Antibiotics in ICU -- A guide for the Perplexed

Rx TablesRationaleResistance PatternsSites of InfectionFun Topics

An Approach to Antibiotic Prescription in ICU

Our approach is:

  1. Ask how well the patient is!
  2. In the gravely ill patient (as opposed to the 'not-so-seriously ill'), there is little time for delay, and an error in choice of antibiotics may well cost the patient his/her life. Prolonged ventilation and prior antibiotic use (especially of broad-spectrum agents) predispose to resistance.

  3. Know the organism
  4. Your benchmark for treatment should be treating a known organism with an appropriate dose of antibiotic to which that organism is likely to respond, based on sensitivity testing. This ideal will often not be met. Sometimes you will obtain an organism and its sensitivity on routine microbiological surveillance and then the patient will show features of infection likely to be due to that organism. More often, you will have to rely on empiric therapy. ( See also: [Am J Med 1991 301 165-72] )

  5. Know the environment
  6. Know the patterns of resistance, and the organisms prevalent in your ICU environment. This helps with antibiotic choice.

  7. Identify the site of infection
  8. Positive blood cultures are simply not good enough. Identify the site of infection (e.g. respiratory tract, urinary tract, a subdiaphragmatic collection, or whatever) and address any surgically remediable pathology right away. The primary treatment of an abscess, for example, is immediate drainage, not antibiotics.

  9. Don't overtreat
  10. Never treat a "fever" or a "leukocytosis" with antibiotics. Assess the patient as a whole, including their predisposition to infection, and likely sites of infection. Ask whether the patient is sick enough to justify antibiotics, rather than treating laboratory values! If you are going to start 'empiric' therapy, first obtain microbiological specimens for culture. Document your reasons for starting therapy, and choose as narrow an antibiotic spectrum as you can reasonably 'get away with'. When you get the results of ID + sensitivity testing, revise your treatment to 'narrow-down' the spectrum as far as possible.

  11. Don't delay
  12. If the patient clearly needs treatment, treat. Do NOT wait for sensitivity results - if the patient is ill and needs treatment now, sensitivity results will make a very poor epitaph.

  13. Don't undertreat
  14. Even more important than giving adequate doses of an antimicrobial is not to give an agent that has a substantial likelihood of failure . In a critically ill patient , you won't get a second chance. If you have antibiotic X and antibiotic Y, and in your unit there is a 35% incidence of resistance to X, but 3% resistance to Y, it's clear which one you should use, even if Y costs three times as much as X, and is regarded as a reserved agent !

  15. Know how critical illness interacts with the antibiotic
  16. The pharmacokinetics of antimicrobials is often substantially altered in the critically ill.

  17. In vitro response is not the same as in vivo
  18. There are some agents that appear to be effective in vitro, but will not work in vivo. Always look at sensitivity results in the light of your knowledge of the microbe and the patient (and especially the site of infection!).

  19. Don't treat for too long
  20. We usually give antibiotics for too long. In our opinion there are very few circumstances where very prolonged 'therapy' is desirable, although many current recommendations for treatment of nosocomial pneumonia suggest treatment be continued for two weeks, or even more. Without good evidence either way, we think this is often far too long. If the patient has responded dramatically, is clinically much improved, and leukocytosis and fever has subsided for 24 to 48 hours, we think that cessation of antibiotic therapy is a good idea. There are notable exceptions to this guideline - infective endocarditis and deep-seated Staphylococcus aureus infections, for example, must be treated for prolonged periods (at least 4 weeks with deep-seated Staph. infection).

  21. Establish treatment guidelines
  22. Each unit should have antibiotic guidelines ('for the obedience of fools and the guidance of wise men')!

  23. Discuss your treatment with experts
  24. Most microbiologists are very keen to advise you. Listen to them - they usually know their subject far better than clinicians. (The Infectious Diseases Society of America also recommends computer-based monitoring with feedback , combined with the use of benchmarking data to tell you if you're being silly (or good) in your antibiotic prescribing).

