Answers are grounded in this paper and cite their sources.
Sepsis and septic shock are true medical emergencies responsible for >1.5 million hospitalizations and 250000 deaths in the US each year (1). Emergency care practitioners recognize that reduction of morbidity and mortality from sepsis is directly dependent on early identification and rapid initiation of therapy, which includes fluid resuscitation and antimicrobial therapy. In the US the initial evaluation for most septic patients, including those admitted to the intensive care unit, occurs in emergency department (ED)3 settings (2). As such, emergency physicians work on the front line of sepsis care, frequently making critical diagnostic and therapeutic decisions that profoundly affect overall care trajectory and patient outcomes. The continuum of sepsis care delivered in the ED can be divided into 2 domains: (a) detection and diagnosis and (b) treatment initiation and ongoing evaluation coupled with resuscitation. Decision-making within each of these domains is complicated by the uniquely challenging practice environment of the ED, where a wide variety of acute and undifferentiated diseases are managed, typically with limited information and always under intensive time pressure. Despite extensive research efforts over the past decade, there has been relatively little increase in the amount of objective data available to emergency physicians during the critical time window of initial encounter to guide key decision-making for patients with suspected or confirmed sepsis. The standardized approach still relies on syndromic-based diagnosis and the nearly uniform algorithmic approach to treatment that includes empiric initiation of broad-spectrum antibiotics, with decisions about continuation or deescalation occurring much later in the patent's care trajectory, based on culture-based pathogen detection. Biomarkers and emerging molecular detection tools hold enormous potential for improving ED-based diagnosis, treatment, and disposition of septic patients. Research and development in this arena, from academic medical centers, government, and industry, has been dramatic over the past decade. At present, there are nearly 5000 indexed articles describing the role of biomarkers in sepsis and over 100 clinical trials currently enrolling patients for biomarker testing in sepsis. Still, in current day-to-day practice, there is no biomarker, set of biomarkers, or molecular test that is universally accepted for sepsis diagnosis and treatment. In this article, we outline current challenges to ED management of sepsis and discuss 2 broad categories of existing and emerging diagnostic platforms, host response and direct pathogen techniques that we believe will revolutionize early management of sepsis. Sepsis is caused by a dysregulated host response to infection, making it difficult to define and diagnose clinically. This is especially true early in its course which is also where it can be most effectively treated. Indeed, consensus definitions for sepsis have undergone numerous iterations over the past 2 decades. Under the most recent iteration, Sepsis-3, sepsis is defined as life-threatening organ dysfunction caused by a host response to infection; recommended screening tools for identification of organ dysfunction include either the sequential organ failure assessment score (SOFA), an objective 6-item scale scored from 0 to 20, or an abbreviated 3-item scale (qSOFA) with higher scores on either suggestive of multisystem organ dysfunction and portending worse outcomes (3). Notably, studies comparing the performance of SOFA, qSOFA, and systemic inflammatory response syndrome (SIRS) criteria (used for the diagnosis of sepsis under Sepsis-2) are conflicting, with some of these suggesting that SIRS is superior for early sepsis identification, especially in the ED setting (4). Regardless of the specific tool used, it is important to recognize that sepsis scales and consensus definitions are highly nonspecific. In practice this means that patients with inflammatory response or organ dysfunction secondary to noninfectious etiologies often screen positive. Further, definitive determination of the presence of infection, still depends on microbiologic culture results which typically lag days behind the initial ED presentation, and are rarely available to the emergency physician. Accordingly, it is not surprising that diagnosis and treatment of sepsis in the ED remains a challenge and continues to be delayed for a large proportion of patients (5). The major challenges to sepsis detection and diagnosis in the ED include (a) distinguishing a dysregulated host response to infection from an inflammatory or hemodynamic response to disease states other than sepsis, (b) identifying patients at early and intervenable points in their disease trajectory, when clinical manifestations of sepsis are less dramatic, and (c) differentiating between bacterial, fungal, and viral infections during the ED encounter. Emergency physicians are tasked with making these determinations for patients with whom they have no longitudinal relationship and often limited access to historical data (from which to draw conclusions about variance of current clinical status from baseline) in a very short timeframe. While biomarkers have revolutionized the approach to ED diagnosis of other diseases (e.g., troponin for diagnosis of acute coronary syndrome), their effect on diagnosis and care of patients with