Curbing Methicillin-Resistant Staphylococcus aureus in 38 French Hospitals Through a 15-Year Institutional Control Program FREE

The rates of methicillin-resistant Staphylococcus aureus (MRSA) have reached high levels during recent decades in many countries all over the world, although some have succeeded in containing the spread of this multidrug-resistant organism.^1- 3 Controlling MRSA is a challenge for medical institutions in countries with a high MRSA burden. Because MRSA spread in hospitals is driven by multiple factors such as cross-transmission, invasive procedures, and intensive antibiotic use, all guidelines for MRSA control include bundled measures such as identification and isolation of MRSA carriers, hand hygiene, and antibiotic policy.^2,4- 7

In France, the proportion of MRSA in clinical strains of S aureus increased continuously during the 1980s and reached approximately 35% including in blood cultures.^8- 14 Such rates became unacceptable after comparative European data were published.¹⁵ In response, a long-term program for controlling MRSA was launched in 1993 in the 38 hospitals of the Assistance Publique–Hôpitaux de Paris (AP-HP), the largest public medical institution in France. We present herein the results of an observational incidence study that assessed the positive impact of this large-scale initiative on MRSA rates over a period of 15 years (1993-2007), showing that high endemic rates can be markedly reduced by an intensive long-lasting program.

The AP-HP is a public institution that administers 38 teaching hospitals (23 acute care hospitals [ACHs] and 15 rehabilitation and long-term care hospitals [RLTCHs]) scattered across Paris (n = 15) and surrounding suburban counties (n = 23) in France. The AP-HP comprises 23 000 beds (14 300 in ACHs and 8700 in RLTCHs), representing 36% of all the hospital beds in the Paris area (“Île de France”; 11.6 million inhabitants) and 5% of all the beds in France, and admits approximately 1 million patients per year and employs 19 000 physicians, 18 500 nurses, and 29 800 assistant nurses. AP-HP hospitals are managed locally by administrators and medical committees, but decisions on large investments and medical developments are made by the central administration. A local team is in charge of prevention and surveillance of nosocomial infections in each hospital, but actions of foremost importance for the whole institution, such as campaigns for hygiene, are coordinated centrally.

A program aimed at controlling MRSA was launched in 1993, based on the following objectives:

Since 1998, our institution has further stimulated the efforts by asking each hospital to report annually to the central administration the following selected information on the program, as part of a wider process to promote the quality at AP-HP: (1) size of the infection control team, (2) implementation of the program, (3) organization of audits (eg, on hand hygiene), and (4) feedback. The progress of the initiative has been annually presented during meetings of infection control teams and bacteriologists from all AP-HP hospitals, yielding opportunities to discuss the results, and disseminated within each hospital and central administration.

To evaluate the impact of the program, a prospective MRSA survey was organized annually from 1993 onwards in AP-HP laboratories, following a standardized protocol. The survey was conducted during the same 6-week period (April to mid-May) from 1993 to 2001, then over 2 full months (April-May) from 2002 to 2004 and over 3 months (April-June) since 2005, to counterbalance the progressive decrease in the number of MRSA cases, given the need to accurately assess the rates of susceptibility to non–β-lactam antibiotics among MRSA strains and the distribution of MRSA cases by type of specimens and wards.

All nonduplicate strains of S aureus isolated from clinical specimens in patients hospitalized for at least 24 hours during the survey period were included. Susceptibility tests were performed using national guidelines.¹⁶ Duplicate strains, defined as strains with the same antibiotic susceptibility pattern as a strain already counted for the same patient, were excluded.^17- 18 An MRSA case was defined as a patient with at least 1 clinical specimen positive for MRSA during the survey period. Strains of MRSA recovered from nares swabs (screening) were not included in the calculation of MRSA rates.

For each strain of S aureus included, the following data were recorded: date of sampling, type of ward, type of specimen, and antibiotic susceptibility pattern. From 1996 onwards, the number of patient-days and admissions was recorded for the calculation of MRSA incidence rates.