If you disagree with the above, and have constructive comments about how we can improve this approach, email us! Please note that the information and ideas contained in this document should not be used to guide clinical decision-making. If you are unsure about what agent to use in clinical patient management, consult a human expert, not our web-page! We will not be held responsible for any consequences of your clinical management decisions.

Quick Tables of Organisms, Sites and Treatment

The following tables are not meant to be definitive, and should be read in conjunction with the above guidelines.

Organisms and their (tentative) treatment
(In the table 'Quinolone' always means a fluoroquinolone)
Organism Rx if 'Naive' Alternative Rx if 'v. Resistant' Avoid
Acinetobacter spp. Quinolone
OR cefepime
OR imipenem (? + aminoglycoside)
Bacteroides fragilis Metronidazole ß-lactam+inhibitor eg. amoxycillin + clavulanate (uncommon) all cephalosporins, penicillin, aminoglycosides
Enterobacter carbapenem
OR cefepime
OR (?) high dose ß-lactam+inhibitor
1st, 2nd, 3rd gen cephalosporins
Enterococcus faecalis Vancomycin + aminoglycoside quinolones, cephalosporins, ampicillin!
Enterococcus faecium Ampicillin + aminoglycoside - Vancomycin + aminoglycoside (unless VRE) quinolones, cephalosporins,
E. faecium is resistant to carbapenems
Escherichia coli Quinolone Co-trimoxazole ß-lactam+inhibitor
OR carbapenem
OR cefepime
Ampicillin, 3rd gen cephalosporins
Klebsiella spp. Cefotaxime Quinolone
OR ? Cefuroxime + aminoglycoside
OR carbapenem
OR cefepime
Organism Rx if 'Naive' Alternative Rx if 'v. Resistant' Avoid
Proteus mirabilis Quinolone OR cotrimoxazole ? ampicillin (resistance now common) 3gen Cephalosporin + aminoglycoside
OR piperacillin + tazobactam
Proteus (other) Quinolone ? 3rd gen. cephalosporin 3gen Cephalosporin + aminoglycoside
OR piperacillin + tazobactam
Pseudomonas aeruginosa Antipseudomonal penicillin (piperacillin, mezlocillin, azlocillin, ticarcillin) ? + aminoglycoside Antipseudomonal cephalosporin (eg. ceftazidime) ? + aminoglycoside Quinolone
OR cefepime
OR imipenem (? + aminoglycoside)
NB. if piperacillin resistant, adding a ß-lactamase inhibitor won't help!
Staphylococcus aureus Cloxacillin First-generation cephalosporin (eg cefazolin) Vancomycin quinolones, penicillin, 3rd gen cephalosporins,
MRSA are resistant to imipenem
Staphylococci - coagulase negative (CNS, S. epidermidis) vancomycin (if pathogenic)
(rarely, the organism is sensitive to cloxacillin, 1st gen. cephalosporins. Do NOT bank on this!)
Streptococcus pneumoniae Penicillin (2MU 4 hourly) Macrolides Cefotaxime
(OR ceftriaxone
OR vancomycin)
most quinolones
Organism Rx if 'Naive' Alternative Rx if 'v. Resistant' Avoid


Looking at antibiotic therapy in ICU from the point of view of a perplexed physician, there seem to be two broad schools of opinion among the "experts" in the field. We will call these the:

  1. Boring Old Standard Hypothesis (BOSH), to which I still subscribe;
  2. The "Thorough Elimination of Microbes Prevents Trouble" theory (which we will call TEMPT).

Needless to say, the terms and abbreviations are entirely my own! We will first explore BOSH, which I see as follows:


"We are in the age of bacteria, which started about 3.5 billion years ago, and still shows no signs of ending. Other non-bacterial organisms (which, from the bacterial point of view, are merely nutrient-rich broth in a flimsy package) have two choices:

  1. Be eaten now; or
  2. Find ways of co-operating with the bacteria, or at least coexisting fairly amicably (Be eaten later).

