Pseudomonas aeruginosa outbreaks in the neonatal intensive care unit – a systematic review of risk factors and environmental sources

Correspondence Johanna Jefferies J.Jefferies@southampton.ac.uk Molecular Microbiology Group, Academic Unit of Clinical & Experimental Sciences, University of Southampton Faculty of Medicine, University Hospital Southampton NHS Foundation Trust, Southampton, UK Health Protection Agency, Southampton, UK Infection Prevention Team, University Hospital Southampton NHS Foundation Trust, Southampton, UK Department of Infection, University Hospital Southampton NHS Foundation Trust, Southampton, UK


Background
Pseudomonas aeruginosa is a Gram-negative bacterium commonly occurring in soil and water.It is an opportunistic pathogen, causing disease in vulnerable individuals, such as the immunocompromised, those where host defences have been breached, such as burns patients, and infants in whom the immune system is not yet developed.Most commonly, P. aeruginosa infections occur in the urinary tract, in burn wounds and as chronic lung infections in cystic-fibrosis patients.However, the organism is also an important cause of healthcare-associated infections, particularly among infants in neonatal intensive care units (NICUs) (Jarvis et al., 1987) owing to their underdeveloped immune system and the fact that such infants are often intubated, catheterized with intravascular catheters/devices in-situ, and/or receiving parenteral nutrition, which may increase risk of infection (Gaynes et al., 2005).Unfortunately, such infections in neonates can be fatal, as was the case in a recent UK outbreak in which four babies died (Wise, 2012).The ubiquitous nature of P. aeruginosa in the environment makes the sources of such outbreaks difficult to identify.
Several outbreaks of P. aeruginosa colonization and infection in NICUs have been published.Simon et al. (2008) published a systematic review of confirmed outbreaks in both neonatal and paediatric patients that were identified using the PubMed database and the web-based register for nosocomial outbreaks.That review did not search for reports of outbreaks listed in the Embase database and so may have been biased by the omission of relevant data.Additionally, the review itself is not listed in PubMed/ Downloaded from www.microbiologyresearch.orgby IP: 54.70.40.11On: Sat, 02 Feb 2019 19:12:43 MEDLINE and so is not readily available to a wide readership.Therefore, we set out to provide an updated review of the literature reporting on P. aeruginosa colonization and infection in NICUs using both Embase and MEDLINE.We have performed a systematic review of P. aeruginosa outbreak reports and prospective surveillance studies following outbreaks of P. aeruginosa infections in neonates, specifically in the NICU setting.

