Time to Antibiotics in Sepsis: A Metric Not Supported by “High Quality” Evidence

Background: Some of the major take home points from the sepsis trilogy of studies recently published (ProCESS, ARISE, and ProMISe) was that early identification of patients with sepsis, early intravenous fluids, and timely, appropriate broad-spectrum antibiotics is key to decreasing morbidity and mortality. In 2006 a study by Kumar et al [3] showed a 7.6% increase in mortality in patients with sepsis for every hour of delay after the onset of shock, but this finding has not been reproduced. In fact, the results of timing of antibiotic administration on outcomes have been all over the map. Regardless, the Surviving Sepsis Campaign still has very specific recommendations regarding the timing of antibiotics. And even more painful is that metrics for the quality of care of patients with severe sepsis and septic shock are now recognizing these recommendations as core measures.

Current Surviving Sepsis Guideline Recommendations (2013) [2]:

  • Administer antibiotics within 3 hours of ED triage
  • Administer antibiotics within 1 hour of severe sepsis (Level 1B)/septic shock (Level 1C) recognition

What Article is Being Reviewed?

Sterling SA et al. The Impact of Timing of Antibiotics on Outcomes in Severe Sepsis and Septic Shock: A Systematic Review and Meta-Analysis. Critical Care Medicine 2015; 43(9): 1907 – 15. PMID: 26121073

What They Did:

  • Meta-analysis of 11 publications (16,178 patients) looking at the association between timing of antibiotic administration and mortality in severe sepsis and septic shock
  • Antibiotic timing defined as:
    • ≤3 hours vs >3 hours from triage
    • ≤1 hour vs > 1 hour from shock/severe sepsis recognition
  • Also performed sensitivity analysis of the effect of time to antibiotics from severe sepsis/shock recognition in hourly increments
  • Excluded: Neutropenic and Immunosuppressed patients

Primary Outcome:

  • Mortality


  • Antibiotic timing from triage (6 of 11 studies):
    • ≤3 hours –> 10,208 patients –> 2,574 died
    • >3 hours –> 5,970 patients –> 1,793 died
    • Pooled OR 1.16 (95% CI, 0.92 – 1.46; p = 0.21)
  • Antibiotic timing from shock/severe sepsis recognition (8 of 11 studies):
    • ≤1 hour –> 335 patients –> 1,174 died
    • > 1hour –> 7,682 patients –> 3,581 died
    • Pooled OR 1.46 (95% CI, 0.89 – 2.40; p = 0.13)
  • Sensitivity Analysis of the Effect of time to Antibiotics from severe sepsis/shock recognition
    • <1 hour –> 2,318 patients –> 848 deaths
    • 1 – 2 hours –> 1,298 patients –> 471 deaths
    • 2 – 3 hours –> 853 patients –> 323 deaths
    • 3 – 4 hours –> 615 patients –> 245 deaths
    • 4 – 5 hours –> 453 patients –> 193 deaths
    • > 5 hours –> 2,386 patients –>1,537 deaths
    • No statistically significant increase in the pooled ORs for each hourly incremental delay in antibiotic administration


  • Utilized a scoring system to determine study quality
  • All studies included were considered moderate to high quality


  • As with any meta-analysis the pooled results are only as good as the individual studies included in the analysis. Remember crap in = crap out!!
  • The study did not include any randomized controlled trials (highest level of evidence) as none have ever been performed.
  • Several studies were excluded, as they did not contain all the data necessary to evaluate time of antibiotic administration on mortality. The authors did attempt to contact original authors three different times, but only received responses from half the authors. Having these results, could potential alter the results of this study
  • Studies included were not limited to studies of appropriate or effective antibiotic use. In other words, included patients may have received inappropriate antibiotics or had resistant organisms, which could also alter results of this study.


  • It is obvious that failure to administer effective broad-spectrum antibiotics will be detrimental to patient outcomes, but the exact time when this occurs is still not known, most likely because sepsis has a complex pathophysiology that has a spectrum of severity as opposed to actual categories of disease.
  • SIRS criteria was created in 1992 as a consensus statement from the Society of Critical Care Medicine and American College of Chest Physicians. Using SIRS criteria alone misclassifies 23.4% of patients [4] and 15% of patients will not meet definitions of severe sepsis and septic shock until after 3 hours from ED arrival [5]. If this holds true then is time to antibiotics from triage a reliable quality metric?
  • It is also important to state that, this study is not saying early antibiotic administration is not important in severe sepsis and septic shock, but instead that this arbitrary, non-evidence based time to antibiotic administration time of 3 hours (triage) or 1 hour (severe sepsis/shock) should not be a metric for quality of care.

Comments from Twitter:

Screen Shot 2015-09-22 at 1.28.40 PM

Tweet and image from Michael Allison (@doc_mga)

Clinical Take Home Point: There is no data driven or evidence based research that demonstrates time to antibiotic administration in severe sepsis/septic shock as a reliable quality metric. Sepsis is a heterogeneous spectrum of illness and as such, one size does not fit all.


  1. Sterling SA et al. The Impact of Timing of Antibiotics on Outcomes in Severe Sepsis and Septic Shock: A Systematic Review and Meta-Analysis. Critical Care Medicine 2015; 43(9): 1907 – 15. PMID: 26121073
  2. Dellinger RP et al. Surviving Sepsis Campaign: International Guidelines for Management of Severe Sepsis and Septic Shock: 2012. Crit care Med 2013; 41(2): 580 – 637. PMID: 23353941
  3. Kumar A et al. Duration of Hypotension Before Initiation of Effective Antimicrobial Therapy is the Critical Determinant of Survival in Human Septic Shock. Crit Care Med 2006; 34: 1589 – 96. PMID: 16625125
  4. Venkatesh et al. Time to Antibiotics for Septic Shock: Evaluating a Proposed Performance Measure. Am J Emerg Med 2013; 31: 680 – 683. PMID: 23380106
  5. Villar et al. Many Emergency Department Patients with Severe Sepsis and Septic Shock do not Meet Diagnostic Crteria Within 3 Hours of Arrival. Ann Emerg Med 2014; 64: 48 -54. PMID: 24680548

For More Thoughts on This Checkout:

Post Peer Reviewed By: Anand Swaminathan (Twitter: @EMSwami)

Cite this article as: Salim Rezaie, "Time to Antibiotics in Sepsis: A Metric Not Supported by “High Quality” Evidence", REBEL EM blog, September 21, 2015. Available at: https://rebelem.com/time-to-antibiotics-in-sepsis-a-metric-not-supported-by-evidence/.
The following two tabs change content below.

Salim Rezaie

Emergency Physician at Greater San Antonio Emergency Physicians (GSEP)
Creator & Founder of REBEL EM

Like this article?

Share on facebook
Share on Facebook
Share on twitter
Share on Twitter
Share on linkedin
Share on Linkdin
Share on email
Share via Email

Want to support rebelem?

8 thoughts on “Time to Antibiotics in Sepsis: A Metric Not Supported by “High Quality” Evidence”

  1. A major weakness of this meta-analysis is that they seem to have used – from what I can tell – unadjusted odds ratios. It is likely that patients who get early antibiotics are sicker (hence prompting their healthcare providers to act sooner), so are predisposed to having worse outcomes; this is an example of confounding by indication. In observational studies of treatment efficacy, adjusting for disease severity is *really* important for this reason. Compare for example the Ferrer et al. paper (ref 8 in the meta-analysis), http://www.ncbi.nlm.nih.gov/pubmed/24717459, the biggest study included in the analysis. Table 2 of Ferrer shows the adjusted OR for mortality with each hour of treatment delay, showing a clear benefit of early antibiotics. In table 5 of the meta-analysis, however, it looks like unadjusted ORs are used, now showing no benefit. This use of unadjusted measures makes this meta-analysis, to my eyes, fundamentally flawed.

    • Hello Robert,
      Really appreciate you taking the time to read and even more to leave your comments and thoughts. One of the major issues I am trying to flush out in this write up is exactly what you are saying, “adjusting for disease severity is *really* important.” I 100% agree that the sicker the patient, the earlier the APPROPRIATE empiric abx become. There are some other flaws with this study as well that are important in my eyes. For example, just because patients got antibiotics does not mean they were APPROPRIATE!!! This could also skew results. I think one of the issues with time to intervention trials is that septic patients didn’t become septic the second they hit the door of the ER, just like STEMI patients didn’t have their MI just as they hit the door. I am NOT advocating for later abx, instead what I am saying is that the sicker the patient, the earlier the abx…but the Surviving Sepsis Campaign is trying to lump all septic patients into one lump sum when the disease is truly a spectrum. ProCESS, ARISE, and ProMISe did all show us something and that is that we are so much more proactive with sepsis now and not reactive as we were before the Rivers et al study 1999.
      Again, TY for putting your eyes on the blog and leaving your thoughts. I 100% agree with them.


  2. I too decry that kpi s based on eminence rather than evidence are frustrating…
    But we all know that the absence of evidence is not evidence of absence…
    If nothing else, it gets us to look at the patient sooner… and that may be what really matters…
    Agree that ODDS should be adjusted, but i must wonder whether they (OR) should be used at all..
    At least not univariate ODDS…
    Perhaps the best and most clear way to study the time question is to look at the data
    graphically…and keep it continuous… avoid the artificial cut points, as Salim rightly suggests, using increments of 60 minutes is itself utterly “artificial”… It would be both refreshing and perhaps more enlightening to see these data presented as continuous data; pooling when not too hetero (where possible) and see what it looks like in pictures…(see Tufte on how to show data)
    Time to abx/mortality curves can be generated overall and then separately for different risk strata–(perhaps using some mixture of Age, APACHE, MAP, lactate, ScVO2, etc).
    This type of work should be done and the old, oft quoted “data” seriously challenged.
    Thanks for raising this.

    • Hello Greg,
      TY for reading and leaving your comments and thoughts. I think you bring up a great point,”the absence of evidence is not evidence of absence.” I do believe that more proactive care is important in these patients, but I am just not sure we have the right answer yet. Speaking pathophysiologically, yes earlier, appropriate abx is most likely good for patients, but as someone showed very nicely in one of their talks (I received permission from them to put their slide on the post)….if the large funded trials can’t get abx into septic patients in 3 hours or less….how is an overcrowded, resource limited, high stressed emergency department expected to do the same? I would love to see a study that avoids the artificial sepsis definitions and looks at sepsis in a spectrum of disease as you propose but also keeps in mind the reality behind long wait times and boarding issues of many emergency departments (at least in the US).


  3. Pingback: Time to Antibiotics in Sepsis: A Metric Not Supported by Evidence - R.E.B.E.L. EM - Emergency Medicine Blog - Admitologist

Leave a Comment

Time limit is exhausted. Please reload CAPTCHA.