thus, a combination of inflammatory (systemic inflammatory response syndrome) and organ dysfunction (Sequential Organ Failure Assessment) criteria may enhance timely electronic health record-based sepsis identification.

The role of steroids in severe CAP

Community-acquired pneumonia (CAP) is a leading cause of morbidity and mortality despite adequate antibiotic therapy. It is the single most common cause of infection-related mortality in the United States. An exaggerated host inflammatory response can potentially be harmful to both the lung and host, and has been associated with treatment failure and mortality. Modulation of inflammatory response may, therefore, be theoretically beneficial. The anti-inflammatory and immunosuppressive effects of steroids seem an attractive therapeutic option in severe CAP patients. Available datapoint to overall shorter time to clinical stability and decreased length-of-stay in CAP patients, with a potential mortality benefit in severe CAP. The level of evidence is, however, low to moderate regarding mortality due to high heterogeneity and insufficient power of data. Furthermore, steroids were deleterious in influenza pneumonia and in patients with pneumococcal pneumonia data suggest a lack of efficacy and potential harm. Both European and American guidelines recommend not using corticosteroids in CAP. Patients who might benefit and those that can be harmed from steroids remain to be clearly identified, as does the ideal steroid for CAP patients, based on pharmacokinetic and pharmacodynamic properties. It is essential for future studies to avoid the same methodological bias present in the available data so that high-quality evidence on the true role of steroids in CAP can be provided.

Accuracy and Effects of Clinical Decision Support Systems Integrated With BMJ Best Practice-Aided Diagnosis: Interrupted Time Series Study

BACKGROUND

Clinical decision support systems (CDSS) are an integral component of health information technologies and can assist disease interpretation, diagnosis, treatment, and prognosis. However, the utility of CDSS in the clinic remains controversial.

OBJECTIVE

The aim is to assess the effects of CDSS integrated with British Medical Journal (BMJ) Best Practice-aided diagnosis in real-world research.

METHODS

This was a retrospective, longitudinal observational study using routinely collected clinical diagnosis data from electronic medical records. A total of 34,113 hospitalized patient records were successively selected from December 2016 to February 2019 in six clinical departments. The diagnostic accuracy of the CDSS was verified before its implementation. A self-controlled comparison was then applied to detect the effects of CDSS implementation. Multivariable logistic regression and single-group interrupted time series analysis were used to explore the effects of CDSS. The sensitivity analysis was conducted using the subgroup data from January 2018 to February 2019.

RESULTS

The total accuracy rates of the recommended diagnosis from CDSS were 75.46% in the first-rank diagnosis, 83.94% in the top-2 diagnosis, and 87.53% in the top-3 diagnosis in the data before CDSS implementation. Higher consistency was observed between admission and discharge diagnoses, shorter confirmed diagnosis times, and shorter hospitalization days after the CDSS implementation (all P<.001). Multivariable logistic regression analysis showed that the consistency rates after CDSS implementation (OR 1.078, 95% CI 1.015-1.144) and the proportion of hospitalization time 7 days or less (OR 1.688, 95% CI 1.592-1.789) both increased. The interrupted time series analysis showed that the consistency rates significantly increased by 6.722% (95% CI 2.433%-11.012%, P=.002) after CDSS implementation. The proportion of hospitalization time 7 days or less significantly increased by 7.837% (95% CI 1.798%-13.876%, P=.01). Similar results were obtained in the subgroup analysis.

CONCLUSIONS

The CDSS integrated with BMJ Best Practice improved the accuracy of clinicians' diagnoses. Shorter confirmed diagnosis times and hospitalization days were also found to be associated with CDSS implementation in retrospective real-world studies. These findings highlight the utility of artificial intelligence-based CDSS to improve diagnosis efficiency, but these results require confirmation in future randomized controlled trials.

Anticoagulant therapy for septic coagulopathy and disseminated intravascular coagulation: where do KyberSept and SCARLET leave us?