Over the last several billion years, organisms have evolved wonderfully complex ways of talking to bacteria, and modulating their behaviour. Likewise, bacteria have evolved wonderfully complex ways of talking to other organisms, and modulating their behaviour, sometimes terminally. Such signalling is exemplified by the recent insights we have gained about bacterial and host interactions in the human bowel. In other words, wherever there are bacteria, there is a complex ecology.

One small component of this ecology is antibiotics. Unfortunately, doctors (and vets & farmers) have seized upon this one small component as if it were the Holy Grail. They have, either for reasons of 'doing good' or for profit, used this component enthusiastically and relentlessly, and have consequently modified the ecology. Such modification is not necessarily a good thing , especially as the major modification has been a compensatory increase in the variety and numbers of bacteria that find such antibiotics inoffensive, or even occasionally, tasty!" [Me, 2001]

Physicians such as myself, who adhere to the BOSH school of thought, advise caution in administering antibiotics, lest one muddles up the ecology even further, especially in the long term. Details of this approach will be explored later.


I think there is at least one other approach, which although not perhaps widely acknowledged or admitted, is fairly prevalent. This approach seems to me to be to "nail the bugs before they do harm". Here are several examples of the TEMPT approach:

It can be seen that the TEMPT approach is heterogeneous. It also fulfills a deep psychological need in the attending doctor, to "do something". Even die-hard adherents to the BOSH approach (such as myself) have to admit that in some circumstances, the TEMPT approach is entirely correct. In others, we believe that it is totally wrong. The grey areas are the interesting ones.

Infection in ICU

There is no doubt that infection is a major association of ICU morbidity and mortality. There is also good evidence that antibiotic resistance is widespread, and an enormous problem. For example, the 1992 EPIC study, which looked at point prevalence of infection and bacterial resistance showed that 45% of 10 038 patients were infected (21% of these infections presumably nosocomial), and that there was widespread resistance of major pathogens to important antibiotics. [EPIC was published in JAMA 1995 274 639-44; For an overview, see Int. Care Med. 2000 26 S3-8, J-L Vincent].

Of even more concern is the emergence of difficult-to-treat (and sometimes, impossible-to-treat) pathogens such as vancomycin-resistant Enterococci, and multiresistant strains of Pseudomonas aeruginosa and Acinetobacter spp.

Why is there increasing resistance?

It is intuitively obvious that in rapidly changing microbial ecologies, selection pressure is necessary if an antibiotic-resistant bacterium is to achieve prominence. In other words, an antibiotic that 'decreases the competition' must be given, and if the bacterium is to remain prominent, an antibiotic must be given repeatedly. This is a necessary criterion for the emergence of resistance to antimicrobials.

In order that a bacterium (resistant to an antimicrobial) can attack a particular patient, there are several other obvious requirements:

Equally clearly, if any one of these steps or predisposing states is removed, a bacterium or fungus will have a torrid time in trying to attack the patient. We therefore have several strategies we can employ in preventing such onslaught. We can:

  1. Wash our hands. This simple practice, first espoused by Semmelweiss in the century before last, is still not ahered to, even in ICUs that preach this gospel (Semmelweiss was hounded to death by his colleagues);
  2. Ensure adequate patient nutrition. Another major failing of medicine - a substantial proportion of hospital patients (and especially, intensive care patients) are either grossly nutritionally compromised, or "at risk";
  3. Minimise suppression of endogenous patient flora. More of this later;
  4. Minimise invasive (and often unnecessary) breaches in the patient's integument, and where such breaches are absolutely necessary, minimise their duration, and manage them "aseptically" as far as is possible.

An important note in assessing studies

At this point we should pause to consider the implications of the above obvious measures. We have already mentioned the enthusiastic meta-analytical admonition to use antibiotics 'prophylactically' in ICU (SDD). Let's look at this in more detail.

