Why the Best Antibiotic is Sometimes the One You Don’t Write

Looking back on the past month of my antimicrobial stewardship series, Lessons from the Bedside—the unnecessary antibiotics, the Gram stains that quietly changed management, the surgeons who chose drainage over drugs, and the delicate timing of HIV care—I began to notice a pattern I had not fully appreciated before. We are not just treating infections; we are often treating our own discomfort with uncertainty. And too often, antibiotics become the tool we reach for.

The call came from the ICU. A young woman had been admitted with diabetic ketoacidosis. She was intubated, with a Foley catheter in place, now on day five of admission. Her laboratory results were concerning: a white cell count of 20,000, a markedly elevated CRP, and a urine culture that had returned positive. The clinical team had already moved to the next step in their minds. “Which antibiotic should we start?” the resident asked. Instead of answering directly, I asked three simple questions: Was she febrile? Was she hypotensive? Did she have flank pain? The answer to all three was no. The recommendation was therefore equally simple: none. Treat the DKA. The Foley catheter was colonized, as they often are. The inflammatory markers reflected the metabolic storm of ketoacidosis, not an underlying infection. There was hesitation—because choosing not to act often feels riskier than acting—but she improved and recovered without receiving a single dose of antibiotics.

That case reinforced a lesson that is rarely emphasized in textbooks: the most difficult prescription is not selecting the correct antibiotic, but deciding when not to prescribe one at all. Early in my practice, I believed that good infectious diseases management meant mastering drug selection—knowing the right agent, the correct dose, and the appropriate duration for every organism. But clinical experience quickly complicated that belief. I saw patients who improved despite antibiotics, and others who deteriorated because of them. It became clear that knowledge alone is insufficient; what matters is judgment—specifically, the ability to distinguish between situations that require intervention and those that do not.

This distinction is especially critical when dealing with colonization. Patients with long-term urinary catheters, for instance, frequently have positive cultures. One such patient’s chart read like an antimicrobial rotation: levofloxacin, followed by piperacillin–tazobactam, then linezolid, and eventually meropenem. Yet the patient himself was stable—afebrile, hemodynamically normal, sitting up in bed and eating. The reluctance to stop antibiotics did not come from clinical evidence but from fear: what if he deteriorates? That question, more than any guideline, drives unnecessary prescribing. But colonization is not infection, and treating it exposes patients to harm without benefit—promoting resistance, increasing toxicity, and reinforcing a false sense of security.

The scale of this issue is significant. The WHO African Region reports the highest global burden of antimicrobial resistance, with 24 deaths per 100,000 people. In Kenya, studies suggest that nearly 40% of antibiotic prescriptions in tertiary hospitals are unnecessary. This is not simply a matter of individual error; it reflects a broader system that equates action with competence. Ordering tests, starting antibiotics, and escalating therapy are often perceived as markers of good care, while restraint is misinterpreted as inaction.

Part of this dynamic is driven by ingrained clinical reflexes. One of my mentors used to describe Staphylococcus in stark terms: it kills, it adheres, it recurs, and it spreads. While accurate, such framing can create a reflexive tendency to “cover Staph” even in the absence of supporting evidence. It is not uncommon to see vancomycin initiated for minimal clinical findings—mild erythema without systemic signs or microbiological confirmation—simply to mitigate uncertainty. In these moments, the decision is less about the patient and more about alleviating clinician anxiety.

Amid increasingly sophisticated diagnostics, it is worth remembering the enduring value of simple tools. The Gram stain, often overlooked, remains highly informative. In one ICU case, a patient developed fever and worsening oxygen requirements on day six of admission, raising suspicion for ventilator-associated pneumonia. Cefepime had already been initiated. When the laboratory reported gram-positive cocci in clusters, the implication was clear: likely Staphylococcus. Vancomycin was added—not as part of broad escalation, but as a targeted response. No carbapenems were introduced. The patient improved. In this case, a basic microbiological observation provided sufficient direction, avoiding unnecessary expansion of therapy.

Equally important is the principle of source control. Antibiotics are limited by physiology; they do not effectively penetrate poorly perfused collections such as abscesses. These environments, characterized by high bacterial density and limited immune access, render antimicrobial therapy alone insufficient. Escalating antibiotics in such contexts is ineffective. Drainage is not an adjunct to treatment—it is the treatment. Without it, antibiotics are unlikely to succeed.

The importance of timing is further illustrated in HIV care. In patients with advanced disease, multiple opportunistic infections often present simultaneously. The instinct to initiate antiretroviral therapy immediately is understandable but can be harmful in certain conditions, such as cryptococcal meningitis. Early initiation of ART in this context can precipitate immune reconstitution inflammatory syndrome, leading to clinical deterioration and increased mortality. Evidence demonstrates that delaying ART initiation by several weeks after antifungal therapy significantly improves outcomes. Here, restraint is not a passive decision but a deliberate and evidence-based intervention.

Developing this level of clinical judgment requires reflection. I maintain a record not of successes, but of decisions that warrant reconsideration—cases where antibiotics were used unnecessarily, where laboratory results were overinterpreted, or where uncertainty drove intervention. This process is uncomfortable but essential. Improvement in clinical practice does not come from being correct, but from understanding the reasons behind incorrect decisions.

One patient, an elderly farmer with a long-term urinary catheter, exemplified this approach. Over the course of a year, he had been treated for seven presumed urinary tract infections, receiving seven courses of antibiotics. Despite this, he remained clinically stable. The decision was made to discontinue antibiotics and remove the catheter. Over the following six months, he required no further treatment. His improvement did not come from additional therapy, but from removing the underlying source of the problem.

Modern healthcare systems often incentivize intervention. There is little consequence for prescribing additional antibiotics, but significant perceived risk in withholding them. As a result, clinicians are conditioned to act, even when action is not indicated. However, patients do not always benefit from more intervention. In many cases, they require careful assessment, targeted management, and, importantly, the discipline to avoid unnecessary treatment.

Ultimately, antibiotics are highly effective when used appropriately. However, they do not address all clinical abnormalities. They do not treat anxiety, colonization, or isolated laboratory findings. They treat infection. Distinguishing between these scenarios is one of the most challenging aspects of clinical practice. It requires not only knowledge, but the confidence to apply restraint in the face of uncertainty.

Anyone can prescribe an antibiotic. That is straightforward and often reassuring. The greater challenge lies in recognizing when not to. Restraint is not inaction; it is the application of expertise under pressure. And in many cases, the best antibiotic decision is the one that is never written.