Show Summary Details
Page of

Antimicrobial selection policies in the ICU 

Antimicrobial selection policies in the ICU
Antimicrobial selection policies in the ICU

David L. Paterson

and Yoshiro Hayashi

Page of

PRINTED FROM OXFORD MEDICINE ONLINE ( © Oxford University Press, 2022. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Medicine Online for personal use (for details see Privacy Policy and Legal Notice).

date: 05 July 2022

Key points

  • Consider the potential source of infection when making empiric antibiotic choices.

  • Local epidemiology in your intensive care unit (ICU) may differ from that published elsewhere.

  • De-escalate antibiotics to the narrowest possible spectrum, when susceptibilities allow.

  • Super bugs are now widespread in many ICUs—colistin is the last-line antibiotic and should be used prudently.

  • Antibiotic stewardship is important in ICUs and pertains not just to initiation of antibiotics, but also to limiting duration of antibiotic courses.


The selection of an antibiotic for an individual critically-ill patient depends on a number of factors. These include the patient’s allergies and renal function, the likely bacterial and fungal pathogens typically found at the potential site of infection, prior isolation of antibiotic resistant pathogens, prior antibiotic use, the local epidemiology of antibiotic resistance, costs of comparator drugs, and the potential ecological effect of prescribed antibiotics. Five key policy issues determine antibiotic selection:

  • The microbiological differential diagnosis.

  • Use of epidemiology data.

  • Adjusting antibiotic therapy with the aid of microbiology results.

  • Special considerations for resistant organisms.

  • Antimicrobial stewardship.

Microbiological differential diagnosis based on suspected source of infection and empiric regimens

Initial empiric antimicrobial therapy can be defined as the use of an antimicrobial for a patient with symptoms and signs of a serious infection prior to identification of the bacteria and the availability of susceptibilities. When making an initial antibiotic choice, definition of the likely location of infection based on clinical symptoms and signs and basic diagnostic testing is important because the microbiological differential diagnosis differs depending on the infection site (Table 286.1). For example, all of Staphylococcus aureus, Enterococcus faecalis and Candida spp. are frequent pathogens in intensive care units (ICU), but S. aureus is an extremely infrequent pathogen for urinary tract infection and neither E. faecalis nor Candida spp. typically cause pneumonia.

Table 286.1 Microbiological differential diagnosis based on source of infection and suggested empiric antimicrobial regimen

Source of infection

Microbiological differential diagnosis

Possible empiric antimicrobial regimen

Community-acquired pneumonia

  • No risk of multidrug resistant organisms:S. pneumoniae, H. influenzae, Legionella spp., Mycoplasma pneumoniae

  • Risk of multidrug resistant organisms: nosocomial GNB, such as P. aeruginosa in addition to the above

  • If CA-MRSA is of concern

  • Ceftriaxone + azithromycin

  • Penicillin + doxycycline

  • Piperacillin/tazobactam + azithromycin

  • Cefepime + azithromycin

  • Add vancomycin, clindamycin or linezolid to the above

Ventilator-associated pneumonia, hospital-acquired pneumonia, health care-associated pneumonia

  • No risk of multidrug resistant organisms:S. pneumoniae, H. Influenzae, MSSA, sensitive E. coli and K. pneumoniae

  • Risk of multidrug resistant organisms: nosocomial organisms such as P. aeruginosa, ESBL-producing E. coli and K. pneumoniae, Enterobacter spp., Serratia spp., and MRSA in addition to the above

  • Ceftriaxone (applicable to some northern European countries only).

  • Piperacillin/tazobactam ± vancomycin ± amikacin.

  • Cefepime ± vancomycin ± amikacin.

  • Meropenem ± vancomycin ± amikacin.

Community-acquired urinary tract infection

Mostly E. coli

  • Ampicillin + gentamicin.

  • Ceftriaxone.

  • Ertapenem.

Complicated urinary tract infection

  • E. coli including ESBL producer

  • P. aeruginosa

  • Piperacillin/tazobactam ±gentamicin.

  • Ertapenem.

  • Meropenem.

Catheter-associated blood stream infection

  • S. epidermidis

  • S. aureus, including MRSA

  • Enterococcus spp.

  • Nosocomial GNB including: P. aeruginosa, Candida spp.

Vancomycin + piperacillin/tazobactam ± fluconazole or echinocandin.