The use of antithrombin and thrombomodulin to restore impaired anticoagulant pathways in septic coagulopathy has been shown to significantly increase the resolution rate of disseminated intravascular coagulation. In KyberSept and SCARLET, two large, international, randomized controlled trials in patients with sepsis, these anticoagulants have not shown significantly reduced mortality. The aim of this assessment was to compare the heterogeneity in responses to treatment in the two trials according to different patient phenotypes. Results of the KyberSept and SCARLET trials reported in original and post-hoc publications were analyzed and directly compared for treatment effects in various patient subgroups. In both KyberSept and SCARLET, the septic coagulopathy phenotype that benefited most from endogenous anticoagulant supplementation showed markers of excessive activation of coagulation. Interaction between concomitant thromboprophylactic heparin and the endogenous anticoagulants abrogated the efficacy of both antithrombin and thrombomodulin. In both trials, higher disease severity was associated with better treatment outcome. In conclusion, in two landmark studies of endogenous anticoagulants in patients with sepsis, similar findings of beneficial effects in the coagulopathy phenotype and interactions with heparin comedication and disease severity support the potential roles that thrombomodulin and antithrombin might play in treating septic coagulopathy and disseminated intravascular coagulation. Further prospective validation is warranted. Future trial designs to definitively establish the therapeutic relevance of antithrombin and thrombomodulin in septic coagulopathy should focus on involvement of patients characterized by coagulopathy and disease severity as well as interactions between endogenous anticoagulants and exogenous heparin.

The damage response framework and infection prevention: From concept to bedside

Hospital-acquired infections remain a common cause of morbidity and mortality despite advances in infection prevention through use of bundles, environmental cleaning, antimicrobial stewardship, and other best practices. Current prevention strategies and further hospital-acquired infection reduction are limited by lack of recognition of the role that host-microbe interactions play in susceptibility and by the inability to analyze multiple risk factors in real time to accurately predict the likelihood of a hospital-acquired infection before it occurs and to inform medical decision making. Herein, we examine the value of incorporating the damage-response framework and host attributes that determine susceptibility to infectious diseases known by the acronym MISTEACHING (ie, microbiome, immunity, sex, temperature, environment, age, chance, history, inoculum, nutrition, genetics) into infection prevention strategies using machine learning to drive decision support and patient-specific interventions.

GDF3 Protects Mice against Sepsis-Induced Cardiac Dysfunction and Mortality by Suppression of Macrophage Pro-Inflammatory Phenotype

Macrophages are critical for regulation of inflammatory response during endotoxemia and septic shock. However, the mediators underlying their regulatory function remain obscure. Growth differentiation factor 3 (GDF3), a member of transforming growth factor beta (TGF-β) superfamily, has been implicated in inflammatory response. Nonetheless, the role of GDF3 in macrophage-regulated endotoxemia/sepsis is unknown. Here, we show that serum GDF3 levels in septic patients are elevated and strongly correlate with severity of sepsis and 28-day mortality. Interestingly, macrophages treated with recombinant GDF3 protein (rGDF3) exhibit greatly reduced production of pro-inflammatory cytokines, comparing to controls upon endotoxin challenge. Moreover, acute administration of rGDF3 to endotoxin-treated mice suppresses macrophage infiltration to the heart, attenuates systemic and cardiac inflammation with less pro-inflammatory macrophages (M1) and more anti-inflammatory macrophages (M2), as well as prolongs mouse survival. Mechanistically, GDF3 is able to activate Smad2/Smad3 phosphorylation, and consequently inhibits the expression of nod-like receptor protein-3 (NLRP3) in macrophages. Accordingly, blockade of Smad2/Smad3 phosphorylation with SB431542 significantly offsets rGDF3-mediated anti-inflammatory effects. Taken together, this study uncovers that GDF3, as a novel sepsis-associated factor, may have a dual role in the pathophysiology of sepsis. Acute administration of rGDF3 into endotoxic shock mice could increase survival outcome and improve cardiac function through anti-inflammatory response by suppression of M1 macrophage phenotype. However, constitutive high levels of GDF3 in human sepsis patients are associated with lethality, suggesting that GDF3 may promote macrophage polarization toward M2 phenotype which could lead to immunosuppression.

