Impact of infection control interventions and antibiotic use on hospital MRSA: a multivariate interrupted time-series analysis

International Journal of Antimicrobial Agents 30 (2007) 169–176

Impact of infection control interventions and antibiotic use on hospital MRSA: a multivariate interrupted time-series analysis夽 A. Mahamat a , F.M. MacKenzie b , K. Brooker c , D.L. Monnet d , J.P. Daures a , I.M. Gould b,∗ b d

a Laboratory of Epidemiology, Clinical Research Institute, Montpellier, France Department of Medical Microbiology, Aberdeen Royal Infirmary, Foresterhill, Aberdeen AB25 2ZN, UK c Dr Gray’s Hospital, Elgin, UK National Center for Antimicrobials and Infection Control, Statens Serum Institut, Copenhagen, Denmark

Received 13 December 2006; accepted 5 April 2007

Abstract Hospitals in the northeast of Scotland have experienced methicillin-resistant Staphylococcus aureus (MRSA) outbreaks since 1997. Several infection control measures were introduced sequentially to control MRSA, and antibiotic use has been monitored. From January 1997 to December 2004, data on the monthly percentage of non-duplicate MRSA infections (%MRSA) were collated from an intervention hospital (IH) and a control hospital (CH). Both hospitals introduced the use of alcohol hand gel in November 2002. Furthermore, the IH introduced an environmental MRSA swabbing programme in March 2001, chlorine disinfection of the environment in September 2001, discharge screening in December 2001, admission screening in November 2003 and environmental audits in March 2004. Multivariate dynamic regression analysis was used to evaluate the longitudinal effects of these interventions as measured by new clinical cases of MRSA. At the IH, the %MRSA increased between January 1998 and January 2001 and then decreased. At the CH, the %MRSA increased from January 1997 to December 2004. Introduction of alcohol hand gel was associated with an absolute decrease in %MRSA of 21% and 30%, respectively, for the IH and CH. At the IH, introduction of chlorine disinfection and environmental swabbing were, respectively, associated with a decrease in %MRSA of 27% immediately and 32% 3 months later. Discharge screening and environmental audit did not significantly affect %MRSA, whereas admission screening was associated with a 22% decrease in %MRSA 4 months later. Increasing macrolide use was associated with increasing %MRSA in both hospitals, and increasing quinolone use was associated with increasing %MRSA in the CH. Implementation of stepwise infection control measures was associated with a decrease in %MRSA in the IH. Introduction of an alcohol gel for hand hygiene was associated with a decrease in %MRSA in both hospitals. Antibiotic use also affects %MRSA, in particular that of macrolides and quinolones. © 2007 Elsevier B.V. and the International Society of Chemotherapy. All rights reserved. Keywords: MRSA; Infection control; Antibiotic policy; Quasi-experimental study

1. Introduction The UK has had one of the highest methicillin-resistant Staphylococcus aureus (MRSA) rates of any European country for the past few years [1], with rates approaching those in the USA [2]. The epidemic spread to the Grampian region of Scotland in 1996 [3]. So far, community-acquired MRSA 夽 This original work has been presented in part as a poster at the 16th European Congress of Clinical Microbiology and Infectious Diseases (ECCMID), 1–4 April 2006, Nice, France. ∗ Corresponding author. Tel.: +44 1224 554 954; fax: +44 1224 550 632. E-mail address: [email protected] (I.M. Gould).

remains sporadic and the great majority of MRSA is hospital acquired [2]. It is believed that this MRSA epidemic has its origins in many different problems with National Health Service (NHS) hospitals and consequently the solutions are likely to be multifactorial [4,5]. Fighting epidemic MRSA involves re-enforcement of infection control measures as well as rational use of antimicrobials. However, robust evidence of which control measures are most effective remains largely elusive [2,4,5]. Hand hygiene is widely believed to be crucial, but its effective long-term implementation has proved problematic [2]. Variations on the search-and-destroy policy favoured by the Dutch and Scandinavians are widely practiced [6], particu-

0924-8579/$ – see front matter © 2007 Elsevier B.V. and the International Society of Chemotherapy. All rights reserved. doi:10.1016/j.ijantimicag.2007.04.005

