SCCM is performing maintenance on its websites. For the best browsing experience, please use Microsoft Edge or Safari. Those using Chrome or Firefox may experience access issues at this time.

SCCM Pod-439 The Association Between Antibiotic Delay and Hospital Mortality

visual bubble
visual bubble

Rapid delivery of antibiotics is a cornerstone of sepsis therapy, although time targets for specific components of antibiotic delivery are unknown. Host Ludwig H. Lin, MD, is joined by Stephanie Parks Taylor, MD, to discuss the significance of time lead for suspected sepsis patients, how to use a generous time window wisely, and new and reconfigured technologies opportunities (Taylor SP, et al. Crit Care Med. 2021 May;49:741-747). 

Dr. Taylor is an associate professor in the Department of Internal Medicine at Atrium Health’s Carolinas Medical Center in Charlotte, North Carolina, USA. This podcast is sponsored by Biofire

*If you are unable to play the podcast please click here to download the file.

Category: CCM Podcast


This podcast is sponsored by BioFire. High-acuity patients require time-sensitive specialized care. As a critical care physician, you need rapid reliable test results to make informed intervention decisions. The BioFire Film Array System utilizes a syndromic approach, simultaneously testing for different pathogens that can cause similar symptoms to deliver actionable results in about an hour. BioFire helps you quickly identify specific infectious agents, allowing you to begin targeted treatments sooner. Learn more about solutions from the leader in syndromic testing at

Dr. Lin: Hello and welcome to the Society of Critical Care Medicine’s iCritical Care podcast. I’m your host, Ludwig H. Lin, MD. Today, I will be speaking with Stephanie Taylor, MD, on the article “The Association Between Antibiotic Delay Intervals and Hospital Mortality Among Patients Treated in the Emergency Department for Suspected Sepsis,” published in the May 2021 issue of Critical Care Medicine. To access the full article, please visit Dr. Taylor is an associate professor of internal medicine at Atrium Health in Charlotte, North Carolina. I really appreciate her taking the time to speak on this very important topic that is very much on the mind of anybody caring for patients with sepsis. Welcome, Dr. Taylor. Do you have any disclosures to report?

Dr. Taylor: I do not.

Dr. Lin: Thank you. SCCM was very excited about having this article published. The topic of timely antibiotic administration and its impact on controlling sepsis cleaves very closely with the Surviving Sepsis Campaign. Dr. Taylor, let’s start off broadly and review why this is such an important topic. What is the significance of timely antibiotic administration for patients with sepsis?

Dr. Taylor: I think that antibiotic timing targets in sepsis has become a really polarized topic. There are some really strong opinions on both sides. But regardless of where people stand on the issue, we all have the same goal of providing life-saving treatment for patients with sepsis. And we all agree that increased sepsis awareness and attention to treatment times has improved sepsis mortality over the past several decades. The controversial part that remains is how long do you really have to give antibiotics? Is there truly a golden hour or can you wait until the silver hour or the bronze hour? This is really important to know because sepsis diagnosis is challenging. It takes time to gather the diagnostic information needed to distinguish infectious versus noninfectious causes of illness. We felt like that question was still unanswered, which is why we conducted this study of patients in our own healthcare system to try to learn more about the association of antibiotic delays with mortality among patients with suspected sepsis.

Dr. Lin: You studied a huge number of patients. Could you tell us briefly about your demographics?

Dr. Taylor: Yes. I am a researcher at Atrium Health, which is a large integrated healthcare system that recently integrated with Wake Forest Baptist Health. So our system has become larger and has some good academic collaboration. We have a huge patient population—something like 40 hospitals— and we studied all sepsis encounters across our system from 2014 to 2017. Our patients were about 60 years old, about 53% female, and a quarter of them African American. They had a lot of comorbidities, with an average comorbidity score of 5 and, compared to some sepsis cohorts, relatively less sick. Eight percent of them were intubated, with about 10% mortality in hospital. I think this was a fairly representative sample of patients who present to the emergency department (ED) with suspected sepsis.