Organ Dysfunction in Sepsis: An Ominous Trajectory From Infection To Death

Sepsis is a highly complex and lethal syndrome with highly heterogeneous clinical manifestations that makes it difficult to detect and treat. It is also one of the major and most urgent global public health challenges. More than 30 million people are diagnosed with sepsis each year, with 5 million attributable deaths and long-term sequalae among survivors. The current international consensus defines sepsis as a life-threatening organ dysfunction caused by a dysregulated host response to an infection. Over the past decades substantial research has increased the understanding of its pathophysiology. The immune response induces a severe macro and microcirculatory dysfunction that leads to a profound global hypoperfusion, injuring multiple organs. Consequently, patients with sepsis might present dysfunction of virtually any system, regardless of the site of infection. The organs more frequently affected are kidneys, liver, lungs, heart, central nervous system, and hematologic system. This multiple organ failure is the hallmark of sepsis and determines patients' course from infection to recovery or death. There are tools to assess the severity of the disease that can also help to guide treatment, like the Sequential Organ Failure Assessment (SOFA) score. However, sepsis disease process is vastly heterogeneous, which could explain why interventions targeted to directly intervene its mechanisms have shown unsuccessful results and predicting outcomes with accuracy is still elusive. Thus, it is required to implement strong public health strategies and leverage novel technologies in research to improve outcomes and mitigate the burden of sepsis and septic shock worldwide.

Data-driven multivariate population subgrouping via lipoprotein phenotypes versus apolipoprotein B in the risk assessment of coronary heart disease

BACKGROUND AND AIMS

Population subgrouping has been suggested as means to improve coronary heart disease (CHD) risk assessment. We explored here how unsupervised data-driven metabolic subgrouping, based on comprehensive lipoprotein subclass data, would work in large-scale population cohorts.

METHODS

We applied a self-organizing map (SOM) artificial intelligence methodology to define subgroups based on detailed lipoprotein profiles in a population-based cohort (n = 5789) and utilised the trained SOM in an independent cohort (n = 7607). We identified four SOM-based subgroups of individuals with distinct lipoprotein profiles and CHD risk and compared those to univariate subgrouping by apolipoprotein B quartiles.

RESULTS

The SOM-based subgroup with highest concentrations for non-HDL measures had the highest, and the subgroup with lowest concentrations, the lowest risk for CHD. However, apolipoprotein B quartiles produced better resolution of risk than the SOM-based subgroups and also striking dose-response behaviour.

CONCLUSIONS

These results suggest that the majority of lipoprotein-mediated CHD risk is explained by apolipoprotein B-containing lipoprotein particles. Therefore, even advanced multivariate subgrouping, with comprehensive data on lipoprotein metabolism, may not advance CHD risk assessment.

Ethical considerations about artificial intelligence for prognostication in intensive care

Background

Prognosticating the course of diseases to inform decision-making is a key component of intensive care medicine. For several applications in medicine, new methods from the field of artificial intelligence (AI) and machine learning have already outperformed conventional prediction models. Due to their technical characteristics, these methods will present new ethical challenges to the intensivist.

Results

In addition to the standards of data stewardship in medicine, the selection of datasets and algorithms to create AI prognostication models must involve extensive scrutiny to avoid biases and, consequently, injustice against individuals or groups of patients. Assessment of these models for compliance with the ethical principles of beneficence and non-maleficence should also include quantification of predictive uncertainty. Respect for patients’ autonomy during decision-making requires transparency of the data processing by AI models to explain the predictions derived from these models. Moreover, a system of continuous oversight can help to maintain public trust in this technology. Based on these considerations as well as recent guidelines, we propose a pathway to an ethical implementation of AI-based prognostication. It includes a checklist for new AI models that deals with medical and technical topics as well as patient- and system-centered issues.

Conclusion

AI models for prognostication will become valuable tools in intensive care. However, they require technical refinement and a careful implementation according to the standards of medical ethics.

