Epidemiology and Transmission of Carbapenemase-Producing Enterobacteriaceae in a Health Care Network of an Acute-Care Hospital and Its Affiliated Intermediate- and Long-Term-Care Facilities in Singapore

ABSTRACT Movement of patients in a health care network poses challenges for the control of carbapenemase-producing Enterobacteriaceae (CPE). We aimed to identify intra- and interfacility transmission events and facility type-specific risk factors of CPE in an acute-care hospital (ACH) and its intermediate-term and long-term-care facilities (ILTCFs). Serial cross-sectional studies were conducted in June and July of 2014 to 2016 to screen for CPE. Whole-genome sequencing was done to identify strain relatedness and CPE genes (blaIMI, blaIMP-1, blaKPC-2, blaNDM-1, and blaOXA-48). Multivariable logistic regression models, stratified by facility type, were used to determine independent risk factors. Of 5,357 patients, half (55%) were from the ACH. CPE prevalence was 1.3% in the ACH and 0.7% in ILTCFs (P = 0.029). After adjusting for sociodemographics, screening year, and facility type, the odds of CPE colonization increased significantly with a hospital stay of ≥3 weeks (adjusted odds ratio [aOR], 2.67; 95% confidence interval [CI], 1.17 to 6.05), penicillin use (aOR, 3.00; 95% CI, 1.05 to 8.56), proton pump inhibitor use (aOR, 3.20; 95% CI, 1.05 to 9.80), dementia (aOR, 3.42; 95% CI, 1.38 to 8.49), connective tissue disease (aOR, 5.10; 95% CI, 1.19 to 21.81), and prior carbapenem-resistant Enterobacteriaceae (CRE) carriage (aOR, 109.02; 95% CI, 28.47 to 417.44) in the ACH. For ILTCFs, presence of wounds (aOR, 5.30; 95% CI, 1.01 to 27.72), respiratory procedures (aOR, 4.97; 95% CI, 1.09 to 22.71), vancomycin-resistant enterococcus carriage (aOR, 16.42; 95% CI, 1.52 to 177.48), and CRE carriage (aOR, 758.30; 95% CI, 33.86 to 16,982.52) showed significant association. Genomic analysis revealed only possible intra-ACH transmission and no evidence for ACH-to-ILTCF transmission. Although CPE colonization was predominantly in the ACH, risk factors varied between facilities. Targeted screening and precautionary measures are warranted.


