Clinical implementation of routine whole-genome sequencing for hospital infection control of multi-drug resistant pathogens

whole geno me sequencin g, infection prev ention and contr ol, multi-resistant organisms, healthcare-associated in fe ctions, cl inical implementation Running title : Routine WGS for hospi tal infe ction contr ol Summary (40 words): We initiated a program of routine sequencing of multi-drug resistant organisms. A custom analysis pipe line was used to aut o mate reporting by in corpora ting cli nical meta-data with genomics to define c luste rs and supp ort inf ecti on cont rol in terven tions. Word count: 2,797 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 3

Background: Prospective whol e-genome sequencing (WGS)-based surveillance may be the optimal approach to rapidly ident ify t ransmission of multi-drug resis tant ( MDR) bacter ia in th e healthcare s ett ing. Materials/methods: We prospectivel y colle cted methic ill in-resistant Staphylococcus aureus (MRSA), vancomycin-resistant ente ro cocci (VRE), carbapenem-resistan t Acinetobacter baumannii (CRAB), extended-spec tru m beta-lactamase (ESBL -E) and carbapenemase-producing Enterobactera les (CPE) isolated from blood cul tures, ste rile s ites or scre eni ng specimens across three large t ertia ry re ferral hospita ls (2 adult, 1 paediatri c) in Brisbane, Au stralia. WGS was used to determine in silico multi-locus seq uence typing (MSLT) and resis tance gene profi ling via a bespoke genomic analysis p ipelin e. Putative transmission eve nts were ide ntified by comparison of core genome single nucl eotid e pol ymorphisms (SNPs). Relevant clini cal meta-data were combined with genomic analyses via customised automation, col lated in to hospital-speci fic reports regularly dis tribu ted to infe cti on contro l teams.

Over four years (Apri l 2017 to July 2021) 2,660 isolates were sequ enced. This included MDR gram -negative bacilli (n=293 CP E, n=1309 ESBL), MRS A (n=620) and V RE (n=433). A total of 379 clinical rep orts were issued. Core genome SN P data identifi ed that 33 % of isolat es formed 76 distinct c luste rs. Of th e 76 cluste rs , 43 were contained to the th ree targ et hospita ls, suggesting ongoing trans mission with in the cl inical envir onment. The remaining 33 clusters represent ed possib le int er-hospital tr ansmission events o r strai ns cir culati ng in the community. In one hospital, pr oven negligible tra nsmission of non-multi-resistan t MRSA enabled changes to infect ion con trol po licy. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 4

I mplementation of ro utine WGS for MDR pathogens in clinica l laborato ries i s feasible and can enable targeted in fe ction pre venti on and cont rol in terven tions. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 5

Healthcare-associated in fect ions (HAIs) are common and associated with significant morbidit y and excess healthcar e-related co st [1-3 ]. Each year in Australia, more than 165,0 00 patients experience HAIs [4] . Increasing rates of ant imicrobial res istance (AMR) ex acerbates the impact of thes e inf ecti ons [5] and adds co nsiderable cost burden to th e heal thcare sys tem [3]. In Australia, 30-day mortality rates for h ospital-onset vancomycin resi stant Enterococcus (VRE) and methicillin-res istant Staphylococcus aureus (MRSA) bloodstream infec tions were 20% and 1 4.9% respecti vely [6], with mortality fr om extended-spectrum beta-lactamase (ESBL)-producing E. coli bloodstream infe ction s up to 18.6% [7 ]. Traditional molecular meth ods used to des crib e potentia l genetic relat ionships betwe en organisms lack res olut ion, may be time consuming and are not read ily ava ilable o utside spec ialist laborato ries [8]. Whole genome sequencing (WGS) is now established as the optimal method to analyse outbr eaks and explore transmissio n dynamics of bacter ial pathogens [9] and could conce ivably ac t as a fron tli ne tool in the analys is and manage ment of most pathogens that repr esent a th reat to human health [10]. There has been limited diagnosti c laborato ry capacity to trac k path ogens causing thes e infect ions o r det ect transmission eve nts in the hea lthcare set ting. This st udy aimed to de velo p a clinical WGS workfl ow, able to pro spective ly dete ct t ransmission even t s before th ey become established, guide in fect ion prev entio n and control inte rvent ions and resp o nd to outbr eaks.