Dr. Lin: The study design is a retrospective study. Did you consciously design the study to make it different from other studies that were conducted previously about the timing of antibiotic administration or did you want to use your very broad patient base to see if you could duplicate the outcome?

Dr. Taylor: It was a little bit of both. Let me start by summarizing the current literature of time to antibiotics in sepsis. There have been a lot of studies—maybe 40 or 50 studies—on time to antibiotics in sepsis, and the vast majority of these were observational. They vary in terms of population and sepsis syndrome definitions and some methodologic differences. There is quite a bit of variability in the results of these studies.
The best known and most cited study is probably Kumar’s 2006 study, which showed about a 7.5% increase in mortality for every hour of delay after the onset of hypotension. But keep in mind that this was a carefully selected population of patients with hospital-onset shock, and time zero was defined as the time of shock onset. So I think that hourly increase is probably extrapolated a little too generously to other populations. Subsequent studies have had mixed results. One 2015 meta-analysis reported no overall association between time to antibiotics and mortality. But a 2017 meta-analysis reported a strong association. So there has been some back-and-forth.

More recently, in 2017, Seymour and his team evaluated New York state data. In New York, there is a state mandate to protocolize and report sepsis treatment. This study had rigorous risk adjustment and showed a small effect of time to antibiotics on mortality. But it appeared that the effect probably did not emerge until about five hours of delay. After that, there was a rapid increase in mortality. Also, in this study, similar to other studies, the mortality effect was present in the subgroup of patients requiring vasopressors but not in the subgroup of patients without shock.

In this landscape of existing literature, we specifically designed our study to address a slightly different question. One limitation of the existing studies is they asked the question, Do rapid antibiotics improve outcomes for patients with sepsis? These studies selected patients who, in retrospect, ultimately have a diagnosis of sepsis. They weren’t at the beside deciding how fast to give antibiotics. We don’t have the luxury of knowing who has sepsis and who doesn’t. So that’s not really the clinically relevant question. The clinically relevant question is more like, Do antibiotics within one or three or six hours improve outcomes in patients with suspected sepsis or in patients who you would consider treating to those targets?

Imagine you were designing a randomized trial comparing early versus conservative antibiotic administration;
you couldn’t use future information about whether the patient ultimately had sepsis to determine eligibility. You would have to enroll patients based on the information you had on the front end. So we specifically designed our study to include those patients who ultimately didn’t have sepsis but were treated as if they had sepsis to address the pragmatic and clinically relevant research question.

Dr. Lin: It’s almost like the concept of intention-to-treat. When you have a patient who might have sepsis, you give antibiotics and what happens?

Dr. Taylor: Right. In the randomized controlled trial analogy, it’s trying to avoid post-randomization exclusions.

Dr. Lin: Right. I like that a lot. Let’s talk about the outcomes. What did you find? Were you surprised by what you found? What were your takeaway points?

Dr. Taylor: We conducted a retrospective study of 24,093 encounters among adults who had a digital signature of infection, a combination of culture orders and antibiotics and concurrent organ dysfunction across 12 EDs in our healthcare system. We conceptualized a total antibiotic delay into two parts—a recognition delay, which was the time from ED triage to the placement of an antibiotic order, and the administration delay, the time from antibiotic order to its infusion.

Unlike other observational studies that evaluate time as a continuous variable, which linearizes a nonlinear outcome, we assessed delay in terms of discreet time intervals that aligned with currently recommended targets. There were two primary findings. The first was that recognition delay accounted for the majority of the interval from triage to infusion, which makes sense. In the study, the media and recognition delay was 2.7 hours, and the administration delay was 0.6 hours. And the big finding was that, after adjustment for a prespecified set of potential confounders, recognition delay was associated with increased mortality but only for delays greater than six hours. There was also an association between administration delay and mortality for delays greater than 1.5 hours.