Methods
Search terms were constructed as shown in Fig. S1 (Available in JMM Online).Searches were performed between 28th and 29th January 2012 and were initially run on the MEDLINE database followed by the Embase database.Search terms were entered as MeSH terms rather than text terms.Only articles in the English language were included in the final list of studies to review.Fig. S1 shows the number of studies retrieved for each of the search terms and the methods used to select and reject studies.
Fifteen studies met the inclusion criteria and were included in the systematic review.In order to analyse the combined results of the studies, the framework outlined in the ORION (Outbreak Reports and Intervention Studies of Nosocomial Infection) statement (Stone et al., 2007) was used to review the studies.Outbreak reports and subsequent investigations, including non-randomized studies, form the main body of evidence regarding risk factors and clinical importance of healthcare-associated infections.Such studies often lack details on study design and basic information such as case definitions, selection of controls and denominator information.The ORION guidelines seek to build on the CONSORT (Begg et al., 1996) and TREND (Des Jarlais et al., 2004) guidelines and aim 'to raise the standards of research and publication in hospital epidemiology, to facilitate synthesis of evidence and promote transparency of reporting, to enable readers to relate studies to their own experience and assess the degree to which results can be generalised'.ORION consists of a 22-item checklist and although it was designed for reporting of new outbreak and intervention studies, it provided an appropriate framework for the systematic analysis of a number of studies as presented here.Seventeen ORION criteria were used to screen studies and withdraw information.Criteria that were not used to screen studies for the present review were economic outcomes, sample size calculations, recruitment, ancillary analysis and harms.These criteria were not used as they apply only to economic evaluations or to intervention studies and not to case reports or outbreak investigations as reviewed here.
Although none of the studies set out to test the efficacy of interventions as the primary outcome, interventions to control outbreaks were reported in 13 studies; these were wide-ranging and are described in Table S1.Of the 13 studies that reported interventions (Garcia et al., 1989;Verweij et al., 1993;Garland et al., 1996;Archibald et al., 1998;Muyldermans et al., 1998;Foca et al., 2000;Gras-Le Guen et al., 2003;Majumdar et al., 2004;Schutze et al., 2004;Zabel et al., 2004;Crivaro et al., 2009;Sa ´nchez-Carrillo et al., 2009;Naze et al., 2010), 10 studies reported the intervention as being successful (Garcia et al., 1989;Garland et al., 1996;Muyldermans et al., 1998;Gras-Le Guen et al., 2003;Majumdar et al., 2004;Schutze et al., 2004;Zabel et al., 2004;Crivaro et al., 2009;Sa ´nchez-Carrillo et al., 2009;Naze et al., 2010) and one study reported temporary success (Foca et al., 2000).Two studies did not report the outcome of the intervention (Archibald et al., 1998;Moolenaar et al., 2000).Table S5 details the authors' interpretation of the results of interventions.Interestingly, only one study reported coordinated efforts of NICU staff and staff from related units as a successful approach to controlling and preventing outbreaks.One study used covert surveillance to assess the effect of an educational intervention on hand washing behaviour, reporting before and after results (Crivaro et al., 2009).The authors concluded that this intervention had been successful but full details of the intervention, which would enable it to be repeated elsewhere, were not included in the study.
Six studies (Archibald et al., 1998;Muyldermans et al., 1998;Foca et al., 2000;Moolenaar et al., 2000;Loureiro et al., 2002;Gras-Le Guen et al., 2003) included a statistical analysis of risk factors (Table S4).The risk factors identified were use of antimicrobial drugs before positive blood culture, the number of days of antimicrobial therapy prescribed before positive blood culture (Loureiro et al., 2002) and exposure to two particular HCWs (Moolenaar et al., 2000).In one study, infants with P. aeruginosa sepsis were found to have been given a mean of 3.6 different antimicrobials prior to providing a blood culture that was positive for P. aeruginosa compared to the mean use of 2.1 antimicrobials in non-case infants (Loureiro et al., 2002).Transfusion of fresh frozen plasma or human albumin also appeared to increase the risk for P. aeruginosa infection but did not reach significance (Muyldermans et al., 1998).Exposure to a particular HCW, and intravenous delivery of total parenteral nutrition simultaneously with glucose/ electrolyte solution were significantly associated with P. aeruginosa bacteraemia.Exposure to umbilical venous catheters also showed evidence of an association with bloodstream infections (Archibald et al., 1998).Increasing age and the use of fingernail wraps (artificial fingernails) were deemed risk factors for colonization of hands of HCWs (Foca et al., 2000).In another study, infants exposed to a particular (colonized) HCW and those exposed to a second un-colonized HCW were also at increased risk of infection (Foca et al., 2000).Low birth weight pre-term infants were at greater risk of mortality from P. aeruginosa infection than older infants (Gras-Le Guen et al., 2003).Measures of association for all risk factors are listed in Table S4.Two studies did not include a formal risk factor analysis but used molecular typing to identify outbreak sources as commercially bottled mineral water (Crivaro et al., 2009) and contamination of a feeding-bottle preparation room (Sa ´nchez-Carrillo et al., 2009).
In general, the studies reviewed here provide evidence that P. aeruginosa can be introduced to the NICU via a number of routes, including environmental contamination, transmission by HCWs, transfer of colonized patients and through the use of contaminated water to prepare milk or other nutrition.The studies also demonstrate that molecular typing is a valuable tool in the investigation of P. aeruginosa outbreaks and for the surveillance of environmental contamination.The overall evidence is shown in Table S5.