Host-Response Subphenotypes Offer Prognostic Enrichment in Patients With or at Risk for Acute Respiratory Distress Syndrome

OBJECTIVES

Classification of patients with acute respiratory distress syndrome into hyper- and hypoinflammatory subphenotypes using plasma biomarkers may facilitate more effective targeted therapy. We examined whether established subphenotypes are present not only in patients with acute respiratory distress syndrome but also in patients at risk for acute respiratory distress syndrome (ARFA) and then assessed the prognostic information of baseline subphenotyping on the evolution of host-response biomarkers and clinical outcomes.

DESIGN

Prospective, observational cohort study.

SETTING

Medical ICU at a tertiary academic medical center.

PATIENTS

Mechanically ventilated patients with acute respiratory distress syndrome or ARFA.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

We performed longitudinal measurements of 10 plasma biomarkers of host injury and inflammation. We applied unsupervised latent class analysis methods utilizing baseline clinical and biomarker variables and demonstrated that two-class models (hyper- vs hypoinflammatory subphenotypes) offered improved fit compared with one-class models in both patients with acute respiratory distress syndrome and ARFA. Baseline assignment to the hyperinflammatory subphenotype (39/104 [38%] acute respiratory distress syndrome and 30/108 [28%] ARFA patients) was associated with higher severity of illness by Sequential Organ Failure Assessment scores and incidence of acute kidney injury in patients with acute respiratory distress syndrome, as well as higher 30-day mortality and longer duration of mechanical ventilation in ARFA patients (p < 0.0001). Hyperinflammatory patients exhibited persistent elevation of biomarkers of innate immunity for up to 2 weeks postintubation.

CONCLUSIONS

Our results suggest that two distinct subphenotypes are present not only in patients with established acute respiratory distress syndrome but also in patients at risk for its development. Hyperinflammatory classification at baseline is associated with higher severity of illness, worse clinical outcomes, and trajectories of persistently elevated biomarkers of host injury and inflammation during acute critical illness compared with hypoinflammatory patients. Our findings provide strong rationale for examining treatment effect modifications by subphenotypes in randomized clinical trials to inform precision therapeutic approaches in critical care.

Murine sepsis phenotypes and differential treatment effects in a randomized trial of prompt antibiotics and fluids

BACKGROUND

Clinical and biologic phenotypes of sepsis are proposed in human studies, yet it is unknown whether prognostic or drug response phenotypes are present in animal models of sepsis. Using a biotelemetry-enhanced, murine cecal ligation and puncture (CLP) model, we determined phenotypes of polymicrobial sepsis prior to physiologic deterioration, and the association between phenotypes and outcome in a randomized trial of prompt or delayed antibiotics and fluids.

METHODS

We performed a secondary analysis of male C57BL/6J mice in two observational cohorts and two randomized, laboratory animal experimental trials. In cohort 1, mice (n = 118) underwent biotelemetry-enhanced CLP, and we applied latent class mixed models to determine optimal number of phenotypes using clinical data collected between injury and physiologic deterioration. In cohort 2 (N = 73 mice), inflammatory cytokines measured at 24 h after deterioration were explored by phenotype. In a subset of 46 mice enrolled in two trials from cohort 1, we tested the association of phenotypes with the response to immediate (0 h) vs. delayed (2 to 4 h) antibiotics or fluids initiated after physiologic deterioration.

RESULTS

Latent class mixture modeling derived a two-class model in cohort 1. Class 2 (N = 97) demonstrated a shorter time to deterioration (mean SD 7.3 (0.9) vs. 9.7 (3.2) h, p < 0.001) and lower heart rate at 7 h after injury (mean (SD) 564 (55) vs. 626 (35) beats per minute, p < 0.001). Overall mortality was similar between phenotypes (p = 0.75). In cohort 2 used for biomarker measurement, class 2 mice had greater plasma concentrations of IL6 and IL10 at 24 h after CLP (p = 0.05). In pilot randomized trials, the effects of sepsis treatment (immediate vs. delayed antibiotics) differed by phenotype (p = 0.03), with immediate treatment associated with greater survival in class 2 mice only. Similar differential treatment effect by class was observed in the trial of immediate vs. delayed fluids (p = 0.02).

