There is increasing concern about the effects of antibiotics on the microflora of both animals and man. The House of Lords Select Committee Report recommended prudent antibiotic use in animal husbandry and astute antibiotic use in hospitals and general practice.1,2 Control of antibiotic use in hospital by restrictive policies can be unpopular, but attempts to monitor or control antibiotic use are now considered to be of primary importance both for the financial benefits and for reducing emergent resistance. Many reports have been published over the years on all aspects of surveillance and antibiotic resistance in hospitals.3-7 Although these are necessary for producing national or international data on predominant pathogens found and on their resistance patterns, it is clear that local knowledge of prevalent pathogens and susceptibilities are of primary importance when deciding appropriate antimicrobial therapy.8,9 In intensive care units (ICUs), Pseudomonas aeruginosa, Enterobacter species, Escherichia coli, Klebsiella species, and Acinetobacter species are the most prevalent Gram-negative bacteria isolated. Staphylococcus aureus, coagulase-negative staphylococci, and Enterococcus species are usually the most prevalent Gram-positive bacteria isolated.10-12 A patient with suspected infection on an ICU is often immediately prescribed with empirical antibiotics that apply to these organisms. Third-generation cephalosporins are often the antibiotic of choice either as part of a combination approach or, sometimes, as monotherapy. Unfortunately, with the increasing prevalence of Gram-negative bacteria that produce both plasmid-mediated and chromosomally mediated extended-spectrum β-lactamases, clinicians are now having to consider alternative antimicrobial drugs; the quinolones and carbapenems are a commonly used option. Is there any evidence that these alternatives are any better than cephalosporins? The use of antibiotics in hospitals has long been thought to be associated with the occurrence of resistant bacteria,13-16 but few studies have looked at the effect of antibiotic use in the ICU environment17-19 or at the more specific modification of antibiotic use and the possible effect on bacterial resistance patterns in the ICU.20-22 We have undertaken a prospective, crossover study of 2 regimens of therapy in the ICUs of 2 large hospitals in the United Kingdom. The usage of ciprofloxacin and third-generation cephalosporins during each arm of the study was compared, with the epidemiology of the bacteria isolated and resistance patterns found.
PATIENTS AND METHODS
The study was conducted in the 2 general ICUs of the General Infirmary at Leeds and the City Hospital, Birmingham, from October 1995 to December 1997. Both ICUs dealt with general medical and surgical patients but did not deal with specialist surgical patients. In particular, there were no cardiac surgical patients in either center. October 1995 to December 1995 was a baseline period before starting the study protocol. The 2 phases of the study were conducted from January to December 1996 and from January to December 1997. There were no changes in infection control procedures in either unit for the duration of the study. All patient information and microbiological data were stored on an Access (Microsoft Corporation, Redmond, WA) database. The study was approved by the research ethics committees of both hospitals.
The ICUs were visited twice weekly throughout the study periods. Demographic data (age, APACHE II score, length of stay) were obtained, and the patients present in the ICU on these visits had microbiological surveillance specimens collected. Only the patients present on surveillance days or those who presented with positive clinical microbiology cultures were included in the study.
During the study, no antibiotic restrictions were operational on the units; however, throughout the first year (1996), the preferred first-line antibiotic used was cefotaxime (Leeds) and ceftazidime (Birmingham). At both centers, these antibiotics had been used as the predominant empirical agent for several years. In the second year (1997), this changed to ciprofloxacin at both centers. All other antibiotics were prescribed as clinically indicated. The switch in antibiotic prescription was brought about by the consultation with clinicians working on the ICUs and those referring patients and by the regular attendance of clinical microbiologists on ICU rounds. Antibiotic data for the ICUs were obtained from pharmacy records. Usage was expressed as the number of defined daily doses (in grams) per 100 patient bed days.
Surveillance specimens, which were either endotracheal aspirates or mouth swabs, were plated onto heated blood agar and MacConkey agar, and then incubated overnight at 37°C. Clinical isolates from diagnostic specimens were also collected from the microbiology laboratory. Aerobic Gram-negative bacteria (Enterobacteriaceae, Acinetobacter species, and Pseudomonas species) and Gram-positive bacteria (S. aureus, coagulase-negative staphylococci, and Enterococcus species) were saved from both the surveillance and the clinical specimens; however, repeat isolates of the same species with the same sensitivity pattern from the same patient in any 2-week period were excluded. Isolates from diagnostic samples were from a range of sites and include endotracheal aspirates, wound swabs, and blood. Each isolate was identified using standard biochemical tests and a modified MAST-ID system or API strips (BioMerieux, Basingstoke, UK).
