JAC Advance Access originally published online on August 18, 2006
Journal of Antimicrobial Chemotherapy 2006 58(4):878-881; doi:10.1093/jac/dkl343
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Effect of individual- and group-level antibiotic exposure on MRSA isolation: a multilevel analysis
1 Service d'Hygiène hospitalière et d'Epidémiologie moléculaire, Centre Hospitalier Universitaire Jean Minjoz 25030 Besançon, France 2 Département d'information médicale, Centre Hospitalier Universitaire Jean Minjoz Besançon, France 3 Tayside Centre for General Practice, University of Dundee UK 4 Service de Prévention et Contrôle de l'Infection Hôpitaux Universitaires de Genève, Switzerland
*Corresponding author. Tel: +33-363082240; Fax: +33-381668914; E-mail: xavier.bertrand{at}univ-fcomte.fr
Received 16 May 2006; returned 20 July 2006; revised 27 July 2006; accepted 31 July 2006
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
Objectives: To observe the relative role of individual and group-level antimicrobial selective pressure on subsequent methicillin-resistant Staphylococcus aureus (MRSA) isolation in a university hospital.
Methods: For this purpose, 18 596 patients were included in a retrospective statistical analysis, applying multilevel modelling with discrete time intervals at the lowest level. Individual-level and hospital group variables on antimicrobial exposure and MRSA colonization pressure were collected from computerized databases.
Results: The simultaneous hospital group- and individual-level analysis showed individual exposure to fluoroquinolones and collective exposure to penicillins to be associated with MRSA isolation after adjustment for colonization pressure and other potential confounders.
Conclusions: These results support efforts to reduce prescriptions of selected antimicrobial drug classes such as fluoroquinolones and show the added value of multilevel analysis for research on the adverse outcomes of antibiotic prescribing.
Keywords: methicillin-resistant Staphylococcus aureus , antimicrobial use , individual exposure , ecological bias
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
Recent studies have highlighted the importance of antibiotic exposure as a significant risk factor for the acquisition and transmission of methicillin-resistant Staphylococcus aureus (MRSA).1 However, the effect of this individual-level antibiotic exposure can be decreased or amplified as a result of an interaction between the individual and the group effect.2 This group-level effect (also called ecological effect) may be particularly important for Gram-positive pathogens, such as MRSA. In a recent article, Monnet et al.3 have reported that, at a hospital level, use of antimicrobial drugs may be an important factor in perpetuating a hospital-wide MRSA outbreak. We have also demonstrated a relationship between antimicrobial use and MRSA spread at the hospital unit (HU) level.4 Notwithstanding their many pitfalls, ecological studies provide a potentially useful function in studies of infectious agents, because they allow measurement of the global effect of an exposure. This is important, because the global effects of antibiotics encompass not just the direct effects on the individual who receives the antibiotic but also the indirect effects mediated by effects on transmissibility or on the likelihood of transmission of susceptible organisms.5 The present study was specifically designed to determine the relative part of individual- and group-level (HU) antimicrobial pressure on subsequent MRSA isolation. For this purpose, we used advanced statistical multilevel modelling, which takes account of factors at the individual and group level simultaneously.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
Setting, study period and patients
The Besançon Hospital is a French university-affiliated hospital. Data for year 2001 were collected for the following departments: medicine, surgery and adult intensive care. Psychiatric, paediatric and gynaecology-obstetric units were excluded. All patients admitted for more than 48 h were included in the study. For each patient, only the first hospitalization during the study period and only the time spent in the HU where the patient was present 48 h after admission were retained for the analysis.
MRSA status
MRSA-positive patients were defined as patients who had a clinical specimen (requested for diagnostic work-up) that yielded MRSA and who were not known to be positive for MRSA during the previous 12 months. Screening samples were not considered for ascertainment of the MRSA status. Patients in whom MRSA was isolated within 48 h after admission (imported MRSA cases) were excluded.
Data structure
Level 1: Time.. The lowest level was that of discrete 24 h time periods nested within individual patients. The history of each patient in the HU was treated as a series of discrete 24 h periods, excluding the first 2 days of hospitalization. For individual colonization pressure and antibiotic exposure variables, each day of hospitalization was coded as negative before occurrence and as positive from the first day it occurred to the last day of the follow-up. Patients without MRSA isolation were censored at the time of HU discharge or in-hospital death.
Level 2: Individual..
Individual patients constituted the second level of the data. Individual-data variables on age (> or 
60 years), sex and individual MRSA colonization pressure (presence or absence in the HU of at least one patient known to be positive for MRSA during the hospitalization) were collected from computerized databases. Individual exposure to each of the four antimicrobial classes of interest defined (see below) was expressed as a binary, time-varying covariate (exposed/non-exposed on a given day).
