Victories and Opportunities in the Surviving Sepsis Campaign’s Antibiotic Timing Guidance*

Timely and appropriate antibiotics are the single most effective treatment for patients with sepsis or septic shock (1,2). Bedside clinicians charged with treating sepsis, however, face the challenge that a third or more of patients presenting with syndromes that look like sepsis have nonbacterial infections (i.e., viruses) or noninfectious conditions (3,4). Clinicians must therefore weigh the necessity of early and appropriate antibacterial treatment for patients who do in fact have bacterial sepsis against the patient and population risks of unnecessary antibiotics for those who do not. The first four versions of the Surviving Sepsis Campaign published 2004–2017 did not acknowledge this clinical challenge but simply recommended treating all patients with sepsis or septic shock within 1 hour of recognition. This approach prompted considerable controversy for fear that it promoted overuse of antibiotics for uninfected patients and premature diagnostic closure. The Infectious Disease Society of America withdrew their endorsement of the Surviving Sepsis Campaign guidelines in 2017 and the American College of Emergency Physicians opted to write their own sepsis guidelines in 2021 (5,6). The 2021 version of the Surviving Sepsis Campaign Guidelines lowered the volume on this controversy by shifting the paradigm. Instead of providing a single uniform recommendation for all patients with sepsis or septic shock, the 2021 Guidelines recommended customizing antibiotic management for each patient based on their probability of bacterial infection and severity of illness (7). For patients with definite sepsis or possible septic shock, the Guidelines continue to recommend treating within 1 hour of recognition. For patients with possible sepsis and no shock, however, the Guidelines suggest rapid, active investigation to clarify the patient's diagnosis before giving antibiotics. This could include targeted diagnostics (e.g., cross-sectional imaging, respiratory viral panels, bloodwork) and/or observing the patient's response to treatments for noninfectious conditions (e.g., diuretics, fluids, analgesics, opioid antagonists, bronchodilators, rate control, anticoagulation, etc. as appropriate). Antibiotics are only recommended if concern for infection persists beyond 3 hours. For patients with a low likelihood of infection and no shock, the Guidelines suggest deferring antimicrobials altogether while continuing to monitor the patient in case their probabilities shift. The revised recommendations were based on a large body of literature that report strong associations between time-to-antibiotics and mortality for patients with septic shock but little or no association within the first 3–5 hours after arrival for patients with possible sepsis but no shock (8–10). No study to date, however, has directly assessed the Surviving Sepsis Campaign guidelines' revised framework in which patients are stratified by likelihood of infection and the presence vs. absence of septic shock. A new study in Critical Care Medicine does exactly this. Taylor et al (11) retrospectively evaluated 166,556 patients with orders for cultures and antibiotics in the emergency departments of 12 hospitals affiliated with Atrium Health between 2017 and 2021. They calculated the probability of infection for each patient using a machine learning model that predicted receipt of 4 or more days of antibiotics vs. 1–3 days of antibiotics. Patients with a 50% or greater likelihood of receiving 4 or more days of antibiotics were classified as having "probable infection," whereas those with less than 50% chance were classified as having "possible infection." They then divided these two groups into patients with shock (defined by hypotension, elevated lactate levels, or vasopressor requirement) vs. patients without shock. This effectively arrayed the population into four groups: probable infection with shock, probable infection without shock, possible infection with shock, and possible infection without shock. They then reported on each group's size, median time-to-antibiotics, and crude hospital mortality rate. The investigators' key finding was that patients with possible infection but no shock constituted 30% of the population treated for infection in their emergency departments. These patients received antibiotics a median of 5.5 hours (interquartile range, 3.2–9.8 hr) after arrival, yet their mortality rate was only 1.9%. This mortality rate was substantially less than the mortality rates of the other three groups: 2.8% for probable infection without shock, 12% for possible infection with shock, and 17% for probable infection with shock. The authors concluded that the updated Surviving Sepsis Campaign recommendations appropriately identify a large group of patients that can safely tolerate longer intervals until antibiotics, thus allowing clinicians more time to work on clarifying a diagnosis before starting antibacterials, and therein hopefully reducing unnecessary antibiotic courses and overly broad antibiotic regimens, and promoting more diagnostic clarity. The study by Taylor et al (11) provides several useful contributions. First, it reminds us that antibiotic