Crowdsourcing Critical Care: When Do You Stop Antibiotics?
Tapping the wisdom of the crowd (yours) to tackle a thorny clinical problem
This is the first of a new experimental series on PulmCCM called “Crowdsourcing Critical Care.” Please share in the comments your personal approach to this frequent clinical question, or email editors@pulmccm.org. Answers will be analyzed using a qualitative research methodology and shared back to the community at a later date. Thank you for sharing your experience and advice! -Ed.
It was recently argued here that a physician’s duty to her individual patient outweighs any imagined duty to “society” to such an overwhelming degree that it renders any societal duty effectively irrelevant to clinical decision-making.
Should intensivists prescribe fewer antibiotics?
Antibiotics are easier to start than stop, especially in the ICU. Critically ill patients have a high prevalence of infection on presentation, which is usually impossible to identify or rule out definitively at first.
Physicians (and medicine itself) function best when the healer has clarity in his role as a fiduciary for the patient. If any tension arises between the interests of the patient and the collective, he puts the patient’s interests first, suppressing any cognitive dissonance or moral distress that arise.
In the context of very high use of broad-spectrum antibiotics in the ICU, this can place treating physicians in tension with the loose confederation of agents, policies, and other professional influences collectively termed antibiotic stewardship.
“Stewardship” is a good name for these efforts’ worthy purpose, which is to conserve the effectiveness of antibiotics by ensuring they are used in the right spectrum, at the right doses, for only the appropriate amount of time that they are beneficial to the patient.
In practice, though, antibiotic stewardship efforts can fail to meaningfully integrate that last and most important criterion (“…beneficial to the patient”), because the agent, policy, or algorithm lacks the information, context, training, or fear of a bad outcome to fully consider the balance of risks for that particular patient.
Generally speaking, stewardship relies on default heuristics (e.g., culture negative at 48 hours —> stop antibiotics; MRSA nares swab PCR negative —> stop vancomycin). These can be very useful and almost always result in shorter courses of antibiotics with net neutral (or slight benefit) to the individual patient.
Heuristic-driven stewardship can result in literal tons of fewer antibiotics used collectively, generating less selective pressure for resistance, which really is good for “society” (i.e., future individual patients, possibly including you and me).
But theoretically, if indiscriminately applied, it could also result in occasional undertreated or late-treated infections and poor outcomes, including death.
Although working within an interdisciplinary team, the treating physician is usually in the best position to consider the totality of the patient’s context, and bears the greatest responsibility for antibiotic decisions and their consequences. This means selecting the patients for whom the risks of stopping antibiotics might outweigh the benefits of stewardship.
“Might,” because it’s usually no more than an intuition grounded in experience and informed by a smattering of often-conflicting physiologic and lab data. There’s virtually no evidence available to guide these decisions.
Patient-Centered Antibiotic Stewardship Strategies
Intensivists manage the tension between clinical uncertainty and pressure for stewardship in a variety of ways.
Antibiotic “Time-Outs”
As two-thirds of ICU patients are receiving empiric antibiotics for suspected infections on or soon after arrival, it’s a common practice to reassess the indications after 48 hours. Some health systems incorporate a formal process for this review.
This may be the most high-yield and high-risk process. Patients with a clear diagnosis and low likelihood of infection should have their antibiotics stopped, but many critically ill patients haven’t “declared themselves” as infected or not in the first two days.
Procalcitonin (and other biomarkers)
Randomized trials using procalcitonin to guide antibiotic discontinuation in ICU patients show a trend toward lower mortality in the intervention arms (with reduced antibiotic courses by ~1-2 days), suggesting safety in submitting to a structured “de-escalation” strategy, broadly speaking.
Is procalcitonin "safe" to guide antibiotic use in patients with sepsis?
Many randomized trials have tested the biomarker procalcitonin as a guide to de-escalate or stop antibiotic therapy in patients with known or suspected infection. A large proportion have concluded PCT is a safe and effective method to shorten antibiotic courses, including in patients with sepsis.
However, these trials virtually all allowed clinicians to deviate from protocols to treat worrisome patients for longer courses—and to varying degrees, they did.