Community-acquired intra-abdominal infection

  • Anaerobes, such as Bacteroides spp.

  • GNB such as E. coli

  • Ceftriaxone + metronidazole.

  • Ertapenem.

Hospital-acquired intra-abdominal infection

Nosocomial organisms such as P. aeruginosa, ESBL-producing E. coli and K. pneumoniae, Enterobacter spp., Serratia spp., MRSA, Enterococcus spp., including VRE and Candida spp. in addition to the above

  • Cefepime + metronidazole ± vancomycin ± fluconazole or echinocandin.

  • Piperacillin/tazobactam ± vancomycin ± fluconazole or echinocandin.

  • Meropenem ± vancomycin ± fluconazole or echinocandin.

Complicated skin and soft tissue infection

  • Community-acquired:S. pyogenes, S. aureus (CA-MRSA where endemic), Clostridium perfringens

  • History of exposure to seawater/freshwater:Aeromonas hydrophila, Vibrio vulnificus

  • Diabetic foot infection: nosocomial organisms including MRSA and P. aeruginosa

  • Penicillin G + clindamycin.

  • Meropenem + ciprofloxacin.

  • Piperacillin/tazobactam ± linezolid.

  • Meropenem ± linezolid.

Community-acquired meningitis

  • S. pneumoniae, Neisseria meningitidis

  • If immunocompromised or age > 50:Listeria monocytogenes

  • High dose ceftriaxone + high dose vancomycin + aciclovir (until HSV negative).

  • Add high dose ampicillin.

Post-neurosurgical meningitis

S. aureus including MRSA, S. epidermidis, nosocomial GNB, including P. aeruginosa

  • High dose vancomycin + high dose meropenem.

  • High dose vancomycin + high dose cefepime.

Neutropenic fever

Nosocomial GNB including P. aeruginosa, S. aureus, including MRSA

  • Cefepime + vancomycin.

  • Piperacillin/tazobactam + vancomycin.

  • Meropenem + vancomycin.

Community-acquired undifferentiated sepsis

S. pneumoniae, N. meningitidis, sensitive GNB such as E. coli

  • If meningitis is possible: high dose ceftriaxone + high dose vancomycin + aciclovir ± gentamicin.

  • If meningitis is unlikely: piperacillin/tazobactam.

Hospital-acquired undifferentiated sepsis

Nosocomial GNB, including P. aeruginosa, S. aureus, including MRSA, Candida spp. if risk factors present

  • Piperacillin/tazobactam + vancomycin ± amikacin ± fluconazole or echinocandin.

  • Cefepime + vancomycin ± amikacin ± fluconazole or echinocandin.

  • Meropenem + vancomycin ± amikacin ± fluconazole or echinocandin.

It is important to decide whether Pseudomonas aeruginosa, carbapenem resistant Enterobacteriaceae/Acinetobacter spp., MRSA, and/or Candida spp. need to be covered. As a general rule, most ICU-acquired infections such as late-onset ventilator-associated pneumonia (VAP) require both anti-pseudomonal and anti-MRSA cover, while most community-acquired infections do not except for special situations (e.g. MRSA cover for severe community acquired pneumonia where community-associated MRSA is endemic, especially following influenza infection). Fever in neutropenic patients requires anti-pseudomonal cover and most cases of undifferentiated ICU-acquired infections need both anti-pseudomonal and anti-MRSA cover in the initial therapy. The addition of antibiotics active against carbapenem-resistant Enterobacteriaceae or Acinetobacter spp. will depend on local epidemiology.

Before the initiation of any antibiotic treatment, obtaining appropriate specimens for culture, especially blood cultures, is crucial to facilitate subsequent important decisions such as antibiotic discontinuation, antibiotic rationalization, and duration of antibiotics (e.g. Staphylococcus aureus blood stream infection may require longer duration of antibiotic therapy).

Clinicians should be aware that some patients with symptoms and signs of a serious infection do not, in fact, have an infection at all. Systemic inflammatory response syndrome (SIRS) may have non-infectious aetiologies (e.g. burns, trauma, post-surgery, acute pancreatitis). Many non-infective pulmonary complications in mechanically ventilated patients (e.g. fluid overload, acute respiratory distress syndrome (ARDS)) mimic ventilator-associated pneumonia [1]‌. In these situations, antibiotic therapy is not only unnecessary, but creates an increased risk of disturbance of endogenous flora, potentially leading to Clostridium difficile infection or colonization with antibiotic-resistant bacteria [2,3]. Unfortunately even in the state-of-the-art practice, differentiation of these conditions is often difficult. Thus, daily review of antibiotic therapy by an experienced clinician is important.