Setting . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 6 The project was in itiat ed in three tert iary-refer ral hosp itals wi thin metropolitan Brisbane , Australia, to prov ide regular WGS re ports to infec tion c ontro l teams. Sequencing services we re also availab le on requ est for region al faci lit ies s erved by th e sta te-wide laborat ory networ k (Pathology Queensland). Recommendations fo r the management of mul ti-resistant organisms (MROs) in Queen sland Health facil itie s are based on sta te-wide guidelines [ 11]. Sample collection Patient inclus ion cr iter ia were de fine d as: 1. Any patient hospi talised at three te rtiary care faci lit ies ( Royal Bris bane and Women’s Hospital [RBWH ], Princess Alexandra Hospital [PA H] and Queensland Children’s Hospital [QC H]) with a MRO screening culture positive fo r any of the fol lowing: MRSA, VRE, ESBL - producing Enterobacteral es (ESBL-E ), carbapenem-resistant Acinetobacter baumannii (CRAB) or carbapenemase-producing Enterobactera les (CPE). 2. Any patient admitted at RBWH, PA H or QCH with a positive cultur e for a t arget pathogen (ESBL/C PE, MR SA, VR E or CRAB), from Sterile Site [Blood, t issues, flu ids or CSF ] cultures. 3. Any patient hospi talised at outly i ng metropolitan Brisbane hospi tal s or regional Queensland Health Facilities with CR AB or C PE referred to the cent ral refe rral laborato ry for molecular test ing. WG S was a vailable on reques t for other bac te rial species or resistance pheno types in the c ontext of su spected outb reaks or un explai ned resis tanc e phenotypes (e.g. suspected carbape nemase production, without carbap enemase genes detected b y PCR). Data collection . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 7 Basic patien t demographic data (age, hospital / ward locati on, dates of a dmission) and sample data (date of col lecti on, species ide ntifica tion, resis tance pro file) were automatically ent ere d into a REDCap electronic data captur e system [2 0] usin g HL7 V 2 messagin g originating from the Pathology Queensland laboratory in formation system (AUSLAB, Citadel Health, Melbourne). The data were then in tegrated int o th e genomic analysis pipel ine (Figure 1) . Microbiology All col lected iso lates we subjec ted to cultu ring, ident ifica tion, and re porting acco rding t o routine laborato ry pro toco ls at Path ology Queensland (Supplementar y material). Acceptab le screening samples f or MRSA inc lud ed skin/mucosal swabs, resp irato ry samples or urin e; fo r VRE, swabs (rectal/perineal ) or fae ce s; for ESBL/CP E/CRAB , skin/mucosal swabs, urine or faeces. Whole genome sequencing All patient isolat es were submitt ed fo r WGS (Illumina Nextseq 500; 150bp paired-end) at the Queensland Health Forens ic Scient ifi c Services (FSS) laborato ry in wee kly b atches (Supplementary material). Genome analysis pipeline Genomic analysis was undertaken using a custom, in -house microbial geno mic analysis pipeline , SnapperRocks ( ht tps://github.com/F ordeGenomics/SnapperRocks ). SnapperRocks us es Conda (https://anaconda.org/) for tool manage ment, and the Nex tFlow workflow engin e (https://www.nextflow.io/) to achie ve reproducib ili ty and easy deploy ment across personal computers or high-performance computing cluste rs (Figure 1). Pipeline workflow is d iscussed i n detail in the supp lementary material. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 8 Figure 1: Schematic representation of the workflow SNP profiling and Clustering Sequence Reads for each isolate w ere mapped to a strain-specific rep resentat ive refe rence genome using SNPdra gon ( https://github.com/Forde Gen omics/SNPdragon ). In brief, trimmed reads were aligned to a referenc e using BWA -me m. Sa mtools [12] mpileup was used to calculate pe r base coverage and fre ebayes (h ttps:/ /github.com/freebayes/freebayes) to call varian ts. Pair-wise core genome SNP differences were used to iden tify c luster s of genetically s imilar isolates where l ocal transmission w as like ly. Three o r more Isola tes we re deemed to form a probable transmission clust er if the p air-wise SNP differ ences between th em was no more tha n 5 core genome SNPs /Mb. Isolates were defin ed as community-associated if they were colle cted within 2 days of hospi tal admission a nd hospita l-associated if coll ected >2 days pos t-admission [13 ]. The mean collection time of all isolat es within a clus ter was used to determine if a particular clust er was community associated (CA) or hosp ital ass ociat ed (HA).

Sample collection Between 19th April 2017 and 1st July 2021, a total of 2, 660 b acterial is olates were included, cultured from samples obtained fro m the participa ting hospita ls. The is olates were coll ecte d from 2336 p atients, with 259 patients providing multiple iso lates. A furt her 64 isolates were obtained from environmental sour ces within participa ting hospitals. Of the 2596 patient- derived i solates, 1535 (58.8%) were collect ed from scr eening specimens (e .g. rectal swabs), 278 (10.6%) from invasive clinical infec ti ons (blood cul tures [n=277] and cerebrospinal f luid [n=1]), . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 9 772 (29.6 %) from non-invasive clini cal infec tions (Table S1). Samples w ere coll ected week ly, with an average of 8 samples (min=0; max=39) sent for whole genome sequencing (WGS) each week. In silico taxonomic profiling and multi-locus sequence typing (MLST) MRO isolates co llec ted over the cou rse of this projec t included r epresen tatives of 50 bacter ia l species from 22 genera. Dominant species acc ounted for 93% of coll ected i solates (Table 1) [14]. The remaining 173 isolates we assigned to 37 different speci es, includin g Citrobacter freundii (n=14), Serratia marcescens (n=8) and Pseudomonas aeruginosa (n =30) (Table S1). Table 1 : Lineage diversity among sequenced isolates MLST profiling r evealed high le vels o f spec ies d ivers ity among 5 most abundant spec ies (Table 1). E. coli were dominated by S T13 1 (n=276; 37.3%) , which accounted for 10% of all clinica l isolates (Figure 2). In contrast, the r e was no single dominant lineage among K. pneumoniae . However, lineages associated with the spread of carbapenem resis tance were present; specifi cally, ST11, S T15, S T101 and ST 258 [15 ]. Dominant Enterobacter spp. lineages ( E. hormaechei ST90 and ST830) and A. baumannii lineages (ST1050) were rela ted to pr evious ly described outb reaks [16]. S. aureus w ere mainly ST93 , a co m munity associ ated clone frequ ently identif ied in Queensland [17, 18]. All prominent E. faecium l ineages belon ged to clonal complex 17 (CC1 7), responsible fo r a signifi can t burden of healthca re-associated inf ections [19]. Figure 2: Distribution of clonal lineages among the 5 most prevalent species Antibiotic resistance genes . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 10 In addition to ESBL and other beta-la ctamases, GNBs carried a diverse range of genes conveying resistance to numerous classes of antib ioti cs, includ ing aminoglycosides, tet racycl ines, sulphonamides and quinolon es (Tabl e S2). Carriage of carbapenamase genes was identified in 289 isolates from 21 Enterobacterale s species (Figure 3; Table S1). bla IMP-4 was most commonly observe d a mong C PEs (n=170), consisten t with prev ious report s [20, 21]. Five differen t NDM homolo gues were iden tif ied: bla NDM-1 (n=9), bla NDM-4 (n=2), bla NDM-5 (n= 25), bla NDM7 (n=1) and bla NDM-9 (n =2) (Table S1 & Table S2). Other ca rbapenemases inc luded OXA-types (e.g. OXA-48 [n=10] , OXA-23 [n=31] , OXA- 181 [n=11]), K PC -2 (n=5) and IM I-1 (n=1). Isolates carrying both NDM and OXA carbapenemases were detec ted in 7 pa tients (Table S1). Two carbapenemase-producing Pseudomonas aeruginosa carr ying IMP-1 and DIM- 1 were iden tifi ed. O ther carbapene mases intrinsi c to les s common species were occasiona lly id entifi ed (e.g. GOB in Elizabethkingia anophelis ). The mec gene ( mecA), which mediat es methicillin resi stance, was identifi ed in 92.5 % (56 9/615) of S. aureus isolate s (Figure 3). The absence of mec (n=46 isolates) usua lly ref lected e rrors in phenotypic testi ng due to calculat e d microbroth MICs and fluctuat ions around interp retat ive breakpoint s. In enterococc i, vanA (n = 239) was m ore common than vanB (n=159), reflecting both global an d national t rends [22]. Ho wever, rates may reflect sampling bias towards th e more resistant vanA gene. Additionally, 33 van comycin-susceptibl e E. faecium (VSEfm) were se quenced to det ermine whether these we re clona lly r elated t o circula ting healthca re-acquired VRE (Figure 3). . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 11 Figure 3: Temporal distribution and frequency of resistance-associated genotypes identified among clinical Isolates. Clustering and Transmission Characterising the genomic re lation ship (SNP distance) be tween iso late s revealed that, 33% (864/266 0) of samples were found to clus ter wi th two or more isolate s of the same clona l lineage (Figure 4). Overall, a total of 76 clusters were identif ied across th e six most frequently identif ied spec ies (Table 2). Table 2: Clustering of predominant species Average cluster size was 4.5 isola t es, with 77.6% of clu sters conta ini ng <10 isolates. Two clusters, Ab1050-A1 and Eh90 -A2 , w ere lin ked to p revious ly desc ribed no socomial outbrea ks o f A. baumannii [16] and E. hormaechei [20 ] , respectivel y. Finally, 79 single t ransmission events (2 isolate c luste rs) were al so dete cted, without fur ther onward transmission (Table S1). The majority (43/76) of our defined clusters wer e geographically con fine d to a single hospita l. The remaining 31 clusters had a wid er geographic dist ributi on with isolat es coll ected from tw o or more diffe rent facil itie s, suggesting the possibil ity o f inte r-facili ty t rans mission networks. Figure 4: Epidemiological curve of all isolates collected during this study. As a maker of heal thcare-associa ted ( HA) transmission the time between h ospital admission an d sampling was inspected. For 35 of ou r clusters, the mean period be tween patient admission an d isolate co llec tion (Del taAS) was foun d to be ≤ 2 days, su ggesting that the se clusters cou ld hav e arisen thro ugh community rather th an nosocomial transmission (Figures S1- S6, Supple mentary . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 12 material), although we cannot rule t he possibi lity of transfe r fr om other unsampled healthcar e facili ties. Colle ctiv ely, Escherichia and Staphylococcus species, account ed for 70% of suspecte d CA-clustering with lower rates o f s uspected community clus tering among other frequ entl y identif ied spec ies (Figure 5). Figure 5: Community versus healthcare-association by genus. Genomics reporting and integration with infection control During the stud y per iod, a to tal o f 37 9 genomics reports wer e issued to inf ection contro l t eams. The median time for sample proc es sing (collec tion to re ferra l fo r WGS) was 10 days. 41% of isolates arrive d for sequenc ing ≥ 11 days post c ollec tion and includ ed >3 00 “historical” isola tes collec ted >30 days (min 30; ma x 102 5) before process ing (Figure S7B, Sup plementary material). The median turnaround time fo r se quencing was 7 days, with 29% of s amples taking ≥ 8 day s (min 8; max 48) (Fi gure S7 C, Supplem entary material). The median turnaround time for genomic analysis and repor t generation was 6 days (min 0; max 1 20), with 43 % of s amples taking ≥ 8 days (Figure S7D, Supple mentary material ). Overall, the median dura tion fro m sample collection t o final genomic report ing was 33 days (+/-1 standard d eviati on: 21 to 45 d ays), with the fastest reports de live red within 10 days (Figure 6; Figure S7A, Supplementary mat erial). Reports were dis cussed during regular hospital in fect ion cont rol committee meetings and presented to hospi tal administ rators rout inely and in resp onse to ou tbre aks. The design o f th e reports (see supplementary material ) aimed to concis ely summarise th e key chara cter istics o f collec ted organisms, alert the team t o any li kely clus tering (be tween cu rr ent and/or pr evious ly sequenced iso lates and es tablished clusters ) and describ e key r esista nce genes. Additiona l . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 13 dynamic exploration o f the r elati ons hips between iso lates cou ld be achi eved by accessing th e online data visualisa tion porta l CAT H AI ( https://cathai.fordelab.com/ ), wh ich can also generate line-listings o f patien t meta-data and epidemic curves o f def ined clu sters. Figure 6: Turn-around times for genomic reports Implementation of genomic surveillance into clinical practice. To enhance the transla tional impact of the su rvei llance pr ogram, w e involved infec tion preventio n and control nur se special ists with an inter est in the appl icati on of genomics. Their roles were essent ial for l inking the genomic reports to the clini cal envir onment and ensuring suspected clust ers were investigate d and subsequent action occurred. Overall, the genomic surveillan ce program provided sever al key advantages when complemen ting routine inf ection control prac tice. Specif icall y, the reports allowed unequivo cal determinati on of whether isolate s were link ed, and the li kelih ood o f t r ansmission between pa tient s or heal thcare env ironments. Conversely, it was use ful i n dispro vin g transmission events in suspec ted o utbreaks when i solates with similar pheno types wer e iden tif i ed in pati ents wi thin shared loca tion s. Criticall y, it a llowed identif icat ion o f ear ly c lusters to initiat e more-targeted in fect ion co ntrol measures (e.g. intensive sc reening of clus ter-spec ific con tacts, environmental sampl ing and cleaning of locations l inked to clus ters). WGS r eporting also identi fied organisms associated with hospita l outbreaks in o ther Australian s tate s (e.g. S T14 24 VRE) and could identify paral lel clus ter s amongst phenotypically similar organisms, allowing more refined prioritisat ion of infe ctio n control resou rces. As Pathology Q ueensland pr ovides a sta te-wide microbiology servi ce, a genomics program such as this off er ed a networ ked su rveil lance system able t o de tect inte r- . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 14 hospital transmissi on events that wo uld be otherwise invis ible. Such a co-ordinated respo nse to AMR threats has been ident ified as a criti cal step for pub lic-health p revent i on [23].

The application of genomics to characteris e transmission routes an d enhance infecti on preventio n and contro l practi ces dur ing clinical outb reaks is now well es t ablished [14, 20 , 24 ]. However, these studies are largely r etrospec tive, of ten temporall y and spatially c onfined, an d generally limited in scope with regards to species of in teres t. Additional ly , the use of genomics as real time surveillance and infec ti on prevention too l is rare. Here we describe the clini cal implementation of continuous, genomic surveillance to iden tify, tra ck and interrupt th e transmission of mult idrug resistan t p athogenic bacter ia. Coupling genomic and epidemiological data revea led tha t 37 identi fie d clusters had arise n through suspect ed community ra the r than h ospital tran smission ev ents. However, comparison of SNP distances suggests tha t, even among CA -clusters, within-hospi tal t r ansmission cannot be completely ruled out. Gen omic-based transmission and cl uster ing in formation suppo rted existing prac tice to not use i solat ion procedur es f or pa tients co lonised with ESBL- E. coli , and contribu ted t o the deci sion a t one hospital to remove is olati on pre ca utions for no n-multi- resistant MRSA, resulting in est imated savings of $690,864 (A UD) over one year [25]. High levels of community transmissi on of MDR organisms is of grea t co ncern. Legislati on t o designate high-risk MDR organisms as not ifiab le t o publ ic hea lth c ould bring add itiona l resources t o bear and help reduce the impact of community reservoi rs of AMR. However, without a better under standing of t he epidemiological facto rs that pro mote dissemination of . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 15 MROs in the wid er c ommunity, it i s unclear what in terven tions would be most e ffec tive t o interrup t these transmission n etwork s. Timely provision of report s is cruc ial for an eff ectiv e survei llance p rogram . While we were able to provide repor ts within 10 days of sample collect ion, the mean report t urnaround time of 33 days stretches the boundar ies of cli nical relevanc e. However, delays as sociated with sample transport to th e central labo ratory, culture-based process ing, the lack of onsite or dedica ted WGS infrastruct ure, continuous deve lopment of the analysis pipeline, tra nsfer of isolat es to an off-site sequenc ing facilit y, batched sequencing, geno mic analysis and report generation al l contribu ted to lengthy r eport ing periods. However, with workflow ref ine ments and structura l re-organisation many of these delays could be minimised. Prior to commencement of ou r geno mic surveil lance p rogram, major gra m-ne gative outbrea ks were recorded i n Brisbane ho spitals in 2015 (IMP -4 carbapenem ase-producing E. hormaechei ST90), 2 016 (carbapenem-resistant A. baumannii ST1050) and 2017 (OXA- 181 carbapenemase- producing E. coli ST38 and ESBL-pro ducing K. michiganensis ) [14, 16, 20 , 26] . Tar geted WGS played a critica l role in th e resolu tio n of these outbrea ks but was largely instigated only onc e extensive transmission had b een de tected. The added hea lth and e con omic benefit of WGS- enhanced inte rventi ons is suppor ted by recent simulat ion modelling, which estimates a 55.9% reduction in infe ctions and a co rresp onding 59.9% decrease in associated healthcare c osts ove r the course of an outbreak [26]. H ere, the identif ication of puta tive transmission cluste rs Ab1050 -A 1 and Eh90 -A 2, which w ere linked back to two of these previous outbreaks , demonstrates that continu ed surveil l ance is required to ensure that outb reaks are completel y resolved. Consequ ently, the true be nefit of WGS can only be r ealised when it is d eployed as a . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 16 prospecti ve surv eil lance t ool. A re cent eco nomic evaluat ion o f our surveillan ce program estimated that prospe ctive WGS will result in ~36,000 fewer H AIs, ~6 50 fewer deaths and overall c ost saving of $30million/y ear [27] . Despite th e success o f this p roject t here were limitat ions. Although rar e, available metadata were not always consis tent b etwe en sites resul ting in i nformation g aps which impacted downstream interpreta tions. The importance of i ntegrating genomic and epidemiological data has been well established as the l atter prov ides cri tical con textual in formation, includin g temporal and spatial dynamics. To facilita te the coupl ing of data in real-time, our pipeline wil l need to be fully integrated with Quee nsland Health’s laborat ory in formatio n system. This in tur n allows for the publ icati on of clin ic ally meaningful repo rts in actionab le timeframes whil e ensuring that pat ient privacy is pr otected. Seco ndly, thi s prospe ctiv e surveil lance pr ogram focused sole ly on multidrug resistan t bacteria. Consequently, the hea lth a nd economic impacts associated with the transmissi on of antibiotic susc eptib le disease-causi ng organis ms was not considered

The routine app licat ion o f prosp ect ive genomic survei llance for MROs encounte red in th e hospital env ironment is feasib le and can be implemented within a clinical diagnostic laborat ory . Such an approach p rovides earl y notif icat ion o f iso late clust ering and like ly i n-hospital transmission event s, thus facil itat ing targeted infec tion c ontro l responses. Despite the c omplex challenges of WGS pipeline develop ment, integrating appropriate computational in frastru cture within exist ing systems in the heal thcare set ting and succinc t commu nication o f comple x genomics information to clinicians, WGS -informed infection preven tion and control strategie s . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 17 are li kely become the new bench mark. While the costs of running such a serv ice a re considerabl e, they are justifi ed by the potential savings to the heal th system as a whole and enhanced preven tion o f health care-acquired in fect ions in vu lnerabl e patie nts. Ethics Ethical overs ight was provided by The Forensic and Scienti fic Serv ices Human Ethics Co mmittee (refer ence HEC17_17) as a Low and Negligible Risk approval, with p rov ision fo r a waiver o f individual pa tient consent. Funding This work was suppor ted by fundin g from the Queensland Genomics Health Alliance (no w Queensland Genomics), Queensland Health, the Queensland Government. P.N .A. H. was supported b y a Na tional Health and Medical Resear ch Counci l Early Career Fell owship Gran t (GNT11575 30). Conflicts of Interest P.N.A. H. reports resear ch grants f ro m Merck, Sandoz and Shionogi, outsi de the submit ted work ; has served on adv isory boa rds for Sandoz and Merck, and has receiv ed speaker’s fees from Pfizer, Sandoz and Sumitomo. D.L.P reports resear ch grants f rom Merc k, Pfizer and Shion ogi outside the submitted wo rk; has r eceived honorar ia f or adv isory boa rd membership from Merck, Pfize r, Shionogi, GSK, QPex , Entasis, VenatoRx, BioM erieux and Accelerat e. Othe r authors have no confl icts o f in terest t o declare. Acknowledgments . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 18 We thank the sci entis ts at Pathology Queensland and Forensic and Scienti f ic Services for their assistance in the labo rator y work and the infe ction contr ol teams at th e pa rticipa ting hospital s for the ir invo lvement and cl inical inp ut.

