This is the first study to our knowledge that provides longitudinal data on colonisation with AROs and risk factors for colonisation in a LTCF. The need for such studies has been recognised as urgent in a recent editorial on E. coli ST131 [31]. Our study yielded nine key findings. First, ESBL-EC were detected in more than 50% of participants of the study LTCF. Second, MRSA were detected in more than 25% of participants, and co-colonisation of ESBL-EC and MRSA was common. Third, care area (corresponding with level of dependency) and antimicrobial use were risk factors for carriage of ESBL-EC and MRSA. Fourth, all ESBL-EC isolates belonged to the H30-Rx ST131 subclone. Fifth, PFGE divided the H30-Rx subclone into two clusters, one of which included UK Strain A and a representative of the international PFGE812 group, the other included UK strain D. Sixth, all ESBL-EC and ESBL-KP contained bla_{CTX-M group 1}. Seventh, once subjects acquired ESBL-EC they tended to remain colonised with the same strain. Eight, we observed that although ESBL-producing Enterobacteriaceae were recovered from a much higher proportion of participants than MRSA, ESBL-producing Enterobacteriaceae was not detected in the environment, whereas detection of MRSA was common. Finally, 39% of residents did not become colonised with ESBL-EC or MRSA at any time, despite sharing the same LTCF with colonised residents from up to a year. These findings have significant implications for our understanding of the relationship between LTCF and the problem of acquired resistance to antimicrobial agents.

Several studies have assessed the prevalence of colonisation with ESBL-EC among LTCF participants. Our findings are broadly consistent with those of a previous study in Northern Ireland. However, our observed colonisation prevalence with ESBL-EC (56%) is higher than that study’s (41%) [20], and is much higher than noted in previous international LTCF surveillance studies, including studies from 2008–2010 and 2003–2012 from Sweden identified ESBL- EC in just 1/268 (0.004%) and 2/1131 (0.002%) residents respectively. Similarly, studies performed in 2010 and 2011 from Melbourne identified ESBL-EC in 14/119 (12%) and 12/115 (10%) of LTCF residents respectively [41-44].

MRSA colonisation was detected at some time in 27% of participants, which is similar to a previous study from Northern Ireland (23% of residents in 45 LTCFs), but higher than an earlier study performed in six LTCFs in Ireland between 1995 and 1996 (9% and 10% respectively) [23,45]. It is possible that 27% represents an underestimate of colonisation, since practical constraints limited our sampling to nasal swabs. A low prevalence of MRSA colonisation among LTCF residents was reported from Sweden, the Netherlands, and Denmark [46-48]. However, these countries have a low proportion of MRSA bloodstream infections compared with Ireland [17], which suggests that MRSA is less prevalent overall as a healthcare-associated organism in those countries compared with Ireland.

Co-colonisation with ESBL-producing Enterobacteriaceae and MRSA was common (39%). Although VRE colonisation was uncommon (3%), it likewise was associated with ESBL-EC and MRSA colonisation. Elsewhere, VRE colonisation of LTCF residents was not detected in Belgium, but has been reported in Australia (2%), Israel (10%), and in the USA (4 to 45%) [42,49-52]. Similarly, Donskey and colleagues (2003) found that, in a Department of Veterans Affairs medical centre (which includes an acute care facility and a separate nursing home), colonisation with MRSA and ceftazidime-resistant Gram-negative bacilli was more prevalent among VRE-positive than VRE-negative patients [53]. Thus our findings of co-colonisation are consistent with existing literature.

Colonisation with ESBL-producing Enterobacteriaceae and MRSA were associated with residence in particular care areas within the LTCF and with antimicrobial consumption. Specifically, participants in care areas 1 and 2 were 3.9 odds more likely to be ESBL colonised, and 6.9 odds more likely to be MRSA colonised. The association with care area may plausibly reflect the interaction of a number of factors associated with providing care for the most dependent residents, including greater resident vulnerability, and more frequent contact with healthcare workers. This reflects the challenges of containing spread of AROs while providing appropriate social, psychological, and physical care for highly dependent residents in a LTCF.

Participants also had a 2.4-fold higher odds of ESBL colonisation and 3.1-fold higher odds of MRSA colonisation if they received antimicrobial in the three months before sampling. Fifty-two percent of participants received antimicrobials during the one-year study period. Broad spectrum beta-lactam antimicrobials (included amoxicillin, co-amoxiclav, and cefuroxime) were the most frequently consumed antimicrobials, with amoxicillin-clavulanic acid accounting for the majority. This is consistent with the European Surveillance of Antimicrobial Consumption point prevalence studies, which identified beta-lactam antimicrobials and penicillins as the most common antimicrobials prescribed in Irish LTCFs [54]. The 2013 HALT survey for Ireland reported 9% and 11% respectively as the median overall antimicrobial use prevalence in general nursing homes (n = 103) and mixed care facilities (n = 26), both with length-of-stay > 12 months [55]. Likewise, in 722 European LTCFs the most commonly prescribed agents were amoxicillin-clavulanic acid (12.7%), nitrofurantoin (10.4%), trimethoprim (9.9%), amoxicillin (7.3%), and ciprofloxacin (6.9%) [24]. A higher point prevalence of antimicrobial prescribing (16%) has been reported in LTCFs in other European studies [56,57]. Benoit et al. [58] reported 42% of residents in 73 LTCFs as receiving antimicrobials in the previous six months in the USA.

