An epidemiological and molecular study regarding the spread of vancomycin-resistant Enterococcus faecium in a teaching hospital in Bogotá, Colombia 2016

This was an observational, cross-sectional descriptive study. It was aimed at describing nosocomial VREfm pattern within the hospital, integrating epidemiological and molecular typing methods through a TPS algorithm. The study lasted 5 months, from May to September 2016 (i.e. enabling the study of outbreak and post-outbreak strains). Case definition followed CDC HAI criteria [28].

The study took place in an 802-bed tertiary-care teaching hospital, in Bogotá, Colombia. The hospital is divided into 2 healthcare facilities and the VREfm outbreak had place in largest one. It has a 506-bed capacity in adult general wards; there are 118 beds in its adult intensive care unit (ICU), 10 beds in its coronary care unit (CCU), 22 beds in its adult intermediate care unit (IMCU) and 25 beds in its neonatal intensive care unit (NICU). Efm has been considered an endemic microorganism in the Méderi teaching hospital since its first isolation in 2001; however, VREfm caused few infections until 2016. After a sudden increase in VREfm isolates in February 2016 (Fig. 1), the hospital’s Epidemiology Department started a healthcare–associated infection outbreak investigation and a VREfm-related outbreak was confirmed for May 2016. Nevertheless, some cases from the end of April and first days of June were also included within the outbreak investigation according to the analysis made by the hospital’s epidemiology department. Outbreak confirmation was based on the index of VREfm-related infections compared to the number of patients discharged per month, in at least 24 months. If this index was above the third standard deviation (SD) an outbreak was confirmed following criteria established by the Bogotá District Health Secretariat [29]. These criteria include:

This index also confirmed outbreak control in June 2016, following primary multidisciplinary interventions (Fig. 2). A post-outbreak study was made from July to September 2016.

Enterococcus faecium species were identified by Vitek 2 system mass spectrometer (software version 1.02, bioMérieux). Isolates’ in vitro susceptibility to antimicrobial drugs was determined by automated Vitek 2XLS card, based on Clinical Laboratory Standards Institute (CLSI) 2016 [30] criteria for Enterococcus. The first confirmatory manual method for vancomycin-resistance identification involved the E-test gradient diffusion method. Additional manual microdilution to quantify minimum inhibitory concentrations (MICs) for vancomycin and teicoplanin resistance was made by the Universidad Nacional de Colombia’s Microbiology Department, using previously identified vancomycin-resistant Enterococcus faecium strains. Enterococcus faecalis ATCC 29212 (NCTC 12697) was used as control strain and CLSI M100-S24 was used for interpreting the results.

vanA gene detection and molecular typing was performed for all 33 recovered strains; PCR was used for evaluating vanA detection. Primer sequences were based on the genes published for Efm [31, 32]. The reactions were performed with AmpliTaq Gold DNA polymerase hot-start enzyme with Buffer I (Applied Biosystems, Foster City, CA, USA); PCR fragments were visualised on 2% agarose gel.

Molecular typing involved VREfm isolates being grown in LB broth overnight with dextrose supplement (5 g/L) at 37 °C with shaking. A Wizard Genomic DNA Purification Kit (Madison, Wisconsin, USA) was used for extracting DNA, following the gram-positive bacteria protocol; 120 μL lysozyme (10 mg/mL) was used for cell lysis. Variable-number tandem-repeat (VNTR) analysis was chosen for determining clonality; VNTR-1, VNTR-7, VNTR-8, VNTR-9 and VNTR-10 were amplified since there were no consistent and/or reproducible results for VNTR-2 (required for multilocus variable number tandem-repeat analysis (MLVA)) [33]. The 5 VNTR loci were processed as described by Top et al., for MLVA, with some minor modifications [34].

Briefly, VNTR-1 PCR conditions were modified, involving 30 cycles. A touchdown PCR was used for VNTR-7, VNTR-8, VNTR-9 and VNTR-10, using the same conditions described by Top et al., except for initial touchdown (TD) temperature for VNTR-9 which was 65 °C, decreasing to 55 °C. The reactions were performed in 10 μL volume with an AmpliTaq Gold DNA Polymerase hot-start enzyme with Buffer I (Applied Biosystems, Foster City, CA USA). PCR fragments were visualised on 2% (w/v) agarose gel using GelRed nucleic acid gel stain (Biotium). These 5 VNTRs’ grouping results defined our clonal profiles.