Any study that purports to be a meaningful evaluation of the use of antibiotics in ICU, but that hasn't stuck to these "rules" should be regarded with grave suspicion. For example, let's say we have a high prevalence of infection in ICU X, and we successfully decrease the infection rate by whacking everyone on a new, expensive, "broad spectrum" antibiotic, or combination of antibiotics. We might be tempted to praise this antibiotic as the new wonder drug, and rush around administering it willy-nilly to all of our patients.

Not so. For the study will not tell us whether, in say two years time, prevalent microbes will have emerged that have high levels of resistance to our new wonder-drug. (We know from past experience that this will likely be the case). The study will almost certainly not have looked at the effect introduction of the agent has on the ecology of the ICU, the hospital, or even the community.

But even more important than these cautions is the possibility that the same results (minus the expense and risk of the antibiotics) may have been achieved by ensuring adequate handwashing, as well as other lesser measures such as limiting the dwell time of intravenous cannulae, and optimising patient nutrition, all with no adverse effect on microbial ecology, and other important beneficial effects!

There is another less obvious 'confounding variable' when it comes to assessing such studies. Let's say that in the general wards of a hospital, it is common practice to lash out with antibiotics at the first sign of a temperature, white cell count, or whatever. Let us also (for the sake of argument) assume that such antibiotic therapy is often 'standardised' ("homogenous antibiotic prescribing") and prolonged, suppressing the patient's normal flora, and encouraging colonisation by resistant organisms. It's clear that in such circumstances (but not of course in our hospital, he cried!), patients who are admitted to ICU will often be colonised by resistant organisms on admission, and normal host flora will be suppressed, with their 'receptors' on the host occupied by harmful pathogens. Such patients will be predisposed to aggressive infection. If we now administer potent antibiotics early on to these patients, we might in the short term see a decrease in infection, leading us to believe that early, aggressive and profligate antibiotic therapy in ICU is the right thing!

Patterns of Resistance in Specific Organisms

Sites and types of infection

"Blood borne infections"

As we said above, it's always a good idea to look diligently for the site of origin of microbes in the blood. Karam & Heffner have summarised the common causes of blood borne infection, based on CDC and other data. Coagulase negative staphylococci come out tops {how many of these were contaminants?}, followed by Staph. aureus and Enterococci, a surprisingly high percentage are Candidal (5 to 11%), and E. coli and Klebsiella make up some of the remainder. If there is no other source for infection, think about that intravenous catheter you have left in for "just one more day"!


Patients that end up in ICU with community acquired pneumonias may be infected with a variety of organisms, including S. pneumoniae , Haemophilus , Klebsiella , Legionella , and even Mycoplasma , Chlamydia , and so on. ICU- and ventilator-associated pneumonias (VAP) are difficult to diagnose and manage, and are commonly due to multiresistant gram-negative organisms, although recently, resistant gram positives have become prominent. VAP is by far the most important infection in ICU. Think Pseudomonas , Klebsiella , Acinetobacter , and also S. aureus . One possible solution to overuse of antibiotics is short course quinolone therapy, with reassessment at 3 days [Am J Respir Crit Care Med 2000 Aug;162(2 Pt 1):505-11]. There is scant evidence that invasive assessment of VAP alters outcome. See for example [Am J Respir Crit Care Med 2000 Jul;162(1):119-25]. Gram stain of sputum in VAP is of mimimal value. Causative organisms of VAP vary widely from ICU to ICU.

Urinary tract infection

While community-acquired UTIs are often due to E. coli , in hospital the usual nosocomial gram negatives are also often responsible.

Intra-abdominal infections

Here too, E. coli is important, but a host of other gram negatives may participate, enterococci often add to the problem, and anaerobes are extremely important, especially Bacteroides fragilis . Remember that infections are often polymicrobial.

Surgical wound infection

Both staphylococci and gram negatives (often hospital-acquired) are important.