170

A. Mahamat et al. / International Journal of Antimicrobial Agents 30 (2007) 169–176

larly for contacts of known cases, but admission screening [7] is not yet routinely performed on a hospital-wide basis. In the UK, isolation is often problematic owing to shortage of single rooms in most NHS hospitals, and there is no consensus about the role of environmental MRSA contamination and the need for disinfection of the close patient environment rather than simple cleaning with detergents [8]. This paper uses interrupted time-series analysis to evaluate the longitudinal effects of infection control interventions and antibiotic use on MRSA rates in two Grampian hospitals: one where several interventions were implemented to control MRSA such as environmental disinfection and admission screening; and the other where only standard precautions were used with the exception of contact tracing for clinical MRSA cases. Both hospitals introduced the same alcohol hand gel during the study period.

2. Materials and methods 2.1. Settings The intervention hospital (IH) is a 200-bed district general hospital with most specialties, located 75 miles from Aberdeen in the northeast of Scotland. The control hospital (CH) is a 300-bed hospital for medicine for the elderly and elective orthopaedic surgery situated in Aberdeen. 2.2. Microbiology data Monthly data for all S. aureus collected from clinical diagnostic specimens were exported from the clinical microbiology information system into a Microsoft Access database (Microsoft, Redmond, WA). Only the first clinical S. aureus isolate from each patient within 7 days was included in the analysis. Information stored included patient identifier, hospital, ward, specimen type and antimicrobial susceptibility pattern. All S. aureus obtained were screened for susceptibility to oxacillin and tested for susceptibility to a range of other antibiotics by the comparative disk susceptibility test method on nutrient agar at 30 ◦ C with 48 h incubation [9] up to January 2001; after that Clinical Laboratory Standards Institute agar disk diffusion was introduced [10]. The monthly prevalence of MRSA isolates was expressed as a percentage (%MRSA), where the denominator was the total number of S. aureus tested for methicillin resistance. Screening swabs were processed as previously described [11]. Briefly, swabs were enriched in salt broth up until October 2001 and subsequent to this were plated directly on ORSAB agar (Oxoid Ltd., Basingstoke, UK). Since 1999, most MRSA index isolates were investigated by the Scottish MRSA Reference Laboratory, which conducted independent confirmation and genotyping. The reference laboratory carried out multiplex polymerase chain reaction with primers for mecA, nuc, rRNA and 16S rRNA

[12–14] and pulsed-field gel electrophoresis typing of SmaIdigested DNA [15]. 2.3. Pharmacy data From 1 January 1997 to 31 December 2004, monthly quantities of all antimicrobial drugs were obtained from the pharmacy stock order computerised database. Similarly, numbers of occupied bed-days (OBD) per month were obtained from each hospital’s admission department. All data were stored in a Microsoft Access database. Antimicrobial drug use was expressed as the number of defined daily doses (DDD) per 1000 OBD, where the DDD for each antimicrobial drug was defined by the World Health Organization (WHO) [16]. As in most hospitals, data on individual patient exposure to antimicrobial drugs were not available. However, for a specific antimicrobial drug class the number of DDD approximates the average number of patients exposed to an antimicrobial drug from this class each day. This measurement is the unit the WHO recommends to express ecological pressure attributable to antimicrobial drugs [17,18]. 2.4. Interventions Both the IH and the CH introduced the use of the same alcohol gel for staff hand hygiene in November 2002. Furthermore, the IH introduced specific MRSA control measures, including environmental swabbing for MRSA in March 2001, terminal chlorine disinfection of the environment of MRSA isolation rooms and cohort areas in September 2001, discharge screening for MRSA carriage by full body swabs in December 2001, admission screening for MRSA carriage by nasal swabs in November 2003 and environmental audits in March 2004. Environmental swabbing for MRSA involved taking 20 swabs each week in two different wards/areas using Copan charcoal swabs (Sterlin, Stone, UK) moistened in sterile water. Items of patient care equipment, areas that were frequently touched such as telephones and door handles, and areas where dust may accumulate were swabbed. If any swabs were found to be positive, appropriate cleaning or disinfection was recommended and the area was re-swabbed to check the effectiveness of this cleaning. Terminal chlorine disinfection of the environment in isolation rooms and cohort areas was carried out upon discharge of known MRSA-positive patients by application of 1:1000 sodium hypochlorite (bleach) in place of detergent. Discharge screening for MRSA carriage was started in the same six wards in December 2001 and discontinued in November 2003 when admission screening was introduced. Each month, these same six wards would randomly screen six patients on discharge with a full body screen. Admission screening for MRSA carriage was introduced into the wards with the main MRSA problem, i.e. the six adult medical and surgical wards, in November 2003. All admissions were risk assessed and a bilateral anterior nares swab