Dr. Lin: So those produced some slightly more generous intervals for the clinicians than previously. I was curious about whether the sum of the two—total time from patient arrival to antibiotic administration—was a third variable and whether, for example, a short recognition time but a long administration time, kicking it over a certain limit, would also result in creased mortality. What did you find?

Dr. Taylor: If you add those two up, the median total delay was about 3.4 hours. Four-fifths of that is recognition delay, and we had about the same type of findings. A total delay of greater than six hours was associated with increased mortality.

Dr. Lin: It sounds like a total time of six hours is the magic window, but in fact most of that time is contributed to by the recognition delay. So the two are almost equal.

Dr. Taylor: Yes. And how generalizable this is to other healthcare systems is a question. A matter of 0.6 hours for administration delay is quite fast. There are a lot of steps. There is a process involved between placing the antibiotic order and actually having it infused at the bedside. At Atrium Health, we have spent a lot of quality improvement time in shortening that interval. In other studies, that interval has been longer and that may have a greater impact on total delay in other healthcare systems.

Dr. Lin: 0.6 hours is an amazing time.

Dr. Taylor: I was impressed when I saw that.

Dr. Lin: Yes, I think your institution is doing an amazing job and that speaks a lot to your quality improvement process. I think we could all learn from that. This brings up a couple of extra questions about the process you were discussing. The study collected data from 2014 to 2017. During this time period, were protocols already in place to encourage timing administration or was this just happening on its own?

Dr. Taylor: The data collected from this study was in the setting of the maturation of our sepsis quality improvement efforts. It’s a cyclical process, with rapid improvement. Multiple interventions during this period overall were pushing toward faster antibiotics and greater recognition of patients with sepsis over this time period.

Dr. Lin: What was the setup in terms of sepsis recognition? Was a team overseeing these patients or overseeing the process? Was it an electronic medical record flag? What was driving it?

Dr. Taylor: It was a variety of those factors. Sepsis coordinators would screen patients early on. They were phased out during that time period and then other initiatives came through. There are always education processes. There are electronic health alerts. There are processes to carry out the orders once they are placed. So, yes, lots of different things were going on over this time period.

Dr. Lin: It sounds like you had different processes and checks in place to really encourage people to do this well and your data showed that people were doing a really good job.

Dr. Taylor: Yes, prestocking antibiotics in the ED was a big factor in decreasing the infusion time. This was a really straightforward intervention that reduced the infusion times pretty quickly.

Dr. Lin: Let’s discuss your findings a little bit more. It looks like you found that people have a more generous time window in which to have antibiotics administered. Would you agree or disagree with that?

Dr. Taylor: Yes. Keeping in mind that this is all observational research, I think the evidence has mounted a bit to start to suggest that current time to antibiotic metrics that target super-fast—like one hour or even three hours for all patients with suspected sepsis—is not strongly evidence based. And if that’s true, if we can safely withhold antibiotics for up to six hours for many of these patients, that’s a lot of time to allow further workup to either rule in infection or evaluate other life-threatening etiologies.

Dr. Lin: Yes, it’s so true. And there’s a lot of pressure in the ED and it’s not like each ED physician has 100% time for each patient. So this probably makes everybody breathe a sigh of relief. Were you surprised by this?

Dr. Taylor: I was not all that surprised. I think a careful evaluation of the existing literature suggests this. When studies report that there’s an hour-to-hour association between time to antibiotics and mortality, it’s a little misleading because they have made time a continuous variable, and when you do that, it averages out the big mortality effects of delays like at 12 hours, with the small or minimal effects early on, like zero to six hours, and it ends up looking like there’s an hour-to-hour linear effect, when really the data probably suggest more that there’s a threshold early on between three and six hours where the harm of delay is minimal. There is still room to individualize and to use your own art of diagnosis because patients who are really sick probably have less time to lose. So you need to get antibiotics in faster. Patients who are less sick probably have a little bit longer for you to make that diagnosis.