A retrospective outbreak analysis was made, mainly for identifying transmission routes rather than the source(s) of infection. This perspective was considered, taking into account that we could not clearly identify the index case, due to Enterococcus faecium being classified as an endemic pathogen since its first isolation in 2001 and an increased number of isolates having been observed during the first months of 2016. Transmission analysis was then performed by adapting an algorithm previously described for a long-term outbreak of Pseudomonas aeruginosa in Germany from 2002 to 2015 [27], having successfully identified transmission routes. The modifications involved using VNTR analysis rather than whole genome sequencing (WGS) for assessing clonality. The outbreak involved 16 patients: 3 patients from April, 9 from May and 4 from June. All available outbreak strains (13 isolates) were included for this analysis since isolates from the first three patients could not be recovered.

Transmission was hypothetically considered if VREfm had first been detected in patient “A” before VREfm was first detected in patient “B”. Patients’ epidemiological tracking data and clonal profiles detected from outbreak isolates were combined to establish four criteria: criterion 1 was fulfilled if patients A and B were located in the same ward (24 h minimum overlap before VREfm was first detected in patient B), criterion 2 was fulfilled if patient B was located in the same room which patient A had occupied a maximum 2 weeks before patient B, criterion 3 concerned both patient A and patient B having stayed in the same room (minimum 24 h overlap before VREfm was detected in patient B) and criterion 4 concerned close genetic relatedness between VREfm isolates from patients A and B.

Possible transmission was considered when criterion 1 was fulfilled; probable transmission was considered when either 2 or 3 were fulfilled, as well as fulfilment of 4 alone. Predicted transmission was determined when 4 was fulfilled for patients A and B in combination with any of the three epidemiological criteria (1, 2 or 3); if no criteria were fulfilled, the probability of transmission was considered unknown [27].

Estimator-predictor epidemiology significance pointed out some deficiencies regarding cleaning and disinfection. Possible transmission (criterion 1 – same ward) could have occurred via healthcare personnel (patients fulfilling this criterion never shared a room but the same ward, receiving medical attention from the same healthcare personnel). Probable transmission (criterion 2, 3 – same room or criterion 4 – genetic relatedness) could have occurred via direct contact or environmental contamination. Predictable transmission (combination of any epidemiological criterion and genetic relatedness) included the previous explanations.

The Endemic Index confirmed that the outbreak occurred in May 2016, but patients from the end of April, May and the beginning of June were also included, according to the Epidemiology Department’s analysis. Different outbreak control strategies were introduced in line with Bogotá Territorial Health Department and Society for Healthcare Epidemiology of America (SHEA) recommendations for preventing healthcare-associated infections [29, 35] (Additional file 1: Figure S1).

The VREfm outbreak involved 16 patients. Upon reanalysis made by the infectious diseases department by retrospectively verifying case definitions according to the CDC criteria [28], four patients ended up being classified as having been colonised. One clinical infection was identified as having been acquired before admission to Méderi (external sources). After 6 patients died and 4 deaths were estimated as being attributable to VREfm infection, the VREfm outbreak became controlled in June 2016 (just 2 VREfm-related infections during this month) (Figs. 1 and 2). Nevertheless, post-outbreak assessment was pursued following national and international recommendations to assess outbreak intervention efficacy and verify outbreak control; 16 VREfm isolates recovered from inpatients from June to September were preserved for these purposes.

Environmental samples were obtained as quality control for cleaning and disinfection during the outbreak; 33 healthcare surfaces were evaluated on June 2nd, 2016, including some healthcare providers’ hands, bed rails, infusion pumps, monitoring equipment, computer keyboards and bedside nurse call buttons. These cultures revealed VREfm growth on three surfaces: 2 bed rails from different rooms in one general ward and one ICU infusion pump. Disinfection procedures were reinforced and followed-up on June 13th; two of the three surfaces proved negative and a VREfm isolate was recovered from the remaining bed-rail on the fifth floor. This surface received a second follow-up on June 20th (no VREfm isolation).