OBD: Occupied bed-days. a The prevalence of each MRSA type is based on isolates sent to the reference laboratory for typing. This included screening and clinical isolates. Prior to 2000, only a random selection of strains were sent for analysis. Typing by pulsed-field gel electrophoresis was introduced in 2003. EMRSA denotes epidemic MRSA strains 15 and 16, which are the two main epidemic hospital clones in the UK.

0.0 50.0 50.0 6

0.0 98.0 2.0 50

8.8 91.2 0.0 57

59.4 39.1 1.5 64

59.1 37.5 3.4 88

77.5 15.0 7.5 80

0.0 96.0 4.0 25

1.2 95.3 3.5 86

10.2 85.2 4.6 88

38.5 53.8 7.7 91

54.6 44.5 0.9 110

67.0 22.7 10.3 185

171

MRSA prevalencea EMRSA-15 (%)a EMRSA-16 (%)a Other MRSA (%)a No. of typed MRSAa

325 6780 102901 19.0 325 7337 102579 16.9 310 6893 97518 18.1 302 6771 96157 18.1 311 7259 97057 17.2 332 7424 98019 16.6 181 14255 48453 5.0 174 13611 48996 5.1 174 13580 49415 5.3 190 14861 49230 4.9 190 14895 49891 5.2

2003 2002 2001 2000 1999 2000

2001

2002

2003

2004 1999

190 14277 48187 5.2 No. of beds No. of admissions No. of OBD Average length of hospital stay (days)

A comparison of the demographics of the two hospitals is shown in Table 1. From January 1997 to December 2004, 1271 non-duplicate clinical S. aureus were isolated from patients at the IH and the observed median monthly percentage of MRSA (%MRSA) was 14.2% (range 0.0–71.4%). At the CH during the same study period, the observed median monthly %MRSA was 30.2% (range 0.0–66.6%) among 2484 non-duplicate clinical S. aureus isolates. Among 345 and 585 MRSA clinical and screening isolates from the IH and CH, respectively, that were submitted for genotyping to the Scottish MRSA Reference Laboratory, the dominant epidemic clones were EMRSA-15 (41.5%)

Control hospital

3. Results

Intervention hospital

Interrupted time series was used, which is considered the strongest quasi-experimental approach [19–23] to evaluate longitudinal effects of infection control interventions on the %MRSA at the IH and the CH. For this purpose, segmented and dynamic regression (DR) analysis was performed with linear transfer function models built according to the DR model-building strategy proposed by Pankratz [24] and presented elsewhere [2,25]. These models are appropriate for studying the effect of interventions, taking account of data autocorrelation and exploring relationships between antimicrobial drug use and new clinical cases of MRSA (expressed as %MRSA; see Appendix A). Statistical analysis was performed with SAS/ETS (software v.8.1; SAS Institute Inc., Cary, NC).

Hospital characteristic

2.5. Statistical analysis

Table 1 Hospital demographics and methicillin-resistant Staphylococcus aureus (MRSA) characteristics at the intervention and control hospitals, 1999–2004

was taken if they fitted certain criteria, such as previous hospital admission in the last year or recent antibiotic therapy. A catheter specimen of urine was also taken, if appropriate, and swabs were taken of any wound/exit sites. High-risk patients were nursed in isolation, wherever possible, until results were known. If the initial nasal screen was found to be positive, a full body MRSA screen, including throat, axilla and groin, was performed. Environmental audits were carried out monthly throughout the whole hospital, with each ward/department being audited on a yearly basis. There were 13 sections, including waste, linen and patient bed areas. All areas within the ward/department were looked at to see whether there were any infection control issues, and a copy of the completed audit was fed back to the ward manager, hotel services manager, estates manager, head nurse and clinical nurse manager. The report includes an overall score (%) and a list of issues raised. During the study, the CH continued to rely on contact tracing and barrier nursing—in single rooms if possible—of known MRSA patients as its only MRSA-specific control measure. The number of single rooms was 36 in the IH and 76 in the CH. Both hospitals used a full set of body swabs for all contacts.

