KK: Clinical decision instruments based on associations—broad associations from large groups of patients, which are then extrapolated to one patient at the bedside—are frequently flawed.
Explore This Issue
ACEP Now: Vol 35 – No 07 – July 2016DY: I think the Seymour paper [on the guidelines] comes right out and says, “Don’t use this now; it needs more testing prospectively.” I am certain we are going to see a deluge of papers in different settings looking at how this performs.
TO: This was based on retrospective EHR [electronic health record] data from more than 250,000 patients. However, there were some issues associated with that data that made them less than perfect. I want to make sure a couple of things are clear. The authors themselves say the definitions need prospective validation before being implemented. When we talk about qSOFA, we’re talking about a hypotensive, altered patient with tachypnea; it can be any two of those three things. You can take out tachypnea; we’ve already discussed the issues with respiratory rate. If you have an infected, hypotensive patient, or an infected, altered patient, that patient isn’t right; the janitor can tell you that the patient has a problem. When the authors say, “screening for patients likely to have sepsis,” they’re really defining sepsis as “people with infection or organ dysfunction.” qSOFA is easy to measure, doesn’t require labs, and is highly sensitive and highly specific for a bad outcome. But is this where we should start to risk stratify or where we should start our screening?”
KK: I want to hear from each of you about why you personally haven’t endorsed the definitions.
DY: It isn’t clear that this will improve care for patients in the early, undifferentiated phase of sepsis. It’s the front end not having a strong sensitivity value and, while being simpler, may have been too restrictive.
TS: For me, taking lactate and using it only in septic shock can confuse the general emergency physician, who can think, “Maybe lactate isn’t so important.” From a practical point of view, you cannot even use these “new definitions” to fulfill your SEP-1 (sepsis) quality measure or ICD10 coding requirements.
TO: Until this is prospectively validated, I think it has the potential to hurt patients; it definitely can hurt EM physicians and hospitals. National quality metrics often use mortality ratios, which equal observed mortality over expected mortality. Observed mortality is very easy; you can’t fake death. Expected mortality is based on diagnosis. So, if you’re using the new definition for sepsis, in which a hypotensive, septic patient would fit their definition of sepsis and not septic shock, and the government is using the previous definition for sepsis, your observed mortality is going to be much higher than expected.
2 Responses to “A Marriage of Old Data and New Concepts: New Sepsis Definitions Raise Concerns about Accuracy, Usefulness in Emergency Medicine”
July 26, 2016
Lawrence LynnWe were encouraged that sepsis science was trying to move toward evidenced based crteria from the guessed SIRS of the past. The use of the guessed SIRS criteria in clinical trials was not scientific and was producing runaway inflation of the sepsis diagnosis, inflating the perceived benefit of intervention and rendering any positive benefit of RCT (in comparison with a control population) nonreproducible. SIRS had to be abandoned as the standard definition used in sepsis research.
However the needs of clinical medicine in a vacuum of objective data are different than the needs of scientists in that same vacuum. It is therefore not surprising that ER physcians have been disappointed with qSOFA which was derived from an effort to improve the scientific study of sepsis.
The problem is that there was no determination of when, on the timeline of the different dynamic relational patterns of common sepsis phenotypes, qSOFA criteria are met.
To illustrate this, consider the case of Rory Staunton. He was alert, non hypotensive, yet he had tachypnea, over 50% bands (evolving neutrophilic failure), a low platelet count (although not down to 100) and evidence of infection. Later (perhaps too late) he had hypotension and mental status changes.
When we called for a new sepsis definition we expected that the data from large trials would be reviewed to determine the dynamic relational patterns of the common sepsis phenotypes and then EARLY markers components of the patterns selected as a screening definition with the addition of the time patterns in the scientific definition. Alas, in the statistical search for correlates, time was, once again, overlooked. ER docs never overlook time as the essence of the word “emergency” is time.
The problem with setting up late criteria is the risk of a false sense of security. Altered mental state and hypotension are proven markers of all sepsis phenotypes but so is death. All three are often late findings and markers of diagnostic delay.
July 31, 2016
Mike GertzThe recommendations by the surviving sepsis campaign have always been to only screen patients who are “seriously ill appearing.” As a specialty, we know what that means. The problem is that we are trying to screen everyone with a lot of resultant noise. In the end, sepsis is like pornography, difficult to define but we all know it when we see it.