Antimicrobial Susceptibility Testing
Each isolate was tested against a range of antibiotics using an agar incorporation method and multipoint inoculation. The breakpoints recommended in the British Society of Antimicrobial Chemotherapy guidelines for sensitivity testing were used.23 Gram-negative isolates were tested against cefuroxime (concentration, 4 mg/L); cefotaxime (concentration, 1 mg/L); ceftazidime (concentration, 2 mg/L); imipenem (concentration, 4 mg/L); ciprofloxacin (concentrations, 1 mg/L and 4 mg/L); gentamicin (concentration, 1 mg/L); and piperacillin/tazobactam (concentration, 16 mg/L). Gram-positive isolates were tested against ciprofloxacin at a concentration of 4 mg/L. Staphylococcus aureus and coagulase-negative staphylococci were also tested against methicillin (concentration, 4 mg/L). The control strains used were P. aeruginosa NCTC 10662, E. coli NCTC 10418, and S. aureus NCTC 6571, as appropriate.
Molecular Strain Typing Methods
All isolates of Enterobacteriaceae and Pseudomonas species, which were collected during the study period from the Leeds center and were resistant to one or more among ciprofloxacin, cefotaxime, or ceftazidime, were typed using a pulsed-field gel electrophoresis (PFGE) method.24 The Acinetobacter species collected from both centers that were resistant to ciprofloxacin were typed using repetitive extragenic palindromic polymerase chain reaction.25 A representative sample of methicillin-resistant Staphylococcus aureus (MRSA)26 Isolates of Enterococcus species collected at the Birmingham center were typed using PFGE.27
The χ2 test was used to compare the data from the first year of the study with the data from the second year of the study. Contingency tables (size, 2 × 2) were used to compare the proportion of sensitive and resistant bacteria from each year. Yates correction was applied where necessary. Differences were considered statistically significant at P ≤ 0.05. Linear regression was performed by straight-line fit using the least squares model (Excel for Windows '95; Microsoft Corporation).
The demographic data from both centers are shown in Table 1. The number and the type of patients were comparable between the 2 years at each center. Birmingham and Leeds are the second and sixth largest cities in England, about 140 miles apart in 2 separate healthcare systems, thus making an exchange of patients, physicians, and nurses unlikely.
The antibiotics used during the study period are shown in Table 2 and Figure 1. During the baseline period, the usage of ciprofloxacin and cefotaxime/ceftazidime was comparable with that used during the separate quarters of the first year. As per the study protocol in the second year, there was a significant increase in the use of ciprofloxacin at both centers (P = 0.008 for Leeds, P = 0.02 for Birmingham); likewise, there was a significant decrease in the use of cefotaxime/ceftazidime (P = 0.03 for Leeds, P = 0.0004 for Birmingham). The usage of other antibiotics remained the same between the 2 years except that flucloxacillin usage in Leeds increased significantly (P = 0.0006) and meropenem usage in Birmingham decreased significantly (P = 0.007) in the second year.
Bacterial Isolates and Antimicrobial Susceptibility Testing
The numbers and percentage of Gram-negative strains resistant to antibiotics is shown in Table 3. At both centers, E. coli, Enterobacter species, and Klebsiella species were the among the most frequent bacterial isolates of Enterobacteriaceae found. Among the Enterobacteriaceae species, no significant difference in resistance to ciprofloxacin was found at either center between the 2 years (P > 0.05). Cefotaxime resistance was high-with more than a quarter of the isolates exhibiting resistance-but, again, there was no significant difference between the 2 years at either center (P > 0.05). Ceftazidime resistance in the Enterobacteriaceae species decreased significantly at both centers in the second year of the study (P < 0.05). There was no significant difference in resistance to ciprofloxacin, cefotaxime, and ceftazidime found in Acinetobacter species isolated in Leeds between the 2 years of the study (P > 0.05). This species was not isolated in Birmingham during the first year of the study; therefore, all results from the second year are considered significant. There was no significant difference in the resistance to ciprofloxacin found in Pseudomonas species isolated from patients in Leeds or Birmingham between the 2 years (P > 0.05). Resistance to cefotaxime remained at a high level at both centers, but there was no significant difference between the 2 years (P > 0.05). Although there was a decrease in ceftazidime resistance in Pseudomonas species at both centers, the difference was not significant (P > 0.05).