In adults the main organisms are Neisseria meningitidis, and Streptococcus pneumoniae. Long-term neurological sequelae are common, if the patient survives. If the person is immune compromised, think Gram -ve bacilli, Listeria monocytogenes , fungal infection, and mycobacteria. It is not uncommon for doctors to mis-diagnose tuberculous meningitis as an acute bacterial meningitis because (a) they haven't taken a decent history and (b) the initial leukocytosis in the CSF may confuse them. Pseudomonas meningitis is uncommon but difficult to treat, and outcome is often poor. Imipenem should be avoided as it may cause seizures, but meropenem is safe, although an antipseudomonal penicillin (such as ceftazidime) is perhaps preferable, unless resistance is suspected.

Immune compromise

We will not here discuss the immune-compromised patient in any detail. Suffice it to say that many ICU patients are subtly or even overtly immune compromised, due to their poor nutritional status. There are others who may be on corticosteroids, and a small subset on potent immunosuppressives, or with underlying disease (such as AIDS) which predisposes to attack by a host of 'normal pathogens', as well as numerous fungi (like Pneumocystis and Candida), parasites, and opportunistic bacteria. In neutropaenic sepsis, aggressive and above all urgent management for presumed Gram negative infection will be life-saving.

Topics of Interest

What is an integron ?

Integrons are very important, because they are the main mechanism for dissemination of resistance genes in Gram negative bacteria. Let's start by describing the structure of an integron. An integron has:

The basic idea is that the integrase catalyzes insertion or deletion of resistance genes, and these are then vigorously expressed due to the strong promoter site. Resistance genes can spread aggressively between bacteria. These genes that can be clipped out of one integron and inserted into another are called gene cassettes (Something like taking a tape recorder cassette and playing it on somebody else's tape deck)! The cassettes are inserted at the attI site, which is recognised by the integrase. Up to five (or possibly even more) resistance genes may be contained in a single integron. There are over 60 gene cassettes described, including those that code for ESBLs and carbapenemases. Other cassettes code for resistance to aminoglycosides, trimethoprim, chloramphenicol, and even antiseptic agents such as quaternary ammonium compounds and mercury!

Different intl genes have been described. There are at least six, with classes 1, 2 and 3 being considered most important in spread of antibiotic resistance. Integrons have been around for a long time - we just haven't been really aware of them until recently. (See the review in [Clin Chem Lab Med 2000 Jun;38(6):483-7] ).

Most integrons have been reported from gram negatives (especially Enterobacteriaceae). "Super-integrons" have also been described, harbouring hundreds of genes, for example in Vibrio species.

Thoughts about predisposition to development of resistance

It makes sense that the larger the population of bacteria, and the longer they are exposed, the more likely they are to develop resistance to a particular antimicrobial. Remembering that the largest natural reservoir of bacteria in man is the bowel, it then comes as no surprise that agents that are extensively excreted into the bowel should promote ready resistance, especially if they persist for long periods of time (eg. rifampicin). Likewise, oral administration of vancomycin, a silly idea which should be avoided if at all possible, will probably promote vancomycin resistance, while intravenous administration should be far less likely to do so, as the drug is then renally excreted.

Mechanisms of Resistance

We have discussed this elsewhere.

Do ICUs export resistant bugs?

There is some evidence suggesting this is the case. See for example [ Clin Infect Dis 1999 29 1411-18 Lucet et al ]. ICUs are often jam-packed with resistant micro-organisms, accounting for up to a quarter of all nosocomial infections (despite constituting under 5% of beds in most hospitals).

Does initial appropriate therapy lower mortality?

Yes. See [ Chest 1999 115 462-74, Kollef et al].

Does good empiric therapy prevent drug resistance?

Yes. See [ Ann Intern Med 1996 124 884-90, Pestotnik et al].

Bactericidal vs bacteriostatic antibiotics?

It is often recommended (without support from a vast amount of research) that bactericidal antibiotics are preferable to bacteriostatic ones, with severe ICU infections. Examples of bactericidal antibiotics are penicillins, cephalosporins, aminoglycosides, carbapenems, and fluoroquinolones.