2004

A. Mahamat et al. / International Journal of Antimicrobial Agents 30 (2007) 169–176

172

A. Mahamat et al. / International Journal of Antimicrobial Agents 30 (2007) 169–176

Table 2 Characteristics of monthly antimicrobial use at the intervention and control hospitals, January 1997 to December 2004 Antimicrobial drug class (ATC code)a

Intervention hospital

Combinations of penicillins with ␤-lactamase inhibitors (J01CR) Macrolides (J01FA) Cephalosporins (J01DB-DE) Penicillins with extended spectrum (J01CA) ␤-Lactamase-resistant penicillins (J01CF) Fluoroquinolones (J01MA) Combinations of sulfonamides and trimethoprim (J01EE) Imidazoles (J01XD) ␤-Lactamase-sensitive penicillins (J01CE) Aminoglycosides (J01GB) Glycopeptides (J01XA) Tetracyclines (J01AA) Lincosamides (J01FF) Carbapenems (J01DH) Antibacterials for systemic use (J01, total)

Control hospital

Median (IQR) monthly use (DDD/1000 OBD)

Trend

Median (IQR) monthly use (DDD/1000 OBD)

Trend

207.2 (176.8–238.5) 109.7 (74.5–149.7) 57.1 (49.4–68.8) 58.7 (40.2–79.6) 113.9 (99.3–142.1) 30.6 (19.9–40.6) 57.3 (43.4–70.4) 24.3 (17.8–38.7) 36.8 (26.8–48.5) 3.6 (1.3–6.8) 1.9 (0.0–3.3) 24.8 (15.4–35.3) 2.7 (1.2–6.1) 0.5 (0.1–1.8) 749.4 (667.9–845.9)

Upwardb No Nob Upward Nob Upwardb Upward Upward Upward No Upward Downward No No Upwardb

110.1 (87.5–138.4) 31.5 (23.6–38.5) 30.1 (22.9–37.7) 27.8 (20.0–35.4) 35.1 (27.6–44.9) 28.2 (21.9–35.5) 47.4 (33.3–58.4) 6.4 (4.3–10.9) 11.2 (27.6–44.9) 1.6 (0.5–4.1) 1.8 (0.5–4.1) 3.5 (1.0–6.2) 1.8 (0.4–3.2) 0 353.9 (328.4–391.9)

Downward No Downward No No Upward No Upward No No Upward Upward No – Upwardb

IQR: Interquartile range; DDD: Defined daily doses; OBD: Occupied bed-days. a According to ATC classification. b Significant seasonal variation.

and EMRSA-16 (53.2%) (Table 1). Both EMRSA-16 and EMRSA-15 were co-resistant to all ␤-lactams, fluoroquinolones and macrolides, however they both remained susceptible to rifampicin, fusidic acid, gentamicin and amikacin. EMRSA-16 was more often co-resistant to mupirocin (9.6%) than EMRSA-15 (0%) (P < 0.0001). During this period, monthly use of all JO1 antibacterial agents (antibacterials for systemic use) showed an upward trend as well as seasonal variation with increased winter use both in the IH and the CH (Table 2). The graphic evolution of the %MRSA series for both hospitals is shown in Fig. 1. At the introduction of the new antibiotic policy in January 2000, both hospitals had a similar %MRSA (ca. 30% of all clinical S. aureus isolates) and this percentage showed a similar increase of 0.11% per month until January 2001. In the IH, the graphic evolution showed

a spring seasonal variation with two distinct epidemic periods. During the first period (January 1997–January 2001), the %MRSA showed a sustained absolute increase with marked peaks observed in February 2000 (38.4%) and January 2001 (71.4%). A sustained absolute decrease of the %MRSA began in the second period (February 2001–December 2004). In the CH, a sustained absolute increase of %MRSA was observed during the whole study period. Audit of the use of single isolation rooms in the IH from January to December 2004 showed that of 177 MRSA (63 new cases), 151 were nursed in single rooms (48 new cases). The rest were cohorted in bays. During 2004, 7 (0.7%) of 1064 environmental swabs were found to be positive for MRSA, i.e. 2 public telephones, 2 medical notes trolleys, a commode, a drug trolley and a patient’s wheelchair. Corresponding figures for pre-