The numbers and percentage of antibiotic-resistant strains of Gram-positive bacteria that were isolated are shown in Table 4. Ciprofloxacin resistance in methicillin-sensitive S. aureus (MSSA) did not change significantly at either center between the 2 years (P > 0.05). There was no significant change in the number of ciprofloxacin-resistant MRSA at either center between the 2 years, although there was an upward trend (P > 0.05). This may be associated with the overall modest activity of ciprofloxacin against several Gram-positive species, such as staphylococci and streptococcal species. In Leeds, the small increase in nonduplicate MRSA isolated from the ICU was associated with an increase in MRSA on the 4 wards from which most referrals to the ICU were made (Fig. 2). There was a significant increase in the numbers of methicillin-sensitive, ciprofloxacin-resistant, coagulase-negative staphylococci in Leeds; however, in Birmingham, the small increase in numbers was not significant. There was no significant difference (P > 0.05) between the 2 years in the numbers of methicillin-resistant, ciprofloxacin-resistant, coagulase-negative staphylococci at either center but, again, the trend was upward. There was no significant difference in ciprofloxacin resistance found in isolates of Enterococcus species at either center (P > 0.05).
Molecular Strain Typing
No evidence was found for any major clusters of cross-infecting resistant Gram-negative bacteria during the study. However, a number of small, brief episodes were identified as being caused by Enterobacter species and Acinetobacter species. During a 14-month period, 8 individuals were infected with a strain of Enterobacter species that seemed indistinguishable using PFGE. There was also a strain of Acinetobacter species, isolated from 12 individuals during the 2 years of the study, that seemed to be indistinguishable using repetitive extragenic palindromic polymerase chain reaction. Analysis of the MRSA showed that 93% of the isolates tested had PFGE profiles that were indistinguishable from that of EMRSA-15. The PFGE typing demonstrated that there was proliferation of at least 3 clones of Enterococcus faecalis during the period of the study in the Birmingham center.
The appearance of antibiotic-resistant bacteria in the ICU is the result of continuous interaction between the mechanisms of resistance, direct and indirect selective pressure, and cross-infection/colonization by bacteria. Understanding these complex relationships requires as many variables as possible to be controlled.28,29 However, most epidemiological studies are undertaken in response to an outbreak of cross-infection by antibiotic-resistant bacteria. An important aspect of our investigation is the detailed, extended study of the effects of changes in empirical antibiotic therapy on both infecting and colonizing antibiotic-resistant/sensitive bacteria in 2 comparable but geographically separate ICUs. We have shown that the modification of an antibiotic policy in ICUs is associated with the changes in the antimicrobial resistance found in common ICU pathogens. In our study, ciprofloxacin replaced cefotaxime and ceftazidime as the preferred first-line agent for 1 year in 2 ICUs and, although the usage of this agent increased significantly in both centers, there was no statistically significant increase in the numbers of ciprofloxacin-resistant Enterobacteriaceae or Pseudomonas species isolated. The numbers of ceftazidime-resistant Enterobacteriaceae species decreased significantly at both centers largely because of a decrease in the numbers of ceftazidime-resistant Enterobacter species.
Other studies have also found that the modification of an antibiotic policy on intensive care is associated with a reduction in antibiotic-resistant, Gram-negative bacteria. An 18-month study conducted by Struelens et al20 and reported in abstract also found that the prevalence and spread of ceftazidime/ciprofloxacin-resistant Enterobacteriaceae was reduced after the restriction of third-generation cephalosporins and the introduction of cefepime. Unlike in our study, the use of ciprofloxacin was also restricted; however, it seems more likely that the reduction of third-generation cephalosporins eased the selective pressure. De Champs et al21 studied neonates receiving antibiotics and also found that the modification of an antibiotic policy was associated with a reduction of antibiotic-resistant Gram-negative bacteria; however, in this study, a switch from gentamicin to amikacin/ampicillin was implemented. The change in antibiotic policy implemented in both of these studies was an attempt to control an outbreak of antibiotic-resistant bacteria. A study by Gruson et al30 also investigated the effect of controlling antibiotic use in an ICU. In contrast with the 2 studies mentioned, this study was not in response to an outbreak, although it was designed to establish whether the problem of antimicrobial resistance could be combated. The findings demonstrated that controlling ceftazidime and ciprofloxacin use was associated not only with a reduction in the prevalence of antibiotic-resistant, Gram-negative bacilli but also seemed to result in a reduction in the incidence rate of ventilator-associated pneumonia.