The data were computerized annually in each laboratory using Epi-info 6.04 (Centers for Disease Control and Prevention, Atlanta, Georgia) and then centralized and merged in a MySQL database (Sun Microsystems Inc, Santa Clara, California) using Perl software. The results were given as proportion (expressed as percentage) of MRSA among clinical isolates of S aureus and MRSA incidence rates per 1000 hospital days (HDs) and, for ACHs, per 100 admissions. Relative changes in MRSA rates were calculated as percentage over all the study period and per year. The susceptibility of MRSA strains to non–β-lactam antibiotics was also assessed.

Following the launch in 2001 of the ABHR campaign, the annual ABHR solution use was recorded in liters per 1000 HDs using the data provided by the central pharmacy.

The χ² test for trend was used to analyze changes over time in the percentage of MRSA cases among clinical isolates of S aureus and of susceptibility to non–β-lactam antibiotics among MRSA strains, with R software (http://www.r-project.org). Poisson regression analysis was used to analyze changes over time in MRSA incidence rates, with SAS 8.02 software (SAS, Cary, North Carolina). Poisson confidence intervals were compared to analyze changes in MRSA incidence rates between the first and the last years of the survey using StatsDirect software (http://www.statsdirect.com). P < .05 was considered significant for all tests.

By 1995, 16 of the 23 ACHs and 5 of the 15 RLTCHs (P = .04) had implemented the program launched in 1993. Almost all the ACHs (19 of 23), but only 7 of the 15 RLTCHs, had done so in 1996 (P = .02). Reports sent from 1998 onwards by each hospital showed that notification of patients with MRSA infections or colonizations by laboratories had been implemented in all hospitals by 1998 and that active surveillance, targeting mainly ICUs, was implemented in almost all ACHs by 1998 and in all by 1999. Screening compliance was not recorded at central level but reached approximately 80% of the targeted patients in 12 ICUs of 5 hospitals that championed the initiative. By 2001, all ACHs, but a third of RLTCHs, had organized audits (eg, on hand hygiene), following local procedures and timing. Feedback information had been formalized by all hospitals by 2001.

Following the ABHR campaign, use of ABHR solution increased regularly between 2000 and 2007 (Figure 1) from 2 L to 21 L per 1000 HDs but reached a higher level in ACHs than in RLTCHs (26 L and 10 L per 1000 HDs, respectively).

The proportion of MRSA among S aureus strains decreased from 39.4% in 1993 to 21.6% in 2007 in ACHs (relative change, −45%; P < .001), interrupted by a temporary rebound in 2000-2001 (Figure 2). The decrease was sharper in ICUs (from 55.1% to 22.4%; relative change, −59%; P < .001) and in surgical wards (from 38.7% to 21.7%; relative change, −44%; P < .001) than in medical wards (from 33.1% to 22.5%; relative change, −32%; P < .001) (Figure 3). The decline was significant and similar for 3 strata of ACHs (400-500, 800-1000, and >1000 beds). The decrease was also sharp in the strains isolated from blood cultures (from 45.3% to 24.2%; P < .001) and from pus and serous fluids (from 40.4% to 18.0%; P < .001).

In RLTCHs, the proportion of MRSA among S aureus strains remained high throughout the survey, ranging between 54% and 73%. As a result, the global decrease at AP-HP as a whole was less marked than in ACHs (from 41.0% in 1993 to 26.6% in 2007; relative change, −35%; P < .001).

The incidence of MRSA cases decreased from 1.16 per 1000 HDs in 1996 to 0.57 in 2007 (relative change, −51%; P < .001) (Figure 2 and Figure 4). The decrease was sharper in ICUs (from 2.95 to 1.23 per 1000 HDs; relative change, −58%; P = .001) and in surgical wards (from 1.52 to 0.72 per 1000 HDs; −52%; P = .001) than in medical wards (from 0.70 to 0.48 per 1000 HDs; relative change, −31%; P = .04). When expressed per 100 admissions, the incidence of MRSA decreased in ACHs from 0.90 in 1996 to 0.44 in 2007 (P = .001).

In RLTCHs (Figure 4) the incidence rate initially increased from 0.49 to 0.94 per 1000 HDs between 1996 and 2001 (P = .001) and began to decrease only after 2001, down to 0.54 in 2007 (P = .001), ie, returning to the level observed in 1996. As a result, the global decrease in MRSA incidence in AP-HP as a whole was less marked than in ACHs, from 0.86 to 0.56 per 1000 HDs (relative change, −35%; P = .001).

