Two years ago, a panel appointed by the Society of Critical Care Medicine and the European Society of Intensive Care Medicine, referred to as a consensus conference, proposed a new definition for sepsis and new diagnostic criteria for sepsis and septic shock, known as Sepsis-3 (Singer M, et al. JAMA. 2016;315:801). The panel proposed that sepsis be defined as life-threatening organ dysfunction due to a dysregulated host response to infection. Upon reflection, one could see that what we had called definitions of sepsis, severe sepsis, and septic shock for over 2 decades actually represented diagnostic criteria more than concise definitions. In that regard, a concise definition is a useful addition in the tool kit for training all health-care professionals to recognize sepsis and to treat it early and aggressively.
However, the diagnostic criteria leave something to be desired, in terms of both practicality and sensitivity for detecting patients whose infection has made them seriously ill. Those who participate in quality improvement efforts in their own hospitals will recognize that to promote change and to achieve a goal of better, higher quality care, it is important to remove obstacles in the system and to structure it so that doing the right thing is easier than not doing it. For sepsis, the first step in the process, recognizing that sepsis is present, has always been complex enough that it has been the bane of the enterprise. As many as two-thirds of patients with sepsis presenting to the ED with severe sepsis never receive that diagnosis while in the hospital. (Deis AS, et al. Chest. 2018;153:39). As any sepsis core measure coordinator can attest, diagnostic criteria that are readily visible on retrospective examination are often unnoticed or misinterpreted in real time.
The crux of this issue is that the very entity of sepsis is not a definite thing but a not-quite-focused idea. Much is known of pathophysiologic features that seem to be important, but there is no one unifying pathologic condition. Contrast that with another critical illness, myocardial infarction. The very name states the unifying pathology. Our predecessors were able to work backward from an understanding that acute blockage of a small artery led to ischemia and infarction, in order to identify methods to detect it while it is happening—measuring enzymes and evaluating an ECG. For sepsis, we don’t even understand why patients are sick or why they die. There is a complex interaction of inflammation, microcirculatory thrombosis, mitochondrial dysfunction, immune suppression, but there is no one combination of those things that is yet understood in a way that lends itself to diagnostic testing. The best we can say is that the patient reacted to their infection in a way that was detrimental to their own body’s functioning. Rather than recognizing a few symptoms and sending a confirmatory test, with sepsis, we must tote up the signs and symptoms in the domains of recognizing infection and recognizing organ dysfunction, then determine whether they are present in sufficient amounts; it is an exercise that requires mental discipline.
If the diagnostic criteria we use, whether Sepsis-1, 2, or 3, are all gross descriptions of complex internal interactions that are not specific, then the syndrome that any of these criteria identifies is also not specific for anything particular. It falls to the medical community, as a whole, to determine exactly what it is that we desire a given syndrome to be indicative of. The Sepsis-3 authors decided that the appropriate syndrome should predict death or prolonged ICU stay. They used several large data sets to develop and validate infection-associated variables that would have good predictive ability for that outcome, and they compared what they found with sepsis by the Sepsis-1 definition, infection plus SIRS (Seymour C, et al. JAMA. 2016;315:762). Infection + SIRS is a strawman in this comparison, because they tested its predictive ability for the outcome against that of the Sequential Organ Failure Assessment (SOFA) and the Logistic Organ Dysfunction Score (LODS). These two scoring systems were developed as severity of injury scales and validated as mortality predictors; the higher the score, the likelier mortality, whereas SIRS clearly contains no information about organ dysfunction. The comparator of interest for this outcome is actually severe sepsis, infection plus SIRS plus organ dysfunction.
Although the criteria the Sepsis-3 investigators used for defining patients with suspected infection were novel and reasonable, we lack additional important information about the patients they studied. They did not report the spectrum of treatments for sepsis in their cohort, whether early or late, adequate or inadequate, so it is impossible to determine whether the criteria address patients who are undertreated, patients who are treated late, patients who will die regardless of adequate therapy, or some combination. In other words, there is no way to tell whether patients who were recognized early in their course via Sepsis-1 criteria and treated aggressively and effectively may have avoided shock, ICU admission, and death. It is, of course, the business of physicians and nurses to help patients avoid exactly those things. Multiple studies have now demonstrated that SIRS criteria are more sensitive than SOFA-based screens, specifically qSOFA, for identifying infection with organ dysfunction, and that qSOFA is more specific for mortality (Serafim, et al. Chest. 2017; http://dx.doi.org/10.1016/j.chest.2017.12.015).
In contrast, the Sepsis-1 authors proposed infection plus SIRS as a sensitive screening tool that could warn of the possibility of an associated organ dysfunction (Sprung, et al. Crit Care Med. 2017;45:1564). Previous to the Sepsis-1 conference, Bone and colleagues had defined the sepsis syndrome, which incorporated both SIRS and organ dysfunction (Bone, et al. Crit Care Med. 1989;17:389). It was the collective insight of the Sepsis-1 participants to recognize that SIRS induced by infection could be a harbinger of organ failure. The Sepsis-3 authors believe that SIRS is a “normal and adaptive” part of infection and that it is “not useful” in the diagnosis of sepsis. That analysis neglects a couple of important things about SIRS. First, numerous studies demonstrate that infection with SIRS is associated with a mortality rate of 7% to 9%, which is by no means trivial (Rangel-Frausto MS, et al. JAMA. 1995;273:117). Second, the components of SIRS have been recognized as representative of serious illness for millennia; the assertion that the Sepsis-1 definitions are not evidence-based is mistaken and discounts the collective experience of the medical profession.
Finally, SIRS is criticized on the basis of being nonspecific. “If I climb a flight of stairs, I get SIRS.” This is clearly a true statement. In fact, one could propose that the name could more accurately be Systemic Stress Response Syndrome, though “scissors” is certainly less catchy than “sirs” when one says it aloud. However, the critique neglects an important concept, encapsulated in Bayes’ Theorem. The value of any positive test result is largely dependent on the prevalence of the disease being tested for in the population being tested. It is unlikely that the prevalence of sepsis is very high among patients whose SIRS is induced by climbing a flight of stairs. On the other hand, tachycardia and tachypnea in a patient who is indulging in no activity while lying on a bed feeling miserable should prompt a search for both the infection that could be causing it and the organ dysfunction that could be associated with it. The specificity of SIRS derives from the population in which it is witnessed, and its sensitivity is to be respected.
To quote a friend, the remarkable CEO of a small Kansas hospital, “If a patient with an infection feels bad enough that they climb up on that gurney and place themselves at our mercy, we owe it to them to prove why they don’t have sepsis, rather than why they do.”