use is very common in emergency departments but infection is only confirmed in a fraction of treated patients. In this cohort, 30% of patients admitted to hospital via the emergency department were given antibiotics for suspected bacterial infection but concern for infection persisted for greater than or equal to 4 days in only two-thirds. This mirrors the findings of others, suggests ample opportunities to improve antibiotic stewardship, and underscores our need for rapid diagnostic tests for bacterial infection (4). Second, the study by Taylor et al (11) provides additional support for the principle that not all patients in whom the possibility of sepsis is being entertained need to be given antibiotics within 1 hour. A sizable proportion of patients can tolerate longer intervals until antibiotics, an important advantage that allows clinicians time to try to clarify the diagnosis before starting antibiotics. Third, clinicians are already risk-stratifying patients and treating them differently depending on their level of suspicion for serious infection. Patients with possible infection in the study by Taylor et al (11) were treated with antibiotics a median of 5.5–6.9 hours after arrival, whereas those with probable infections were treated a median 2.7–3.2 hours after arrival. This helps confirm that the Surviving Sepsis Campaign recommendations to risk stratify patients are feasible and align with existing clinical practice. There are, of course, important limitations to this article. First and foremost, it was a retrospective analysis of observational data, not a prospective randomized comparison of different management strategies. The study by Taylor et al (11) reported reassuringly low mortality rates for patients with possible infection without shock despite receiving antibiotics a median of 5.5 hours after arrival, but we do not know if outcomes could have been even better if these patients had been treated sooner. The Surviving Sepsis Campaign framework for risk-stratifying antibiotic treatment decisions ideally needs a prospective, randomized trial to confirm safety and to quantify its impact on antibiotic use, appropriateness, adverse effects, and patient outcomes. Second, the study reported time-to-antibiotics relative to emergency department arrival rather than sepsis recognition and so the findings are not perfectly translatable to the Surviving Sepsis Campaign's guidance. This is important because many patients do not meet diagnostic criteria for sepsis until several hours after arrival (12). Third, the investigators used a good but imperfect reference standard for probable infection ("≥ 4 d of antibiotics"). This measure has a sensitivity of about 55% and positive predictive value of about 85% (13). In addition, the prediction algorithm performed well but not perfectly so may have added further misclassifications. Fourth, the study imputed emergency department clinicians' level of suspicion based upon patients' future treatment courses. We do not know what clinicians were thinking in real-time when patients first presented; some syndromes may have looked very suspicious for infection but subsequently turned out to be noninfectious or vice versa. This reminds us that calculating infection probabilities in real-time can be very challenging and the Surviving Sepsis Campaign framework requires clinicians to exercise considerable judgment. Reassuringly, the study by Taylor et al (11) suggests that clinicians already do this and are generally able to differentiate low- vs. high-risk patients. Finally, mortality rates were also very low for patients with probable infection but no shock (3%) despite a median of 3.2 hours until antibiotics. This suggests requiring clinicians to treat this group within 1 hour may also be unnecessarily constraining (although we do not know how many of these patients had true sepsis with organ dysfunction rather than uncomplicated infection alone). More broadly, both the Surviving Sepsis Campaign antibiotic guidance and the study by Taylor et al (11) affirm that there is considerable heterogeneity among patients presenting to the emergency department with possible sepsis. The Surviving Sepsis Campaign stratifies patients across two dimensions (likelihood of infection and presence vs. absence of shock) but these groups still encompass a very wide range of patients and presentations. It is reasonable to suspect there may be other clinical variables associated with meaningful differences in outcomes depending on time-to-antibiotics. For example, patients with possible infection and multiple organ failure, severe immunocompromise, or possible meningitis may also benefit from immediate antibiotics even if they do not have shock (14,15). All told, the current Surviving Sepsis Campaign guidance on antibiotic timing is a great step forward but likely has room for further refinement. Possibilities include extending the time window for investigation for patients with possible infection but no shock and shortening the time window to treat selected high-risk patients without shock. We hope that additional investigators will follow the lead of Taylor et al (11) to further test and help refine antibiotic guidance for patients with possible but unconfirmed serious infections.