When introducing 2 periods (period 1, up to the ABHR campaign launch; period 2, after the campaign launch) in a multivariate model, the period was a significant variable of the decrease in MRSA incidence, but not the year, and interaction between the study period and year was established, justifying a separate analysis for each period. In ACHs, the MRSA rate decreased less sharply in period 1 (relative change, −2% per year) than in period 2 (relative change, −4.7% per year). The contrast was particularly apparent in medicine, where the percentage increased in period 1 (relative change, +2.3% per year) but decreased in period 2 (relative change, −6.2% per year). In surgery, the relative decrease was −1.8% per year in period 1 and −4.7% per year in period 2, but remained steady over both periods in ICUs.

Between 1993 and 2007, the susceptibility rates among MRSA strains significantly increased (P < .001) for gentamicin (from 12% to 95%), erythromycin (from 8% to 63%), and rifampicin (from 27% to 92%). In contrast, the susceptibility rates remained unchanged for fluoroquinolones (≤10%) and fusidic acid (>85%). The proportion of MRSA strains with decreased susceptibility to vancomycin (vancomycin intermediate-resistant S aureus, defined as vancomycin minimum inhibitory concentration >4 mg/L) remained below 1%.

An MRSA control program was launched in 1993 in the 38 hospitals of the AP-HP institution in response to the results of the first European MRSA comparative study conducted in 1990 showing high rates (approximately 35% of MRSA among the clinical isolates of S aureus) in France, Italy, and Spain, contrasting with low rates (<2%) in northern European countries.¹⁵ These results triggered the decision to take prompt action in our institution. Our MRSA control program was similar to, and largely based on, the guidelines published by national agencies that were available at that time^2,4,19 and to which few changes have been made since.^5- 7 The program focused on a bundle of measures aimed at decreasing cross-transmission including single room placement, promotion of hand hygiene, active surveillance in high-risk patients, quick notification of cases, and feedback. Decolonization by mupirocin was not recommended due to the risk of resistance²⁰ and was used scarcely. We did not close units where MRSA rates were particularly prevalent. Controlling antibiotic use was not attempted at the institution level during the study, and the overall use of antibiotics at AP-HP remained stable at approximately 450 defined daily doses per 1000 HDs.²¹ The MRSA control program was progressively implemented at AP-HP, more readily and completely in ACHs than in RLTCHs. The campaign launched halfway through the study to promote the use of ABHR solutions, whereas classic hand washing was advocated in the first phase of the study, led to a dramatic increase in ABHR solution use, sharper in ACHs than in RLTCHs.

The main result of the present study was the clear decrease in MRSA burden (−35%) between 1993 and 2007, whether expressed as a proportion of MRSA among clinical isolates of S aureus or the incidence of MRSA. The decline was steady and sharp in ICUs (−59%), where control measures have been readily implemented, and to a lesser extent in surgical wards (−44%) that are strongly connected with ICUs (many ICU patients are admitted for postsurgical complications and return to surgical wards when improving). In medical wards, where patients with MRSA infections or colonizations were scattered over a large number of units, likely hindering measure application, the decrease was less marked (−32%) and was delayed until the launch of the ABHR campaign in 2001. During the early years of the study, there was a trend to sharper decline in a subset of ACHs that led the initiative and generally applied more readily the program (results not shown). In RLTCHs, the decrease was delayed until 2001 and only counterbalanced the increase that initially occurred in 1993 to 2001, likely due to the difficulty to implement complex bundle measures and to hidden MRSA reservoirs in such setting. The decrease of MRSA in blood cultures (−47%) or pus and serous fluids (−55%) added clinical relevance to the results, based on all clinical specimens and not targeting specific infections.