Table 3 Impact of infection control interventions and antimicrobial use on monthly percentage of non-duplicate methicillin-resistant Staphylococcus aureus infections (%MRSA) in a multivariate dynamic regression at the intervention hospital Type of intervention

Effect decaya

P-value

3

Step

Wave

<0.0001

−26.6

0

Step

Exponential

<0.0001

November 2002

−21.0

0

Step

Wave

<0.05

November 2003

−21.4

4

Step

Wave

<0.05

– – – – –

– – – – –

<0.05 <0.001 <0.0001 <0.0001 <0.0001

Variable

Date or period

Direction and size of absolute effect on %MRSA

Environmental swabbing for MRSA Chlorine disinfection of environment Introduction of alcohol hand gel Admission screening for MRSA carriage Macrolides Autoregressive Autoregressive Seasonal autoregressive Seasonal autoregressive

March 2001

−31.5

October 2001

January 1997–December 2004 January 1997–December 2004 January 1997–December 2004 January 1997–December 2004 January 1997–December 2004

+0.4 −0.27 −0.87 −0.83 −0.61

Average delay for effect on %MRSA (months)

2 1 2 12 24

a The effect decay pattern controls how the effect of the intervention dissipates over time. In an exponential case, the effect declines at an exponential rate. Wave specifies that the effect declines like an exponential damped sine wave. In others case, the effect ends immediately.

A. Mahamat et al. / International Journal of Antimicrobial Agents 30 (2007) 169–176

173

41 (6%) to 83 (15%), with the average number being 58 (10%). Capture audits were also carried out to see whether any screens had been missed. In January and February 2004, 60 patients were audited, of which 14 were screened and 9 others (20%) should have been screened but were missed. In August 2004, 59 patients were audited, of which 16 were screened and 1 other (2%) should have been screened but was missed. In November and December 2004, 59 patients were audited, of which 14 were screened and 8 (18%) others should have been screened but were missed. All of those missed fell into the category of previous hospital admissions within the last year. Results for the discharge screening from June 2002 to May 2003 showed 25 new cases (5%) and 37 known positives from 504 patients screened. 3.1. Impact of interventions

Fig. 1. Smoothed (3-month moving average) monthly percentage of nonduplicate methicillin-resistant Staphylococcus aureus infections (%MRSA) series and interventions at (A) the intervention hospital and (B) the control hospital, January 1997–December 2004.

vious years were 1.6% of 1057 specimens in 2003, 1.3% of 940 specimens in 2002 and 2.9% of 723 specimens in 2001. Audit of the admission screen process in 2004 showed that the number of patients screened each month ranged from

Using interrupted time-series analysis of %MRSA in the IH, the slope (S.E.) was +0.006 (0.003) for the preintervention period and −0.009 (0.003) for the intervention period. Thus, the change in slope was −0.014 (P < 0.005) and the change in level was −17.0% for the IH. In the multivariate analysis, introduction of alcohol hand gel for hand hygiene was associated with a 21% and 30% decrease in %MRSA at the IH and CH, respectively. At the IH, environmental swabbing for MRSA and introduction of chlorine disinfection of the environment were, respectively, associated with a decrease in the %MRSA of 32% after 3 months and a decrease of 27% immediately. Discharge screening for MRSA carriage and environmental audit did not significantly affect the %MRSA, whereas admission screening was associated with a 22% decrease in the %MRSA with a delay of 4 months. Additionally, macrolide use was independently associated with an increase in %MRSA at the IH, i.e. an increase in macrolide use of 10 DDD per 1000 OBD was associated with an increase in %MRSA of 0.4% 2 months later (Table 3). At the CH, macrolide and fluoroquinolone use were independently associated with increase in %MRSA, i.e. increased use of 10 DDD per 1000 OBD of macrolides and fluoroquinolones resulted in an increase in the %MRSA of 3.1% and 3.2%, respectively, with a respective average delay of 5 and 2 months (Table 4).