DISCUSSION
Prevalence and transmission of CPE in the health care network. In Singapore, different types of carbapenemase genes have been identified (12)(13)(14)(15)(16); this is suspected to be a consequence of its being a highly connected international travel hub (1, 17). In our study, the proportion of CPE among meropenem-resistant Enterobacteriaceae in the ACH was 47/78 (60.3%) which was similar to the CPE prevalence of 64.7% (2010 to 2015) in six public hospitals reported by the Carbapenemase-Producing Enterobacteriaceae in Singapore (CaPES) Study Group (12). Our study further observed that three of the ertapenem-susceptible isolates harbored CPE genes (two bla IMP-1 and one bla KPC-2 ). bla OXA-48 was also identified in one meropenem-susceptible isolate. Additionally, we identified the presence of plasmid pNDM-ECS01 in two E. cloacae strains, which has not been   previously reported. It is therefore prudent to include the use of molecular and genomic methods, in addition to conventional cultures, for the active surveillance of CPE in all facilities.
The ACH was the main reservoir of CPE in the health care network, with patients in the ACH being 1.3 and 4 times as likely to be CPE colonized as patients in ITCFs and LTCFs, respectively. ACH patients with $3 weeks of hospital stay were 2.7 times as likely to be CPE colonized as those with shorter stays (aOR, 2.67; 95% CI, 1.17 to 6.05).
In contrast, ILTCFs patients with a ,3-week stay were 53% more likely to be CPE colonized (aOR, 1.53; 95% CI, 0.37 to 6.41), suggesting possible colonization due to recent admissions to ACHs. The effect of duration of stay on CPE colonization differed by the type of health care facility (18). Our findings support CPE screening when long stayers are transferred from the ACH to ILTCFs to prevent interfacility transmission of CPE (19). Moreover, the gradual increase in CPE prevalence in both the ACH and ILTCFs over the 3 years highlights the need for enhancing infection prevention and control strategies in all health care facilities. Interfacility transmission of CPE would result in hospital outbreaks (20,21). The estimated cost for a single CPE outbreak (assuming it affects 40 patients) within a network of health care facilities is 1 to 1.6 million U.S. dollars (22). Of 983 patients who overlapped based on year of admission in the same facility, we found only one potential example of a phenotypically identical and genotypically nearly identical strain isolated from two patients. These patients were on different wards, which could suggest a widespread distribution of this strain; alternatively, as we found no additional closely related CPE strains from that facility that year, this could indicate that any intrafacility spread, if present, was limited and sporadic. There was one additional cluster of five closely related E. cloacae strains, all isolated from the ACH. The SNP distances for the strains isolated in 2015 were suggestive of direct transmission within the ACH, with potential persistence of that strain at least until 2016.
Acquisition of CPE could happen within a facility as well as between the different types of facilities during step-up or step-down treatments (23). We found only one potential example of two E. coli strains (with the same carbapenem resistance phenotype) that were genomically separated by 22 SNPs. These were isolated in different years, and the isolate from 2015 was from an LTCF, while the isolate from 2016 was from the ACH. Therefore, we had no evidence of direct interfacility transmission and concluded that such transmission events must be too rare for us to have captured them in our data set. Patients transferred from ACHs to long-term acute-care hospitals (LTACHs) in the United States tended to be more severely ill and possibly more prone to CPE colonization (19). Although only half of the LTACH residents had prior ACH admissions, carbapenem resistance in LTACHs was 9 times higher than that in ACHs, suggesting that ACH-to-LTACH transmission was uncommon (11). As LTACHs provide just as intense clinical care as ACHs and for a much longer duration (average length of stay in LTACHs is $25 days [24]), it is no surprise that they are major reservoirs of carbapenem and other antibiotic resistance. Unlike LTACHs in the United States, which provide rigorous clinical care and observation, including the prolonged use of ventilators, the ILTCFs in Singapore provide more rehabilitative and much less intense clinical care, as well as long-term residential nursing care for their patients and residents. As such, ILTCFs are likely to play a much smaller role in the CPE epidemic. Moreover, the prevalence of CPE in ILTCFs was very low, supporting our suggestion that ACH-to-ILTCF transmission is rare. Prior hospital admission was also not found to be associated with CPE colonization in ILTCFs.
Risk factors for CPE colonization. Several risk factors for colonization of carbapenem-resistant Enterobacteriaceae have been previously identified (6,25,26). As a wellknown risk factor for antibiotic resistance (27,28), history of use of any antibiotics increased the prevalence of CPE colonization in all health care facilities overall. In multivariate analysis, we observed that exposure to penicillins tripled the odds of CPE colonization in ACH patients (29,30). Hence, antibiotic stewardship is crucial in the control of CPE in ACHs. We observed that prior PPI exposure tripled the odds of CPE colonization in ACH patients; interestingly, this association was not observed among ILTCF patients. This could possibly be due to the combined effect of PPI and antibiotics on the gut bioflora reducing the population of commensal bacteria and increasing the risk of colonization with pathogenic ones (31). We found only a marginal interaction between penicillins and PPI (data not shown).
We also observed that a history of wounds increased the odds of CPE colonization in ILTCFs by 5 times. Presence of wound in patients indirectly reflects functional status, underlying medical conditions and nursing care required by a patient. Prevention of wounds and proper wound care in ILTCFs cannot be overemphasized (32). The finding of an association between dementia and higher CPE colonization in the ACH might be due to dementia patients requiring intensive nursing care, which could increase their risk for acquisition of CPE in ACHs. We further identified an association of connective tissue diseases with CPE colonization in the ACH, possibly due to the immunosuppressive effects of their medications.
We found that prior exposures to respiratory procedures were independent risk factors for CPE colonization in ILTCFs. Although respiratory procedures were likely to have been performed in ACHs, care for these devices continues even after patients are transferred to ILTCFs. Hence, the proper handling of medical devices is more important than the type of medical devices used, irrespective of the health care facility (27). Training of health care staff in the proper handling and cleaning of devices, good hand hygiene after handling of devices, and contact precautions of patients after identification of patients with CPE are recommended for all types of health care facility (11).
In our study population, as with those in other studies (33), prior CRE carriage was the factor most strongly associated with current CPE colonization, regardless of facility type. The time from prior CRE colonization to current CPE colonization ranged from 4 to 493 days (median, 24; IQR, 10 to 66). We further observed that prior VRE carriage was an independent risk factor for CPE colonization in ILTCFs. This could be due to the similar mode of transmission by VRE and CRE. These findings support our current hospital policy of isolation and contact precautions for prior VRE and CRE carriers from the point of admission (34).
Strengths and limitations. A major strength of our study was the inclusion of a large sample of patients hospitalized at various health care facility types, representing a participation rate of 87%. Hence, any selection bias was likely to be minimal. Moreover, the comprehensive, systematic, and standardized manner in which the rectal swabs/stool samples and clinical data were collected reduced any potential measurement error. Furthermore, any potential confounding was adjusted for in the multivariable regression models. Nonetheless, the study had some limitations. Antibiotic exposures outside the respective health care facilities, if not documented in the medical records, would have been missed. However, any information bias was likely to be nondifferential, thereby attenuating observed effects; hence, the associations observed in our study are likely to be conservative estimates. Another limitation is that the health care system-specific risk factors in the Singapore population, particularly under the provision of subsidized care, might not be generalizable to other health care systems. Regardless, the advanced medical care provided to patients receiving subsidized care makes the study findings applicable to other developed health care systems. Unfortunately, the epidemiology and risk factors for specific carbapenemase genes could not be examined in our study population due to the small sample size.
Conclusion. In conclusion, CPE colonization was low in the ACH and very low in ILTCFs. Indications of CPE transmission within the ACH were seen, but no examples of transfer between ACH and ILTCFs were observed. As CPE prevalence increases with time, CPE screening of long stayers being transferred from ACHs to ILTCFs can prevent interfacility transmission. Furthermore, preemptive isolation and contact precautions for prior CRE carriers should be undertaken on admission to any health care facility.

MATERIALS AND METHODS
Study design and setting. We conducted serial cross-sectional studies in an ACH and three each of its closely affiliated ITCFs and LTCFs over a 6-week period during June and July in 2014 to 2016. Over each 6week period annually, the study was conducted serially in the ACH and ILTCFs. The ACH was a 1,700-bed tertiary-care hospital which provided emergency, inpatient, and intensive care services for adults in general medicine, infectious diseases, cancer chemo-and radiotherapies, and general surgery, as well as trauma, neurosurgical, and spinal cord injury care. The ITCFs involved were a 100-bed rehabilitation center, a 116-bed community hospital, and a 360-bed community hospital. These facilities provide rehabilitative and subacute care for a period of 1 to 2 months for patients who require such care after admission to acute-care hospitals. In comparison, LTCFs are residential facilities that provide long-term nursing care for individuals who are unable to be cared for in their own homes. The LTCFs included in the study comprised a 164-bed chronic sick unit and two