Level 3: Hospital unit (HU).. The highest level of data was at the HU level. The HUs were classified into three types: medicine (n = 31), surgery (n = 17) and intensive care (n = 2). The average antibiotic use was 718.5 DDD per 1000 patient-days, ranging from 535.4 in surgical units to 808.6 in medical units and 1517.7 in ICUs. Antibiotics were grouped into five classes: penicillins, cephalosporins, fluoroquinolones (96% inactive against MRSA), glycopeptides (100% active against MRSA) and other antibiotics (including macrolides, lincosamides, aminoglycosides) with variable activity against MRSA. The distribution of antibiotic use by class was penicillins, 51.3%; cephalosporins, 10.9%; fluoroquinolones, 18.6%; glycopeptides, 3.6%; and other antibiotics, 15.6%. We only retained penicillins, cephalosporins, fluoroquinolones and glycopeptides for the analysis. The HUs were divided into two categories using the median cut-point for each of the four retained antimicrobial classes (weak versus high consumer). In addition, the mean colonization pressure was calculated for each unit from the ratio of the number of patient-days of MRSA patients to the total number of patient-days.
Statistical analysis
We analysed data using a multilevel discrete-time logistic regression model with MRSA ascertainment as binary outcome.6 This approach allowed us to consider a nested hierarchical structure of the data. Initially, each variable was tested in a univariate model adjusted for the time at risk. Second, a reference model was built by introducing the variables with P < 0.20 in the univariate analysis, with the exception of antimicrobial exposure. Both individual and collective antimicrobial exposure variables were simultaneously forced into the reference model. Modelling was performed using MlwiN V2.0 software.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
In 2001, 41 790 patients (excluding re-admissions) were admitted, of whom 18 596 were included in our study. We identified 59 MRSA-positive patients according to our selection criteria. Characteristics of the patients are shown in Table 1. In univariate analysis, group-level MRSA colonization pressure (P = 0.002), group-level penicillin use (P = 0.001) and individual fluoroquinolone exposure (P < 0.001) were associated with MRSA isolation. In Figure 1, MRSA isolation is plotted by individual antibiotic exposure and level of antibiotic use in the HU. Figure 1(a) suggests a unique increase with the individual fluoroquinolone exposure regardless of the level of fluoroquinolone use in the HU. Conversely, Figure 1(b) suggests a unique increase with high level of penicillin use in the HU regardless of the individual penicillin exposure.
|
|
The simultaneous hospital group- and individual-level analysis showed individual exposure to fluoroquinolones and collective exposure to penicillins to be associated with MRSA isolation after adjustment for colonization pressure and other potential confounders (Table 2). For fluoroquinolones, the effect of individual exposure was not modified when adjusted on collective exposure. Conversely, for penicillins, the observed collective effect was not influenced by individual exposure.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
The results reported here demonstrate a significant association between antibiotic exposure and subsequent isolation of MRSA in our hospital. This relationship persists when other identified risk factors such as age, sex, MRSA colonization pressure and the type of unit are taken into account. The advanced multilevel analysis fits the complex structure of the data and allows differentiation between individual and collective antibiotic exposure for each antimicrobial class.
Regarding the fluoroquinolone effect, our results are concordant with those of Weber et al.,1 who demonstrated that this class has a specific individual effect on MRSA. Our multilevel model shows that this effect is observed regardless of the amount of fluoroquinolones used at the HU level. It suggests that the ecological effect of fluoroquinolones reported by previous studies may just reflect the sum of individual effects.3,4,7 Fluoroquinolones, which are frequently ineffective against nosocomial MRSA,8 have an excellent tissue diffusion, which could promote the acquisition of MRSA by eradicating susceptible microorganisms, such as methicillin-susceptible Staphylococcus aureus (MSSA). Moreover, Bisognano et al.9 have demonstrated that exposure to subinhibitory levels of ciprofloxacin results in increased expression of adherence factors promoting host colonization. It seems that the combination of the two mechanisms gives a plausible explanation for the specific effect of fluoroquinolones on MRSA: fluoroquinolone exposure would promote Staphylococcus aureus colonization while selectively eradicating MSSA strains.
We observed a group-level effect of penicillin use on MRSA isolation and this could not be explained by an ecological fallacy because both individual- and group-level antibiotic exposure was considered using adequate statistical modelling. Thus, an ecological effect purely explained by the aggregated effect of individual exposures can be excluded. The observed ecological effect could be due either to a confounding factor or to a real effect. Penicillins are the most frequently prescribed antibiotic class in our hospital, making it possible that an ecological effect would only manifest above a certain threshold of use.
Some limitations of our study have to be addressed. First, our findings are supported by data collected in a single hospital. It would be of interest to apply our multilevel model to other settings. Second, due to statistical complexity, we used a nested multilevel model which implies that one level is related to only one upper level. So, we only retained the first unit-stay for analysis and consequently did not consider the entire spectrum of MRSA colonization occurring in our hospital. Third, our microbiological data, i.e. isolation of MRSA from clinical sample, were laboratory-based. We did not collect clinical information to confirm MRSA infection. MRSA screening on admission was performed for 15% of the patients admitted. So, we have not evidenced a specific association of antibiotic exposure with MRSA acquisition or with progression from MRSA colonization towards infection but an association between antibiotic exposure and a mix of these two stages.