In our study, there were no obvious outbreaks of antibiotic-resistant bacteria. However, by using molecular typing methods, the repeated importation of Enterobacter species resistant to third-generation cephalosporins was identified at 1 unit. Four of these patients were from a local district general hospital that regularly referred patients to the Leeds unit. On 1 occasion, there was a probable cross-infection of this strain to another patient; however, the 3 other individuals who were also infected with this strain were found to be on the unit on separate unrelated occasions. The existing infection-control procedures probably contained the spread of this clone, but this is likely to be an unidentified source of antibiotic-resistant bacteria in ICU patients that is rarely, if ever, addressed in other studies. Acinetobacter species that were largely resistant to ciprofloxacin were isolated in Birmingham in year 2; no Acinetobacter species had been isolated in the previous year. However, it is unlikely that selection by ciprofloxacin alone accounts for their presence because they were also largely resistant to third-generation cephalosporins. There also seemed to be a clone of Acinetobacter species that was sporadically isolated from the Leeds center. This strain had a characteristic profile and, although patients were admitted from various wards and hospitals, it is possible that this strain could have survived not only in the ICU but also in the general hospital environment. The reported tolerance to desiccation of Acinetobacter species31 could account for such survival. The number of isolates and the prevalence of resistance in Klebsiella species were too low to indicate significant effects on selection by antibiotic choice.
The policy in our units was not to prescribe ciprofloxacin for infections caused by Gram-positive bacteria. However, the use of any antibiotic will affect all infecting or colonizing bacteria, and our results show that there was an upward trend in the isolation rate of ciprofloxacin-resistant, coagulase-negative staphylococci, and MRSA. As reported by other authors,32 these increases may be caused by the dissemination of 1 or 2 clones and not as a direct result of ciprofloxacin use. To support this hypothesis, we saw a marked and parallel increase in the isolation of MRSA in other wards of both hospitals (Fig. 2). This was also in line with nationally observed increases.33 The PFGE typing indicated that most were indistinguishable from EMRSA-15. This clone and EMRSA-16 represent the most prevalent clones of MRSA in England and Wales.34 The proportions of MRSA that were ciprofloxacin resistant were high at both centers. Similarly, high proportions (74%) of MRSA that were ciprofloxacin resistant have been reported in a UK study of isolates causing invasive disease.30 The selective pressure for ciprofloxacin-resistant MRSA in our centers would have been essentially unaltered by the switch from cephalosporins to ciprofloxacin. In other hospitals, the effect of switching antibiotic on the selection of S. aureus, whether MRSA or not, may be different. Interestingly, there was an increase in the number of isolates of enterococci in year 2 in Birmingham only. This could result from the increased use of ciprofloxacin; however, it was not observed in Leeds or in either center during the period of increased cephalosporin use. This suggests that selective pressure was not the reason for this increase; furthermore, molecular typing suggested cross-infection as the cause.
Although it has been reported that quinolone use can be a risk factor for acquisition of a quinolone-resistant, Gram-negative isolate,14,35 reports of outbreaks resulting from large-scale use of ciprofloxacin are rare. However, caution must be exercised because this property of ciprofloxacin may not be reflected in other quinolones.36 Furthermore, cephalosporins also vary in their selective ability.37 In contrast with the rarity of outbreaks of quinolone-resistant organisms, the epidemics caused by bacteria producing extended-spectrum β-lactamases and derepressed AmpC β-lactamases are widely reported38,39 and the selective pressure caused by third-generation cephalosporins can lead to rapid rise in resistance rates. A study by Kolleff et al,40 which examined antibiotic class cycling in a cardiac ICU, demonstrated the differences between 2 classes (cephalosporins and quinolones) in the rate of resistance development. Of particular interest is the finding that a switch from ceftazidime to ciprofloxacin was associated with a reduction in ventilator-associated pneumonia due to antibiotic-resistant, Gram-negative organisms. This finding supports the observations from our study. Kolleff et al40 also showed increased treatment costs associated with this increased resistance. One cannot be certain that there really is a class difference in propensity to select resistance between cephalosporins and quinolones, although our findings and the findings of other investigators mentioned suggest that this may be so. Probably, prolonged ciprofloxacin use is associated with an increase in ciprofloxacin resistance; however, the follow-up period in our study was too short to demonstrate this. There is now a need for coordinated prospective, multicenter studies where defined empirical antibiotic regimens are implemented and compared with the isolation rates and where the species/strain types of antibiotic-resistant bacteria are isolated. Changes in preferred antibiotic choice should be made on the basis of changes in resistance rates in the more prevalent pathogens as indicated by routine surveillance. In this challenging climate of economic restraint and increasing antibiotic resistance worldwide, continued improvement of antibiotic policies, together with the monitoring of bacterial resistance, is a rational way of controlling the major healthcare problems of antibiotic resistance.