The change in MRSA resistance patterns (increased susceptibility to gentamicin), already reported in France,^9,22 was related to the rapid spread of new hospital clones with higher fitness that derived from older gentamicin-resistant clones,^23- 24 whereas clones producing the Panton-Valentine leukocidin (“community-acquired MRSA”), first identified in the early 2000s in French hospitals,²⁵ represented less than 2% of our MRSA cases during the survey.²⁶ Thus, virtually all our MRSA cases were hospital acquired, either locally or imported from other hospitals through previous stays or transfers, often within AP-HP, and were all recorded to assess the global MRSA burden in our institution. Vancomycin intermediate-resistant S aureus strains remained scarce as reported.²⁷

The major limitation of our study is its observational design and lack of a control group. Indeed, the aim was to curb as quickly as possible the high MRSA rates prevailing in the early 1990s, and the study was conducted as a pragmatic intervention program in our whole institution. Consequently, we should be cautious in concluding that the statistically significant association between the MRSA decline and our control program implies a causal association. However, several facts support such an association. First, previous reports from French hospitals, including those of AP-HP, showed that the proportion of MRSA among S aureus strains increased steadily from approximately 15% in the 1970s to approximately 35% in the early 1990s,^8- 14,22 indicating that MRSA burden was not declining when our program was initiated. In fact, our program was launched in reaction to increasing MRSA rates. These reports actually provide “preintervention” measurements, which allow the present study to be considered as a “pre-post intervention” study.^28- 29 Unfortunately, the number of points of measurements (1 per year) was too low to allow a time series analysis.³⁰ Second, the survey was carried out at the same time each year, avoiding seasonal differences, and used continuously the same definitions. The structure and recruitment of AP-HP hospitals did not change much either. Third, the efforts made to detect MRSA did not decline over time, since (1) volumes of clinical specimens did not decrease, allowing an overall inclusion of approximately 32 000 strains of S aureus, and (2) national guidelines published in 2000 introduced new sensitive tests to detect oxacillin resistance.¹⁶

Regression to the mean or local maturation effect are unlikely to explain the clear and sustained decline in MRSA rates observed in the present study that involved 38 hospitals covering a large geographical area.^28- 29,31 Interestingly, the decrease in MRSA burden at AP-HP during period 1 of the study (1993-2001) contrasts with a marked increase in MRSA rates that occurred between 1996 and 2000 in 23 hospitals scattered over the same part of France,³² but where MRSA control programs were introduced only in the early 2000s.³³ Furthermore, the decrease measured during our period 2 (2001-2007) is consistent with the results of the European Antimicrobial Resistance Surveillance System (EARSS), which showed a decrease in MRSA prevalence in blood cultures between 2001 and 2006 in a set of hospitals scattered over France,³ after the introduction of the national MRSA control program.³³ In contrast, EARSS data showed that MRSA rates were still on the increase during the years 2001 to 2006 in several countries with high MRSA burden, whereas they started to decline lately in several countries such as in the United Kingdom and Belgium.³

Effective control of MRSA in epidemic settings has been extensively demonstrated,^{2,5,7,19,31,34- 37} but the feasibility of controlling endemic situations with high MRSA rates has been questioned.^35- 37 Our study demonstrates that this objective can be reached at the scale of a large medical institution. Partial control of MRSA was achieved in Belgian hospitals in the mid-1990s³⁸ but was not maintained over time.³ Incidence of MRSA central line–associated bacteremia decreased in US ICUs between 1997 and 2007, likely due to bacteremia prevention, whereas the proportion of MRSA in S aureus strains actually increased.³⁹

Modeling studies showed that only bundled measures confer efficacy in controlling MRSA,⁴⁰ and all recommendations include such measures.^{5- 7,34,36- 37} The efficacy of single measures is difficult to evaluate separately,^{7,31,34- 36,41} particularly in our study owing to its design. However, it is noteworthy that MRSA rates declined first in ICUs where active surveillance (screening) has been already shown to improve patient contact isolation at AP-HP.^42- 44 Screening is widely recommended^7,34- 37 and allows early contact isolation and reduction in MRSA bacteremia in ICUs.⁴⁵ We did not include screening results in MRSA rates because the variability in number of beds targeted would have introduce bias.⁴⁶ Importantly, MRSA rates began to decline in medical wards and RLTCHs only during period 2 of the study, ie, after launching our ABHR campaign. Extensive use of ABHR solutions is known to increase hand hygiene compliance and MRSA control^35,47- 48 and ABHR use is now used as a crude marker of hand hygiene in France.⁴⁹ Audits on hand hygiene organized at the local level likely played a role to improve compliance but were not recorded at central level owing to variability in methodology. At present, ABHR solutions are part of international actions in health care settings.⁵⁰

Other measures likely played a role such as signaling procedures that helped to better identify patients with MRSA infections or colonizations¹⁹ and feedback that stimulated the efforts.^36,51 Finally, a sustained leadership involved intensivists, bacteriologists, and infection control practitioners who championed the initiative, and a strong commitment of AP-HP administration provided an efficient incentive.