Given that third-generation cephalosporins and fluoroquinolones were used infrequently in this facility, it appears that selective pressure from these agents is not likely to be a major factor in the high prevalence of colonisation with ESBL-producing Enterobacteriaceae or MRSA. However, we note that in this LTCF the dominant ESBL-Enterobacteriaceae and MRSA are resistant to broad-spectrum beta-lactams, the most commonly consumed antimicrobials in the LTCF; thus, the overall association of colonisation with antimicrobial consumption is not surprising from a microbiological perspective.

It was surprising that nitrofurantoin was independently associated with ESBL colonisation (OR 3.4 (1.1-11.3)), since it is generally considered to have limited selective impact on gut flora [59]. However, record review indicated that 8/9 participants who received nitrofurantoin prior to sampling had already been identified as ESBL-positive before the nitrofurantoin was prescribed, based on clinically-submitted urine samples. Since nitrofurantoin use in these patients was most likely guided by prior susceptibility test reports, and since nitrofurantoin is one of few oral antimicrobials active against ESBL-EC, this association can be explained on that basis.

The finding that all ESBL-EC isolates represented the ST131 subclone contributes to the evidence that dissemination of ST131 is a global public health problem in LTCFs. Banerjee et al. [60] reported that, among consecutive E. coli clinical isolates from Rochester, Minnesota, ST131 accounted for 80/299 (27%) of isolates from healthcare or community-associated infections, but for fully 28/37 (76%) of isolates from LTCFs. Rooney et al. and Dhanji et al. reported ST131 representing 99% of ESBL-EC collected from residents of LTCFs in Northern Ireland [20,61]. The proportion of ST131 we encountered among faecal ESBL-EC was similar to these reports, and higher than in other international studies of LTCF residents [20,60,61]. The predominance of ST131 in LTCFs suggests that such facilities may serve as reservoirs for dissemination of ST131. This problem is likely to be exacerbated by growing number of LTCF residents, which currently is approximately 19,800 (2012) in Ireland, 3.7 million in the EU (2010), and 1.5 million in the United States (2009) [24,25,31].

Genomic diversity within ST131 has been described previously based on sequencing the fimH, gyrA, and parC genes, PFGE, and whole genome analysis [3,6,7]. The H30 subclone and its H30-Rx subset have been identified as the major lineages of E. coli associated with antimicrobial-resistant infections. Colpan et al. [38] found that the H30 subclone as a whole accounted for ≥ 95% of fluoroquinolone-resistant and ESBL-producing ST131 isolates from US veterans. Price et al. [7] and Banerjee et al. [5] reported a high proportion of ESBL-EC as belonging to the H30-Rx subset; 91% and 92% respectively [5,7]. Similarly, in the present study all ESBL-EC belonged to the H30-Rx subclone. The H30-Rx subclone has a reported association with sepsis, suggesting that virulence may be contributing to its epidemiologic success [5,7].

Johnson et al. [3] and Colpan et al. [38] identified various pulsotypes within ST131, with PFGE968, PFGE800, and PFGE812 being the most prevalent [3,38]. Here, 91/98 (93%) of the H30-Rx isolates formed a single PFGE cluster, at the ≥ 85% similarity level. This group includes the ST131 UK epidemic strain A and a representative of international PFGE812 (Figure 1). The latter two reference strains represent, respectively, the most prevalent ST131 variant in the UK [40,62] and the most prevalent European-associated ST131 pulsotype in a large private PFGE library (unpublished data, J. R. Johnson) [3]. Thus, the predominance of this clonal group within the study LTCF is part of the international dissemination of the H30Rx ST131 subclone. However, six of the remaining eight H30-Rx isolates, from two participants, formed a distinct PFGE cluster, Cluster B (Figure 1). It is of interest that this variant was not entirely displaced by the predominant variant; thus, two variants of the H30-Rx subclone were circulating in the LTCF.

Our finding of high prevalence of bla_CTX-M group 1 is consistent with a previous reports from Banjeree et al. [5] and Price et al. [7] demonstrating a high prevalence (91%) of bla_CTX-M-15 within the H30-Rx subclone [5,7]. Additionally, we found that most isolates (87%) also carried bla_OXA-1, which was associated with co-amoxiclav resistance, reflecting inefficient inhibition of OXA enzymes by clavulanic acid [63]. This finding may be relevant to the success of this clonal group given the frequent consumption of co-amoxiclav.