But I think what we showed is that if you’re going to apply a rule that antibiotics need to go in by a certain hour, one and three hours might be a little too aggressive and something like six hours might be better for a quality check for all patients who have suspected sepsis.

Dr. Lin: One of your first goals was to look at this as an “all comers” thing—giving antibiotics to all patients with suspected sepsis. For those patients who eventually wound up not carrying a sepsis diagnosis, did your team find any adverse outcomes caused by aggressive antibiotic administration or the lack thereof?

Dr. Taylor: That’s a really important area for future study. We didn’t specifically look at harm, but it may be that the mortality benefit is not apparent because of either lack of benefit to the patients who didn’t ultimately have sepsis or harm to those patients. What we learned from the aggressive pneumonia antibiotic targets was that these type of quality metrics increased both appropriate and inappropriate treatment. What we don’t know is how harmful that is. There’s not a lot of evidence to answer that question. At SCCM’s 2020 Congress, Sarah Seelye, PhD, from the Ann Arbor VA, presented some reassuring data that antibiotic use did not increase overall despite accelerating time to antibiotics across the VA system. But there is evidence to suggest that maybe broad-spectrum antibiotic use has increased.

A recent study led by Ian Barbash, MD, used a really cool time series design to show that one measure was associated with 10% increase in broad-spectrum antibiotic use. Antibiotic use is a big concern for both the patient and public health, but probably antibiotic stewardship efforts and de-escalation protocols can mitigate that harm for the most part. What I worry about with these aggressive targets is the opportunity cost of a hypervigilant focus on rigid sepsis metrics. This is why our emergency medicine colleagues tend to be more wary of aggressive targets. It’s resource intensive both in terms of personnel and cognitively so they end up in a zero-sum situation. If so much effort is focused on rigid adherence to aggressive sepsis metrics, what other evidence-based practices are skimped on? What alternative diagnoses are missed in our patients when we’re pushed to anchor on a sepsis diagnosis?

These types of harms and unintended consequences or opportunity costs of being too aggressive with sepsis targets are really hard to measure but they make up an important part of the benefit/harm equation.

Dr. Lin: Yes, and I think it’s important for us to always ask questions and to reassess, which is what this study does. Let’s say that we do have a more generous time window to give antibiotics to our sepsis patients, how do we use that time wisely? Let’s say you are an emergency medicine physician or a hospitalist or an intensivist admitting these patients. Maybe you have two or three more hours than you thought you had, but how are you going to use that time to figure out whether this patient should get antibiotics or not? Are there going to be new technologies or is there a way for us to reconfigure our systems to try to figure this out better?

Dr. Taylor: Yes, that’s a decent amount of time. We already have some tests that can come back pretty quickly. For example, you can see if someone responds to diuresis and distinguish heart failure from sepsis. You can get some other diagnostic tests back. Ultimately, the holy grail that we all want is a rapid point-of-care test that rules sepsis in or out to take the uncertainty out of the equation. I’m pretty sure that even those most cantankerous against early antibiotic targets agree that it’s good and noble to give antibiotics quickly to patients who definitely or almost definitely have sepsis. I think we really need to push science forward toward the development of a rapid diagnostic point-of-care test that can identify sepsis early enough to inform these treatment decisions.

In the meantime, I think we just need to get innovative with our predictive analytics. We need to work on diagnostic tools that are not only accurate but clinically useful in the ranges of disease probability where there’s uncertainty. There are a lot of great teams working on artificial intelligence approaches for sepsis diagnosis so I think we can do this well if we’re thoughtful and creative about how we handle diagnostic uncertainty and if we can generate good science to quantify the risks and benefits of potential treatments.

Dr. Lin: I think it’s always good for us to pose questions, to ask ourselves the same questions, and reassess after a period of time. How has this changed your institution’s practice?