CONCLUSIONS

We identified two sepsis phenotypes in a murine cecal ligation and puncture model, one of which is characterized by faster deterioration and more severe inflammation. Response to treatment in a randomized trial of immediate versus delayed antibiotics and fluids differed on the basis of phenotype.

Flow Cytometric Analysis of Monocytes Polarization and Reprogramming From Inflammatory to Immunosuppressive Phase During Sepsis

Sepsis outcome is determined by a balance between inflammation and immune suppression. We aimed to evaluate monocytes polarization and reprogramming during these processes. We analyzed 93 patients with procalcitonin level >0.5 ng/mL (hPCT) and suspected/confirmed sepsis, and 84 controls by analysis of CD14, CD16 and HLA-DR expression on blood monocytes using fluorescent labeled monoclonal antibodies and BD FACS CANTO II. Complete blood cell count, procalcitonin and other biochemical markers were evaluated. Intermediate monocytes CD14⁺⁺CD16⁺ increased in hPCT patients (including both positive and negative culture) compared to controls (13.6% ± 0.8 vs 6.2% ± 0.3, p<0.001), while classical monocytes CD14⁺⁺CD16^-were significantly reduced (72.5% ± 1.6 vs 82.6% ± 0.7, p<0.001). Among hPCT patients having positive microbial culture, the percentage of intermediate monocytes was significantly higher in septic compared with non-septic/localized-infection patients (17.4% vs 11.5%; p<0.05) whilst the percentage of classical monocytes was lower (68.0% vs 74.5%). Three-four days following the diagnosis of sepsis, HLA-DR expression on monocyte (mHLA-DR) was lower (94.3%) compared to controls (99.4%) (p<0.05). Septic patients with the worst clinical conditions showed higher incidence of secondary infections, longtime hospitalization and lower HLA-DR+ monocytes compared to septic patients with better clinical outcome (88.4% vs 98.6%, p=0.05). The dynamic nature of sepsis correlates with monocytes functional polarization and reprogramming from a pro-inflammatory CD14⁺⁺CD16⁺ phenotype in non-septic hPCT patients to a decrease of HLA-DR surface expression in hPCT patients with confirmed sepsis, making HLA-DR reduction a marker of immune-paralysis and sepsis outcome. Analysis of monocytes plasticity opens to new mechanisms responsible for pro/anti-inflammatory responses during sepsis, and new immunotherapies.

A Systematic Summary of Systematic Reviews on Anticoagulant Therapy in Sepsis

Many systematic reviews have been published regarding anticoagulant therapy in sepsis, among which there is substantial heterogeneity. This study aimed to provide an overview of existing systematic reviews of randomized controlled trials by using a comprehensive search method. We searched MEDLINE, EMBASE, and Cochrane Database of Systematic Reviews. Of 895 records screened, 19 systematic reviews were included. The target agent was as follows: antithrombin (n = 4), recombinant thrombomodulin (n = 3), heparin (n = 3), recombinant activated protein C (n = 8), and all anticoagulants (n = 1). Antithrombin did not improve mortality in critically ill patients but indicated a beneficial effect in sepsis-induced disseminated intravascular coagulation (DIC), although the certainty of evidence was judged as low. Recombinant thrombomodulin was associated with a trend in reduced mortality in sepsis with coagulopathy with no increased risk of bleeding, although the difference was not statistically significant and the required information size for any declarative judgement insufficient. Although three systematic reviews showed potential survival benefits of unfractionated heparin and low-molecular-weight heparin in patients with sepsis, trials with low risk of bias were lacking, and the overall impact remains unclear. None of the meta-analyses of recombinant activated protein C showed beneficial effects in sepsis. In summary, a beneficial effect was not observed in overall sepsis in poorly characterized patient groups but was observed in sepsis-induced DIC or sepsis with coagulopathy in more specific patient groups. This umbrella review of anticoagulant therapy suggests that characteristics of the target populations resulted in heterogeneity among the systematic reviews.