Discussion
Water as a reservoir of P. aeruginosa The 15 studies reviewed here represent outbreaks of P. aeruginosa infection and/or colonization in NICUs from a wide range of developed and middle-income countries.
Water as an environmental source of infection was a common factor in among studies which identified point source outbreaks of P. aeruginosa infections.Two studies identified contaminated water baths as reservoirs (Muyldermans et al., 1998;Gras-Le Guen et al., 2003); other 'wet' reservoirs were a pasteurizer used to sterilize milk (Gras-Le Guen et al., 2003), humidifying equipment for ventilators (Zabel et al., 2004) and a tap in a neighbouring high-dependency unit (Majumdar et al., 2004).One study also found P. aeruginosa unrelated to the outbreak strain in a sink drain (Moolenaar et al., 2000).One study identified multi-dose vials of parenteral nutrition and a further study found that feeding bottles were likely to have become contaminated as a result of water-contamination in a preparation area whilst an additional study traced the outbreak to commercially bottled mineral water used to prepare multidose milk.P. aeruginosa is primarily an environmental organism that is adapted to survive in numerous conditions and is particularly welladapted to wet or damp conditions.Water systems are wellrecognized as potential reservoirs of this organism for healthcare-associated infections (Trautmann et al., 2005).
The World Health Organization has published guidelines for safe water quality that state that all healthcare facilities should have specific water safety plans to control waterborne healthcare-associated infections including P. aeruginosa as part of their infection control programme (WHO, 2006).Point-of-use filtration using 0.2 mm filtration units was associated with a significant reduction in chronically endemic P. aeruginosa colonizations/infections on a surgical ICU where P. aeruginosa was cultured from 113/117 (97 %) of tap water samples and was also shown to be cost-effective when changed weekly by designated personnel (Trautmann et al., 2008).However, the filtration units do not eradicate the organism from the system, but only prevent discharge to the environment from the filtered outlet.In the Health Technical Memorandum 04-0 (Department of Health, 2006), continuous long-term use of point-of-use filtration is not recommended, except where there is no effective alternative to eradicate the organism from the water system.Such an intervention may also be useful in cases of endemic P. aeruginosa in NICU settings.Whilst filtration has proved to be useful in situations where tap water has been shown to harbour P. aeruginosa, contamination of such filters from environmental P. aeruginosa in the sink below the filter unit is a possibility.It is recognized that strains of P. aeruginosa within sink traps may not be the same as those that cause infection.In the present systematic review, 11 of the 15 studies identified the outbreak strain in the environment.Nevertheless, environmental reservoirs such as sinks have the potential to lead to outbreaks.An outbreak due to splash-back from contaminated sink drains was reported from the ICU and transplant unit of a Canadian hospital in 2009 (Hota et al., 2009).In the UK, the EPIC2 guidelines of the Hospital Infection Society were published in February 2007 (Pratt et al., 2007).These guidelines are wide-ranging and cover many aspects of infection control; however, they do not refer to design of sinks and other water systems.NHS guidelines for water system specifications state that 'swanneck outlets should be avoided' and recommend that 'in existing facilities, when such an outlet has become damaged and is due for repair that it be completely replaced by an appropriate, alternative component as described' (Ayliffe et al., 2000).Other guidelines also state that 'taps discharging into a shallow sink or directly into a drain hole can cause splashing which disperses contaminated aerosols.Thus, the tap outlet flow should not point directly into the sink outlet' (SHFN, 2007).Oxidising biocides such as chlorine dioxide or ionisation systems that release copper and silver ions can be used as control measures.These methods have been shown to be effective in the control of organisms in water systems (Department of Health, 2006).

Biofilm formation and antibiotic resistance
The potential of P. aeruginosa to form biofilms on environmental surfaces is well known and causes difficulties in eradicating this organism.Biofilms can be defined as structured microbial communities of single or multiple species attached to a surface and surrounded by a matrix of extracellular polymers or 'slime'.Bacteria grown as biofilms display tolerance to antibiotics in the absence of known resistance determinants (Costerton et al., 1987) and may also be hard to eradicate using environmental antibacterial agents due to their complex structure and protective matrix.Recently, it has been demonstrated that biofilms generate extensive genetic diversity, both among resident bacteria and among cells that detach and disperse from biofilms (Molin & Tolker-Nielsen, 2003;Webb et al., 2003).This phenomenon is thought to enhance the ability of biofilm cells to persist and spread under diverse environmental stresses, including antibiotic use.Antibiotic-resistant strains of P. aeruginosa were associated with five of the outbreaks reported here and multiple resistance was reported in three studies.Emergence of antibiotic-resistant strains is occurring in Europe and elsewhere (Souli et al., 2008;Lautenbach & Polk, 2007).The use of antibiotics and antimicrobial compounds may provide a driving mechanism for the emergence of antibiotic resistant P. aeruginosa.This hypothesis is strengthened by the study of Loureiro et al. (2002), which demonstrated that prior antibiotic use was associated with bloodstream infections in the NICU.Recently, a case-control study showed that multi-drug resistant (MDR) P. aeruginosa caused significantly higher mortality rates among patients with chronic obstructive pulmonary disease than did infection with non-MDR strains (Montero et al., 2009).MDR strains are also of particular concern in vulnerable infants, such as those being treated in the NICU; Crivaro et al. (2009) reported that both deaths associated with P. aeruginosa infections in the NICU outbreak reviewed here were associated with MDR isolates of P. aeruginosa.
Recently, studies regarding control of biofilms in water systems have been undertaken.Such work includes studies of the materials making up the water system itself, as well as studies regarding the use of various anti-biofilm agents.Such agents include various electrochemically activated solutions (Coleman et al., 2009), sublethal concentrations of the anti-bactericidal agents piperacillin-tazobactam (Fonseca et al., 2004) and low (sublethal) concentrations of nitric oxide (NO), allowing antibacterial agents to attack Pseudomonas bacteria in the planktonic (non-biofilm) mode of growth.Barraud et al. (2006) showed that treatments using low concentrations of NO together with antimicrobials against P. aeruginosa were highly effective in removing P. aeruginosa biofilms.Such novel methods may be important in the control and prevention of biofilmassociated infections.However, use of chemicals directly in the water system could have significant COSHH implications especially when the water is used in highly vulnerable infants in a NICU setting.