Thus, the trials are better described as testing “clinician judgment influenced by procalcitonin,” rather than the performance of procalcitonin itself as a discriminator.
In fact, the estimated ~5% absolute mortality reduction when using procalcitonin can be viewed as a signal for unmeasured bias in these trials, as it is exceptionally unlikely that 1-2 fewer days of antibiotics save the lives of one in 20 patients thus treated.
Different schemes have been used in RCTs, incorporating absolute thresholds for procalcitonin (e.g., <0.5 µg/L —> encourage stopping antibiotics) and relative change (e.g., decrease by ≥80% of peak level—>encourage stopping).
Although C-reactive protein has been suggested to identify patients with severe community-acquired pneumonia most likely to benefit from systemic corticosteroids, evidence has not yet suggested CRP is useful to guide antibiotic discontinuation.
Numerous other biomarkers are under development or at various stages of marketing as “sepsis tests”, but none can be recommended as a reliable tool to guide antibiotic use (either initiating or discontinuing).
Narrowing / Targeting Antibiotics to Isolated Pathogens
When a pathogen is identified that is clearly causing an infection, and its sensitivities are learned, broad-spectrum antibiotics can be switched to a narrower-spectrum agent covering that organism.
This is appropriate and works beautifully if that’s the only pathogen causing an infection.
Stopping Vancomycin for Negative MRSA Screens
Studies generally show that the negative predictive value of a negative PCR nasal swab for infection by methicillin-resistant Staphylococcus aureus in another organ system is high (95-98%).
These impressive figures are highly dependent on the prevalence of MRSA colonization in the studied population, however.
As an example, in the largest meta-analysis (Parente et al Clin Infect Dis 2018), the negative predictive value of MRSA nasal PCR was 98% for MRSA community-acquired pneumonia, but only 95% for ventilator-associated pneumonia with MRSA. That’s an uncomfortable ~5% false negative rate for MRSA screening among all patients with VAP, with a higher false negative rate expected in those with more MRSA risk factors (e.g. nursing home and dialysis patients, prolonged ICU stays, etc).
Pneumonia is the best-validated illness for an MRSA-swab rule-out; the data for other conditions is less robust. The negative predictive value of a negative MRSA screen for skin and soft tissue infections may be only ~80% (a ~20% false negative rate).
Arbitrarily Selecting Short Treatment Courses
In patients whose suspicion of infection results in treatment with antibiotics past the 48-hour “time out,” but still have neither positive cultures nor clear improvement, an arbitrary treatment course may be selected (e.g., 5 days).
Recommendations for such courses are beyond the scope of this review—or any review, probably.
Increasing evidence from RCTs has suggested that shorter courses of antibiotics may be equally effective for patients with a variety of infections:
Gram-negative bacteremia (7 days instead of 14)
Community-acquired pneumonia (5 days instead of 10-14)
Complicated UTI/pyelonephritis (5-7 days instead of 10-14)
Intra-abdominal infection (4-8 days instead of 10-14)
Very few studies have examined ICU populations specifically. Short courses for equivocal VAP in the ICU were found to be adequate in one small, probably underpowered study—way back in 1999.
This may give some confidence in reducing empiric antibiotic courses, as well. Again, one should assume that in all these studies, patients who “looked infected” in the shorter-course arm were nevertheless getting longer courses, off-protocol.
Outsourcing Uncertainty to ID
Infectious disease consultation can result in an illusion of definitive management, with unpredictable results.
At some hospitals in the community, ID physicians routinely treat ICU patients with extended periods of antibiotics beyond which even the most risk-averse intensivist would choose.
Meanwhile, the occasional academic ID physician still confidently advocates in flagship journals to withhold antibiotics from patients with suspected infection and sepsis because “the data [for empiric antibiotics] are equivocal for patients who have sepsis without shock” and also don’t worry, because if they get sicker, it will increase the diagnostic yield of cultures.
Is it better for your patient to outsource the discomfort of uncertainty to either of these consultants?