Although the decline in MRSA burden at AP-HP is very encouraging and led to generalize our initiative at the national level,^33,49 the rates remained high in 2007 compared with those in northern Europe. Patients with MRSA infections or colonizations discharged from AP-HP still represent a source of secondary cases in household contacts.⁵² We should therefore maintain our efforts, particularly through antibiotic policy that is at present part of French national programs.⁴⁹ Specific actions targeting serious infections such as catheter-related bacteremia would also contribute to hasten MRSA control.³⁹

Correspondence: Vincent Jarlier, MD, PhD, Laboratoire de Bactériologie-Hygiène, Groupe Hospitalier Pitié-Salpêtrière, 47 Blvd de l’Hôpital, 75651 Paris CEDEX 13, France (vincent.jarlier@psl.aphp.fr).

Accepted for Publication: October 13, 2009.

Author Contributions: Dr Jarlier had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Jarlier, Marty, Andremont, Carlet, Gottot, Nicolas-Chanoine, and Aggoune. Acquisition of data: Jarlier, Trystram, Fournier, Carbonne, Marty, Andremont, Arlet, Buu-Hoi, Decré, Gutmann, Joly-Guillou, Legrand, Nicolas-Chanoine, Soussy, Wolf, and Brücker. Analysis and interpretation of data: Jarlier, Trystram, Brun-Buisson, Carbonne, Marty, Lucet, Brücker, and Régnier. Drafting of the manuscript: Jarlier, Fournier, and Gutmann. Critical revision of the manuscript for important intellectual content: Jarlier, Trystram, Brun-Buisson, Carbonne, Marty, Andremont, Arlet, Buu-Hoi, Carlet, Decré, Gottot, Joly-Guillou, Legrand, Nicolas-Chanoine, Soussy, Wolf, Lucet, Aggoune, Brücker, and Régnier. Statistical analysis: Trystram, Carbonne, and Marty. Administrative, technical, and material support: Brun-Buisson, Fournier, Carbonne, Marty, Andremont, Buu-Hoi, Gutmann, Joly-Guillou, Legrand, Soussy, Wolf, Lucet, Aggoune, and Brücker. Study supervision: Jarlier, Nicolas-Chanoine, and Régnier.

Financial Disclosure: None reported.

Additional Information: Dr Jarlier coordinated the survey and wrote the manuscript. Drs Trystram, Carbonne, and Marty were the successive data managers of the survey. Dr Brun-Buisson participated in the program design and headed the AP-HP infection control committee between 2000 and 2007. Drs Fournier, Lucet, and Brücker and Ms Aggoune were successively members of the infection control team at AP-HP central administration. Drs Andremont, Arlet, Buu-Hoi, Carlet, Decré, Gottot, Gutmann, Joly-Guillou, Legrand, Nicolas-Chanoine, Soussy, and Wolf participated in the program design. Drs Andremont, Gutmann, Nicolas-Chanoine, and Soussy participated in launching the bacteriological survey. Dr Régnier initiated the program and headed the AP-HP infection control committee between 1993 and 2000.

Additional Contributions: Stéphanie Depeigne, PharmD (central AP-HP pharmacy), provided data on alcohol-based hand-rub solutions and antibiotic use; Michelle Huang (intensive care nurse, AP-HP Central Infection Control Team) analyzed AP-HP hospital reports to the central administration; Yves Péan, MD, and Luis Stuardii, MD (Institut Mutualiste Monsouris), assisted in database design; and Pascal Astagneau, MD, PhD (Public Health Department, Université Pierre et Marie Curie, Paris), and Isabelle Arnaud, MSc (CCLIN Paris-Nord), assisted in statistical analysis.