We thank Jenny Andrews for performing the antibiotic sensitivity testing of all Gram-positive isolates.
1. House of Lords Select Committee on Science and Technology. Resistance to Antibiotics and Other Antimicrobial Agents
. London, England: The Stationery Office; 1998. 7th Report. Session 1997-98.
2. Hawkey PM. Action against antibiotic resistance: no time to lose. Lancet
3. Verbist L. Incidence of multi-resistance in Gram-negative bacterial isolates from intensive care units in Belgium: a surveillance study. Scand J Infect Dis
4. Spencer RC. Cross-susceptibility of cefpirome and four other β-lactams against isolates from haematology/oncology and intensive care units. Scand J Infect Dis
5. Burwen DR, Banerjee SN, Gaynes RP. Ceftazidime resistance among selected nosocomial Gram-negative bacilli in the United States. National Nosocomial Infections Surveillance System. J Infect Dis
6. Vincent JL, Bihari DJ, Suter PM, et al. The prevalence of nosocomial infection in intensive care units in Europe. JAMA
7. Piddock LJV, Walters RN, Jin YF, et al. Prevalence and mechanism of resistance to 'third-generation' cephalosporins in clinically relevant isolates of Enterobacteriaceae from 43 hospitals in the UK, 1990-1991. J Antimicrob Chemother
8. Shanahan PMA, Thomson CJ, Miles RS, et al. Antibiotic resistance in nosocomial isolates in Scotland. Int J Antimicrob Agents
9. O'Brien TF. The global epidemic nature of antimicrobial resistance and the need to monitor and manage it locally. Clin Infect Dis
. 1997;24(suppl 1):S2-S8.
10. Verbist L. Epidemiology and sensitivity of 8625 ICU and hematology/oncology bacterial isolates in Europe. Scand J Infect Dis
11. Bryce EA, Smith JA. Focused microbiological surveillance and Gram-negative β-lactamase-mediated resistance in an intensive care unit. Infect Control Hosp Epidemiol
12. Spencer RC. Predominant pathogens found in the European prevalence of infection in intensive care study. Eur J Clin Microbiol Infect Dis
13. Mulgrave L. The changing ecology of hospital bacteria and the selective role of cephalosporins. Epidemiol Infect
14. Richard P, Delangle MH, Merrien D, et al. Fluoroquinolone use and fluoroquinolone resistance: is there an association? Clin Infect Dis
15. Segreti J, Connelly R. Effect of quinolone use on antimicrobial susceptibility patterns over a 5-year period. Drugs
. 1995;49(suppl 2):185-187.
16. Jacobson KL, Cohen SH, Inciardi JF, et al. The relationship between antecedent antibiotic use and resistance to extended-spectrum cephalosporins in Group I β-lactamase-producing organisms. Clin Infect Dis
17. Burman LG, Haeggman S, Kuistila M, et al. Epidemiology of plasmid-mediated β-lactamases in Enterobacteria in Swedish neonatal wards and relation to antimicrobial therapy. Antimicrob Agents Chemother
18. Hanberger H, Hoffmann M, Lindgren S, et al. High incidence of antibiotic resistance among bacteria in 4 intensive care units at a university hospital in Sweden. Scand J Infect Dis
19. White RL, Friedrich LV, Mihm LB, et al. Assessment of the relationship between antimicrobial usage and susceptibility: differences between the hospital and specific patient areas. Clin Infect Dis
20. Struelens MJ, Byl B, Govaerts D, et al. Modification of antibiotic policy associated with decrease in antibiotic-resistant Gram-negative bacilli in an intensive care unit. In: Program and Abstracts of the 38th Interscience Conference on Antimicrobial Agents and Chemotherapy, San Diego, CA, 1998. Washington, DC: American Society for Microbiology; 1998:502. Abstract K-12.
