Antimicrobial stewardship in the emergency department — right drug, dose, route and duration, syndromic empirical therapy, de-escalation and resistance

The emergency department is the largest empirical-prescribing environment in the hospital, and the antibiotics started at the front door set the trajectory for the entire admission — and increasingly for the patient's lifetime of microbiome and resistance risk. Antimicrobial stewardship is the coordinated set of interventions designed to optimise antimicrobial use so that every patient receives the right drug, at the right dose, by the right route, for the right duration, with de-escalation when culture and susceptibility data return. The Fellowship candidate is expected to apply the four Ds at the bedside, to deliver empirical therapy within the Surviving Sepsis Campaign Hour-1 window, to choose a syndromic regimen that matches the likely organism and the local resistance pattern, to dose for the patient's weight and renal function and for the pharmacokinetic and pharmacodynamic behaviour of the drug, and to de-escalate when the laboratory returns. The global burden of antimicrobial resistance is now beyond reasonable dispute: the Antimicrobial Resistance Collaborators estimated 4.95 million deaths associated with bacterial resistance in 2019 and 1.27 million deaths attributable to it, surpassing the global toll of HIV and malaria.[16] The emergency department cannot solve resistance, but it can stop feeding it.

Definition and scope — what stewardship is, and why the emergency department is the leverage point

Antimicrobial stewardship is the discipline of using antimicrobials only when they are needed, and using them well when they are. The Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America define an institutional stewardship programme around accountability, drug expertise, and a set of interventions — pre-authorisation, prospective audit and feedback, dose optimisation, intravenous-to-oral switch, and the avoidance of redundant or duplicate therapy — delivered by a multidisciplinary team led by an infectious diseases physician and a clinical pharmacist.[15] In the emergency department, the programme translates into a culture: the right empirical choice at the moment of presentation, the culture sent before the dose when it can be, the dose matched to weight and renal function, the duration planned at the moment of prescribing, and the handover that carries the de-escalation plan forward.

The emergency department matters because it is where empirical therapy begins. Between 30 and 50 per cent of antibiotics prescribed in the emergency department are inappropriate — either unnecessary, too broad, too narrow, wrongly dosed, or wrongly prolonged — and the consequences land both on the individual (Clostridioides difficile infection, adverse drug events, lengthened stay) and on the population (selection of resistance).[15][17] Every antibiotic started at triage is also a decision the ward inherits, because the empirical regimen tends to be continued long after the cultures return, and the duration set at the front door is the duration the patient finishes. Good emergency department stewardship is therefore an upstream intervention with downstream leverage: a single empirical choice, made well, can prevent days of unnecessary therapy and a cascading series of complications.

The framework — the four Ds plus route

The candidate's mental scaffold for every empirical prescription is the four Ds — Drug, Dose, Duration and De-escalation — with Route as the operational fifth, and Allergy as the safety check. Each D is a discrete decision the prescriber makes consciously, never by default. [1]

Drug is chosen by matching the syndrome to the likely organism, the likely organism to the local antibiogram, and the antibiogram to the patient's allergies and comorbidities. A community-acquired pneumonia in a previously well adult is almost always a pneumococcus, a haemophilus or an atypical bacterium, and a beta-lactam plus a macrolide covers them; a nursing-home pneumonia in a patient with a recent urinary catheter raises the spectre of an ESBL-producing Enterobacterales, and ceftriaxone alone may fail. Dose is chosen by weight, renal and hepatic function, the severity of the infection, and the pharmacokinetic and pharmacodynamic behaviour of the drug. Duration is planned at the moment of prescribing, not deferred to the ward, and is set to the shortest effective course the evidence supports. De-escalation is the active narrowing of therapy when culture and susceptibility data return, or the cessation of therapy when a non-infective diagnosis is confirmed. Route is the choice between intravenous and oral — intravenous for the septic, the haemodynamically unstable, the patient with a poorly-absorbed oral option or with no functioning gut, oral for the stable patient in whom the chosen agent has bioavailability above 90 per cent. Allergy is the safety net that turns a routine prescription into a preventable anaphylaxis if it is not asked. [1]

Cultures before antibiotics — the timing rule, and why it has limits

The single most important procedural rule in emergency department stewardship is to send cultures before the antibiotic dose — but never to delay the antibiotic beyond one hour to do so. Kumar and colleagues' cohort study of septic shock established that the duration of hypotension before effective antimicrobial therapy is the critical determinant of survival, with mortality rising approximately 7.6 percentage points for every hour of delay in the first six hours.[1] Seymour and colleagues' evaluation of the New York State mandated sepsis care confirmed the finding at scale: among patients with sepsis and septic shock, initiation of antibiotics within one hour was associated with lower mortality, and the survival benefit was greatest in the most severely ill.[2] The 2021 Surviving Sepsis Campaign guideline operationalises the rule: antibiotics within one hour for sepsis or septic shock, with two blood cultures from separate sites sent before the dose when feasible without delaying therapy.[3]

The candidate must reconcile two truths that seem to contradict. First, cultures drawn after antibiotics are far less likely to grow the organism — blood culture yield falls by half within an hour of antibiotics, and a sterile culture forecloses the chance to de-escalate. Second, in septic shock, every minute spent on cultures is a minute of untreated hypotension, and untreated hypotension kills. The reconciliation is a hierarchy: in suspected septic shock or high-likelihood sepsis, draw blood cultures rapidly — two separate sites, two bottles each, before the antibiotic — but if the one-hour window is at risk, give the antibiotic and draw what cultures you can. Source control is the parallel and equally time-critical imperative — Reitz and colleagues linked every additional hour to source control in sepsis to higher 90-day mortality, and the debridement, drainage or device removal must accompany, not follow, the antibiotic.[4]

Syndromic empirical therapy — the emergency department matrix

The empirical choice is made by syndrome, because at the moment of presentation the organism is unknown and the culture is hours to days away. The matrix below is the Australasian and IDSA-aligned consensus for the immunocompetent adult; each line is a starting point, to be adjusted for severity, allergy, pregnancy, renal and hepatic function, age, recent healthcare exposure, recent antibiotics, and the local antibiogram. [1]

Pharmacokinetic and pharmacodynamic dosing — why the dose matters as much as the drug

The dose is not a formality — it is the variable that determines whether the drug reaches the site of infection at a concentration that kills the organism without poisoning the patient. The governing principle is the pharmacokinetic and pharmacodynamic index of the drug class, and the candidate is expected to know the three patterns and the agents that follow each. [1]

Time-dependent killing is the pattern of the beta-lactams (penicillins, cephalosporins, carbapenems), the macrolides and the lincosamides. The killing is proportional to the proportion of the dosing interval during which the free drug concentration exceeds the minimum inhibitory concentration of the organism — written as the fT greater than MIC. The clinical implication is that beta-lactams are more effective when given by prolonged or continuous infusion in severe sepsis, and that missed or delayed doses erode efficacy. Concentration-dependent killing is the pattern of the aminoglycosides, the fluoroquinolones, the polymyxins and the daptomycin. The killing is proportional to the peak concentration relative to the MIC — written as the Cmax over MIC — and the clinical implication is that high once-daily dosing (gentamicin 5 to 7 mg per kilogram once daily) is both more effective and less nephrotoxic than divided dosing. Concentration-dependent killing with time-dependence is the pattern of vancomycin and the glycopeptides, and the index is the area under the concentration–time curve divided by the MIC — the AUC over MIC — with the consensus target for serious MRSA infection set at 400 to 600 (the 2020 consensus guideline replaced trough-only monitoring with AUC-guided dosing).[10]

Augmented renal clearance is the sepsis-specific trap. The hyperdynamic state of early sepsis, especially in the younger, the previously well and the patient with normal or supranormal renal function, accelerates the renal clearance of hydrophilic antibiotics (beta-lactams, aminoglycosides, vancomycin, linezolid) and produces subtherapeutic plasma concentrations at standard doses. The candidate should consider higher doses, prolonged infusions, or therapeutic drug monitoring for the critically ill patient with augmented renal clearance — the standard dose of piperacillin–tazobactam may be inadequate for the young septic patient with a creatinine clearance above 130 mL per minute. The opposite trap is the patient with unrecognised renal impairment — the gentamicin dose that is therapeutic in normal renal function is nephrotoxic in a clearance below 30 mL per minute — which is why the dose is always matched to a measured or estimated creatinine clearance and the trough is monitored. [1]

De-escalation — when the cultures return

De-escalation is the active narrowing of empirical therapy when the pathogen and its susceptibilities are identified, and the cessation of therapy when the working diagnosis of infection is overturned. It is the single most powerful stewardship intervention available to the inpatient team, and the emergency department sets it up by sending the right cultures at the right time and by writing the de-escalation plan into the chart at the moment of prescribing. The de-escalation conversation at 48 to 72 hours — the antibiotic time-out — is where the empirical regimen is reviewed against the culture, the clinical response and the working diagnosis. [1]

Three de-escalation moves are recognised. Narrow to the narrowest-spectrum agent active against the identified organism: an E. coli susceptible to trimethoprim in the urine is treated with trimethoprim, not ceftriaxone; a methicillin-susceptible Staphylococcus aureus in the blood is treated with flucloxacillin (cephalothin equivalent), not vancomycin, because the beta-lactam kills faster and clears bacteraemia sooner. Stop when a non-infective diagnosis is confirmed — the chest X-ray that showed pulmonary oedema, not pneumonia; the abdominal pain that was biliary colic, not cholecystitis; the confusion that was hyponatraemia, not encephalitis. Switch from intravenous to oral when the patient is haemodynamically stable, afebrile for 24 to 48 hours, has a functioning gastrointestinal tract, has an oral agent with bioavailability greater than 90 per cent available, and is improving clinically. The intravenous-to-oral switch shortens stay, reduces line-associated infection and reduces cost without harming outcomes; the agents with high enough bioavailability are the fluoroquinolones, doxycycline, trimethoprim–sulfamethoxazole, linezolid, metronidazole and fluconazole. [1]

Differential — when empirical therapy appears to fail

Empirical therapy that fails at 48 hours demands a structured differential, because the failure is rarely solved by simply broadening the spectrum — a reflexive escalation that hides the real cause and amplifies resistance selection. The candidate must work through five families of cause, because each has a different remedy. [1]

Resistance patterns — what they mean for the empirical choice

The four resistance patterns the candidate must know — MRSA, VRE, ESBL and CPE — are not abstractions; each translates into a specific empirical and definitive choice, and each is the answer to the question "why did ceftriaxone fail?". [1]

Methicillin-resistant Staphylococcus aureus (MRSA) is staphylococcus that has acquired the mecA gene, altering its penicillin-binding protein so that all beta-lactams (penicillins, cephalosporins, carbapenems) fail. It is community-acquired (skin and soft tissue, post-influenza pneumonia, septic arthritis in the young) and nosocomial (line, surgical, ventilator-associated). Treatment is vancomycin (AUC-guided, loading 25 to 30 mg per kilogram for serious infection), or — by syndrome — clindamycin, linezolid, daptomycin, ceftaroline, or trimethoprim–sulfamethoxazole. The Liu 2011 IDSA MRSA guideline remains the reference standard.[8]

Vancomycin-resistant enterococcus (VRE) is almost always Enterococcus faecium that has acquired the vanA or vanB gene, altering its cell-wall precursor so that vancomycin fails. It is nosocomial, often line-associated, and problematic in bacteraemia, endocarditis and intra-abdominal infection. Treatment is linezolid or daptomycin (high dose, 8 to 10 mg per kilogram); daptomycin is inactivated by pulmonary surfactant and cannot treat pneumonia. [1]

Extended-spectrum beta-lactamase-producing Enterobacterales (ESBL) are E. coli, Klebsiella and Proteus that have acquired a beta-lactamase hydrolysing the third-generation cephalosporins (ceftriaxone, cefotaxime, ceftazidime) but not the carbapenems. They are increasingly community-acquired, especially in urinary tract infection, and pose the inoculum-effect problem — a ceftriaxone-susceptible isolate in vitro may fail in vivo at the high inoculum of severe infection. The Tamma 2024 IDSA guidance recommends a carbapenem for severe ESBL infection (pyelonephritis, bacteraemia, intra-abdominal), and nitrofurantoin, trimethoprim–sulfamethoxazole, ciprofloxacin, or oral fosfomycin for cystitis if susceptible.[9]

Carbapenemase-producing Enterobacterales (CPE, also termed CRE) are Enterobacterales that have acquired a carbapenemase (KPC, NDM, OXA-48, VIM, IMP) and are resistant to the carbapenems as well as the third-generation cephalosporins. They are nosocomial, often in the heavily healthcare-exposed patient, and carry mortality in bacteraemia in excess of 40 per cent. Treatment is with the novel beta-lactam–beta-lactamase inhibitor combinations — ceftazidime–avibactam, meropenem–vaborbactam, imipenem–cilastatin–relebactam — or, where these fail, colistin, tigecycline, or fosfomycin, in mandatory consultation with infectious diseases.[9]

Special populations — the elderly, the neutropenic, the pregnant, the allergic, the renally impaired

The empirical choice is modified by the patient in front of the candidate, and four populations demand specific consideration at the front door. [1]