Table 4 Impact of infection control interventions and antimicrobial use on percentage of non-duplicate methicillin-resistant Staphylococcus aureus infections (%MRSA) in a multivariate dynamic regression at the control hospital Variable

Date or period

Direction and size of absolute effect on %MRSA

Average delay for effect on %MRSA (months)

Type of intervention

Effect decaya

P-value

Introduction of alcohol hand gel Macrolides Fluoroquinolones

November 2002

−29.9

2

Step

Exponential

<0.001

+3.1 +3.2

5 2

– –

– –

<0.001 <0.001

January 1997–December 2004 January 1997–December 2004

a The effect decay pattern controls how the effect of the intervention dissipates over time. In an exponential case, the effect declines at an exponential rate. In others case, the effect ends immediately.

174

A. Mahamat et al. / International Journal of Antimicrobial Agents 30 (2007) 169–176

4. Discussion This study confirms our previous work on the importance of antibiotic prescribing in the evolution of hospital MRSA [3]. However, even if changes in the consumption of certain antibiotics are an independent factor for changes in %MRSA (Tables 3 and 4), it has traditionally proved difficult to reduce antibiotic consumption. Our report also confirms previous work of the immediate benefits of alcohol hand gel to improve hand hygiene and to reduce MRSA [2,5]. However, this was not enough to halt the MRSA epidemic in the CH, highlighting the need for a multifactorial approach to control MRSA [4]. Such an approach was followed in the IH, with sequential introduction of several other measures, most of which proved effective and cumulatively led to the %MRSA approximating that of the community (data not shown), thus suggesting virtual cessation of transmission in the hospital. Whilst feedback of positive environmental culture data to nursing and cleaning staff was felt to be useful in improving performance, it was also felt advisable to add terminal chlorine disinfection of MRSA isolation rooms and cohort areas to our control strategy. In fact, environmental cultures yielded little MRSA contamination of hand touch sites, both in isolation rooms used for MRSA as well as in open ward areas. Currently, however, whilst it is accepted that standard cleaning schedules are poor at removing MRSA from the environment [8], there is no consensus about the use of disinfectants [4]. This seems surprising in view of the potential role of environmental MRSA as a source of infection for patients and staff. Discharge screening for MRSA carriage, performed for most of 2002 and 2003, could not, in retrospect, be expected to lead to any immediate control of hospital MRSA acquisition, particularly in the absence of any properly enforced decolonisation of MRSA-positive patients in the community. It did, however, suggest transmission of MRSA to as many as 5% of inpatients prior to the intervention of admission screening. Actually, this figure of 5% is likely to be an overestimate as admission screening of all patients at another hospital in Grampian, which commenced in September 2003, indicated that ca. 2% of emergency admissions screened by nasal swabs were previously unknown MRSA-positive [26]. This suggests that in 2002–2003 perhaps 2–3% of discharges had acquired MRSA in hospital. There may be a further overestimate here though, as discharge screening involved the more sensitive full body swab screen. When the resources were redirected to nasal screening of risk-assessed admissions, there was an immediate and sustained beneficial effect, also described in the literature [7]. Contributing to this success was the ability to isolate most high-risk patients while awaiting culture results (usually a delay of 72 h or more due to the need to transport specimens over 70 miles to the laboratory). Environmental audits, whilst not having any apparent effect on MRSA since introduction, are still carried out monthly. They allow all areas of the ward/department to be

audited and any infection control issues highlighted. Often a member of staff from the ward will accompany the infection control nurse, allowing informal teaching and giving an alternative viewpoint. As each area is audited yearly, it also provides a comparison both over time and between areas. Meanwhile, in the CH, apart from its lower use of antibiotics (probably due to case-mix differences) and introduction of the same alcohol hand gel, no specific MRSA control measures were instigated, although the standard contact tracing of clinical positive cases was continued unchanged. We firmly believe that the consequence of not having specific anti-MRSA control measures was a much higher MRSA rate. Reliance on positive clinical specimens as the only means of identification of MRSA cases in a hospital seriously underestimates the infection burden or days of exposure that MRSA-negative patients are subjected to [27]. Whilst the study design was primarily observational, it had several strengths. It used segmented DR, which is less prone to bias, it involved interrupted time series with many data points before and after each intervention, and contained a control hospital where the rate and trend in %MRSA were very similar to that of the IH prior to implementation of the infection control interventions. The marked divergence in %MRSA between the two hospitals following these interventions is robust evidence of their efficacy and is in line with current evidence-based guidelines [28]. In conclusion, the success of our approach of multiple interventions seems highly plausible but needs to be tested in other hospitals before it can be advocated widely. Each hospital will have its own issues regarding MRSA, and different emphasis will need to be placed on the sequence and timing of these interventions. Funding: None. Competing interests: None declared. Ethical approval: Not required.

Appendix A A.1. Interventions Three types of intervention were considered: a point, a step or a ramp intervention. A point or pulse intervention is a temporary event that may affect the level of the output variable, i.e. %MRSA. At the end of a point intervention, the level of the output variable might return to its normal level. The ith intervention series Xi,t has a point intervention at time tint when the series takes the value 1 at this time and zero otherwise, i.e.:   1, t = tint Xi,j = 0, otherwise A step intervention may affect during a period a permanent change in the level of the output variable. For a step intervention, before time tint , the ith intervention series Xi,t is

A. Mahamat et al. / International Journal of Antimicrobial Agents 30 (2007) 169–176

zero and then steps to a constant level thereafter, i.e.:   1, t ≥ tint Xi,t = 0, otherwise In a ramp intervention, the level of the output variable increases linearly after the intervention’s beginning. For a ramp intervention, before time tint , the ith intervention series Xi,t is zero and increases linearly thereafter, that is proportionally to time:   t − tint , t ≥ tint Xi,t = 0, otherwise The effect decay pattern controlled how the effect of the intervention dissipates over time. It could be immediate at the end of the intervention’s period or decline as an exponential or a wave. A.2. Segmented regression analysis of interrupted time-series analysis Segmented regression analysis assessed how much an intervention changed an outcome of interest immediately and over time: Yt = β0 + β1 × timet + β2 × X1,t + β3 × time afterX1,t + β4 × X2,t + β5 × time after X2,t + . . . + βx−1 × Xk,t . . . + βk+1 × time after Xk,t + . . . + Nt where Nt is an autoregressive integrate moving average (ARIMA) model, used as proxy to account for the autocorrelation in the disturbance series. Adequacy of the ARIMA model for Nt , the disturbance series, was examined using three diagnostic checks: (1) statistical significance of parameters; (2) checking of autoregressive stationary parameters; and (3) checking of residuals that effectively corresponded to white noise.

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17] [18] [19]

References [1] Johnson AP, Aucken HM, Cavendish S, et al. UK EARSS Participants. Dominance of EMRSA-15 and -16 among MRSA causing nosocomial bacteraemia in the UK: analysis of isolates from the European Antimicrobial Resistance Surveillance System (EARSS). J Antimicrob Chemother 2001;48:143–4. [2] Grundmann H, Aires-de-Sousa M, Boyce J, Tiemersma E. Emergence and resurgence of meticillin-resistant Staphylococcus aureus as a public-health threat. Lancet 2006;368:874–85. [3] Monnet DL, Mackenzie FM, L´opez-Lozano JM, et al. Antimicrobial drug use and methicillin-resistant Staphylococcus aureus, Aberdeen 1996–2000. Emerg Infect Dis 2004;10:1432–41. [4] Gould IM. Control of methicillin-resistant Staphylococcus aureus in the UK. Eur J Clin Microbiol Infect Dis 2005;24:789–93. [5] Marshall C, Wesselingh S, McDonald M, Spelman D. Control of endemic MRSA—what is the evidence? A personal view. J Hosp Infect 2004;56:253–68. [6] Pan A, Carnevale G, Catenazzi P, et al. Trends in methicillinresistant Staphylococcus aureus (MRSA) bloodstream infections:

[20]

[21]

[22]

[23]

[24] [25]