Use of epidemiological data to guide antibiotic choice

Epidemiological data on antibiotic susceptibilities is a useful tool in the selection of empiric antibiotic therapy, because the expected reliability of each agent is significantly variable depending on the location (e.g. country, city, specific ICU). The following three methods can be used to increase the chance of adequate therapy.

Use of data from surveillance studies

Antibiotic resistance surveillance programs assist clinicians to ensure a substantial probability of effectiveness of empiric therapy [4]‌. For example, a Spanish nationwide surveillance of Enterobacteriaceae causing community-acquired urinary tract infection showed 56.7%, 30.3%, and 22.9% of E. coli isolates were non-susceptible to ampicillin, sulfamethoxazole-trimethoprim and ciprofloxacin, respectively [5]. Therefore, those agents would be a poor empiric choice for patients with suspected urosepsis in that country. National and international trends in antibiotic resistance, derived from sequential examination of surveillance data can also alert clinicians to a need to re-examine their empiric antibiotic choices. For example, the emergence of substantial degrees of fluoroquinolone resistance in P. aeruginosa described in national surveillance studies in the United States [6] necessitated discontinuation of use of fluoroquinolones as sole agents to cover Gram-negative bacilli (GNB) in empiric therapy for most of ICU-acquired infections.

Use of ‘unit-based antibiograms’ and ‘combination antibiograms’

Unit-based antibiograms are cumulative antibiotic susceptibility reports from patients in a particular ward of the hospital (e.g. a specific ICU) in a specified time period. They are potentially more useful than international, national, or hospital-specific antibiograms for a prescriber making decisions about antibiotic therapy in a specific ICU. Unit-based antibiograms have been shown to have substantially different antibiotic susceptibility results compared to hospital-wide antibiograms [7,8].

Traditional antibiograms provide susceptibility data for organisms to a range of antibiotics, but do not answer the question of what antibiotic combinations may be optimal for providing coverage for any given organism. Combination antibiotic therapy is necessary as initial empiric antibiotic therapy for infections where P. aeruginosa is prominent, for example, because in most ICUs <90% P. aeruginosa strains are susceptible to any particular antipseudomonal beta-lactam [9]‌. Thus, for example, a traditional antibiogram does not give data as to what aminoglycoside or fluoroquinolone should be used in combination with a core antipseudomonal beta-lactam to ensure that initial empiric cover is optimized. A combination antibiogram provides information on the percentage of isolates susceptible to a particular antibiotic if the isolate is resistant to a core antipseudomonal antibiotic. For example, Bhat et al. [10] showed that, of the P. aeruginosa isolates resistant to cefepime in their institution, 19.5% were resistant to amikacin, 82.9% to ciprofloxacin, 43.9% to gentamicin, 87.8% to levofloxacin, and 39.0% to tobramycin. Therefore, in this case, combinations of cefepime plus amikacin would be much more likely to improve adequacy of initial empiric therapy than a combination of cefepime plus a fluoroquinolone. Unit-based, combination antibiograms can be constructed to complement individualized antibiotic choices.

Review of prior antibiotic use and previously isolated organisms with resistance profiles

Attention needs to be given to prior antibiotic use and previously isolated organisms with their resistance profiles because such knowledge could be useful to improve the adequacy of initial empiric therapy. For example, in one study [10], 37% of patients who received piperacillin/tazobactam in the month prior to the current infection were newly infected with a piperacillin/tazobactam-resistant strain and 64% of patients who had isolation of piperacillin/tazobactam-resistant GNB in the month prior to the current infection were now infected with piperacillin/tazobactam-resistant P. aeruginosa. Therefore, piperacillin/tazobactam was only considered appropriate initial empiric therapy if the patient had neither received the antibiotic nor had isolation of a piperacillin/tazobactam-resistant organism in the past month. Comparable findings were made with respect to cefepime. The policy to avoid antibiotics which were used in the last one month and antibiotics which have not demonstrated in vitro activity against bacteria from the patient in the last month may be useful not only to increase the chance of appropriate empiric treatment, but also to maintain antibiotic heterogeneity in the ICU.

