For most of the modern history of sepsis, estimates of its incidence and prevalence were based on data mined from administrative claims. In other words, our knowledge of incidence and prevalence was dependent on physicians to recognize sepsis and to actually record the presence of sepsis in patients’ medical records, and on coders to identify that physicians had done so and appropriately enter sepsis as a principal or secondary diagnosis on the patient’s discharge face sheet. It should be apparent that there are several ways in which this system could result in either overestimates or underestimates, although data suggest that incidence and prevalence were mostly underestimated by this approach, initially (1). The problem was compounded for years by the absence of coding categories for sepsis and severe sepsis, with only “septicemia,” International Classification of Diseases, 9th Revision (ICD-9) code 038.9, available for use by coders.
In 2002, ICD-9 codes for sepsis and severe sepsis were added. Shortly after, in 2007, the Centers for Medicare and Medicaid Services and other insurers began recognizing the expense associated with sepsis care by reimbursing more for patients diagnosed with sepsis or severe sepsis. Simultaneously, the Surviving Sepsis Campaign and other quality improvement efforts began to spread the awareness of sepsis as an important public health problem beyond the confines of the ICU (2). The incidence of sepsis and severe sepsis was seen to grow rapidly in the ensuing decade (3). An often advanced, rather cynical viewpoint suggests that patients were and are being over-coded with sepsis principally so that hospitals could take advantage of increased reimbursement (4). A more optimistic view is that physicians and other providers, once informed of how to recognize sepsis and severe sepsis and given appropriate tools, began efforts to do so, in the best interests of patients.
Against this backdrop, CMS instituted measures in 2015 that require participating hospitals to report their performance in meeting certain treatment milestones for all septic patients. Again, however, patients are audited based on whether they receive a sepsis diagnostic code, meaning that septic patients who are never specifically diagnosed with sepsis will not be audited. Shortly thereafter, a third consensus conference proposed a definition for sepsis, along with new diagnostic criteria (Sepsis-3) that eliminate using the components of the Systemic Inflammatory Response Syndrome and focus entirely on organ dysfunction, as measured by the Sequential Organ Failure Assessment (SOFA) score (5).
In March of 2018, the Centers for Disease Control and Prevention (CDC) responded to this entire history of changing sepsis criteria and coding requirements by publishing criteria for what they term an Adult Sepsis Event (ASE) (6). This definition recognizes that nearly all U.S. hospitals are now equipped with electronic health records, and the definition is intended to provide a measure of interfacility comparison that has not been previously accomplished. The ASE definition avoids use of vital signs data, deemed subject to measurement variability, and also omits evidence of CNS dysfunction, on the assumption that Glasgow Coma Scale is not measured in many patients. Rhee et al (7) used this definition to provide the first ever epidemiologic description of sepsis based on clinical criteria, rather than on administrative data. Patients with high likelihood of infection and at risk for sepsis are identified by orders for blood cultures, along with receipt of at least 4 days of IV antibiotics.
In this issue of Critical Care Medicine, Rhee et al (8) further their work and present a comparison of ASE criteria with Sepsis-3 criteria, to assess whether similar sets of patients are identified by the two criteria, as a form of validation of the utility of the ASE. The foundation for this work is data from the 111 hospitals participating in Cerner Health Facts from 2013 to 2015, and the results were validated with data from the Emory health system, collected during the same period. Although not a part of either the ASE or the SOFA-based criteria, ICD-9 codes were used to identify the likely source of infection for patients in both groups, for comparison. The authors report that the two definitions do result in identification of similar, though not identical, patient populations. Demographics and comorbidities were similar in both groups, whereas mortality was slightly higher in the ASE cohort. The ASE criteria were 60% sensitive, relative to the Sepsis-3 criteria, not classifying as septic those patients with mild hypoxemia or neurologic dysfunction. As expected, pneumonia and urinary tract were the most common sites of infection in both groups. Respiratory and neurologic dysfunctions were the most common for the SOFA group, whereas increased lactate and doubling of creatinine were the most common organ dysfunctions in the ASE group. This represents a fascinating symmetry, as SOFA does not include lactate criteria, whereas ASE does not include neurologic criteria. The ASE criteria had a slightly higher predictive ability (area under the receiver-operator characteristic curve) for in-hospital mortality than the Sepsis-3 criteria.