A Proposed Framework
Here is an arbitrary approach that considers all of the above:
Patients with suspected or possible infection who are unstable receive broad antibiotics on admission to the ICU.
All patients on antibiotics receive a diligent workup for infection (cultures +/- imaging).
After 48 hours, all patients’ antibiotic courses are re-considered. In those with a clear alternative diagnosis and low suspicion of infection, antibiotics are stopped.
In those with a negative MRSA nasal screen, consideration is given to stopping vancomycin (if provided).
In those whose antibiotics are continued, a differential of possible infection sources is made (respiratory, GI, etc) and an intended treatment course is set (e.g., 5 days).
Procalcitonin and other appropriate studies (e.g., chest films) are obtained on an interval basis. If suspicion of infection declines sufficiently, antibiotics are stopped before the originally intended course.
The Case for Being an “Early Antibiotic Hawk”
Early brief administration of broad-spectrum antibiotics (with or without concomitant nonabsorbable antibiotics instilled into the GI tract) appears to reduce the later development of nosocomial infections and possibly antibiotic resistance. This may occur by reducing the load of bacteria in the gut by orders of magnitude, diminishing the population exposed to selective pressures.
“Gut decontamination” with antibiotics is not a standard practice, nor is citing it here an endorsement. But it does illuminate how antibiotics, bacteria, and the microbiome will tend to ignore (or transcend) the logic of clinical reasoning and antibiotic stewardship policies.
What prevents antibiotic resistance? Antibiotics
Patients receiving mechanical ventilation are particularly vulnerable to hospital-acquired infections, especially by aerobic gram-negative bacteria and yeasts. These pathogens incubate in the stomach and gain resistance after exposure to intermittent antibiotics.
Of course, it’s also good to be hawkish on stopping antibiotics as soon as safe and appropriate, for our current patients (to prevent adverse effects) and to preserve antibiotic efficacy for those in the future (including, quite possibly, ourselves and those we care about).
Your Turn
How do you decide to narrow and stop antibiotics in the ICU?
This is the first of a new experimental series on PulmCCM called “Crowdsourcing Critical Care.”
Please share in the comments your personal approach to this frequent clinical question, or email editors@pulmccm.org.
Answers will be analyzed using a qualitative research methodology and shared back to the community at a later date.
Thank you for sharing your experience and advice!
References
Jantzen A, Woolever N, Treu M, Stakston J, Cai S, Tempelis J, Kujak RC, Dierkhising RA, Dababneh AS, Lessard S. Impact of Methicillin-Resistant Staphylococcus aureus Nasal Polymerase Chain Reaction Screening Tests on Duration of Vancomycin Therapy for Skin and Soft Tissue Infections. Hosp Pharm. 2025 Feb;60(1):83-89. doi: 10.1177/00185787241289281. Epub 2024 Oct 17. PMID: 39544829; PMCID: PMC11559747.
I am so glad that this post essentially reaffirms my practice. Especially the fact that consultants e.g ID at my specific hospital are notorious for extended antibiotic use. For example, a patient with pulmonary edema due to severe mitral valve regurgitation requiring emergent transfer to the intensive care unit remains on prolonged antibiotic therapy due to a chest x-ray that has been read as bilateral pulmonary infiltrates. as intensivist we are recording the movie, consultants are only taking snapshots.
As a nocturnist, I’m the one doing the starting but not the stopping of antibiotics. I work in a community academic center. One of the arguments I make frequently to residents is “what is broad spectrum?” I’m an advocate for narrowest targeted treatment.
— All antibiotics get end dates (on the shorter side) when ordered based on infection site.
— MRSA coverage only if risks or after a positive screen if no risks.
— If this is a community-acquired infection, no need for pseudomonas or ESBL coverage unless patient history of the same.
— If diagnosis of infection is unclear, I’ll send a procalcitonin, although it won’t come back for 3 days, to help my colleagues feel more comfortable de-escalating.
All of this goes out the window for a patient I am “throwing the kitchen sink at” because they are incredibly unstable. I find my daytime rounding colleagues rarely escalate further, and if they do it is usually when the patient is not improving quickly, which is reasonable.