Sepsis Induces Prolonged Epigenetic Modifications in Bone Marrow and Peripheral Macrophages Impairing Inflammation and Wound Healing

OBJECTIVE

Sepsis represents an acute life-threatening disorder resulting from a dysregulated host response. For patients who survive sepsis, there remains long-term consequences, including impaired inflammation, as a result of profound immunosuppression. The mechanisms involved in this long-lasting deficient immune response are poorly defined. Approach and Results: Sepsis was induced using the murine model of cecal ligation and puncture. Following a full recovery period from sepsis physiology, mice were subjected to our wound healing model and wound macrophages (CD11b+, CD3-, CD19-, Ly6G-) were sorted. Post-sepsis mice demonstrated impaired wound healing and decreased reepithelization in comparison to controls. Further, post-sepsis bone marrow-derived macrophages and wound macrophages exhibited decreased expression of inflammatory cytokines vital for wound repair (IL [interleukin]-1β, IL-12, and IL-23). To evaluate if decreased inflammatory gene expression was secondary to epigenetic modification, we conducted chromatin immunoprecipitation on post-sepsis bone marrow-derived macrophages and wound macrophages. This demonstrated decreased expression of Mll1, an epigenetic enzyme, and impaired histone 3 lysine 4 trimethylation (activation mark) at NFκB (nuclear factor kappa-light-chain-enhancer of activated B cells)-binding sites on inflammatory gene promoters in bone marrow-derived macrophages and wound macrophages from postcecal ligation and puncture mice. Bone marrow transplantation studies demonstrated epigenetic modifications initiate in bone marrow progenitor/stem cells following sepsis resulting in lasting impairment in peripheral macrophage function. Importantly, human peripheral blood leukocytes from post-septic patients demonstrate a significant reduction in MLL1 compared with nonseptic controls.

CONCLUSIONS

These data demonstrate that severe sepsis induces stable mixed-lineage leukemia 1-mediated epigenetic modifications in the bone marrow, which are passed to peripheral macrophages resulting in impaired macrophage function and deficient wound healing persisting long after sepsis recovery.

Biomarkers of Infection and Sepsis

The role of biomarkers for detection of sepsis has come a long way. Molecular biomarkers are taking front stage at present, but machine learning and other computational measures using bigdata sets are promising. Clinical research in sepsis is hampered by lack of specificity of the diagnosis; sepsis is a syndrome with no uniformly agreed definition. This lack of diagnostic precision means there is no gold standard for this diagnosis. The final conclusion is expert opinion, which is not bad but not perfect. Perhaps machine learning will displace expert opinion as the final and most accurate definition for sepsis.

Clinical applications of artificial intelligence in sepsis: A narrative review

Many studies have been published on a variety of clinical applications of artificial intelligence (AI) for sepsis, while there is no overview of the literature. The aim of this review is to give an overview of the literature and thereby identify knowledge gaps and prioritize areas with high priority for further research. A literature search was conducted in PubMed from inception to February 2019. Search terms related to AI were combined with terms regarding sepsis. Articles were included when they reported an area under the receiver operator characteristics curve (AUROC) as outcome measure. Fifteen articles on diagnosis of sepsis with AI models were included. The best performing model reached an AUROC of 0.97. There were also seven articles on prognosis, predicting mortality over time with an AUROC of up to 0.895. Finally, there were three articles on assistance of treatment of sepsis, where the use of AI was associated with the lowest mortality rates. Of the articles, twenty-two were judged to be at high risk of bias or had major concerns regarding applicability. This was mostly because predictor variables in these models, such as blood pressure, were also part of the definition of sepsis, which led to overestimation of the performance. We conclude that AI models have great potential for improving early identification of patients who may benefit from administration of antibiotics. Current AI prediction models to diagnose sepsis are at major risks of bias when the diagnosis criteria are part of the predictor variables in the model. Furthermore, generalizability of these models is poor due to overfitting and a lack of standardized protocols for the construction and validation of the models. Until these problems have been resolved, a large gap remains between the creation of an AI algorithm and its implementation in clinical practice.