Hand hygiene and aseptic technique
Exposure to particular HCWs and umbilical catheterization were the most important risk factors identified here.This review confirms that P. aeruginosa can be transmitted from HCWs to patients via contaminated hands (Foca et al., 2000;Moolenaar et al., 2000).Accordingly, improved hand hygiene measures at the point of care, re-enforcement of infection control guidelines and education of healthcare staff were reported as part of successful infection control interventions for the majority of studies evaluated.The importance of HCWs keeping short clean nails, in order to prevent them acting as a reservoir of infection, was also highlighted.A recent study has reported that observing correct hand hygiene, using a protocol based on WHO recommendations (Pittet et al., 2009), minimizes the risk of contamination of hands with P. aeruginosa even when hands are washed in water heavily contaminated with the organism (Jones, 2011).

Molecular typing
This review considered studies that had used molecular typing to provide evidence that outbreak P. aeruginosa strains were genetically related to each other and to demonstrate an association between P. aeruginosa infection/colonization and environmental isolates.Such techniques are now commonplace and can be performed routinely.Of the five different typing methods, namely PFGE, RAPD, inter-repeat PCR plasmid profiling and VNTR-HRMA, reported in 12 of the studies, PFGE appears to be the most discriminatory and widely used method for typing of P. aeruginosa at present.VNTR-HRMA is a promising method that requires further validation with larger numbers of P. aeruginosa isolates in order to confirm discriminatory power.Whilst it is not currently available for small-scale outbreaks, whole genome sequencing using highthroughput 'next generation' sequencing technology is fast becoming available and affordable and has been used successfully to track large outbreaks of food poisoning (Mellmann et al., 2011;Rohde et al., 2011).Such technologies are likely to become the tool of choice in the future, even for local outbreak investigations.

Limitations of the studies
The studies reviewed here are outbreak reports.Seven studies (Garcia et al., 1989;Verweij et al., 1993;Garland et al., 1996;Archibald et al., 1998;Loureiro et al., 2002;Gras-Le Guen et al., 2003;Majumdar et al., 2004) did not include denominator data, meaning that the impact of the outbreak was unclear.Three studies mentioned data in the text that could be used to estimate the denominator; one as the number of beds in the NICU (Schutze et al., 2004) and two as the number of admissions per year (Muyldermans et al., 1998;Zabel et al., 2004).Other limitations were the omission of methodological data or case definitions.The small numbers of cases in a several of the studies prevent a robust analysis of risk factors.Lack of a standardized approach in reporting is a barrier to meta-analysis for the identification of risk factors.

Conclusion
The studies reviewed here have shown that P. aeruginosa, especially antibiotic-resistant strains, are a well-known cause of outbreaks of invasive and non-invasive disease in the NICU setting.Such outbreaks occur in both high-and middle-income countries and can result in both morbidity and mortality.Outbreaks can be long-lasting but can also be effectively brought under control by the use of various interventions, the most important being effective infection prevention practices, especially hand hygiene for staff and visitors; such practices are key to the prevention of P. aeruginosa outbreaks.Studies outside the scope of this review have described a number of environmental modifications as preventative measures, such as those relating to the design of sinks, taps and water-systems, which may help to prevent or reduce environmental contamination with P. aeruginosa.The reviewed studies also highlight the usefulness of molecular typing techniques and demonstrate the importance of environmental sampling for outbreak investigations.The ubiquity of P. aeruginosa and its ability to colonize environmental surfaces and water systems presents a particular challenge, as environmental strains unrelated to the outbreak strain may be present.The data presented in some studies were subject to certain limitations and future studies in this area would benefit from adherence to ORION guidelines in order to provide improved evidence.