Adjusting antibiotic therapy with the aid of microbiology results

Antibiotic rationalization should be considered at the following three stages: when formal identification is not completed, but the provisional identification is available, when the identification of organism is complete, but susceptibilities are awaited (which is now frequent with the advent of matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) mass spectrometry identification systems in many clinical laboratories), and when both identification and susceptibilities are complete. Even in empiric therapy with the best prediction strategy, antibiotic rationalization is often required in order to provide more reliable treatment than the initial therapy, to reduce the risk of antibiotic collateral damage, and to secure treatment for organisms which were initially not considered for the treatment, but revealed to have been the potential causative organism.

Optimizing choice for therapy when the organism is known, but susceptibilities are awaited

The new MALDI-TOF technology provides far more rapid organism identification than semi-automated technology used in laboratories over the last two decades. Close liaison with clinical microbiologists may allow clinical use of identification data to amend the initial regimen, even when susceptibility results are not available. For example, anti-pseudomonal therapy must be ensured if P. aeruginosa is identified. An important role of the microbiology laboratory is in alerting prescribers to the possibility of an organism not routinely covered by standard anti-pseudomonal therapy. For example, the identification of Stenotrophomonas maltophilia from respiratory specimens should allow treatment for this organism with co-trimoxazole, if there is a high clinical likelihood of VAP.

While MALDI-TOF systems in their current state do not provide rapid susceptibility results, they can be used to identify epidemic clones of multiply resistant organisms.

Earlier appropriate antibiotic therapy may be initiated for multi-resistant epidemic organisms, such as vancomycin-resistant enterococci (VRE), carbapenem-resistant A. baumannii or KPC- or extended-spectrum beta-lactamase (ESBL)-producing K. pneumoniae or E. coli, if the characteristic phenotype of these organisms is observed on incubation of specimens on solid media [11].

Direct susceptibility tests, although clearly not standardized, may also provide similar benefit. Bouza et al. demonstrated in randomized controlled trials (RCT) with VAP patients that the rapid reporting of susceptibility testing of respiratory specimens led to decreased antibiotic consumption, decreased rates of C. difficile infection, and fewer days of receiving mechanical ventilation than the standard procedure group [12].

Despite not being final susceptibilities, early amendment of the initial empiric antibiotics with the these interventions may provide some clinical benefit.

‘Fine-tuning’ therapy when identification and susceptibilities are complete

When identification and susceptibility testing results are available for the clinician, antibiotic regimens can be ‘fine-tuned.’ Although it is sometimes not feasible in the treatment of serious infections, de-escalation is a reasonable and favourable strategy if the causative organism is found to be treatable with narrower spectrum antibiotics than used empirically. This may not only minimize antibiotic collateral damage, but also provide more reliable therapy. For example, anti-staphylococcal penicillins such as flucloxacillin or nafcillin have a superior track record in the treatment of serious infections due to meticillin sensitive S. aureus (MSSA) than vancomycin or carbapenems, although these agents may be active against MSSA. Continuous use of broad spectrum antibiotics is more likely to lead to risks of antibiotic collateral damage such as C. difficile associated disease and emergence of carbapenem-resistant GNB than use of an antibiotic only active against MSSA. Antibiotics for the definitive therapy of common organisms in ICU are shown in Table 286.2.

Table 286.2 Suggested antibiotics for definitive therapy

Causative organism

Antibiotics for definitive therapy

Alternative choice

Gram-positive cocci

Enterococcus spp.


Vancomycin (if ampicillin-resistant)

Staphylococcus aureus (MSSA)

Flucloxacillin, nafcillin, oxacillin


Staphylococcus epidermidis


Streptococcus pneumoniae (penicillin sensitive)


Ampicillin, ceftriaxone

Group A, B, C, F, G streptococci


Ampicillin, ceftriaxone

Viridans streptococci


Ampicillin, ceftriaxone, vancomycin

Gram-positive bacilli

Bacillus anthracis


Ampicillin, ciprofloxacin

Corynebacterium jeikeium



Listeria monocytogenes


Nocardia spp.