Finally, our results are consistent with several studies, supporting a relationship between antimicrobial use, particularly individual exposure to fluoroquinolones, and MRSA spread.3,4 Interestingly, hospitals in Nordic European countries, with very low MRSA incidence, use the least fluoroquinolones.10 Our findings support efforts to further study the effect of implementing programmes to control antibiotic use. To conclude, our multilevel analysis shows that exploring the problem of antimicrobial resistance at the individual level or at the collective level alone will miss either of these aspects of the problem.
|
Transparency declarations |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
Conflicts of interests: none.
|
Acknowledgements |
|---|
This work was partially supported by a grant from the French Ministry of Health (regional hospital project of clinical research, 2002).
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
1 Weber SG, Gold HS, Hooper DC, et al. (2003) Fluoroquinolones and the risk for methicillin-resistant Staphylococcus aureus in hospitalized patients. Emerg Infect Dis 9:1415–22.[Web of Science][Medline]
2 Lipsitch M and Samore MH. (2002) Antimicrobial use and antimicrobial resistance: a population perspective. Emerg Infect Dis 8:347–54.[Web of Science][Medline]
3 Monnet DL, MacKenzie FM, Lopez-Lozano JM, et al. (2004) Antimicrobial drug use and methicillin-resistant Staphylococcus aureus, Aberdeen, 1996–2000. Emerg Infect Dis 10:1432–41.[Web of Science][Medline]
4 Muller A, Mauny F, Bertin M, et al. (2003) Relationship between spread of methicillin-resistant Staphylococcus aureus and antimicrobial use in a French university hospital. Clin Infect Dis 36:971–8.[CrossRef][Web of Science][Medline]
5 Harbarth S, Harris AD, Carmeli Y, et al. (2001) Parallel analysis of individual and aggregated data on antibiotic exposure and resistance in gram-negative bacilli. Clin Infect Dis 33:1462–8.[CrossRef][Web of Science][Medline]
6 Goldstein H, Pan H, Bynner J. (2004) A flexible procedure for analysing longitudinal event histories using a multilevel model. Understanding Statistics 3:85–9.[CrossRef]
7 MacDougall C, Harpe SE, Powell JP, et al. (2005) Pseudomonas aeruginosa, Staphylococcus aureus, and fluoroquinolone use. Emerg Infect Dis 11:1197–204.[Web of Science][Medline]
8
Thouverez M, Muller A, Hocquet D, et al. (2003) Relationship between molecular epidemiology and antibiotic susceptibility of methicillin-resistant Staphylococcus aureus (MRSA) in a French teaching hospital. J Med Microbiol 52:801–6.
9
Bisognano C, Vaudaux P, Rohner P, et al. (2000) Induction of fibronectin-binding proteins and increased adhesion of quinolone-resistant Staphylococcus aureus by subinhibitory levels of ciprofloxacin. Antimicrob Agents Chemother 44:1428–37.
10
Muller-Pebody B, Muscat M, Pelle B, et al. (2004) Increase and change in pattern of hospital antimicrobial use, Denmark, 1997–2001. J Antimicrob Chemother 54:1122–6.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  D. Talon, L. Excoffon, M. Tiv, A. Pincon, H. Gbaguidi-Haore, and X. Bertrand Environmental reservoirs of meticillin resistant Staphylococcus aureus in patients' rooms: Potential impact on care practices Journal of Infection Prevention, September 1, 2008; 9(5): 10 - 14. [Abstract] [PDF]
|
|
|
|
 |
|
 M. A. Aldeyab, D. L. Monnet, J. M. Lopez-Lozano, C. M. Hughes, M. G. Scott, M. P. Kearney, F. A. Magee, and J. C. McElnay Modelling the impact of antibiotic use and infection control practices on the incidence of hospital-acquired methicillin-resistant Staphylococcus aureus: a time-series analysis J. Antimicrob. Chemother., September 1, 2008; 62(3): 593 - 600. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Vernaz, H. Sax, D. Pittet, P. Bonnabry, J. Schrenzel, and S. Harbarth Temporal effects of antibiotic use and hand rub consumption on the incidence of MRSA and Clostridium difficile J. Antimicrob. Chemother., September 1, 2008; 62(3): 601 - 607. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Harbarth and M. H. Samore Interventions to control MRSA: high time for time-series analysis? J. Antimicrob. Chemother., September 1, 2008; 62(3): 431 - 433. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Eveillard, P. Charru, P. Rufat, M.-C. Hippeaux, E. Lancien, F. Benselama, and C. Branger Methicillin-resistant Staphylococcus aureus carriage in a long-term care facility: hypothesis about selection and transmission Age Ageing, May 1, 2008; 37(3): 294 - 299. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. J. Dancer The effect of antibiotics on methicillin-resistant Staphylococcus aureus J. Antimicrob. Chemother., February 1, 2008; 61(2): 246 - 253. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||