Endotoxin release by antibiotics

We know that gram negative bacteria release endotoxin from their cell walls when proliferating and when dying, and it is this endotoxin that initiates many cellular events (such as cytokine production) that cause morbidity and mortality. An attractive hypothesis (with little current substantiation or refutation) links administration of some antibiotics, massive bacterial killing, endotoxin release, and patient deterioration. We are not convinced that such endotoxin release is clinically significant.

If I stop using an agent, will resistance to it disappear?

No. Resistance will be suppressed, but the chances are that the resistant organism will still lurk in the background, and reappear quickly in large numbers, once it is encouraged to appear by suppression of the competition (when you start using the agent enthusiastically once more).

Dosing considerations - infusions and stat doses

Aminoglycosides kill bacteria based on high concentrations, and because (unlike most other agents) they have a post-antibiotic effect (PAE) that may last several hours, should probably be given in high doses once a day, rather than smaller doses twice or more per day. Although quinolones don't have a PAE, they too kill depending on concentration, and so area under the plasma concentration-time curve is important in determining bacterial kill rates.

On the contrary, beta-lactam killing of bacteria depends on the amount of time the tissue levels are above the minimum inhibitory concentration (MIC), and (above this level) is concentration-independent. It is therefore logical to give penicillins by continuous infusion, and it is unclear to me why so many people are still giving their penicillins as intermittent push-ins! (Probably just a matter of convenience and tradition flying in the face of reason). See for example Craig & Ebert [Antimicrob Agents Chemother 1992 36 2577-83], and Drusano (1998).

Where can I get consensus guidelines on preventing spread of resistant micro-organisms?


Weber et al also have a lot of detail, especially on management of MRSA outbreaks.

Crop Rotation

Kollef et al from St Louis [Crit Care Med 2000 28.10 3456-64], in the context of increasing incidence of microbial resistance, pursued the idea of scheduled changes in the class of antibiotics used for empirical therapy. (Some have called this "crop rotation", or "heterogeneous antibiotic use"). They rotated (for periods of six months) from a baseline of ceftazidime, through ciprofloxacin, and then cefepime, showing a progressive decline in the primary outcome - incidence of inadequate antimicrobial treatment. This incidence was assessed by isolation of the causative organism, and sensitivity testing where appropriate. Approximately 3/4 of the 3668 patients received antibiotics, including about a quarter who received "post-operative prophylactic antibiotics". 37% of patients had an identified infection, 90% of these being ventilator-associated or "bloodstream" infections. Inadequate antimicrobial therapy (use of a 3rd generation cephalosporin against a resistant organism, and to a lesser extent MRSA, Candida, VRE) was associated with increased in-hospital mortality. The study could perhaps be faulted because there was no simultaneous division of the study population into two groups - one group receiving therapy based on "the current crop", and the other at the discretion of the attending physician. The limitations of the study are well-discussed in the article.

Also note the potential concerns about crop rotation, notably cross-resistance. See [ J Antimicrob Chemother 1992 29 307-12] and [Antimicrob Agent Chemother 1990 34 2142-7] for cross resistance between quinolones and imipenem!


There is a vast literature. Here are just a few articles we found relevant.
(More are included in the text, and as 'subtext' comments in the HTML source of this document).

  1. The CDC guidelines for the prevention of Nosocomial Pneumonia (1997).

  2. Karam GH &Heffer JE.
    Emerging issues in Antibiotic Resistance in Blood-borne Infections
    . Am J Respir Crit Care Med 2000 162 1610-16.

  3. Singh N & Yu VL.
    Rational Empiric Antibiotic Prescription in the ICU
    Chest 2000 117(5) 1496-9.

  4. Weber DJ, Raasch R, Rutala WA.
    Nosocomial Infections in the ICU. The growing importance of antibiotic- resistant pathogens
    Chest 1999 115(3) Supp 34S-41S.

  5. Drusano GL.
    Infection in the Intensive Care Unit. ß-Lactamase-mediated resistance among enterobacteriaceae and optimal antimicrobial dosing.
    Clinical Infectious Diseases 1998 27(S1) S111-6.