Gram-negative cocci

Neisseria meningitidis

Penicillin G

Ampicillin, ceftriaxone

Gram-negative bacilli

Aeromonas hydrophila



Achromobacter xylosoxidans



Acinetobacter baumannii


Meropenem if susceptible

Burkholderia cepacia


Campylobacter jejuni


Citrobacter spp.


Ciprofloxacin, meropenem

Escherichia coli


Meropenem (if ESBL producer)

Enterobacter spp.


Ciprofloxacin, meropenem

Haemophilus influenzae


Ceftriaxone (if ampicillin-resistant)

Klebsiella spp.


Meropenem (if ESBL producer); colistin (if carbapenem resistant)

Legionella pneumophila



Proteus mirabilis



Proteus vulgaris



Pseudomonas aeruginosa

Piperacillin/tazobactam, ceftazidime, cefepime, meropenem, ciprofloxacin


Salmonella spp.



Serratia marcescens


Ciprofloxacin, meropenem

Shigella spp.


Co-trimoxazole, azithromycin

Stenotrophomonas maltophilia



Vibrio cholera



Vibrio vulnificus

Doxycycline + ·ceftazidime

Clostridium difficile

Oral or intravenous metronidazole

Oral vancomycin

Non-difficile Clostridium spp.




Mycoplasma pneumoniae


Levofloxacin, doxycycline

Candida spp.

Candida albicans



Candida tropicalis



Candida parapsilosis



Candida glabrata


Candida krusei


Candida lusitaniae


Special considerations for drug-resistant organisms

Issues of antibiotic resistance in ICU have become increasingly serious across the globe. This pertains not only to relatively resistant bacteria in nature such as P. aeruginosa and A. baumannii, but also those which used to be susceptible to a wide variety of antibiotics, such as S. aureus, E. coli, and K. pneumonia. MRSA and multiply resistant K. pneumoniae and E. coli have already become common issues in many ICUs. At an extreme end of the spectrum are ICUs in some parts of the world where 90% of A. baumannii isolates, > 50% of P. aeruginosa isolates and >75% of K. pneumoniae isolates are resistant to carbapenems. Treatment for resistant organisms may require specialist input from Infectious Diseases Physicians or Clinical Microbiologists. Table 286.3 summarizes antibiotic choice for representative resistant organisms in ICU.

Table 286.3 Antibiotic therapy for drug-resistant organisms

Causative organism

Antibiotics for definitive therapy

Alternative choice

Hospital-associated MRSA


  • Linezolid for pneumonia, and skin and soft tissue infection

  • daptomycin for blood stream infection, but not for pneumonia

Community-associated MRSA


Vancomycin + clindamycin

Vancomycin non-susceptible S. aureus

  • Linezolid for pneumonia and skin and soft tissue infection

  • daptomycin for blood stream infection, but not for pneumonia

Vancomycin -resistant E. faecalis

  • High dose ampicillin (if sensitive)

  • linezolid


Vancomycin-resistant E. faecium



Penicillin-resistant S. pneumoniae (non-meningitis)

High dose ampicillin

  • High dose ceftriaxone

  • vancomycin

Penicillin-intermediate S. pneumoniae (meningitis)

High dose ceftriaxone


Penicillin-resistant S. pneumoniae (meningitis)

High dose vancomycin + high dose ceftriaxone + rifampicin

ESBL-producing Enterobacteriaceae


  • Ciprofloxacin (if sensitive)

  • Amikacin (if sensitive)

Carbapenem-resistant Enterobacteriaceae

High dose meropenem extended infusion + colistin + tigecycline

Amikacin (if sensitive)

Pan-beta-lactam and quinolone resistant P. aeruginosa


Amikacin (if sensitive)

Carbapenem-resistant A. baumannii


High dose ampicillin/sulbactam (if sensitive)

Working in an institutional framework: antimicrobial stewardship programs

Antimicrobial stewardship programs have become an integral component of most large hospitals. As stated in the Infectious Diseases Society of America and Society of Health Care Epidemiologists of America guidelines on antimicrobial stewardship programs, the primary goal of such programs is to optimize clinical outcomes while minimizing unintended consequences of antimicrobial use such as toxicity, the selection of pathogenic organisms (for example, Clostridium difficile) and the emergence of resistance. This goal should be compatible with care of patients with suspected infections in the ICU.