Dr. Taylor: We are still beholden to the metrics and policies that are out there. But it has opened up a lot of eyes to the importance of reevaluating this rather than just devoting so much quality improvement effort into getting these times down as low as we can to as many people as we can, being a little more thoughtful about the denominator, how many false positives are there and is there harm in that group and how can we be a little more accurate in our diagnosis. I don’t think anybody argues that you should get antibiotics quickly once the diagnosis of sepsis is pretty sure. So we’re trying to be more thoughtful about getting that diagnosis right.

Dr. Lin: Sounds great. You have access to a very large patient population, and you’ve already asked some very thought-provoking questions. What are the next steps for you in terms of this area of study, and what other studies would you like to see conducted in this area?

Dr. Taylor: My top priority, which is not my own personal research skill, is working on rapid diagnostic testing. I think that would revolutionize the way we practice in terms of sepsis. So, while I’m not personally working on that, I am very hopeful that the science moves toward that. Our lab is working on more innovative diagnostic techniques, trying to look at risk-stratified treatment targets as opposed to one-size-fits-all treatment targets and evaluating the effects we can see with that approach—a more measured rational approach rather than blasting everyone with antibiotics, getting the rapid antibiotics to the people who need it. So stay tuned.

Dr. Lin: Yes, it’s very interesting. It sounds like, through research and refinement of our approaches, it’s almost approximating those good old days when people used the art of medicine. I’m eagerly awaiting to see what happens next with your lab. Before I wrap this up, do you have any other thoughts you want to make sure our listeners walk away with from our podcast conversation?

Dr. Taylor: Yes, I want to reiterate that the takeaway from our study is not that we can be lazy about sepsis. Our data show that, in our patient population, if it took longer than six hours to recognize sepsis or if it took longer than 90 minutes to get antibiotics infused, people died. Mortality went up. So let’s not see this as, Okay, we can sleep on sepsis. Let’s keep innovating ways to improve sepsis recognition and optimize treatment so that we can get the best outcomes for patients with sepsis and the least harm to the patients without it.

Dr. Lin: I think that is the perfect takeaway message. Do not rest on our laurels. Keep on improving, perhaps improving different ways than trying to get the time as short as possible but keep on working on this. Thank you so much, Dr. Taylor. This concludes another edition of the iCritical Care podcast. And for our team, I’m Ludwig Lin and appreciate your listening. Take care.

This podcast is sponsored by BioFire. High-acuity patients require time-sensitive specialized care. As a critical care physician, you need rapid reliable test results to make informed intervention decisions. The BioFire Film Array System utilizes a syndromic approach, simultaneously testing for different pathogens that can cause similar symptoms, to deliver actionable results in about an hour. BioFire helps you quickly identify specific infectious agents, allowing you to begin targeted treatments sooner. Learn more about solutions from the leader in syndromic testing at

Ludwig H. Lin, MD, is an intensivist and anesthesiologist at Alta Bates Summit Medical Center in San Francisco, California, and is a consulting professor at Stanford University, where he teaches a seminar on the psychosocial and economic ramifications of critical illness. Dr. Lin did his medical training, anesthesia residency, and critical care medicine fellowship at the University of California San Francisco. He has served as faculty at both Stanford University and the University of California San Francisco, where he was a professor and the medical director of critical care at San Francisco General Hospital. His interests are in patient-family communication as well as education. Being an SCCM podcast host reminds Dr. Linn of his undergraduate days as a new broadcaster for his college radio station KZSU.

Join or renew your membership with SCCM, the only multiprofessional society dedicated exclusively to the advancement of critical care. Contact a customer service representative at +1 847 827-6888 or visit for more information.

The iCritical Care podcast is the copyrighted material of the Society of Critical Care Medicine and all rights are reserved. Statements of fact and opinion expressed in this podcast are those of authors and participants and do not imply an opinion or endorsement on the part of the Society of Critical Care Medicine, its officers, volunteers, or members or that of the podcast’s commercial supporter.


Disclaimer Knowledge Area: Quality and Patient Safety Pharmacology