Although few, the ESBL-KP in our study were 91% similar by PFGE (Figure 2). All carried bla_{CTX-M group 1} and bla_OXA-1 and were resistant to co-amoxiclav. The six ESBL-KP strains were 86% similar to other Irish isolates belonging to K. pneumoniae clonal CC43 (i.e., ST48 and ST123) This is consistent with clonal dissemination of ESBL-KP in Ireland, similar to that described elsewhere [64]. Various clonal groups of K. pneumoniae have been reported worldwide, such as CC43, which is commonly associated with resistance enzymes such as TEM-3, SHV-12, and CTX-M-15 [10,12]. We detected ESBL-KP for the first time in this LTCF three months after commencing the study. Given the propensity for dissemination of such strains of K. pneumoniae [65], it seems surprising that after nine months of follow-up only five participants had evidence of colonisation.

The duration of persistence of colonisation with ESBL-producing Enterobacteriaceae has previously been reported. The median time to clearance of ESBL-producing Enterobacteriaceae after hospital discharge was 6.6 months in a French study [66] and 98 days (i.e., approximately 3 months) in a study from Thailand [67]. A Swedish study by Titelman et al. [68] concluded that clearance of ESBL-producing Enterobacteriaceae has not occurred in many patients (43%) at 12 months after infection and that persistence is associated with phylogroup B2 E. coli carrying bla_{CTX-M group 9}. This study also demonstrated that failure to detect the organism on sampling does not reliably establish clearance [68]. However, these studies did not address residents of LTCFs.

We found that spontaneous clearance of colonisation of E. coli carrying bla_{CTX-M group 1} was uncommon; 75% of ESBL-EC colonised participants who had two or more subsequent samples after their initial positive swab remained positive. Moreover, despite the presence within this LTCF of two different PFGE variants of ESBL-EC ST131, the variant identified in a given resident was stable over time. Thus, participants tended to remain colonised with the same ESBL-EC strain for an extended period. Furthermore, two initially ESBL-EC-colonised participants in whom ESBL-EC ST131 was not detected at 3 or 6 months reverted to being positive at 9 and 12 months, in both cases with isolates that were indistinguishable from or ≥ 95% similar to the previous isolates by PFGE. This is consistent with persistence, or auto-reinfection. As no environmental contamination with ESBL-producing Enterobacteriaceae was detected, auto-reinfection does not appear to be from an environmental reservoir but more likely from resident-resident interactions or resident-health care worker interactions. This may also indicate the need for caution in accepting failure to detect colonisation as confirmation of clearance.

Persistent colonisation was also common for MRSA. Of the 17 MRSA-colonised participants 73% of those who had two or more subsequent samples remained positive. Consistent with standard practice in LTCF in Ireland this LTCF had no programme of routine resident MRSA decolonisation.

Regarding environmental contamination, MRSA was recovered much more commonly from the environment than was ESBL-EC, despite being recovered from a smaller proportion of participants, which is consistent with previous findings [69]. The greater frequency of environmental detection of MRSA compared with ESBL-E. coli may reflect more intense shedding of MRSA or better environmental survival. The environmental detection of MRSA compared with no ESBL-EC environmental contamination may reflect more intense shedding of MRSA or better environmental survival.

Interestingly, a large number of residents did not become colonised with ESBL-EC or MRSA throughout the year. This occurred despite an absence of restrictions on social interactions for residents colonised with MRSA or ESBL-Enterobacteriaceae, and frequent environmental contamination with MRSA. It is clear that, as the inverse of risk factors for colonisation, those who do not become colonised were less likely to have consumed antimicrobial agents and to reside in the high-dependency areas. However, other factors may contribute to colonisation resistance. This phenomenon clearly merits further study.

The study has several limitations. First, although initial study enrolment was high and retention was good, both were incomplete, with represents one of many challenges in conducting a longitudinal LTCF-based study. Second, antimicrobial exposure may have been underestimated, since antimicrobial use during hospital admission was not captured. Third, although the observed clonality of the ESBL-EC and ESBL-KP study isolates supports dissemination within this LTCF, it is not possible from the current analysis to differentiate definitively between acquisition of ESBL-producing organisms within the LTCF vs. during episodes of hospitalisation or via other exposures. In future studies, emerging higher-resolution typing methods such as whole-genome sequencing could be used to better clarify transmission pathways [70].

Despite those limitations this study represents a significant contribution to research in this area given the lack of longitudinal studies in LTCFs and there are several novel and important findings. Antimicrobial prescribing in LTCFs is an immediately modifiable risk factor for colonisation with AROs. We demonstrate the clonality and temporal stability of ESBL-producing Enterobacteriaceae within a single LTCF. Although multiple ST131 variants were detected, all belong to the H30-Rx subclone [5,7]. Colonised participants tended to retain the same pulsotype over an extended period. We report the first identified ESBL-KP in an Irish LTCF. We note that although present in 5 participants this organism has not disseminated to the same extent as ESBL-EC (n = 35). Further studies are needed to clarify the acquisition, retention, and dissemination of antimicrobial-resistant clones within LTCFs and to establish if there are other characteristics of such individuals, e.g., the gut microbiota, that may contribute to such colonisation resistance.