However, the execution of antimicrobial stewardship programs in the ICU environment may be quite different from that of a ward environment. ‘Front-end’ approvals, whereby a prescriber seeks approval from a physician or pharmacist from the Antimicrobial Stewardship Program, are rarely employed in ICUs. The major reason for this is the need for no delays in administration of antibiotics to septic patients. However, there is a role for discussion of choice of antibiotic with an Infectious Diseases Physician or Clinical Microbiologist, particularly when the patient is likely to be infected with a multidrug resistant pathogen. Other major contributions of antimicrobial stewardship programs in the ICU are: (1) to help create antibiotic guidelines for empiric choice of antibiotics, and (2) to help limit duration of antibiotic choices.


In summary, clinicians should apply logic in their selection and use of antimicrobials in the severely ill patient, rather than to use a ‘tradition-based’ method of selecting treatment. Tradition is hard to break, but results in patients receiving inappropriate antibiotics as well as contributing to the development of resistance in bacteria through increased selection pressure. Information is available with which to make a more logical choice of antibiotic, and closer liaison with microbiologists, antimicrobial pharmacists and infectious diseases physicians will greatly facilitate the decision-making process.


1. Chastre J, Luyt CE, Combes A, and Trouillet JL. (2006). Use of quantitative cultures and reduced duration of antibiotic regimens for patients with ventilator-associated pneumonia to decrease resistance in the intensive care unit. Clinical Infectious Disease, 43(2), S75–81.Find this resource:

2. Muto CA, Pokrywka M, Shutt K, et al. (2005). A large outbreak of Clostridium difficile-associated disease with an unexpected proportion of deaths and colectomies at a teaching hospital following increased fluoroquinolone use. Infection Control in Hospital Epidemiology, 26(3), 273–80.Find this resource:

3. Donskey CJ. (2006). Antibiotic regimens and intestinal colonization with antibiotic-resistant Gram-negative bacilli. Clinical Infectious Diseases, 43(2), S62–9.Find this resource:

4. Masterton RG. (2000). Surveillance studies: how can they help the management of infection? Journal of Antimicrobial Chemotherapy, 46(2), 53–8.Find this resource:

5. Cuevas O, Cercenado E, Gimeno M, MarÃn M, Coronel P, and Bouza E. (2010). Comparative in vitro activity of cefditoren and other antimicrobials against Enterobacteriaceae causing community-acquired uncomplicated urinary tract infections in women: a Spanish nationwide multicenter study. Diagnostic Microbiology and Infectious Disease, 67(3), 251–60.Find this resource:

6. Neuhauser MM, Weinstein RA, Rydman R, Danziger LH, Karam G, and Quinn JP. (2003). Antibiotic resistance among Gram-negative bacilli in US intensive care units: implications for fluoroquinolone use. Journal of the American Medical Association, 289(7), 885–8.Find this resource:

7. Binkley S, Fishman NO, LaRosa LA, et al. (2006). Comparison of unit-specific and hospital-wide antibiograms: potential implications for selection of empirical antimicrobial therapy. Infection Control in Hospital Epidemiology, 27(7), 682–7.Find this resource:

8. Lamoth F, Wenger A, Prodhom G, et al. (2010). Comparison of hospital-wide and unit-specific cumulative antibiograms in hospital- and community-acquired infection. Infection, 38(4), 249–53.Find this resource:

9. American Thoracic Society; Infectious Diseases Society of America. (2005). Guidelines for the management of adults with hospital-acquired, ventilator-associated, and healthcare-associated pneumonia. American Journal of Respiratory and Critical Care Medicine, 171(4), 388–416.Find this resource:

10. Bhat S, Fujitani S, Potoski BA, et al. (2007). Pseudomonas aeruginosa infections in the Intensive Care Unit: can the adequacy of empirical beta-lactam antibiotic therapy be improved? International Journal of Antimicrobial Agents, 30(5), 458–62.Find this resource:

11. Paterson DL and Bonomo RA. (2005). Extended-spectrum beta-lactamases: a clinical update. Clinical Microbiology Reviews, 18(4), 657–86.Find this resource:

12. Bouza E, Torres MV, Radice C, et al. (2007). Direct E-test (AB Biodisk) of respiratory samples improves antimicrobial use in ventilator-associated pneumonia. Clinical Infectious Diseases, 44(3), 382–7.Find this resource: