The Hospital Water Environment as a Reservoir for Carbapenem-Resistant Organisms Causing Hospital-Acquired Infections—A Systematic Review of the Literature

Abstract

Over the last 20 years there have been 32 reports of carbapenem-resistant organisms in the hospital water environment, with half of these occurring since 2010. The majority of these reports have described associated clinical outbreaks in the intensive care setting, affecting the critically ill and the immunocompromised. Drains, sinks, and faucets were most frequently colonized, and Pseudomonas aeruginosa the predominant organism. Imipenemase (IMP), Klebsiella pneumoniae carbapenemase (KPC), and Verona integron-encoded metallo-β-lactamase (VIM) were the most common carbapenemases found. Molecular typing was performed in almost all studies, with pulse field gel electrophoresis being most commonly used. Seventy-two percent of studies reported controlling outbreaks, of which just more than one-third eliminated the organism from the water environment. A combination of interventions seems to be most successful, including reinforcement of general infection control measures, alongside chemical disinfection. The most appropriate disinfection method remains unclear, however, and it is likely that replacement of colonized water reservoirs may be required for long-term clearance.

carbapenem-resistant, carbapenemase, healthcare-associated infections, outbreak, water.

Over the last 10–15 years, clinically relevant carbapenem-resistant organisms (CROs), such as Pseudomonas aeruginosa, Acinetobacter baumannii, and the Enterobacteriaceae, have disseminated globally [1]. Genes encoding for important carbapenemases, such as Klebsiella pneumoniae carbapenemase (KPC), oxacillinase-48 (OXA-48), and the metallo-β-lactamases, are often transmitted between organisms by mobile genetic elements, such as plasmids, contributing to their spread. Limited treatment options and high mortality in those infected are particularly worrying. Hence, understanding reservoirs and transmission of common CROs and transmissible carbapenemases is a research priority.

Hospitals present a unique opportunity for bacteria to interact, proliferate, and infect vulnerable populations. The healthcare water environment, including potable water, faucets, sink surfaces and wastewater drainage systems (drains, sink/shower traps, toilets, drainage pipes), can be a reservoir for nosocomial pathogens, such as drug-resistant Enterobacteriaceae, Pseudomonas spp., and A. baumannii [2–4]. The prevalence of these multidrug-resistant organisms (resistant to ≥3 antimicrobial classes) is rising [5], and they will increasingly dominate the hospital environmental microbiome. Determining effective infection control (IC) measures to decontaminate environmental reservoirs and prevent cross-transmission of multidrug-resistant organisms between patients and the environment may minimize potentially lethal outbreaks. The aim of this literature review was to summarize studies identifying common CROs in the hospital water environment, the evidence for CRO transmission between this environment and patients, and successful IC interventions to terminate outbreaks and eliminate CROs from this environment.

METHODS

PubMed was searched using the following Medical Subject Headings (MeSH) and text terms: (enterobacter* OR pseudomon* OR acinetobacter) AND (drain OR sink OR shower OR faucet OR hospital water) (search date: 7–9 March, 2016). All abstracts in English, French, Spanish, and German from 1967 to the search date were screened. Articles were excluded if the full article was unavailable through PubMed or the University of Oxford, if the study did not occur in the acute healthcare setting, if it involved hospital wastewater remote to patient care areas, or if the organisms of interest were not identified. Citations in the selected articles were reviewed for additional relevant reports.

Additional searches of conference abstracts were undertaken (9–10 March 2016) for the European Congress of Clinical Microbiology and Infectious Diseases (2005–2015), Interscience Conference on Antimicrobial Agents and Chemotherapy (2013–2014), Infectious Diseases Society of America ID Week (2004–2015), and the Australian Society of Infectious Diseases (ASID) meetings (2013–2015).

All studies involving CROs (organisms phenotypically nonsusceptible to ≥1 carbapenem or producing a carbapenemase) were then included. The ORION (Outbreak Reports and Intervention Studies of Nosocomial Infection) framework was used to assess outbreak reports [6]. Data were collected for the following variables: author, publication year, study design, organism, carbapenemase mechanism, reservoir(s), evidence of environment-patient transmission, and type/success of interventions (Supplementary Table 1). Authors of studies reporting terminating outbreaks were contacted (August 2016) regarding the ongoing success of any interventions.

RESULTS

Study Settings and Populations

Search and screening strategy results are shown in Figure 1. Thirty-two studies were included (Supplementary Table 1). Of these, 27 were outbreak investigations, and 5 were epidemiological surveillance studies [7–11]. Twenty-three were full articles, 6 were conference abstracts [10–15], 2 were letters [16, 17] and 1 was a short report [18]. The studies included data from Europe (n = 16), Asia (n = 7), North America (n = 6), and Australia (n = 3) (Figure 2). Thirty studies occurred between 1996 and 2015, with 50% of these since 2010. Two studies did not document the study time frame [8, 15]. Most were conducted in adult inpatients, with 7 in pediatric/neonatal populations or unspecified (Table 1). Eleven studies involved immunocompromised patients, including those with hematological malignancies [12, 19–22], solid tumors [22], primary immunodeficiencies [22], renal disease [18], burns [7, 8, 23, 24], or unspecified diagnoses [11]. Fifteen studies occurred solely in the intensive care setting [13, 15–18, 23, 25–33], and the others involved various medical/surgical wards, intensive care units, operating theaters, or the whole hospital (Table 1).

Figure 1.

Flowchart of study selection process. Three studies were excluded during data extraction because the outbreak organism was not found in the water environment.

Figure 2.

Geographic and temporal distribution of carbapenem-resistant organisms (CROs). The geographic and temporal distribution of the included studies based on the organism reported. MDR, multidrug-resistant (resistant to ≥3 antimicrobial classes). ^*Study period unknown. ^†Dewi et al [10] report Acinetobacter spp. and Aeromonas spp. along with Enterobacteriaceae.

Table 1.

Study Settings and Populations^a

Patient population	Studies, No. (N = 32)	Intensive Care Unit	High-risk (Hematology, Nephrology, Burns Unit)	Multiple Wards	Otherb
Adult	25	Knoester et al [25], Kotsanas et al [26], Durojaiye et al [16], Wang et al [27], La Forgia et al [28], Wendel et al [29], Bukholm et al [31], Tofteland et al [32], Vergara-López et al [33], Leitner et al [20], Snitkin et al [22], Podnos et al [23], Odom et al [11], Peña et al [35], Pitten et al [36], Biswal et al [8], Landelle et al [37]	Wong et al [19], Leitner et al [20], Breathnach et al [21], Leung et al [24], Betteridge et al [7], Ambrogi et al [18], Odom et al [11]	Peña et al [35], Pitten et al [36], Kouda et al [38], Landelle et al [37], Seara et al [34], Yomoda et al [9], Breathnach et al [21], Biswal et al [8]	None
Pediatric/ neonatal	3	Hong et al [30], Alter et al [15]	None	None	Ito et al [14]
Not given	4	Majumdar et al [17], Kaiser et al [13]	None	Stjarne Aspelund et al [12]	Dewi et al [10]

^aSome studies included multiple different settings, so categories are not mutually exclusive.

^bSingle non–high-risk wards or not specified.

Patient Sampling Strategy and Risk Factors for CRO Infection/Colonization

In total, 926 patients from 31 studies were CRO colonized (n = 184), infected (n = 189), or unspecified (n = 553). In 22 studies, active patient screening was performed, varying in site (swab samples from the rectum, nose, throat, groin, axilla, perineum, perianal, wound, and catheter sites; samples of sputum, urine, stool, tracheal aspirate, blood, and gastric tubes) and frequency (twice weekly, weekly, or fortnightly). The prevalence of patient infection or colonization ranged from 1.6% to 26.7% (reported in 8 studies [8, 14, 19, 22, 24, 30, 31, 33]). Risk factor analysis for colonization/infection was performed in 4 studies [14, 24, 25, 31], and risks included preceding surgery, patient location, prolonged mechanical ventilation, older age, burns, longer hospital stay, and drinking tea from a contaminated dispenser. Length of stay before colonization/infection was documented in only 10 studies, varying from 1 to 134 days [19, 24–26, 28–30, 32–34]. Mortality was assessed in 18 studies, with a mean rate of 25.7% (range, 0%–85%) [12, 16, 19–24, 26, 28–36]. However, 2 studies did not comment on whether deaths were attributable to the study organism(s) [28, 36].

Environmental Sampling Strategies

Hospital water environment investigations included sampling from faucets (n = 18), drainage systems (n = 17), sink surfaces (n = 16), and water (n = 14). Other environmental samples were taken in 26 studies, including medical equipment, patient environment, antiseptic solutions/liquid soaps, enteral nutrition, staff areas, and air samples. Sampling methods varied but typically included moist sterile swab/water samples of varying volumes. Only 5 studies detailed sink and drainage system design [18, 20, 21, 26, 30].

All studies identified the relevant CROs in the water environment, mostly in drains/traps, sink surfaces, and faucets (Table 2). CROs were found in other sites in 15 of 32 studies. The hands of healthcare workers (HCWs) were sampled in 6 studies [8, 17, 23, 27, 33, 36]; additional pharyngeal, rectal, and nasal swab samples were obtained in 2 [8, 33], 1 [33], and 1 [8] study respectively. The study CRO was not colonizing HCWs in any studies.

Table 2.

Water Reservoirs Containing Carbapenem-Resistant Organisms^a

Water Reservoir	Studies, No. (N = 32)	References
Drains/drainage systems	17	Peña et al [35], Kotsanas et al [26], La Forgia et al [28], Betteridge et al [7], Leitner et al [20], Wendel et al [29], Breathnach et al [21], Leung et al [24], Snitkin et al [22], Tofteland et al [32], Vergara-López et al [33], Yomoda et al [9], Stjarne Aspelund et al [12], Odom et al [11], Knoester et al [25], Landelle et al [37], Seara et al [34]
Sink surfaces	14	Betteridge et al [7], Wendel et al [29], Knoester et al [25], Podnos et al [23], Wang et al [27], Biswal et al [8], Hong et al [30], Bukholm et al [31], Kouda et al [38], Landelle et al [37], Dewi et al [10], Kaiser et al [13], Ito et al [14], Leung et al [24]
Faucets	8	Odom et al [11], Knoester et al [25], Majumdar et al [17], Pitten et al [36], Hong et al [30], Bukholm et al [31], Alter et al [15], Leung et al [24]
Water	3	Knoester et al [25], Ambrogi et al [18], Bukholm et al [31]
Inflatable hair wash basin	2	Wendel et al [29], Knoester et al [25]
Sensor mixer taps	1	Durojaiye et al [16]
Water/tea dispenser	2	Wong et al [19], Ito et al [14]
Shower/shower equipment	3	Betteridge et al [7], Leung et al [24], Seara et al [34]
Toilet bowl/brush	2	Breathnach et al [21], Kouda et al [38]

^aSome studies had multiple water reservoirs, so categories are not mutually exclusive.

CROs Investigated

Organisms studied included P. aeruginosa, Pseudomonas fluorescens, Pseudomonas putida, A. baumannii, and various Enterobacteriaceae (Table 3). Molecular carbapenemase identification was performed in 17 studies. The genes identified included bla_IMP (n = 6), bla_KPC (n = 5), bla_VIM (n = 4), bla_NDM (n = 1), and bla_GIM (n = 1) (Table 3). Multiple species of Enterobacteriaceae were involved in 7 studies, suggesting interspecies/intergenus resistance gene transfer [7, 10, 13, 14, 24, 26, 32], specifically demonstrated in vitro for bla_IMP and bla_KPC [7, 9, 32].

Table 3.

Carbapenem-Resistant Organisms and Carbapenemase Type^a

Organism	Studies, No. (N = 32)	Reservoirs	IMP	KPC	VIM	NDM	GIM	MIC/Phenotype only
Pseudomonas aeruginosa	13	Sinks, drains, faucets, hair wash basins, water samples, sensor mixer taps, toilet bowls/ brushes	None	None	Knoester et al [25], Ambrogi et al [18], Breathnach et al [21], Stjarne Aspelund et al [12]	None	Wendel et al [29]	Peña et al [35], Durojaiye et al [16], Majumdar et al [17], Pitten et al [36], Biswal et al [7], Bukholm et al [31], Kouda et al [38], Alter et al [15]
Other Pseudomonas spp.	2	Drinking water dispenser, water pipes	Yomoda et al [9]	None	None	None	None	Wong et al [19]
Acinetobacter baumannii	5	Sinks, drains, faucets	None	None	None	None	None	Podnos et al [23], Wang et al [27], La Forgia et al [28], Hong et al [30], Landelle et al [37]
Klebsiella pneumoniae	7	Drains, faucets, showers, sinks	Kotsanas et al [26], Leung et al [24], Ito et al [14]	Snitkin et al [22], Tofteland et al [32], Odom et al [11]	None	Seara et al [34]	None	None
Klebsiella oxytoca	3	Drains, faucets, showers, sinks	Leung et al [24], Vergara- López et al [33]	Leitner et al [20]	None	None	None	None
Enterobacter spp.	5	Drains, faucets, showers, sinks	Kotsanas et al [26], Betteridge et al [7], Leung et al [33]	Tofteland et al [32]	None	None	None	Dewi et al [10]
Escherichia coli	3	Drains, cold tea dispenser	Kotsanas et al [26], Betteridge et al [7], Ito et al [14]	None	None	None	None	None
Serratia marcescens	3	Drains, faucets, showers, sinks	Kotsanas et al [26], Betteridge et al [7] , Leung et al [24]	None	None	None	None	None
Other (Leclercia spp., Pantoea spp., Citrobacter freundii, Raoutella planticola, Escherichia hermannii, Aeromonas hydrophilia, Proteus mirabilis or not specified)	4	Drains, faucets, showers, sinks	Betteridge et al [7] , Leung et al [24]	Kaiser et al [13]	None	None	None	Dewi et al [10]

Abbreviations: GIM, German imipenemase; IMP, imipenemase; KPC, Klebsiella pneumoniae carbapenemase; MIC, minimum inhibitory concentration; NDM, New Delhi metallo-β-lactamase; VIM, Verona integron-encoded metallo-β-lactamase.

^aSome studies included multiple different species, so categories are not mutually exclusive.

Microbiological and Typing Methods Deployed

Microbiological methods for culture and identification were clearly described for 11 studies [11, 19, 20, 22, 24–26, 29, 32, 33, 37], with no methods given in 7 [12–15, 17, 28, 35]. Methods differed in the media and biochemical tests used. Samples from patients in 2 studies [25, 34] were incubated in an enrichment broth; however, only 1 of these studies supplemented this with a carbapenem disc [34], following Centers for Disease Control and Prevention recommendations [39]. Antimicrobial susceptibility testing methods were reported in 26 studies and included disc diffusion, microbroth dilution, Etest and PCR. Seven of 32 studies did not report susceptibilities [7, 10–14, 24]. CROs were susceptible only to polymixin B/colistin in 8 of 16 studies that reported testing for colistin susceptibility [18, 21, 23, 27, 28, 30, 36, 37]. In 4 studies that did not test for colistin susceptibility, CROs remained susceptible only to amikacin [15, 16, 26, 35]. Variable susceptibility to fluoroquinolones, tigecycline, fosfomycin, other aminoglycosides, and β-lactams was found in the remaining 13 studies. One study reported that the outbreak organism (KPC–Klebsiella pneumoniae) became resistant to all antibiotics after initially being susceptible to gentamicin, tigecycline, and colistin [22]. Molecular typing was performed in all but 2 studies [8, 10], mostly using pulsed-field gel electrophoresis (Table 4).

Table 4.

Bacterial Strain/Mobile Genetic Element Typing Methods^a

Typing Method	Studies, No. (N = 32)	Reference(s)
PFGE	20	Breathnach et al [21], Snitkin et al [22], Tofteland et al [32], Seara et al [34], Wendel et al [29], Wong et al [19], Podnos et al [23], Yomoda et al [9], Stjarne Aspelund et al [12], Kotsanas et al [26], Kouda et al [38], Peña et al [35], Majumdar et al [17], Pitten et al [36], Wang et al [27], Ambrogi et al [18], Landelle et al [37], Vergara-López et al [33], Kaiser et al [13], Alter et al [15]
Resistance gene PCR	8	Knoester et al [25], Seara et al [34], Wendel et al [29], Betteridge et al [7] , Kotsanas et al [26], Kouda et al [38], Leung et al [24], Odom et al [11]
MLST	5	Tofteland et al [32], Seara et al [34], Hong et al [30], Wendel et al [29], Leitner et al [20]
Rep-PCR	4	Snitkin et al [22], Leitner et al [20], Betteridge et al [7], Stjarne Aspelund et al [12]
Plasmid/MGE profiling	3	Tofteland et al [32], Betteridge et al [7] , Yomoda et al [9]
RAPD	2	Wong et al [19], Podnos et al [23]
AFLP	2	Knoester et al [25], Bukholm et al [31]
VNTR	2	Durojaiye et al [16], Breathnach et al [21]
REA	1	La Forgia et al [28]
WGS	2	Ito et al [14], Snitkin et al [22]
Antimicrobial susceptibility profile only	1	Biswal et al [7]

Abbreviations: AFLP, amplified fragment length polymorphism; MGE, mobile genetic element; MLST, multilocus sequence typing; PCR, polymerase chain reaction; PFGE, pulsed-field gel electrophoresis; RAPD, random amplification of polymorphic DNA; REA, restriction enzyme analysis; Rep-PCR, repetitive element sequence-based PCR; VNTR, variable-number tandem repeat; WGS, whole-genome sequencing.

^aSome studies used multiple methods, so categories are not mutually exclusive. One study did not perform any typing (Dewi et al [10]).

CRO Transmission and IC Interventions

The organism under investigation was detected in both patients and the environment in all but 1 study where only the environment was assessed [10]. All studies found evidence of cross-transmission between patients and the environment based on epidemiological links and/or identical antimicrobial susceptibility phenotypes/molecular typing (Table 4).

Nine studies reported IC breaches that probably contributed to outbreaks. These included poor sink design [18, 20, 21, 26, 30], use of sinks for contaminated clinical waste disposal [20, 26, 37], storage of clean patient materials around sinks/sluices [21, 29], reuse of nonsterile surgical drapes and open drainage in the cystoscopy room [35], use of a single brush to clean sinks without between-site disinfection [26], blocked sewage pipes and waste pipe leaks [21], and failure to clean shower drains [21]. Nontouch sensor taps were a reservoir in 1 study [16].

Interventions to eliminate CROs were reported in 27 studies. Twenty-five studies included water environment decolonization interventions, including chemical disinfection (alcohol, chlorination, aldehydes, biguanides, sodium hypochlorite [bleach], acetic acid, hydrogen peroxide, silver nitrate, hot water, and pressurized steam), sterile water for high-risk patient care, assignment of sinks to hand hygiene only, and replacement of contaminated equipment, faucets, sinks or drainage systems (Figure 3). Two studies reported cleaning or disinfecting the water environment, but without details [17, 23]. Two studies described agents used to clean rooms but did not provide specific information on disinfecting the water environment [11, 22].

Figure 3.

Infection control interventions and outcomes. Five studies did not report any interventions (Betteridge et al [7], Biswal et al [7], Kouda et al [38], Yomoda et al [9], Dewi et al [10]), and 5 reported only plumbing interventions (Wong et al [19], Wendel et al [29], Alter et al [15], Peña et al [35], and Ito et al [14]). For duration of follow-up (right axis), months shown in bold denote studies that responded to our communication.

Twenty-two studies reported enhancing general IC measures, including contact isolation, strict hand hygiene, active surveillance, reinforcement of cleaning and disinfection procedures, audits, and education sessions. Of the 25 studies that reported specific environmental interventions, 22 reported success in terminating the clinical outbreak, whereas just more than one-third of these managed to eliminate the organism from environmental reservoirs (Figure 3) [12, 16, 17, 20, 27, 28, 33, 37]. Interventions successful at disinfecting water reservoirs included cleaning of sinks and taps (details not given) [17], daily cleaning of sink surfaces with 0.1% sodium hypochlorite [27], weekly cleaning of sinks and plumbing with acetic acid/hot water [12], transferring all patients to a dedicated isolation unit and hydrogen peroxide vapor disinfection [37], replacing nontouch sensor taps with conventional taps [16], and replacing sinks or drainage systems [20, 33]. Kouda et al [38] reported success without giving details of interventions.

Only 7 of 22 studies reporting success stated the duration of follow-up after the intervention [23, 27–30, 33, 34], which ranged from 2 months to 3.5 years. The authors of 10 studies responded to our communication regarding further follow-up. Seven of these reported no further cases of the outbreak organism [14, 19, 22, 33, 36, 37, 38]. Kouda et al [38] attributed their success to ceasing use of a urinal and changes to mopping practices (details not given). Snitkin et al [22] occasionally isolated CROs from drains, but not the original outbreak organism. Wendel et al [29] reported ongoing sporadic patient cases and environmental isolation despite enhanced IC measures and banning storage of patient material around sinks, however further strain typing has not been done. Seara et al [34] reported 7 additional cases in 2015, but none in 2016. They found hydrogen peroxide vapor to be the most effective disinfectant. Stjarne Aspelund et al [12] reported re-emergence of the outbreak organism after sink replacement, but no further cases or environmental isolation since July 2015, a change thought attributable to weekly flushing of sink drains and waste pipes with acetic acid/hot water.

DISCUSSION

Hospital Water Environment as a Reservoir of CROs

Drains/traps, sinks and faucets were the most common reservoirs of CROs identified. Others included portable hair washing basins, water samples, drink dispensers, toilet bowls/brushes, and shower equipment (Table 2). Initial seeding of these reservoirs was potentially due to contamination from affected patients [20, 22, 25, 32, 33, 36, 37]. Large, complex premise plumbing systems could have areas of stagnation and corrosion, variable nutrient and microbiology loads, and water temperatures ideal for promoting bacterial colonization and biofilm formation [3, 4]. Once colonized, there may be further propagation via the wastewater drainage system to distal sink drains connected to the initial reservoir, and via direct or indirect water contact to other patients [2].

P. aeruginosa was the most frequent organism (41% of studies), and distributed across all water reservoirs. A. baumannii was found predominantly in sink basins, and Enterobacteriaceae most commonly in drains. These findings may reflect biological differences in which Pseudomonas spp. and Acinetobacter spp. are environmental colonizers surviving in low-nutrient conditions, whereas Enterobacteriaceae are predominantly of human origin and their concentration in drains may represent a different ecosystem and inoculation directly from patient waste. However, all these organisms are capable of colonizing water system biofilms [3]. Knowing the predominant site for each organism may help guide surveillance strategies, outbreak investigations, and IC interventions.

Because they can be horizontally transferred on mobile genetic elements, carbapenemase genes represent the most concerning mechanism of carbapenem resistance and may be readily transmitted in environmental reservoirs. Tofteland et al [32] found evidence of possible environmental bla_KPC-plasmid transfer between K. pneumoniae strains, and Betteridge et al [7] concluded there was likely environmental intergenera plasmid exchange between Enterobacteriaceae. Resistance genes and associated mobile genetic elements should therefore be characterized in CRO outbreak investigations. Despite this, only 56% of studies assessed carbapenemase production. The most common enzymes identified were bla_IMP, bla_KPC, and bla_VIM, consistent with previous prevalence reports [1]. Phenotypic surveillance may facilitate the detection of novel carbapenemases.

CRO Transmission Between the Hospital Water Environment and Patients

There was evidence of CRO transmission between the environment and patients based on phenotypic [8] or genotypic methods in all studies assessing this, but most studies used relatively low-resolution methods. Fourteen studies used >1 genetic typing method (Table 4), potentially allowing for greater discrimination. Despite this, transmission routes were difficult to characterize, possibly due to both limited sampling and typing resolution. Whole-genome sequencing (WGS) is a highly discriminatory typing tool increasingly used in outbreak/transmission investigations. Two studies used WGS, with Snitkin et al [22] highlighting its advantages, including confirmation of a monoclonal outbreak, identification of unexpected modes of transmission, and tracking several potential resistance mutations in newly colistin-resistant isolates.

Previous studies have found that HCWs may facilitate nosocomial transmission of multidrug-resistant organisms [3]. Notably, there was no evidence of HCW colonization in the studies reviewed. However, only 19% assessed HCWs, perhaps underestimating the role of HCWs in CRO transmission. Other confounders could include inappropriate timing of sampling and observer effects if staff were aware of surveillance during sampling. Despite the absence of HCW colonization, 10 studies concluded that HCWs were probably implicated in CRO transmission [17, 18, 20, 22, 23, 25, 27, 32, 36, 37]. Colonization of a sink in a medication room attended only by HCWs was cited as evidence for this in 1 study [20]. Most studies also reported a significant reduction in transmission with enhancement of general IC measures, including reinforcement of hand hygiene and contact precautions. However, the relative contribution of HCWs to CRO transmission remains undefined.

Within the hospital there are numerous opportunities for patient exposure to water and drainage reservoirs [4]. Transmission may result from direct or indirect water contact, or from droplets created during water activities, highlighting the importance of water delivery and wastewater design to minimize these risks. Seven studies recognized that poor design or use of sinks, drains, and sluice areas may have contributed to institutional outbreaks [18, 20, 21, 26, 29, 30, 37]. Guidelines for design of hand-wash basins generally include a large basin to contain splashes, taps that are not aligned directly over drains to minimize aerosols, no plugs or overflows, and ensuring that basins are not used for disposal of patient-related waste [40]. This guidance varies by country and application of the recommendations probably varies between institutions.

Effective IC Strategies

IC strategies used mostly included bundled approaches involving enhanced general IC measures, disinfection, and replacing reservoirs, making it difficult to ascertain the relative contribution of individual approaches. Enhancing general IC measures led to a reduction in clinical cases in most studies but often did not completely terminate outbreaks.

Chemical disinfection was most useful for environmental colonization with A. baumannii and P. aeruginosa. Faucet contamination with P. aeruginosa was successfully treated with aldehydes [36] or hot water [31], and Enterobacteriaceae with bleach, hydrogen peroxide vapor or UV light [11]. Eradication of sink colonization with A. baumannii was possible with bleach [27] or hydrogen peroxide vapor [37], however chemical disinfection was not effective for Enterobacteriaceae. In drain colonization, some success with eradicating Enterobacteriaceae and A. baumannii was reported using bleach [11, 22, 28] or hydrogen peroxide vapor [11, 22, 34, 37]. One study reported ongoing suppression of P. aeruginosa with weekly disinfection of sinks and drainage systems with acetic acid/hot water [12]. Use of hydrogen peroxide vapor was most promising, but this method is expensive and was ineffective in liquid and foam form [13, 34]. Bleach, hot water, UV light, and aldehyde-based disinfectants were effective in some studies, but other studies reported failures with these agents. Other agents were also used without success (Figure 3). Notably, the concentrations and deployment of agents varied widely; standardizing approaches would allow for more comparable outcome assessment.

Overall, the most successful intervention was replacing reservoirs (Figure 3). Replacing taps contained 67% of outbreaks associated with P. aeruginosa affecting faucets [15, 16, 18, 36]. Replacement of drains/drainage systems and sinks was always successful for Enterobacteriaceae colonization [20, 32–34] but failed with P. aeruginosa [12, 21, 29]. Other effective strategies included drainage or sink redesign, sterile water for care of high-risk patients, and prohibiting the storage of clean patient material around a sink. These findings may also be relevant for water-environment associated outbreaks with organisms that harbor other resistance mechanisms, such as extended-spectrum- β-lactamases. A list of current recommendations from international and national organizations on the prevention and control of CROs can be found in Supplementary Table 2; no organization makes recommendations regarding environmental sampling or chemical disinfectant.

Review and Study Limitations

General limitations of our study include potential publication bias and the language and literature source restrictions used in this review. Healthcare water environments are probably often neglected in outbreak assessments, which may result in an underestimation of the problem.

The heterogeneity in study methods also limits the generalizability of conclusions. Measuring the clinical impact of outbreaks was difficult; 23 studies did not give prevalence data, and 13 did not assess mortality rates. Screening of patients and the environment varied widely with regard to the methods used, frequency, and site. Rectal swab/fecal samples are recommended for patient CRO colonization screening [5], but 10 studies only sampled other body sites, and 9 did not screen patients at all. Consequently patient colonization may have been underestimated in up to 60% of studies.

Differences in environmental sampling, such as collecting water samples before or after tap flushing and use of varying water volumes, may affect sampling sensitivity. Microbiological methods differed between studies, and although the use of selective culture media may have improved detection, organisms with near-breakpoint minimum inhibitory concentrations could have been missed, underestimating the true prevalence and extent of environmental contamination. Molecular tests may increase sensitivity [29].

CONCLUSIONS

Water environment-associated CRO outbreaks have been reported increasingly in the last 5 years. Critically ill and immunocompromised patients are particularly at risk. Drains/traps, sinks and faucets were most frequently colonized, and P. aeruginosa was the predominant organism. Standardizing patient and environmental screening, and microbiological methods, when investigating CRO outbreaks may allow earlier detection of reservoirs. Carbapenemases should be characterized, and strains typed to confirm relatedness. WGS is a promising tool for determining transmission routes, reservoirs, and resistance mechanisms.

The incorrect design and use of hand-wash basins and other water areas may propagate outbreaks. Various IC interventions have been used with mixed results for each organism and reservoir. A combination of interventions is probably required to terminate outbreaks and eradicate CROs from the environment. Useful approaches include reinforcing general IC measures, combined with chemical disinfection using hydrogen peroxide vapor. Further studies assessing the effectiveness of different chemical disinfection strategies are needed. Replacement and improved management of colonized water reservoirs is also probably an important component of control.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Financial support. This work was supported by the National Institute for Health Research/University of Oxford (Academic Clinical Lectureship to N. S.), the Academy of Medical Sciences (United Kingdom) (Starter Grant for Clinical Lecturers Scheme award to N. S.), the Oxford Biomedical Research Centre (A. S. W., T. E. A. P., and D. W. C.), and the National Institute for Health Research (Senior Investigator awards to T. E. A. P. and D. W. C.).

Potential conflicts of interest. All authors: no reported conflicts. The authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Author notes

Correspondence: A. E. Kizny Gordon, Department of Microbiology (Research), John Radcliffe Hospital, Level 7, Headley Way, Headington OX3 9DU, UK ([email protected]).