Introduction
Vancomycin-resistant enterococci (VRE) are a major cause of healthcare-associated infections, responsible for an estimated 54,500 infections associated with 5,400 deaths annually among hospitalized patients in the United States.1 Mortality, length of stay, and costs of care are higher for patients with VRE infections compared to patients infected with vancomycin-susceptible enterococci.Reference Puchter, Chaberny, Schwab, Vonberg, Bange and Ebadi2,Reference Prematunge, MacDougall and Johnstone3
Patients can become asymptomatically colonized with VRE. Colonized individuals are at increased risk of VRE infectionReference Meschiari, Kaleci and Monte4,Reference Alevizakos, Gaitanidis, Nasioudis, Tori, Flokas and Mylonakis5 and can spread VRE to others. In the hospital setting, patients can become infected or colonized through contact with other patients, healthcare workers, or hospital surfaces and equipmentReference Arias and Murray6; VRE can survive on dry surfaces for months,Reference Wendt, Wiesenthal, Dietz and Rüden7–Reference Porter, Sultan and Mitchell9 and removal requires aggressive disinfection regimens.Reference Byers, Durbin, Simonton, Anglim, Adal and Farr10,Reference Tan, Linkenheld-Struk and Williams11
Detection of suspected transmission events has generally relied on uncovering epidemiologic links among patients from whom organisms of the same species with phenotypic similarities—typically antibiotic resistance profiles—are isolated, often by means of prospective screening and culturing (“traditional surveillance methods”). However, these methods are often insufficient for VRE given the substantial spatiotemporal gaps that frequently characterize transmission events.Reference Permana, Harris and Runnegar12,Reference Abdelbary, Senn, Greub, Chaillou, Moulin and Blanc13 In addition, such investigations cannot distinguish between true transmission events and cases in patients with genetically unrelated strains.
Recent infection control (IC) efforts have focused on applying whole-genome sequencing (WGS) to characterize relatedness of clinical isolates, including VRE.Reference Sherry, Gorrie and Kwong14–Reference Higgs, Sherry and Seemann22 These efforts have demonstrated the utility of WGS in investigating suspected outbreaks. Emerging studies also suggest that prospective WGS applied to all diagnostic and screening isolates can detect otherwise unrecognized transmission events and facilitate targeted interventions.Reference Sherry, Gorrie and Kwong14,Reference Sundermann, Babiker and Marsh20 In this study, we evaluated the utility of WGS for VRE surveillance among oncology and ICU patients in an academic medical center.
Methods
Study setting and isolate recovery
We conducted this study at Brigham and Women’s Hospital (BWH) in Boston, MA. BWH serves primarily adult patients with approximately 50,000 inpatient visits per year across over 800 inpatient beds and includes programs for cancer care, solid organ transplant, and stem cell transplant. The Mass General Brigham IRB approved this study.
Vancomycin-resistant Enterococcus faecium (VREfm) strains were isolated in the BWH clinical microbiology laboratory between June 2022 and September 2023 from screening and diagnostic specimens. Per BWH IC policy, all patients admitted to intensive care and oncology units (Table s1) who are not known to be colonized by or infected with VRE are screened for VRE on admission and weekly thereafter; for additional details, see the supplementary methods. Diagnostic specimens included urine, blood, wounds, fluids, and tissue. All isolates were from unique patients and were assigned numbers in order of sequencing.
Sequencing and genomic analysis
Genomic DNA was extracted using the MasterPure Gram Positive DNA Purification Kit (Epicentre, Madison, WI). Libraries were prepared using the Illumina DNA Prep Kit and sequenced on the Illumina Miseq (Illumina, Inc., San Diego, CA).
Genomic sequence assembly was performed using an in-house bioinformatics pipeline. Quality control was performed on raw reads using Trimmomatic v0.39Reference Bolger, Lohse and Usadel23 and FastQC v0.11.9 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Draft assemblies were generated using SPAdes v3.15.3.Reference Bankevich, Nurk and Antipov24 Assembled genome quality was assessed using QUAST v5.0.3 (Table s2).Reference Gurevich, Saveliev, Vyahhi and Tesler25 Sequences with N50 < 25 kb were excluded. Genus and species were assigned using Kraken v2.1.2 (01/31/2023) and StrainSeeker v1.5Reference Roosaare, Vaher and Kaplinski26. Multilocus sequence typing (MLST) was performed using open-source code (https://github.com/tseemann/mlst) and stringMLST v0.6.3.Reference Gupta, Jordan and Rishishwar27 Minimum spanning trees were generated using SeqSphere+ (Ridom GmbH, Muenster, Germany). All sequences were deposited in the National Center for Biotechnology Information’s (NCBI) Pathogen Detection database (https://www.ncbi.nlm.nih.gov/pathogens/) under BioProject PRJNA278886.
SNP calling (Table s3) was performed using two methods. For whole-genome reference-based alignment, the U.S. Food and Drug Administration Center for Food Safety and Nutrition (CFSAN) SNP pipeline v2.2.1Reference Davis, Pettengill and Luo28 was used with reference strain ZY11 (CP038995.1). For split k-mer analysis (SKA), SKA v1.0Reference Harris29 was used to create k-mer files (k = 15, fastq subcommand, default thresholds for all variables) and calculate whole-genome pairwise SNP distances (ska distance function, default parameters).
Epidemiologic data collection and cluster identification
CFSAN and SKA have both been used for pairwise SNP calling in VREfm with SNP thresholds to trigger investigation for transmission ranging from ≤ 6 to ≤ 24 (median ≤ 10, interquartile range ≤ 7-15).Reference Sherry, Gorrie and Kwong14,Reference Brodrick, Raven and Harrison16–Reference Higgs, Sherry and Seemann22 We selected the upper limit of the interquartile range (≤ 15 SNPs) as an empiric threshold. For all patients whose isolates were ≤ 15 SNPs from another isolate by both methods, we collected information on patient unit assignment for the admission during which the isolate was recovered plus additional admissions as needed. We then identified epidemiologic relationships and sorted them into six categories: “same bed”; “same unit, same time”; “same service, same time”; “same unit, different time”; “same service, different time”; and “same outside facility, same time.” “Unit” refers to a set of rooms in the same physical location within the hospital; patients on a unit typically share the same nursing staff and some equipment. “Service” refers to the admitting service, which determines the patient’s clinician care team. Patients on a single unit might be admitted to different services, and vice versa. No patient pairs in the “same unit, same time” category shared the same room. Patients admitted to both the same unit and same service were classified as “same unit, same time” if concurrent and “same unit, different time” if not concurrent. We defined putative transmission clusters as sets of two or more isolates wherein each had a SNP distance of ≤ 15 from at least one other isolate and wherein each patient had an identifiable epidemiologic relationship with at least one other patient. We prepared color-coded maps to visualize the connections. Statistical tests were performed in GraphPad Prism 10 except the Marascuilo procedure for multiple comparisons, which was performed manually following a significant chi-square test.
Results
Description of the isolates and putative transmission clusters
We analyzed the genomes of 156 VREfm isolates (Table s2). Most (142, 91.0%) were obtained from screening specimens, while the remaining 14 (9.0%) were from diagnostic specimens (Figure 1A). Most (143, 91.7%) were collected between June 2022 and January 2023, including 129 from screening specimens, which represented 60.8% of the total screening isolates recovered from unique patients during that time. MLST sequence types (ST) 203 (35.3%), 117 (30.1%), and 736 (14.1%) were most common (Figure 1B, Figure 2A).
Analysis of genomic relatedness and epidemiologic connections among 156 vancomycin-resistant Enterococcus faecium (VREfm) clinical isolates identifies 16 putative transmission clusters. A) VREfm isolate source. Screening isolates were obtained from rectal swabs or stool samples collected to screen for VRE colonization, while diagnostic isolates were obtained from other specimen types sent to the microbiology laboratory for diagnostic culture. B) Sequence type (ST) distribution of all isolates (n = 156) and isolates involved in putative transmission clusters (n = 41). C) Histograms depicting the distributions of single nucleotide polymorphism (SNP) distances calculated by reference-based alignment (CFSAN) or split k-mer analysis (SKA). The left panel shows comparisons between all isolate pairs except isolate 1416, which was separated from all other isolates by 19,109–20,018 SNPs by CFSAN and 16,738-18,206 SNPs by SKA (asterisk at top of y axis). Three prominent peaks are evident. The peak around 3,000 SNPs primarily represents comparisons between isolates from ST203 or ST117 and isolates from ST736, while the peak around 2,300 SNPs primarily represents comparisons between isolates from ST203 and ST117. The peak centered around 25-50 SNPs is comprised of comparisons between isolates of the same ST. Bins are 50 SNPs in width, and tick marks on the y axis represent the center of each bin. The right panel is an expansion of the smallest SNP distance bin on the left panel, which contains pairs with SNP distances < 50. Bins are 1 SNP in width. The bracket indicates pairs with SNP distances of 15 or fewer, which are quantified in the Venn diagram. Epidemiologic data were collected from the electronic medical record for all patients involved in the 82 pairs with SNP distances of 15 or fewer by both CFSAN and SKA to probe for epidemiologic evidence of transmission. D) and E) SNP distance distributions of isolate pairs for which epidemiologic connections were identified (n = 33, dark blue) compared to those for which connections were not identified (n = 49, light blue). Each circle represents an isolate pair. Black lines behind the dots represent the median and interquartile range. The distributions were compared using the Mann-Whitney test (U = 481.5 for D and 591.5 for E). Exact two-tailed p values are shown. F) and G) The 82 isolate pairs separated by 15 SNPs or fewer by both CFSAN and SKA were organized into four bins of equal width (4 SNPs). The distribution of isolates with epidemiologic connections identified (dark blue) was assessed by the chi-square test (chi-square = 8.575 for F and 5.632 for G, both with three degrees of freedom). A chi-square test for trend was used to test for a linear relationship between SNP distance and the proportion of isolates with epidemiologic connections (chi-square = 8.022 for F and 3.742 for G, both with one degree of freedom). H) Number of isolates in the 16 putative transmission clusters identified based on genomic relatedness and epidemiologic connections. Graphs in parts A and C-H generated using GraphPad Prism 10.
Minimum spanning tree (MST) prepared using core genome multilocus sequence typing (cgMLST) shows relationships among the 156 vancomycin-resistant Enterococcus faecium isolates. The two panels show the same MST color coded by A) sequence type or B) putative transmission cluster as determined by the single nucleotide polymorphism (SNP) distance and epidemiologic connection analysis described in the text. Each circle represents one isolate, or occasionally multiple isolates that share the same cgMLST. Circles are labeled with the isolate number(s). The MST was generated by Ridom SeqSphere+ as a 3D visualization in which the length of each line connecting two isolate circles is linearly correlated to the core allele difference between the isolates. The figure is a 2D flattening of the 3D structure; thus, lengths of the flattened lines do not correlate directly with core allele difference. All core allele differences > 10 are labeled on the connecting lines in panel A. Distances are based on 1,062 columns from E. faecium cgMLST with no missing values and a clustering distance of 20. cgMLST is a method of pairwise comparison that is related to but distinct from CFSAN (reference-based alignment) and SKA (split k-mer analysis); rather than detecting SNPs across the entire genome, cgMLST types isolates based on the alleles present at > 1,000 genes. Overall, the cgMLST relationships depicted in the MST align well with those indicated by pairwise SNP distances with some minor discrepancies. For example, isolate 406 has the same cgMLST sequence type as isolate 1363 but not isolate 1433. However, by pairwise SNP comparison, isolate 406 is more closely related to isolate 1433 (five SNPs different by CFSAN and SKA) than isolate 1363 (seven SNPs different by CFSAN, nine by SKA). Although closely related to isolates 1363 and 1433, isolate 406 is not included in cluster 4 due to lack of identifiable epidemiologic links. For the same reason, isolate 1434 is not included in cluster 2 and isolate 1313 is not included in cluster 3.
We identified 82 isolate pairs separated by ≤ 15 SNPs by both CFSAN and SKA (Figure 1C) and found 33 (40.2%) with epidemiologic links (Figure 1D). The double threshold resulted in flagging of 34 fewer pairs compared to CFSAN alone and one fewer compared to SKA alone and slightly enriched for pairs with identifiable epidemiologic links (Figure 1C, Figure s 1). The number of pairs with epidemiologic links generally decreased with increasing SNP distance (Figure 1D-E). Among pairs separated by ≤ 15 SNPs, CFSAN and SKA SNP distance distributions were significantly different for pairs with epidemiologic links vs pairs without (medians four vs six by CFSAN and five vs eight by SKA) (Figure 1D-E). A chi-square test for trend performed on pairs separated into four SNP distance bins demonstrated a significant linear relationship between CFSAN SNP distance and proportion of pairs with epidemiologic links (p = 0.0046; Figure 1F) with a similar but nonsignificant trend for SKA (P = 0.0531; Figure 1G).
Epidemiologic analysis placed the 33 pairs with epidemiologic links into 16 putative transmission clusters of two to four isolates each (Figure 1H, Figure 2B). These clusters involved 41/156 isolates (26.3%) with a similar sequence type distribution to that of all isolates (Figure 1B, Figure 2).
VREfm WGS supports infection control investigations
Twelve isolates comprising four sets of three were sequenced at IC’s request as part of outbreak investigations (reactive WGS). For IC sets one and two, WGS ruled out clonal relationships (Figure 3B). For IC sets three and four, WGS supported a clonal relationship between two of the isolates (Set 3: 1176 and 1178, Set 4: 1174 and 1177) with the third more distantly related. In IC set three (Figure 3C), the third isolate (1194) belonged to a different ST, while in IC set four (Figure 3D), the third isolate (1175) belonged to the same ST but was separated from the other two isolates by > 70 SNPs.
Reactive and surveillance whole-genome sequencing (WGS) of vancomycin-resistant Enterococcus faecium (VREfm) supports infection control investigations. A) Key to maps of putative transmission clusters shown in Figures 3–6 and Figure s2. In each map, the lefthand side shows the single nucleotide polymorphism (SNP) distances between each member of the cluster (CFSAN, reference-based alignment; SKA, split k-mer anlaysis), while the righthand side shows the epidemiologic connections linking the patients from which the isolates were recovered. Each square represents one day, and sets of contiguous squares represent admissions. The color of each square represents the category of epidemiologic (epi) connection between or among patients in the cluster on that day. Squares in some maps are broken up by sets of five dots against a white background, which represent continuous stretches of time without notable changes to patient location. Patient locations during each admission are shown in white rectangles above the square(s) with the admitting service in parenthesis. Units were assigned arbitrary numbers and are described in further detail in Table s2. Vertical lines marked with day numbers demarcate location changes or vancomycin-resistant enterococci (VRE) screens or isolation events and connect to the relevant patient’s timeline. For VRE screen or isolation events, circles above the vertical line represent the sample type. Sequenced VREfm isolates are indicated by red outlines on the sample type circle, red day numbers, and red vertical lines. B) Infection control (IC) sets 1 and 2. Each set contained three isolates for which sequencing was requested by IC due to suspected involvement in a transmission cluster. Surveillance WGS linked one member of each set to another isolate. C) and D) Maps for IC set 3/cluster 1 and IC set 4/cluster 2. In each set, reactive WGS supported a clonal relationship between two isolates who were admitted to the same unit at the same time and ruled out involvement of a third isolate (patient ID with gray background), while surveillance WGS identified an additional transmission cluster member. GI surg, gastrointestinal surgery. Int, intensive care. NS, neurosurgery. Onc, oncology. PACU, postanesthesia care unit. E) and F) Maps for the reactively sequenced isolates from IC sets 1 and 2 that were linked to other isolates sequenced for surveillance. Card surg, cardiac surgery. ED, emergency department. ES, endoscopy. IR, interventional radiology. Rehab, Brigham and Women’s Hospital (BWH)-affiliated inpatient rehabilitation facility (units numbered separately from BWH units). Thorac surg, thoracic surgery. Vent, ventilator service.
Surveillance WGS combined with epidemiologic analysis identified additional members of IC sets three and four that would have otherwise been missed, likely due to spatiotemporal separation from the other patients (Figure 3C-D). One isolate from each of IC sets one (1445) and two (1363) were also placed into separate transmission clusters (Figure 3B, E-F).
Surveillance WGS of VREfm detects transmission clusters missed by traditional surveillance measures
Surveillance WGS detected 14 putative transmission clusters missed by traditional surveillance measures (Figures 3E-F, 4, 5; Figure s2). Two clusters demonstrate putative transmission between patients admitted to the same service at the same time (Figure 4). The others involve more complicated epidemiologic relationships. For example, the two patients involved in cluster 7 (Figure 5A) were on the same unit at different times (patient 1133 was admitted to unit 21 seven days after patient 1140 had left) and also concurrently on the same services (intensive care then medicine from days 15–23) while on different units. VREfm could have been transmitted between or to the patients from the unit 21 environment or a common provider or piece of equipment, potentially in a chain involving an unsampled third patient.
Surveillance whole-genome sequencing of vancomycin-resistant Enterococcus faecium detects transmission between patients admitted to the same service at the same time. Refer to Figure 3A for key to transmission cluster maps. ED, emergency department. Int, intensive care. Onc, oncology. Thorac surg, thoracic surgery.
Surveillance whole-genome sequencing of vancomycin-resistant Enterococcus faecium detects transmission among patients with complex epidemiologic relationships. Refer to Figure 3A for key to transmission cluster maps. AH, BWH-affiliated partner hospital (units numbered separately from BWH units). BMT, bone marrow transplant. Burn, burn service. Card, cardiology. ED, emergency department. Int, intensive care. Med, medicine. PACU, postanesthesia care unit. VS, vascular surgery. In C, “outside hospital” refers to the same facility. See Figure s2 for additional transmission cluster maps.
Other clusters detected by surveillance WGS included cases where the positive VREfm result for one of the patients occurred during an admission following the one in which transmission likely occurred. In Figure 5B, patients 1134 and 1376 were both admitted to the bone marrow transplant service from days 7–15 and to oncology unit 1 from days 17 to 21, when patient 1376 was discharged. Patient 1376 was readmitted on day 57, and their VRE screen returned positive, possibly reflecting colonization acquired during one of their two previous admissions via contact with patient 1134, common hospital staff or equipment, or the oncology unit environment.
Many putative transmission clusters demonstrated extended temporal separation between isolate recovery. For example, isolate 1424 (Figure 3D) was recovered about three months before the other two members of cluster 2. In Figure 3E, nearly seven months separate recovery of the two isolates with multiple epidemiologic links between the corresponding patients. This is consistent with reports describing prolonged VRE survival on environmental surfaces.Reference Wendt, Wiesenthal, Dietz and Rüden7–Reference Porter, Sultan and Mitchell9
Finally, one cluster of two isolates likely involved transmission at a common outside facility from which both patients were transferred before VRE screening cultures returned positive at our hospital (Figure 5C). Although the patients were concurrently admitted to BWH, they had no clear epidemiologic links during that period. Similarly, in Figure 5D, transmission might have occurred when the three patients were admitted to the same unit at an outside institution, although these patients had additional links during their subsequent overlapping BWH admissions.
Surveillance WGS of VREfm identifies the postanesthesia care unit (PACU) as a transmission locus
A previous report used VRE WGS to uncover a VREfm transmission locus.Reference Sundermann, Babiker and Marsh20 Our study identified PACU exposure as an epidemiologic link for six pairs of patients across four clusters. In cluster 2 (Figure 3D), a surgical tissue culture from patient 1424 grew VREfm. Three months later, two closely related isolates were recovered from patient 1174, who had visited the PACU twice prior to VREfm isolation, and patient 1177, who had no identifiable epidemiologic relationship to patient 1424 but was on the same inpatient unit (oncology unit 10) and service (oncology) as patient 1177. A potential transmission route was from 1424 to 1174 via the PACU and then from 1174 to 1177 via the oncology unit environment or a shared provider.
In another example (Figure 6A), a plausible chain involves contamination of the PACU by patient 1419 on day three with an isolate that was then picked up by patient 621 on day 18. In a third PACU-associated cluster (Figure 6B), patient 538 visited the PACU on day 15, the same day the patient’s VRE screen returned positive, and day 42. Meanwhile, patient 605 visited the PACU on day 43, likely resulting in colonization with VREfm (detected on day 51) from patient 538. Finally, in cluster 13 (Figure 6C), patient 1272 might have become colonized with VREfm left behind in the PACU several months prior by patient 1195, while patient 1387 could have picked up VREfm from either patient 1195 or 1272 in the PACU.
Surveillance whole-genome sequencing of vancomycin-resistant Enterococcus faecium identifies the postanesthesia care unit (PACU) as a locus for transmission. Refer to Figure 3A for key to transmission cluster maps. BMT, bone marrow transplant. Card surg, cardiac surgery. ED, emergency department. ES, endoscopy. Heme, hematology. Med, medicine. Neuro, neurology. NS, neurosurgery. Onc, oncology. Ortho, orthopedic surgery. Thorac surg, thoracic surgery. VS, vascular surgery.
Discussion
Building on previous reports,Reference Sherry, Gorrie and Kwong14,Reference Sundermann, Srinivasa and Mills19,Reference Gouliouris, Coll and Ludden21 our analysis shows that WGS for VREfm surveillance is an effective tool for identifying transmission events in the hospital setting. Fourteen of the 16 transmission clusters we identified were not detected by traditional methods, and surveillance WGS uncovered an additional member of the two recognized clusters. In addition, 24.3% of 144 isolates sequenced for surveillance were part of a cluster, indicating undetected transmission events are common; since only 60.8% of total screening isolates were available for inclusion, this likely underestimates the true extent of transmission. Factors that made these transmission events difficult to detect via traditional methods include temporal separation of up to several months between isolate recovery; recovery during an admission following the transmission event; and transmission in areas not previously suspected to pose a high risk of spreading VREfm, namely the PACU.
The PACU services both outpatients and inpatients and is a high-volume, high-turnover area with dedicated anesthesiology and nursing teams. Although the potential for spread of infectious diseases in the PACU is recognized,Reference Badgwell30 to our knowledge VRE transmission in the PACU has not previously been reported. Many patients requiring surgery are vulnerable to VRE colonization or infection due to severe underlying illness and/or perioperative antibiotic treatment. Features of the PACU that promote transmission among vulnerable patients include high turnover rates, which complicate consistent disinfection between patients, and receipt of care in open bays separated only by curtains with inconsistent availability of hand hygiene supplies, potentially hampering adherence to hand hygiene guidelines. Our findings of multiple VREfm transmission events linked to the PACU could inform targeted interventions, such as increased monitoring of hand hygiene and equipment disinfection adherence rates, more frequent environmental disinfection, and possibly pre-operative VRE screening to trigger contact precautions in the PACU.
Limitations of this study include incomplete sampling and lack of environmental sampling. Sequencing only 60.8% of screening isolates during the main collection period likely limited detection of transmission events and raises the possibility that some of the identified transmission clusters—for example, those with long intervals between cases—are incomplete. Furthermore, while the goal of this study was not to assess risk factors for VRE colonization, the gaps in sampling limited our ability to identify patient factors that might facilitate transmission events; such information could help prioritize patients for genomic screening. In addition, our epidemiologic analysis was restricted to manual analysis of high-level data comprising patient service, unit, and bed assignments during admission to hospitals within our network. Inclusion of more detailed information, such as provider identities and dates of interaction and dates of imaging and bedside procedures involving reusable equipment, could improve our ability to identify clusters. Automated tools to mine this information from the electronic medical recordReference Miller, Chen and Sundermann31,Reference Kanjilal32 could facilitate these investigations.
An additional limitation is the retrospective nature of the study, which precluded communication of the results to IC and clinical teams in real time for immediate intervention. However, the findings provide impetus to engage in prospective surveillance to inform and assess the efficacy of interventions. Newer sequencing platforms, such as the Oxford Nanopore, and automated tools to evaluate epidemiologic linkage could reduce turnaround time for transmission chain identification.
Other challenges include selecting a method for assessing genomic relatedness and determining the degree of genomic relatedness that warrants epidemiologic investigation. This can be particularly challenging for pathogens like VREfm that comprise epidemic healthcare-associated clonesReference Pinholt, Larner-Svensson and Littauer15,Reference Higgs, Sherry and Seemann22,Reference Pecora, Holzbauer and Wang33 ; the highly homogeneous population structures of such clones can be explored using NCBI’s Pathogen Detection database (https://www.ncbi.nlm.nih.gov/pathogens/), which shows that the isolates from our study are most closely related to other isolates from BWH and another institution in Boston. Both CFSAN and SKA have been used for comparison of VREfm genomes, the former more commonly. Here, we implemented both methods with an empiric double threshold of ≤ 15 SNPs. As only one pair was flagged by SKA but not CFSAN, the double threshold was effectively equivalent to a SKA threshold alone. Higgs and colleagues found that compared to CFSAN SNP distances, SKA SNP distances correlated more closely with those calculated by a gold-standard method (pairwise comparison using de novo references).Reference Higgs, Sherry and Seemann22 Thus, SKA SNP distances might better reflect genomic relatedness in VREfm, and pairwise comparison by SKA could enrich pairs flagged for investigation for true transmission events.Reference Maechler, Weber and Schwengers34
Importantly, SNP thresholds are used not to indicate definite evidence of transmission but to identify pairs warranting further epidemiologic investigation. Using our empiric threshold, we demonstrated a linear relationship between SNP distance and proportion of epidemiologically linked pairs, suggesting that establishing SNP ranges that correlate with different probabilities of epidemiologic linkage could be a practical strategy to triage pairs for epidemiologic investigation in WGS surveillance programs. Our initial data suggest a clear divide between 0–11 SNPs and 12–15 SNPs by CFSAN, with pairs in the latter category less likely to have identifiable epidemiologic connections. A larger dataset collected via a comprehensive WGS VRE surveillance program could help further refine these risk categories, including for SNP distances calculated by SKA.
Finally, cost remains a barrier to implementation of hospital-based genomic pathogen surveillance. While a comprehensive cost-benefit analysis is outside the scope of this study, Kumar and colleagues have proposed that bacterial WGS surveillance programs have an 80% chance of being cost effective if hospital systems are willing to pay $2,400 per averted transmission.Reference Kumar, Sundermann and Martin35 One Canadian study estimates that an averted transmission could save about $16,000 on average,Reference Lloyd-Smith, Younger, Lloyd-Smith, Green, Leung and Romney36 reflecting the costs of longer hospital stays, higher mortality, antibiotic treatment, and implementation of contact precautions. Other recent analyses also support a positive cost-to-benefit ratio.Reference Kumar, Sundermann and Martin35,Reference Price, Ngwira and Lewis37–Reference Mellmann, Bletz and Böking39
Overall, our results demonstrate that WGS of diagnostic and screening VREfm isolates can detect transmission clusters missed by traditional IC methods. If obtained prospectively, such information could spur interventions to halt ongoing outbreaks and inform changes to hospital IC strategy to better protect patients from the harmful consequences of VRE colonization and infection.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ice.2025.68.
Acknowledgments
The authors would like to thank the hard-working staff in the BWH microbiology laboratory who recovered the isolates included in this study.
Financial support
This study was funded by the BWH Department of Pathology; a P30 grant from the National Institute of Diabetes and Digestive and Kidney Diseases, National Institues of Health under grant number DK034854; and the Massachusetts Life Sciences Center. The work of Jay Worley was supported by the National Center for Biotechnology Information of the National Library of Medicine and the National Institute of Allergy and Infectious Diseases, National Institutes of Health.
Potential conflicts of interest
All authors report no conflicts of interest relevant to this article.
Competing interests
The authors declare none.