Because there is no pathobiologic gold standard for diagnosing sepsis, the condition is recognized through clinical syndromes, whether prospectively or retrospectively, as the ASE was intended to do (6). It is unclear what syndrome represents the “best” one for characterizing sepsis, and the different available criteria serve different needs or functions, whether they be sensitive early recognition (severe sepsis, machine learning), risk stratification (Sepsis-3, machine learning), or overall surveillance (ASE). The Sepsis-3 criteria were chosen based on their predictive ability for mortality or prolonged ICU stay. Although they were intended for diagnosing sepsis, they were developed by retrospective data analysis and they remain mostly unproven as prospective diagnostic tools. Neither Sepsis-3 criteria nor ASE criteria specify a mechanism for identifying infection, only that an infection is strongly suspected, as evidenced by blood cultures and antibiotics. Because the ASE criteria demonstrate superior predictive ability for mortality, one is forced to wonder whether they could themselves be adapted for prospective recognition.
The study and the ASE have strengths and weaknesses. Principal among the strengths is the reliance on completely objective measures that should be present in every medical record. Principal among the weaknesses is that sepsis-induced encephalopathy, a significant source of morbidity and mortality in sepsis, is ignored, because it is less than objective and not consistently available in the medical record. An additional weakness is that these criteria apply only to adults and further criteria are needed, if we are to better understand the epidemiology of sepsis in children. The authors use the moniker “eSOFA” for the organ dysfunction criteria of the ASE; however, this is a misnomer. SOFA, the SOFA score, has been validated as a mechanism for tracking individual patients’ organ dysfunctions and risk of mortality (9). The organ dysfunction criteria of the ASE have not been so validated and, in fact, are not intended for sequential assessment or monitoring of individual patients. There is little need for further confusing or misleading terminology in the war on sepsis.
It is important to emphasize that the ASE is not currently intended as a diagnostic system, but rather as a means to retrospectively evaluate which hospitalized patients have or had sepsis, regardless of whether they were diagnosed as septic at the time of hospitalization; it is a surveillance mechanism. The presented data indicate that a simple, objective, and automated set of criteria for identifying sepsis can effectively identify a subset of infected patients with a high risk of mortality and can do so with good predictive ability. On its face, this may not seem earth-shaking, but it is potentially revolutionary for our understanding of sepsis epidemiology and our ability to determine whether we are, in fact, altering sepsis mortality. Once adopted by the American healthcare system as the CDC suggests, we will have the ability to move from a sampling methodology to determine these things to a counting methodology that literally tallies in near real time the number of Americans affected by sepsis, along with their outcomes. Furthermore, once the ASE criteria are embedded within electronic medical records (EMRs), they could serve as a real time backup to other sepsis recognition and treatment measures, calling attention to patients whose sepsis may have not been explicitly recognized by their providers. There is tremendous value for patients, for healthcare providers, and for the healthcare system in a relatively simple to implement sepsis surveillance system that is potentially available to all hospitals with an EMR and has predictive validity for poor outcome.
1. Angus DC, Linde-Zwirble WT, Lidicker J, et al. Epidemiology of severe sepsis
in the United States: Analysis of incidence, outcome, and associated costs of care. Crit Care Med 2001; 29:1303–1310.
2. Calandra T, Cohen J, Gea-Banacloche J, et al. Surviving Sepsis
Campaign guidelines for management of severe sepsis
and septic shock. Crit Care Med 2004; 32:858–873.
3. Gaieski DF, Edwards JM, Kallan MJ, et al. Benchmarking the incidence and mortality of severe sepsis
in the United States. Crit Care Med 2013; 41:1167–1174.
4. Gohil SK, Cao C, Phelan M, et al. Impact of policies on the rise in sepsis
incidence, 2000-2010. Clin Infect Dis 2016; 62:695–703.
5. Singer M, Deutschman CS, Seymour CW, et al. The Third International Consensus Definitions for Sepsis
and Septic Shock (Sepsis
-3). JAMA 2016; 315:801–810.
6. Centers for Disease Control and Prevention: Hospital Toolkit for Adult Sepsis Surveillance
. 2018Atlanta, U.S. Department of Health and Human Services.
7. Rhee C, Dantes R, Epstein L, et al. Incidence and trends of sepsis
in US hospitals using clinical vs claims data, 2009–2014. JAMA 2017; 02215:2009–2014.
8. Rhee C, Zhang Z, Kadri SS, et al; for the CDC Prevention Epicenters Program: Sepsis Surveillance
Using Adult Sepsis
Events Simplified eSOFA Criteria Versus Sepsis
-3 Sequential Organ Failure Assessment Criteria. Crit Care Med 2019; 47:307–314.
9. Ferreira FL, Bota DP, Bross A, et al. Serial evaluation of the SOFA score to predict outcome in critically ill patients. JAMA 2001; 286:1754–1758.