175

effect of the MRSA ‘search and isolate’ strategy in a hospital in Italy with hyperendemic MRSA. Infect Control Hosp Epidemiol 2005;26: 127–33. Muto CA, Jernigan JA, Ostrowsky BE, et al. SHEA guideline for preventing nosocomial transmission of multidrug-resistant strains of Staphylococcus aureus and Enterococcus. Infect Control Hosp Epidemiol 2003;24:362–86. French GL, Otter JA, Shannon KP, et al. Tackling contamination of the hospital environment by methicillin-resistant Staphylococcus aureus (MRSA): a comparison between conventional terminal cleaning and hydrogen peroxide vapour decontamination. J Hosp Infect 2004;57:31–7. A guide to sensitivity testing. Report of the Working Party on Antibiotic Sensitivity Testing of the British Society for Antimicrobial Chemotherapy Working Party. J Antimicrob Chemother 1991; 27(Suppl. D):22–45. National Committee for Clinical Laboratory Standards. Performance standards for antimicrobial disk susceptibility testing. 7th ed. M2-A7. Wayne, PA: NCCLS; 2000. Lopez-Lozano JM, Beyaert A, MacKenzie FM, Wilson R, Stuart D, Gould IM. Effect of an intervention programme on the MRSA outbreak in Aberdeen Royal Infirmary. Clin Microbiol Infect 2004;10: 131. Bignardi GE, Woodford N, Chapman A, Johnson AP, Speller DC. Detection of the mec-A gene and phenotypic detection of resistance in Staphylococcus aureus isolates with borderline or low-level methicillin resistance. J Antimicrob Chemother 1996;37:53–63. Brakstad OD, Aasbakk K, Maeland JA. Detection of Staphylococcus aureus by polymerase chain reaction amplification of the nuc gene. J Clin Microbiol 1992;30:1654–60. Kobayashi N, Wu H, Kojima K, et al. Detection of mecA, femA, and femB genes in clinical strains of staphylococci using polymerase chain reaction. Epidemiol Infect 1994;3:259–66. Leonard RB, Mayer J, Sasser M, et al. Comparison of MIDI Sherlock system and pulse-field gel electrophoresis in characterizing strains of methicillin-resistant Staphylococcus aureus from a recent hospital outbreak. J Clin Microbiol 1995;33:2723–7. World Health Organization. Anatomic Therapeutic Chemical (ATC) classification index with defined daily doses (DDDs). Oslo, Norway: WHO Collaborating Centre for Drug Statistics Methodology; 2001. Capella D. Descriptive tools and analysis. WHO Reg Publ Eur Ser 1993;45:55–78. Gould IM, Jappy B. Trends in hospital antimicrobial prescribing after 9 years of stewardship. J Antimicrob Chemother 2000;45:913–7. Gillings D, Mackuc D, Siegel E. Analysis of interrupted time series mortality trends: an example to evaluate regionalized perinatal care. Am J Public Health 1981;71:38–46. Madden JM, Soumerai SB, Lieu TA, Mandl KD, Zhang F, Ross-Degnan D. Effects of a law against early postpartum discharge on newborn follow-up, adverse events, and HMO expenditures. N Engl J Med 2002;347:2031–8. Schneeweiss S, Maclure M, Soumerai SB, Walker AM, Glynn RJ. Quasi-experimental longitudinal designs to evaluate drug benefit policy changes with low policy compliance. J Clin Epidemiol 2002;55:833–41. Soumerai SB, McLaughlin TJ, Ross-Degnan D, Casteris CS, Bollini P. Effects of a limit on Medicaid drug-reimbursement benefits on the use of psychotropic agents and acute mental health services by patients with schizophrenia. N Engl Med 1994;331:650–5. Wagner AK, Soumerai SB, Zhang F, Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther 2002;27:299–309. Pankratz A. Forecasting with dynamic regression models. New York, NY: Wiley; 1991. Mahamat A, Lavigne JP, Fabbro-Peray P, Kinowski JM, Daures JP, Sotto A. Evolution of fluoroquinolone resistance among Escherichia coli urinary tract isolates from a French university hospital: applica-

176

A. Mahamat et al. / International Journal of Antimicrobial Agents 30 (2007) 169–176

tion of the dynamic regression model. Clin Microbiol Infect 2005;11: 301–6. [26] MacKenzie FM, MacLennan G, Stuart D, Gould IM. Evaluation of the effect of MRSA admission screening on MRSA infection rates by segmented regression analysis of interrupted time series. In: Proceedings and Abstracts of the 45th Interscience Conference of Antimicrobial Agents and Chemotherapy (ICAAC); 16–19 December 2005; Washington, DC [abstract K-426-2005].

[27] Harbarth S, Sax H, Fankhauser-Rodriguez C, Schrenzel J, Agostinho A, Pittet D. Evaluating the probability of previously unknown carriage of MRSA at hospital admission. Am J Med 2006;119: e15–23. [28] Ramsay CR, Matowe L, Grilli R, Grimshaw JM, Thomas RE. Interrupted time series designs in health technology assessment: lessons from two systematic reviews of behavior change strategies. Int J Technol Assess Health Care 2003;19:613–23.