Prognostic and predictive enrichment in sepsis

Sepsis is a heterogeneous disease state that is both common and consequential in critically ill patients. Unfortunately, the heterogeneity of sepsis at the individual patient level has hindered advances in the field beyond the current therapeutic standards, which consist of supportive care and antibiotics. This complexity has prompted attempts to develop a precision medicine approach, with research aimed towards stratifying patients into more homogeneous cohorts with shared biological features, potentially facilitating the identification of new therapies. Several investigators have successfully utilized leukocyte-derived mRNA and discovery-based approaches to subgroup patients on the basis of biological similarities defined by transcriptomic signatures. A critical next step is to develop a consensus sepsis subclassification system, which includes transcriptomic signatures as well as other biological and clinical data. This goal will require collaboration among various investigative groups, and validation in both existing data sets and prospective studies. Such studies are required to bring precision medicine to the bedside of critically ill patients with sepsis.

Biomarkers of Infection: Are They Useful in the ICU?

Biomarkers are increasingly used in patients with serious infections in the critical care setting to complement clinical judgment and interpretation of other diagnostic and prognostic tests. The main purposes of such blood markers are (1) to improve infection diagnosis (i.e., differentiation between bacterial vs. viral vs. fungal vs. noninfectious), (2) to help in the early risk stratification and thus provide prognostic information regarding the risk for mortality and other adverse outcomes, and (3) to optimize antibiotic tailoring to individual needs of patients ("antibiotic stewardship").Especially in critically ill patients, in whom sepsis is a major cause of morbidity and mortality, rapid diagnosis is desirable to start timely and specific treatment.Besides some biomarkers, such as procalcitonin, which is well established and has shown positive effects in regard to utilization of antimicrobials and clinical outcomes, there is a growing number of novel markers from different pathophysiological pathways, where the final proof of an added value to clinical judgment and ultimately clinical benefit to patients is still lacking.Without a doubt, the addition of blood biomarkers to clinical medicine has had a strong impact on the way we care for patients today. Recent trials show that as an adjunct to other clinical and laboratory parameters these markers provide important information about risks for bacterial infection and resolution of infection. Moreover, biomarkers can help to optimize management of patients with serious illness in the intensive care unit, thereby offering more individualized treatment courses with overall improvements in clinical outcomes.

Trends in sepsis mortality over time in randomised sepsis trials: a systematic literature review and meta-analysis of mortality in the control arm, 2002-2016

Background

Epidemiologic data have shown an increasing incidence and declining mortality rate in sepsis. However, confounding effects due to differences in disease classification might have contributed to these trends.

To assess if a declining mortality over time could be supported by data derived from high-quality prospective studies, we performed a meta-analysis using data from randomised controlled trials (RCTs) on sepsis. The primary aim was to assess whether the mortality in sepsis trials has changed over time. The secondary aim was to investigate how many of the included trials could show efficacy of the studied intervention regarding 28-day mortality.

Methods

We searched PubMed for RCTs enrolling patients with severe sepsis and septic shock, published between 2002 and 2016. The included trials were assessed for quality and sorted by date of first inclusion. A meta-analysis was performed to synthesise data from the individual sepsis trials.

Results

Of 418 eligible articles, 44 RCTs on sepsis were included in the analysis, enrolling 13,315 patients in the usual care arm between 1991 and 2013. In this time period, mortality decreased by 0.42% annually (p = 0.04) to give a total decline of 9.24%. In subgroup analyses with adjustments for APACHE II, SAPS II and SOFA scores, the observed time trend was not significant (p = 0.45, 0.23 and 0.98 respectively). Only four of the included trials showed any efficacy with regard to mortality.

Conclusions

Data from RCTs show a declining trend in 28-day mortality in severe sepsis and septic shock patients during the years from 1991 to 2013. However, when controlling for severity at study inclusion, there was no significant change in mortality over time. The number of trials presenting new treatment options was low.

Trial registration

PROSPERO CRD42018091100. Registered 27 August 2018.

Electronic supplementary material

The online version of this article (10.1186/s13054-019-2528-0) contains supplementary material, which is available to authorized users.