21. De Champs C, Franchineau P, Gourand JM, et al. Clinical and bacteriological survey after change in aminoglycoside treatment to control an epidemic of Enterobacter cloacae
. J Hosp Infect
22. de Man P, Verhoeven BA, Verbrugh HA, et al. An antibiotic policy to prevent emergence of resistant bacilli. Lancet
23. Working Party of the British Society of Antimicrobial Chemotherapy. A guide to sensitivity testing. J Antimicrob Chemother
. 1991;27(suppl D):1-50.
24. Gautom RK. Rapid pulsed-field gel electrophoresis protocol for typing of Escherichia coli
0157:H7 and other Gram-negative organisms in 1 day. J Clin Microbiol
25. Snelling AM, Gerner-Smidt P, Hawkey PM, et al. Validation of use of whole-cell repetitive extragenic palindromic sequence-based PCR (REP-PCR) for typing strains belonging to the Acinetobacter calcoaceticus-Acinetobacter baumannii
complex and application of the method to the investigation of a hospital outbreak. J Clin Microbiol
26. Bannerman TL, Hancock GA, Tenover FC, et al. Pulsed-field gel electrophoresis as a replacement for bacteriophage typing of Staphylococcus aureus
. J Clin Microbiol
27. Murray BE, Singh KV, Heath JD, et al. Comparison of genomic DNAs of different enterococcal isolates using restriction endonucleases with infrequent recognition sites. J Clin Microbiol
28. Struelens MJ, Baudouin B, Vincent JL. Antibiotic policy: a tool for controlling resistance of hospital pathogens. Clin Microbiol Infect
. 1999;5(suppl 1):S19-S24.
29. McGowan JE. Do intensive hospital antibiotic control programs prevent the spread of antibiotic resistance? Infect Control Hosp Epidemiol
30. Gruson D, Hilbert G, Vargas F, et al. Rotation and restricted use of antibiotics in a medical intensive care unit. Am J Respir Crit Care Med
31. Jawad A, Seifert H, Snelling AM, et al. Survival of Acinetobacter baumannii
on dry surfaces: comparison of outbreak and sporadic isolates. J Clin Microbiol
32. Dalhoff A. Quinolone resistance in Pseudomonas aeruginosa
and Staphylococcus aureus
. Development during therapy and clinical significance. Infection
. 1994;22(suppl 2):S111-S121.
33. Speller DC, Johnson AP, James D, et al. Resistance to methicillin and other antibiotics in isolates of Staphylococcus aureus
from blood and cerebrospinal fluid, England and Wales, 1989-95. Lancet
34. Marples RR, Speller DCE, Cookson BD. Prevalence of mupirocin resistance in Staphylococcus aureus
. J Hosp Infect
35. Muder RR, Brennen C, Goetz AM, et al. Association with prior fluoroquinolone therapy of widespread ciprofloxacin resistance among Gram-negative isolates in a Veteran Affairs medical center. Antimicrob Agents Chemother
36. Rifenburg RP, Paladino JA, Bhavnani SM, et al. Influence of fluoroquinolone expenditures on total antibiotic expenditures and Pseudomonas aeruginosa
susceptibility. Am J Health Syst Pharm
37. Sanders CC. Cefipime: the next generation? Clin Infect Dis
38. Lucet JC, Chevret S, Decre D, et al. Outbreak of multiply resistant Enterobacteriaceae in an intensive care unit: epidemiology and risk factors for acquisition. Clin Infect Dis
39. Meyer KS, Urban C, Eagan JA, et al. Nosocomial outbreak of Klebsiella infection resistant to late-generation cephalosporins. Ann Intern Med
© 2006 Lippincott Williams & Wilkins, Inc.
40. Kolleff MH, Vlasnik J, Sharples L, et al. Scheduled change of antibiotic class: a strategy to decrease the incidence of ventilator-associated pneumonia. Am J Respir Crit Care Med