possible sepsis has been less dramatic. Lactate measurement has been endorsed as a “standard of care” component of sepsis diagnosis and management, and Sepsis-3 lists persistent lactate elevation (>2 mmol/L) in the face of adequate fluid resuscitation as a criterion for septic shock diagnosis (3). Lactate assay is now widely available in emergency settings and can be performed on small volumes of blood, with rapid turnaround time for results (≤30 minutes). However, lactate elevation is an extremely nonspecific marker of physiologic stress that may also be elevated in a wide variety of other clinical syndromes frequently seen in the ED, including but not limited to cardiogenic, obstructive, and hemorrhagic shock, trauma, liver dysfunction, malignancy, seizure, diabetic ketoacidosis, or secondary to numerous medications (6). Several proteins involved in host response pathways to infection, including C-reactive protein, interleukin 6, lipopolysaccharide binding protein, and soluble triggering receptor expressed on myeloid cells 1 have been evaluated for potential use in the clinical setting as biomarkers for sepsis diagnosis. To date, however, their use in clinical practice has been limited owing to lack of sensitivity and specificity (7). Procalcitonin (PCT), reviewed elsewhere in this issue of JALM, appears to have the most promising performance, especially for differentiation of bacterial from viral infection, with 1 ED study showing a negative predictive value of 98% but a positive predictive value of only 17% for positive blood cultures in patients suspected of sepsis (8). While the specific role of PCT in sepsis diagnosis remains unclear, international Surviving Sepsis guidelines endorse its use as an adjunct for antibiotic deescalation decisions (9). The most promising recent advance in host response-based platforms for sepsis diagnosis rely on omics-based methods. The RNA transcriptome is of particular interest, given mounting evidence that differences in RNA production patterns can be exploited to differentiate inflammation from infection and potentially discriminate class of causative pathogens in cases of infection (7). Several promising multigene RNA amplification-based assays are under development for these purposes. In one very recently published study, the 4 gene (CEACAM4, LAMP1, PLA2G7, and PLAC8) RNA-based classifier SeptiCyte LAB was able to reliably discriminate sepsis from SIRS with an area under the receiver-operator characteristics curve of 0.82–0.89 (10). In early 2018, the panel was approved by the US Food and Drug Administration as the first RNA-derived aid for differentiating sepsis from noninfectious SIRS in critically ill patients. This tool was validated with whole blood drawn from patients on their first day in the intensive care unit; further studies are required and are in the planning phases now to evaluate assay performance in the ED setting, where clinical performance may be different, given the differences in the populations in whom the test would be used (including the fact that many of these patients who go on to full blown sepsis will be in an earlier stage of the sepsis continuum). Pilot findings from 2 other (earlier exploratory) RNA transcriptomic-based platforms suggest promise, with results indicating more accurate performance versus PCT in differentiating adult respiratory infections as bacterial or nonbacterial in both the ED (87% versus 78%) and inpatient (93% versus 71%) settings (11, 12). While more extensive validation in the acute care (ED) setting is required, this class of diagnostic platform holds enormous potential to transform our approach to evaluation and diagnosis of patients with suspected sepsis in the ED. Sepsis treatment in the ED is directed toward provision of hemodynamic support while simultaneously attempting to identify the presence (or absence) of infection and achieve source control. As noted above, definitive identification of the source of infection and the specific pathogen has been challenged by a lack of sufficiently rapid, sensitive, and specific diagnostic tests. Traditional culture-based methods require time to result and are rarely (if ever) available to the emergency physician; even more challenging (owing to time requirement of traditional microbiologic methods), antibiotic susceptibility is reported long after antimicrobial therapy is initiated. Given the mortality of sepsis and the lack of pathogen identification tools available in the ED, early initiation of broad-spectrum antibiotics is commonplace and recommended but results in inappropriate antimicrobial therapy for as many as 40%–50% of patients, including many with viral or fungal etiologies of sepsis (1, 13). While there is likely a role for biomarkers of host response (including PCT and RNA transcriptomics-based classification) in guiding the decision to administer or continue antibiotics (1, 7), emerging technologies for direct pathogen detection hold promise for much more specific support of treatment decisions. Nucleic acid amplification technologies (NAATs) were long heralded as the future of rapid and broad-based pathogen detection for sepsis. However, translation into practice has been hampered by varied technical challenges. Amplification and detection of relatively low concentration pathogen nucleic acid from undifferentiated and nonenriched whole blood samples, where they are dwarfed by host nucleic acid, are particularly difficult. Analysis can be further complicated by presence of multiple (polymicrobial) agents (13). Improved accuracy can be achieved when NAATs are applied to blood specimens that have been precultured, but such approaches provide limited benefit for the ED setting, where time to result is of the utmost importance. Several teams have sought to overcome these challenges by optimizing performance of direct pathogen detection platforms via development of advanced sample preparation methods and closely coupling NAATs with other highly sophisticated technologies for rapid pathogen characterization. Of these, the most advanced include SeptiFast (Roche Diagnostics) (14, 15, 16), SepsiTest (Molzyme) (14, 17), Iridica Plex ID (Abbott Molecular) (14, 18), and T2Bacteria and T2Candida assays (T2 Biosystems) (19, 20), permitting simultaneous identification of multiple bacterial and fungal species (25, 358, 780, 6, and 3, respectively). Specific post-PCR detection methods include probe hybridization, DNA melt analysis (21, 22) (SeptiFast), selective lysis and Sanger sequencing (SepsiTest), electrospray ionization mass spectrometry (Iridica) (18), and T2 magnetic resonance technology (T2Bacteria and T2Candida). The potential for clinical impact of these platforms is obvious because rapid and reliable pathogen species identification could have drastic effects on antibiotic prescribing patterns. Further, SeptiFast and Iridica both have capacity for identifying antibiotic resistance genes, increasing their potential for support of antibiotic selection. While these technologies are promising and represent an improvement over traditional NAATs, performance estimates have been highly variable and complicated by comparison with an imperfect gold standard (microbiologic culture). All 5 are approved for commercial use in Europe but only T2Bacteria and T2Candida have been approved to market in the US. Notably, the most broad based of the 5 assays, Iridica, for which the FDA trial was recently completed, was recently pulled from production by Abbott, with that decision driven in part by economics and logistics of system deployment (23). Trials testing the efficacy, acceptability, and added benefit of these technologies in clinical practice are needed, and testing in the ED setting will be particularly important. The greatest potential for improved sepsis diagnosis and treatment may lie at the interface of molecular diagnostics and the rapidly advancing fields of data science and health information technology. Over the past decade, widespread adoption of the electronic health record and application of high content data analytic methods to clinical data has enabled development of data-driven decision support not previously possible. We recently developed and validated an electronic tool that uses supervised machine learning and real-time interaction with the electronic health record to provide risk-driven clinical decision support to nurses assigning triage acuity scores at time of ED arrival (24). This tool (e-triage) is applied to all patients and uses routinely gathered information (basic demographics, arrival mode, vital signs, chief complaint, and medical history) to make reliable probabilistic predictions of need for hospital ward or intensive care unit admission, need for emergency procedure and in-hospital mortality at the individual patient level. These predictions drive ED resource distribution and patient care trajectory, and this machine learning-based approach has been shown to outperform the current gold standard for ED triage in identification of patients at elevated risk for adverse outcomes (24). While none can make assessments as early in the patient's course, a number of machine-learning algorithms with specific capacity for prediction of sepsis and septic shock have also been developed (13, 25–26). Similar to e-triage, these tools generate risk estimates based on data available in the electronic health record but enable clinical decision support for enhanced sepsis detection and sepsis-related decision-making. These tools have independent utility but using them to identify patients most likely to benefit from biomarker measurement or molecular diagnostic assay may maximize benefit for individual patients at lower overall cost than is achievable using less discriminate testing. In turn, incorporation of molecular test results into machine-learning algorithms is likely to improve their own predictive capacity. Recent technological advances in multiple fields have enabled innovation at a level not seen in the past, and transition from syndromic diagnosis and empiric treatment to objectively driven decision-making is underway. Given the fact that EDs serve as the initial gateway to care for patients with sepsis, advances in tools that can be used in this setting to expedite diagnosis and treatment decisions is particularly critical. At this stage, focused research that integrates the expertise of biomedical scientists, microbiologists, data scientists, and acute care translational researchers is needed to facilitate assay optimization and ensure targeting of the most appropriate patient populations. Updated Surviving Sepsis Guidelines CLN, February 2017 emergency department sequential organ failure assessment score quick SOFA systemic inflammatory response syndrome procalcitonin nucleic acid amplification technologies.
No takes yet. Share an insight, caveat, or question.
Jalbout et al. (2018) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: