In critical care practice, there is often a reflexive inclination to initiate broad-spectrum antibiotics at the first sign of infection. However, ICU patients with suspected infection represent a heterogeneous population, ranging from individuals who are critically ill for noninfectious reasons and have no sepsis-related organ dysfunction, to those with sepsis and mild organ dysfunction, to patients in fulminant septic shock with multiorgan failure. Moreover, many ICU patients have noninfectious conditions—such as pancreatitis, drug reactions, postoperative inflammation, or pulmonary embolism—that can mimic infection and trigger unnecessary antimicrobial use. ^, The consequences of inappropriate empiric therapy vary accordingly.

Although judicious antibiotic use is generally preferred from a stewardship standpoint, there is little margin for error in patients with septic shock; observational studies have shown that each hour of delay in administering effective antibiotics is associated with a stepwise increase in mortality in this population. ^,,,,, By contrast, the benefits of immediate broad-spectrum therapy are less well established in patients with sepsis without shock and even more so in infection without sepsis. Practically, this means that not every febrile ICU patient—such as one with isolated fever but no clinical deterioration—warrants immediate empiric broad-spectrum antibiotics.

This distinction is essential, as indiscriminate use of broad-spectrum antibiotics promotes antimicrobial resistance and increases the risk of adverse events, including Clostridioides difficile infection, nephrotoxicity, and opportunistic superinfections. ^,, Excessive empiric therapy has also been associated with higher mortality, likely caused by these complications and disruption of the gut microbiome, which plays a key role in immune regulation. ^,,

Balancing the risks of undertreatment and overtreatment is therefore a central challenge in critical care. Because microbiologic data typically require 24 to 72 hours to return, empiric regimens must serve as a carefully selected bridge, guided by clinical presentation, severity of illness, local resistance patterns, and patient-specific risk factors for multidrug-resistant organisms.

Obtaining blood cultures before initiating antibiotics doubles the likelihood of pathogen recovery and remains the most critical microbiologic test in patients with suspected sepsis. ^,, Although there are few data directly linking blood cultures to improved outcomes, they provide essential diagnostic information to guide immediate treatment decisions and support antimicrobial stewardship. Positive blood cultures facilitate targeted therapy with narrow-spectrum agents potentially reducing adverse effects and antimicrobial resistance. New rapid molecular diagnostic tests, such as polymerase chain reaction (PCR) assays, allow for antimicrobial optimization within a matter of hours compared to days with traditional culture and susceptibility testing. In addition, resistance data from bloodstream isolates are crucial for generating hospital antibiograms and informing empiric prescribing guidelines. When feasible, other site-specific cultures (eg, respiratory, urine, wound) should also be obtained to enhance diagnostic yield and tailor therapy.

Clinicians should make every effort to obtain blood and other appropriate cultures before starting antibiotics, provided this does not meaningfully delay treatment. The Surviving Sepsis Campaign recommends obtaining routine microbiologic cultures before antimicrobial therapy in patients with suspected sepsis or septic shock, if this can be done within 45 minutes.

In patients with central venous catheters in whom a catheter-related bloodstream infection is suspected, it is important to obtain paired blood cultures—one drawn from the central line and one from a peripheral vein. This allows for assessment of differential time to positivity (TTP), which can help determine whether the catheter is the likely source of infection. A central line culture with differential TTP greater than 2 hours (ie, turns positive at least 2 hours before the peripheral culture) has high sensitivity and specificity for diagnosing catheter-related bloodstream infection.

Effective empiric therapy begins with a presumptive diagnosis based on the patient’s presenting syndrome, informed by clinical features, imaging, laboratory data, and risk factors. A syndromic approach helps narrow the list of likely pathogens and guides empiric antibiotic selection while awaiting culture results.

For example, community-acquired pneumonia is most commonly caused by Streptococcus pneumoniae , Haemophilus influenzae , and atypical organisms such as Mycoplasma pneumoniae , Chlamydia pneumoniae , and Legionella species. In contrast, hospital-acquired and ventilator-associated pneumonia are more frequently caused by Pseudomonas aeruginosa and methicillin-resistant Staphylococcus aureus (MRSA). Urinary tract infections in hospitalized patients are often caused by gram-negative bacilli such as Escherichia coli , Klebsiella pneumoniae , and Proteus species, with P aeruginosa and Enterococcus sp. more likely in patients with chronic catheters or recent antibiotic exposure. Intra-abdominal infections are typically polymicrobial, involving a combination of Gram-negative bacilli, anaerobes (eg, Bacteroides fragilis ), occasionally Enterococcus spp., and Candida spp., particularly in healthcare-associated cases. Soft tissue infections also vary by context: non-purulent cellulitis is most often due to streptococci, while abscesses or infections in patients with prior colonization or risk factors for resistance may involve MRSA. Catheter-related bloodstream infections are commonly caused by coagulase-negative staphylococci, S aureus , enteric gram-negative bacilli, and Candida species.

Identifying the likely source is also critical for anticipating antibiotic concentration at the site of infection. For examples, central nervous system infections require agents with reliable blood-brain barrier penetration (eg, ceftriaxone, cefepime, carbapenems), while pneumonia necessitates drugs with high epithelial lining fluid concentrations and excludes use of daptomycin, which is inactivated by pulmonary surfactant. Establishing a focused presumptive diagnosis is therefore the cornerstone of rational empiric antibiotic selection. Common pathogens and recommended empiric regimens for frequently encountered infectious syndromes are summarized in Table 1 .

Local antibiograms report the cumulative susceptibility of various antibiotics to different pathogens throughout an institution. Combined with antimicrobial stewardship programs, antibiograms increase appropriate antibiotic use. In the context of critically ill patients, antibiograms can assist clinicians in selecting empiric antibiotics with favorable susceptibility profiles. Although there are minimal data describing optimal susceptibility cut-offs for infections, survey results show clinicians generally consider a minimum of 90% pathogen susceptibility for severe infections. The 2016 Hospital-acquired and Ventilator-associated Pneumonia guidelines recommend dual anti-pseudomonal therapy in units with greater than 10% resistance to monotherapy options, supporting a 90% susceptibility cut-off for severe infections. Antibiograms can be as broad as an entire health care system or as narrow as a single unit in the hospital. For critically ill patients presenting from the community or transferred from another unit, general hospital antibiograms may be appropriate. However, for optimal management of new infections in patients admitted to the ICU, an antibiogram specific to this population should be used, as ICUs generally have higher resistance rates compared with other units. Antibiograms should only include species with at least 30 isolates available, which may limit usefulness for uncommon pathogens or smaller institutions.

Although antibiograms provide a valuable population-level snapshot of local resistance patterns, they cannot capture patient-specific variables that influence the risk of resistant infections. To select the most appropriate empiric regimen—especially in critically ill patients—it is essential to combine antibiogram data with an assessment of individual risk factors for multidrug-resistant (MDR) organisms. Numerous studies have evaluated risk factors for MDR organisms across different infection syndromes. ^,,,,, Although the specific predictors may vary by site of infection—for example, risk factors for MDR pneumonia may differ from those for urinary tract or bloodstream infections—several clinical features have been consistently associated with increased risk of MDR pathogens across settings. The following factors are among the most robust and commonly identified risk factors.

• •

  Recent health care exposure, including hospitalization or residence in a long-term care facility within the past 90 days

• •

  Hospital-acquired (vs community-acquired) infection

• •

  Recent antibiotic use, particularly broad-spectrum and intravenous agents, within the past 90 days

• •

  Prior colonization or infection with MDR organisms (eg, MRSA, vancomycin-resistant Enterococci [VRE], extended spectrum beta-lactamase [ESBL]-producing Enterobacterales , carbapenem-resistant Enterobacterales [CRE], MDR Pseudomonas )

• •

  Presence of indwelling medical devices, such as central venous catheters, urinary catheters, or tracheostomies

• •

  Immunosuppression, including receipt of chemotherapy, chronic systemic corticosteroids, or solid organ/stem cell transplant

• •

  Septic shock, which independently increases the risk of MDR infection and adverse outcomes

• •

  Mechanical ventilation, which increases the risk of MDR pathogens in pneumonia

• •

  Hemodialysis or other chronic organ support therapies

These risk factors should prompt clinicians to broaden empiric coverage to include resistant pathogens such as MRSA and P aeruginosa , depending on the suspected source of infection.

ESBL-producing Enterobacterales warrant special consideration given controversial efficacy from common first-line antipseudomonal β-lactams (eg, cefepime and piperacillin-tazobactam), necessitating empiric carbapenem therapy when infection is suspected. Key risk factors for ESBL infection include prior colonization or infection with ESBL organisms, recent exposure to later-generation cephalosporins (particularly when sepsis develops while on third- or fourth-generation cephalosporins), recent hospitalization or residence in a long-term care facility, indwelling urinary catheters, and recent travel to regions with high ESBL prevalence. ^, ESBL organisms most commonly cause urinary tract infections, but can also contribute to intra-abdominal infections, bacteremia, and pneumonia in critically ill patients.

Empiric antibiotic selection must also be informed by local and regional resistance patterns, which can vary widely by geography, care setting, and health care infrastructure. For example, resistant gram-negative organisms such as ESBL-producing Enterobacterales , CRE, and Acinetobacter baumannii are relatively uncommon in North America and Northern Europe but are endemic in parts of South Asia and Southeast Asia, Latin America, and the Middle East, where broader empiric coverage—including carbapenems or novel agents—may be warranted, particularly in patients with recent health care exposure. ^,,, MRSA prevalence also varies globally, influencing the need for empiric coverage in pneumonia or skin/soft tissue infections. Within hospitals, resistance rates tend to be highest in ICUs, making ICU-specific or source-specific antibiograms more useful than hospital-wide data. In resource-limited settings, broader empiric therapy may be necessary because of limited diagnostics or surveillance, while institutions with strong stewardship programs may support narrower initial therapy. Ultimately, general guidelines must be adapted to local epidemiology to ensure effective and judicious empiric antibiotic use.

Beta-lactams (BLs) are the backbone of empiric therapy in hospitalized and critically ill patients because of their broad-spectrum activity, favorable safety profile, and bactericidal properties. Anti-pseudomonal BLs such as piperacillin-tazobactam, cefepime, and carbapenems (imipenem/cilastin, meropenem) are commonly used for serious infections like sepsis, pneumonia, and intra-abdominal infections, as they provide reliable coverage against a wide range of gram-negative pathogens, including P aeruginosa and many gram-positive organisms. It is important to note that while ceftazidime is also anti-pseudomonal, it lacks clinically meaningful gram-positive coverage, making it unsuitable for empiric monotherapy when gram-positive pathogens are a concern. The broad-spectrum efficacy and clinical familiarity of most anti-pseudomonal BLs make them essential components of initial empiric regimens in acutely ill patients.

The choice between piperacillin-tazobactam and cefepime for empiric therapy—particularly in combination with vancomycin—has been the focus of ongoing debate. Early observational studies suggested that vancomycin plus piperacillin-tazobactam was associated with a higher risk of acute kidney injury (AKI) compared with vancomycin plus cefepime, likely because of synergistic nephrotoxicity. ^, However, the recent ACORN randomized controlled trial, which enrolled 2511 hospitalized patients started on anti-pseudomonal antibiotics (77% also receiving vancomycin) within 12 hours of presentation, found no significant difference in kidney outcomes between the 2 regimens, challenging prior observational findings. Notably, the median duration of vancomycin therapy in ACORN was only 2 days, leaving uncertainty about nephrotoxic risks with longer courses. Additionally, an updated nationwide analysis of 8427 patients with septic shock also showed no difference in AKI rates with piperacillin-tazobactam or cefepime. A subsequent large observational study using instrumental variable analysis reported higher mortality in sepsis patients treated with piperacillin-tazobactam versus cefepime, potentially because of the adverse effects of unnecessary anti-anaerobic therapy. Yet these findings have been questioned by an analysis attributing the results to collider bias. Balancing this, cefepime is associated with a risk of neurotoxicity—including encephalopathy and seizures—particularly when dosed inappropriately and in older adults, those with renal dysfunction, and critically ill patients. Notably, the ACORN trial reported higher rates of delirium with cefepime compared to piperacillin-tazobactam, although other outcomes of neurotoxicity (such as seizures) were not assessed. Given the totality of evidence, the authors generally consider piperacillin-tazobactam safe for empiric use with vancomycin, particularly when anaerobic coverage is needed or when patients are at higher risk for cefepime-associated neurotoxicity.

Empiric combination therapy with vancomycin and a broad-spectrum BL is commonly prescribed for patients with suspected sepsis. However, MRSA accounts for only a small proportion of sepsis cases, making vancomycin retrospectively unnecessary in most instances. When vancomycin is administered first in patients with limited intravenous access, it can delay BL initiation because of its prolonged infusion time (typically at least 60 minutes to reduce infusion reactions). Two retrospective studies suggest that administering BLs before vancomycin may improve outcomes. ^, Although these findings may be influenced by residual confounding, prioritizing BL administration remains appealing given that these agents are rapidly infused, well tolerated, rapidly bactericidal, and provide coverage against most sepsis pathogens.

Double coverage for P aeruginosa —typically with 2 agents from different antibiotic classes, such as an antipseudomonal BL combined with a fluoroquinolone or an aminoglycoside—is often recommended for patients with septic shock at high risk of multidrug-resistant gram-negative infection. The aim is not synergy but to maximize the chance that at least 1 agent will be active when susceptibility is uncertain, as inappropriate initial therapy is linked to increased mortality, especially in sepsis or septic shock. ^,, In patients with gram-negative bacteremia and sepsis, empiric double coverage has been associated with higher rates of appropriate empiric therapy and improved survival compared with monotherapy.

However, there is little evidence of benefit to continuing double coverage beyond the empiric period, and prolonged use of combination therapy increases the risk of toxicity—particularly nephrotoxicity with aminoglycosides—and contributes to antimicrobial resistance. As such, once susceptibilities are known, definitive therapy should be narrowed to a single active agent.

Double coverage is generally not necessary in patients at low risk for resistant organisms or in those who are clinically stable. Moreover, certain agents commonly used for empiric double coverage (eg, aminoglycosides) have poor tissue penetration for specific infections (eg, pneumonia), which limits their efficacy as monotherapy and emphasizes their role as a temporary supplement during the empiric phase. Ultimately, double coverage should be used selectively and discontinued once culture results and clinical stability allow for de-escalation.

Rising rates of highly resistant gram-negative pathogens over the past decade—including CRE and MDR P aeruginosa —have posed increasing challenges for empiric therapy in critically ill patients. Fortunately, this period has also seen the development of safer and more effective BL/beta-lactamase inhibitor (BLI) combinations, such as ceftazidime-avibactam, meropenem-vaborbactam, and imipenem-relebactam, reducing reliance on older, more toxic agents like colistin or polymyxin B.

Routine empiric use of these novel BL/BLI agents is generally discouraged because of their broad spectrum, high cost, and potential to promote resistance—particularly given the relative rarity of CRE and other multidrug-resistant gram-negative pathogens. However, empiric use may be warranted in select high-risk patients, particularly those with prior colonization or infection with CRE or difficult-to-treat Pseudomonas , recent health care exposure in regions with a high prevalence of resistant organisms, or recent receipt of multiple broad-spectrum antibiotics. In these situations, initiating therapy with a novel BL/BLI may be appropriate while awaiting culture and susceptibility data. When available, rapid diagnostic testing and early infectious diseases consultation should inform prompt de-escalation based on pathogen identification and susceptibility results.

Although most patients treated for suspected sepsis do not have MRSA infections, the high morbidity and mortality associated with MRSA justify empiric coverage in select critically ill patients. Vancomycin is often appropriate in those with known colonization, risk factors for MRSA, suspected pneumonia, severe skin and soft tissue infections (particularly with purulence), or suspected central line-associated bloodstream infections. Empiric therapy should be narrowed promptly once culture and susceptibility results are available.

Nasal swab testing for MRSA has emerged as a valuable tool to guide antibiotic therapy. A recent negative MRSA nasal swab has a high negative predictive value (NPV) for MRSA pneumonia—exceeding 95% in many studies—and can safely support withholding or discontinuing empiric vancomycin in patients with suspected pneumonia. However, the NPV is lower for other infections (eg, bloodstream or soft tissue), and the positive predictive value (PPV) is limited in all contexts. As such, MRSA nasal swabs are best used to rule out MRSA rather than confirm its presence. Several studies have demonstrated that using MRSA swabs in this context can reduce unnecessary vancomycin use in critically ill populations with suspected pneumonia. ^, A negative MRSA nasal PCR result generally remains valid for at least 14 days, which can extend its utility in guiding antibiotic decisions across an episode of care. ^,

Compared with MRSA nasal swabs, VRE rectal swabs are less well-studied for guiding empiric therapy. Known colonization does increase the risk of invasive VRE infection, particularly in immunocompromised patients, those with prolonged hospital stays, or recent broad-spectrum antibiotic use. ^, In these high-risk settings, empiric treatment with linezolid or daptomycin may be appropriate in patients with positive VRE swabs when Enterococcus is a likely pathogen (eg, intra-abdominal or catheter-related infections) and the patient is severely ill. However, most colonized patients do not develop invasive disease, and a positive swab lacks the specificity to justify routine empiric coverage. Furthermore, unlike MRSA, VRE infections often cause more indolent bacteremia, giving clinicians more time to assess the need for targeted therapy. In this context, a prior negative VRE swab can help support the decision to withhold empiric VRE coverage.

While vancomycin is generally used for empiric gram-positive coverage in severe necrotizing skin/soft tissue infections, linezolid is gaining popularity because of its ability to inhibit protein synthesis and suppress endotoxin expression. Compared with clindamycin, which is commonly used for its antitoxin effect, linezolid exhibits higher susceptibility rates for MRSA and Streptococcus species. ^, In a large retrospective target trial emulation study of patients with invasive group A streptococcal infections, linezolid was non-inferior to clindamycin for in-hospital mortality when combined with BL therapy, with no significant differences in length of stay or C difficile infection, supporting its use, particularly amid rising clindamycin resistance. Linezolid may also be given as monotherapy rather than in combination with another agent. A recent quasi-experimental study comparing outcomes before and after updating internal guidelines for necrotizing soft tissue infections from vancomycin plus clindamycin to linezolid found similar efficacy but lower rates of AKI in the linezolid group, likely because of reduced vancomycin use. This benefit may apply to necrotizing pneumonia also, although robust studies are lacking. ^, Outside of necrotizing infections, linezolid has shown similar efficacy compared with vancomycin for other severe MRSA infections, mainly complicated soft-tissue and pulmonary infections. ^,

Candida species are among the most common causes of nosocomial bloodstream infections in the ICU and are associated with very high crude and attributable mortality rates. ^, Empiric antifungal therapy is therefore an important consideration in critically ill patients with risk factors for Candida bloodstream infections, which include recent broad-spectrum antibiotic exposure, central venous catheter use, total parenteral nutrition, abdominal surgery (especially with gastrointestinal perforation or anastomotic leak), prolonged ICU stay, blood transfusion, and immunosuppression. Colonization with Candida at multiple sites may also increase risk, although it is neither sensitive nor specific. Clinical clues such as persistent fever despite broad-spectrum antibacterial therapy or new signs of sepsis without an identified source should raise suspicion. However, several randomized controlled trials of empiric antifungal therapy in non-neutropenic critically ill patients at high risk for invasive candidiasis have not demonstrated improved clinical outcomes, highlighting the importance of careful patient selection. ^,, In cases where empiric antifungal therapy is pursued, an echinocandin is typically preferred for empiric coverage because of its broad activity against most Candida species, including C glabrata and C krusei , and favorable safety profile. Early de-escalation or discontinuation based on culture data and clinical trajectory is essential to avoid overtreatment and minimize antifungal resistance.

Selection of empiric therapy can be complicated by reported allergies. Up to 35% of patients self-report an antimicrobial allergy, although true rates are much lower. Penicillin allergy is documented in up to 25% of US medical records, yet immunoglobulin E (IgE)-mediated reactions are confirmed in only about 5% of cases; true prevalence in the general population is approximately 1% to 3%. ^, Overdocumentation often stems from childhood hypersensitivity reactions, and up to 80% of previously positive patients lose sensitivity within 10 years. Inaccurate allergy labels can lead to suboptimal antibiotic choices, increasing risks of C difficile infection, treatment failure, longer hospital stays, repeat therapy, and mortality. ^,,,

Clinicians should review reported allergies carefully, distinguishing non–immune-mediated effects (eg, diarrhea) or subsequent tolerance after a reported reaction from true hypersensitivity. For confirmed IgE-mediated reactions, cross-reactivity should be considered. In BLs, this is driven mainly by similarities in the R1 side chain: early generation cephalosporins share side chains with aminopenicillins, whereas later generations have minimal overlap. Skin testing shows up to 25% of confirmed penicillin-allergic patients react to first-generation cephalosporins, but only about 2.4% react to third-generation cephalosporins, similar to the background rate. Carbapenem hypersensitivity occurs in fewer than 1% of penicillin-allergic patients. ^, Given these low rates, later-generation cephalosporins should be considered as carbapenem-sparing options in IgE-mediated penicillin allergy. Although test doses, allergy consultation, or desensitization may be appropriate, their utility is limited in the urgent empiric setting.

Optimizing drug dosing in critically ill patients requires an understanding of pharmacokinetics (PK) and pharmacodynamics (PD). PK describes how the body absorbs, distributes, metabolizes, and eliminates the drug, or what the body does to the medication; PD is the relationship between the drug concentration at the site of effect and the resulting effect, or what the drug does to the body. ^, The interaction between PK and PD determines the drug’s effect over time.

Critical illness often causes profound PK alterations, leading to increased or decreased plasma concentrations ( Fig. 1 ). Contributing factors include changes in cardiac output, protein binding, hepatic or renal function, volume status, capillary leak, and drug absorption/distribution.

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

Recognition and Management of Infectious Biothreats and Emerging Pathogens Optimal Approach to Source Control in Intensive Care Unit Patients The Global Intensive Care Unit Intensive Care Unit Management of Meningitis and Encephalitis The Respiratory Triple Pandemic in the Intensive Care Unit Lower Respiratory Tract Infections

Stay updated, free articles. Join our Telegram channel

Join