The 33 isolates were evaluated for antibiotic susceptibility but only the 29 clinical VREfm isolates were included in such analysis. Automated MICs were determined for ampicillin, ciprofloxacin, high-load streptomycin, high-load gentamicin, linezolid, quinupristin-dalfopristin, tetracycline, vancomycin and teicoplanin. Manual microdilution was performed to determine vancomycin and teicoplanin MICs. All isolates (100%) were confirmed as vancomycin- and ciprofloxacin-resistant strains, susceptible to linezolid, high-load gentamicin and quinupristin-dalfopristin; all but one (i.e. 28 isolates) were identified as teicoplanin- and ampicillin-resistant. Twenty-four had resistance to high-load streptomycin and 21 were tetracycline-resistant. Four qualitative antibiotic susceptibility patterns were identified (Table 1); manual methods revealed 26 isolates (89.6%) as having high-level vancomycin resistance (MIC> 128 μg/dL) and 28 isolates (96%) as being teicoplanin-resistant. One isolate was susceptible to ampicillin and teicoplanin, an unusual pattern for this kind of bacteria [16, 36, 37].

PCR identified the vanA gene in 26 of the 29 clinical isolates and in all 4 environment-recovered strains; vanA detection was related to medium and high-level vancomycin resistance (64 to > 512 MIC) and heterogeneous teicoplanin resistance (8 to < 256 MIC) [31, 36, 38, 39]. One of these 26 clinical isolates proving positive for the vanA gene was also identified as the only ampicillin- and teicoplanin-susceptible strain. Four isolates expressed a high-level vancomycin-resistance phenotype despite lack of vanA gene amplification (Table 2).

Molecular typing was based on VNTR-1, VNTR-7, VNTR-8, VNTR-9 and VNTR-10 grouping and was performed for the 33 recovered VREfm isolates (including environmental samples). Examples of the different VNTRs are shown in Additional file 2: Figure S2. Four clonal profiles (A, B, C and D) were identified from the isolates. The most identified profile for all clinical strains (29 isolates) was “A”, in 17 of the strains. The second was profile “B” in 10 of the strains, and “C” and “D” profiles were identified in one strain, each. Outbreak isolates were classified as 9 (69.2%) belonging to clonal profile “A”, 2 (15.4%) to profile “B”, whilst clonalities “C” and “D” were each recognised in just one isolate (7.7%). Profile “C” was also recovered from an environmental surface (bed-rail). Profile “D” was only identified in one external source isolate (having the lowest vancomycin-resistance level and being susceptible to teicoplanin and ampicillin). Only profiles “A” and “B” were identified in the post-outbreak isolates (16 strains), 8 for “A” (50%) and 8 for “B” (50%). No environmental study was implemented during this period. (Table 2a and b). Figure 3 shows the number of isolates displaying each clonal profile, distributed per month from May to September 2016.

A TPS algorithm (previously described for Pseudomonas aeruginosa) [27] was adapted to assess transmission routes (direct transmission or cross-transmission) in the 16 outbreak patients only; data collected from these 16 patients related to the VREfm outbreak was analysed and epidemiological criteria (possible and probable) were applied to all of them; however, genetic relatedness (predicted transmission) was evaluated in 13 out of the 16 samples as no isolates were recovered from the first 3 outbreak patients.

TPS algorithm analysis was used to track transmission routes in the outbreak patients. The epidemiological module was run for all 16 patients whilst epidemiological + genetic relatedness modules were run for 13 of them. P1 was the first patient identified during the outbreak (index case), but P2 was the estimated case who infected the greatest number of patients according to patient flow tracking. Using the adapted algorithm was aimed at investigating whether genetic relatedness based on VNTRs grouping (clonalities) could improve transmission route tracing compared to just conventional epidemiological surveillance, typically involving time (epidemic curve), space (geographic distribution) and person (patient characteristics) [29]. Figure 4a shows possible (criterion 1) and probable (criteria 2 and 3) transmission, and Fig. 4b shows predicted transmission (criterion 4 + 1 or 2 or 3). Four transmissions which had been classified as possible based on epidemiological criteria alone could be predicted with high probability by also applying genetic criterion 4. Looking at these four cases, it is worth noting that patients stayed on the same ward at the same time, but not in the same room (P4 likely transmitted his strain to P7 and P14; P5 likely transmitted his strain to P11; and P7 likely transmitted his strain to P16). Estimated transmission via healthcare personnel was thus likely (Fig. 4b). Patient 9 became infected during her hospital stay, having a different clonal profile, and was not noticeably exposed to previously infected patients (Fig. 4b).

Considering the high mobility between wards and floors that the patients had, it was difficult to trace the transmission between them. Some patients changed rooms four or more times during the outbreak, so representing an epidemiological curve including rooms was difficult, but it was constructed by wards (Additional file 3: Figure S3). Nevertheless, if a detailed map of where each patient was at each time is required, a spreadsheet depicting this can be found as Additional file 4: Table S1.