The elderly patient carries two stewardship burdens — inappropriate prescribing and polypharmacy — and the Beers 2023 criteria and STOPP/START version 2 are the structured filters the candidate is expected to apply.[13][14] The high-yield antimicrobial pitfalls in the elderly are several. Nitrofurantoin is inappropriate when the creatinine clearance is below 30 mL per minute (inadequate urinary concentration, pulmonary toxicity, hepatic toxicity) — both Beers and STOPP flag it. Trimethoprim–sulfamethoxazole with an ACE inhibitor, an angiotensin receptor blocker or spironolactone precipitates hyperkalaemia and acute kidney injury within days in the elderly, and it potentates warfarin, raising the INR and the bleeding risk — both lists flag it. The fluoroquinolones (ciprofloxacin, levofloxacin) raise the risk of tendinopathy and rupture, QT prolongation, aortic aneurysm dissection or rupture, C. difficile infection, delirium and dysglycaemia in the elderly, and Beers 2023 advises avoidance where an alternative exists. Long-term nitrofurantoin for urinary suppression is inappropriate (pulmonary, hepatic and neuropathic toxicity). Asymptomatic bacteriuria is over-diagnosed and over-treated in the elderly — the Nicolle 2019 IDSA guideline explicitly states that asymptomatic bacteriuria in the catheterised, the institutionalised or the confused elderly patient should not be treated, because treatment does not improve outcomes and selects resistance.[11]

The neutropenic patient (neutrophil count below 0.5 × 10⁹ per litre, or below 1.0 with an expected nadir below 0.5) is treated empirically for any fever above 38.3 °C single or 38.0 °C sustained, with an anti-pseudomonal beta-lactam — piperacillin–tazobactam 4.5 g intravenously, cefepime 2 g intravenously, or meropenem 1 g intravenously — within one hour, after blood cultures are drawn. Vancomycin is added only for suspected catheter infection, haemodynamic instability, severe mucositis, or known MRSA colonisation. The Freifeld 2011 IDSA neutropenic fever guideline remains the standard.[12]

The pregnant patient is constrained by teratogenicity and fetal toxicity. Trimethoprim–sulfamethoxazole is avoided in the first trimester (folate antagonist — neural tube defect risk) and the third trimester (kernicterus risk); nitrofurantoin is avoided in the third trimester (haemolytic anaemia in the fetus, especially with G6PD deficiency); the fluoroquinolones and tetracyclines are contraindicated throughout (cartilage and tooth/bone effects). The safe empirical agents in pregnancy are the penicillins, the cephalosporins, azithromycin, clindamycin and metronidazole. The penicillin-allergic patient is stratified by the severity of the history: a non-severe history (delayed rash, childhood intolerance, vague gastrointestinal symptoms) permits a cephalosporin, because the cross-reactivity between penicillin and the third-generation cephalosporins is below 1 per cent and the side chain, not the beta-lactam ring, drives the allergy; a severe history (anaphylaxis, angioedema, Stevens–Johnson syndrome) demands avoidance of all beta-lactams and an alternative regimen — vancomycin plus aztreonam for severe sepsis, clindamycin or doxycycline for skin and soft tissue, aztreonam for urinary tract. The renally impaired patient is dosed against the creatinine clearance — vancomycin, the aminoglycosides, the beta-lactams, the fluoroquinolones and the antifungals all require dose adjustment below specific thresholds, and the dose is reviewed against the most recent renal function, not the historical value. [1]

Common errors and pitfalls

The recurring failures are themselves examinable, because each is the failure mode the stewardship programme is designed to prevent. Treating asymptomatic bacteriuria in the catheterised, the institutionalised or the confused elderly patient — the most common inappropriate antibiotic prescription in the emergency department, and the one Nicolle 2019 explicitly condemns.[11] Continuing empirical therapy unchanged for the whole admission despite a positive culture and clinical improvement — the empirical regimen is continued long after the culture returns, and the chance to narrow is lost; the antibiotic time-out at 48 to 72 hours is the remedy. Using ceftriaxone for everything — a ceftriaxone-treated cellulitis that is MRSA fails; a ceftriaxone-treated pyelonephritis that is enterococcus fails; a ceftriaxone-treated meningitis that is Listeria fails; the empirical choice is matched to the syndrome, the patient and the resistance pattern, not defaulted. Prolonging intravenous therapy when the oral switch criteria are met — the patient who is afebrile, stable and improving on day three stays on intravenous antibiotics for the rest of the admission, with line-associated infection, longer stay and higher cost. Ignoring drug interactions — the warfarin patient who is started on ciprofloxacin, metronidazole, trimethoprim–sulfamethoxazole or a macrolide and returns with an INR above 8 and a bleed within days. Under-dosing the septic patient with augmented renal clearance — the standard beta-lactam dose that is subtherapeutic at a clearance above 130 mL per minute, and the patient who fails to improve until the infusion is prolonged or the dose raised. Over-using vancomycin for empirical MRSA cover when the local MRSA prevalence is low and the risk factors are absent — selecting VRE and C. difficile without benefit. Failing to send cultures before the antibiotic dose, so that the empirical regimen cannot be narrowed when the patient improves. Treating a colonisation rather than an infection — the urine culture that grows a coloniser in a catheterised patient, the sputum culture that grows Candida, the wound swab that grows Staphylococcus epidermidis.

Evidence and regional guidelines

The evidence base for emergency department antimicrobial stewardship rests on three pillars — the timing-and-mortality literature, the syndrome-specific empirical therapy guidelines, and the institutional stewardship programme literature. The timing literature is anchored by Kumar and colleagues' cohort study of septic shock (the duration of hypotension before effective antimicrobial therapy is the critical determinant of survival) and by Seymour and colleagues' evaluation of mandated New York sepsis care (antibiotics within one hour associated with lower mortality).[1][2] Reitz and colleagues extended the timing principle to source control.[4] The empirical therapy guidelines are the 2021 Surviving Sepsis Campaign (Evans), the 2019 ATS/IDSA community-acquired pneumonia guideline (Metlay), the 2014 IDSA skin and soft tissue infection guideline (Stevens), the 2004 IDSA bacterial meningitis guideline (Tunkel), the 2011 IDSA MRSA guideline (Liu), the 2024 IDSA guidance on ESBL-, AmpC- and carbapenemase-producing Enterobacterales (Tamma), the 2020 vancomycin consensus (Rybak), the 2019 IDSA asymptomatic bacteriuria guideline (Nicolle), and the 2011 IDSA neutropenic fever guideline (Freifeld).[3][5][6][7][8][9][10][11][12] The institutional stewardship literature is anchored by the 2016 IDSA/SHEA stewardship guideline (Barlam) and Drekonja's systematic review of outpatient stewardship.[15][17] The burden of resistance is documented by the Antimicrobial Resistance Collaborators' GRAM global burden analysis.[16]

Exam pearls

• The four Ds — Drug, Dose, Duration, De-escalation — plus Route and Allergy. Every empirical prescription is the conscious answer to all six.
• Antibiotics within one hour for sepsis and septic shock. Kumar 2006 (mortality rises 7.6 percentage points per hour of delay), Seymour 2017 (antibiotics within one hour reduce mortality at scale), Surviving Sepsis Campaign 2021 (Hour-1 bundle). Cultures before the dose when you can — never delay the dose for cultures in septic shock.
• Ceftriaxone 2 g intravenously is the workhorse empirical agent for community sepsis, urinary sepsis, community-acquired pneumonia (with azithromycin) and meningitis (with vancomycin, plus ampicillin if Listeria is a risk). It does not cover MRSA, enterococcus, ESBL-producing Enterobacterales, CPE, Listeria, atypicals, or anaerobes — know what it does not do.
• Cellulitis is flucloxacillin 2 g intravenously. Add MRSA cover (vancomycin, clindamycin, TMP-SMX) for purulent infection, severe sepsis, immunocompromise, or failed initial therapy. Drain the abscess.
• Cystitis is trimethoprim or nitrofurantoin. Avoid nitrofurantoin at a creatinine clearance below 30 mL per minute and in the third trimester; avoid trimethoprim in the first trimester. Avoid ciprofloxacin and ceftriaxone as routine empirical cystitis therapy.
• De-escalation is narrow, stop or switch. Narrow when the organism and susceptibilities are known; stop when the diagnosis is overturned; switch intravenous to oral when stable, afebrile, the gut works and an oral agent with greater than 90 per cent bioavailability is available.
• MRSA — vancomycin AUC over MIC 400 to 600 (Rybak 2020). VRE — linezolid or daptomycin (daptomycin not for pneumonia). ESBL — carbapenem for severe infection, nitrofurantoin or TMP-SMX for cystitis if susceptible (Tamma 2024). CPE — ceftazidime–avibactam, meropenem–vaborbactam, consult infectious diseases.
• The elderly patient. Avoid nitrofurantoin at a clearance below 30 mL per minute; avoid TMP-SMX with ACEi/ARB/spironolactone (hyperkalaemia, AKI) and with warfarin (INR potentiation); avoid fluoroquinolones (tendon, QT, aorta, delirium, C. difficile). Do not treat asymptomatic bacteriuria (Nicolle 2019).
• Source control is parallel, not sequential. Reitz 2022 — every hour to source control raises 90-day mortality. The antibiotic is adjunctive to the surgery or the drainage. [1]

Red flags