1. Agodi A, Au xilia F, Ba rchit ta M, e t al. Tr e nds, risk factors and ou tcomes of healt hcare-associa ted infections within th e It alian ne twork SPI N-UTI. J Hosp Infect 2013 ; 84(1): 52-8. 2. Emerson CB, Eyzaguirre LM, Alb recht JS, Comer AC, Harris AD, Furun o JP. He althc are-associa ted infection and hospi tal r eadmission. Infec t Control Hosp Epid emiol 2012 ; 33(6): 539-44. 3. Wozniak TM, Bail ey EJ, Gr aves N. H ealth and economic burd en of antimicrobi al-re sistant infections in Aust ralian hospi tals: a popul ation-base d model. Infect Cont rol Hosp Epidemiol 2019 ; 40(3): 320-7. 4. Australi an commission on safety and qua lity in healthca re . Reducing harm t o pati ents from healthca re associa ted infec tions: A n Aust ralian infecti on preven tion and con trol m odel for acut e hospitals, 2009 . 5. Lydeamore MJ , Mitch ell BG, B ucknall T, Cheng AC, Russo PL, Stewardson AJ . Burd en of five healthca re associa ted infec tions in Aust r alia. medR xiv 2021 . 6. Coombs G BJ, Daley D, Collignon P, Cooley L, Gottlieb T, Ired ell J, W arn er M, Nim mo G, Robson J Australi an Gr oup on An timicrobial R esist ance Sepsis Ou tcome Programs: 2020 Re port Aus trali an Commission on Safety and Quality in Hea lth Care, 2020 . 7. Australi an commission on safety and qua lity in healthca re . AURA 2021: fou rth Aus tralian repo rt on antimicrobi al use and r esistanc e in human health . Sydney, 2021 . 8. Didelot X, Bowden R, Wilson DJ, Peto TE A, Crook DW. Transforming clinical micro biology with bacteri al genome seq uencing. N at R ev G enet 2012 ; 13(9): 601-12. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 19 9. Quainoo S, Coole n JPM, van Hijum S, et a l. Whole- Genom e Seque ncing of Bacteri a l Pathogens: the Fut ure of Nos ocomial Ou tbre ak Anal ysis. Clin Microbiol Rev 2017 ; 30(4): 1015 -63. 10. PHG Foundati on. Path ogen G enomics Int o Practice. Avail able a t: https://www.phgfoundation .org/media/ 113/download/Pathogen_gen omics_pra ctice.pdf?v=1&i nline=1 . 11. Queenslan d Heal th. Ma nagemen t of multi-resista nt organ isms. Available a t: https://www.health .qld.gov. au/__d ata/a ssets/pdf_file/0026/444626/multi-resistant- organisms.pdf . Accesse d 05/11/21. 12. Danecek P, Bonfield JK, Liddle J , et al . Twelve years of SAMt ools and BCFto ols. Gig ascience 2021 ; 10(2). 13. Friedman ND, Kaye KS, St out JE, et al . He alth care--associa ted blo odstr eam infecti ons in adults: a reason t o change th e accept ed defini tion of community-acquired infections. A nn I ntern M ed 2002 ; 137(10): 791-7. 14. Chapman P, Forde BM, R ober ts LW, et al. Genomic Investiga tion R eveals Contami nated Detergen t as th e Source of an Ex tend ed- Spectrum-be ta-Lactamas e-Producing Klebsiella michiganensis Outbr eak in a N eona tal U nit. J Clin Microbi ol 2020 ; 58(5). 15. Munoz-Price LS, Poirel L, Bonomo R A, et al. Clinical epidemiol ogy of the global ex pansion of Klebsiella pneumo niae car bapen emases . Lancet Infect Dis 2013 ; 13(9): 785-96. 16. Rober ts LW, Ford e BM, Hurs t T, et al . G e nomic surveillance, char acte riza tion and interven tion of a polymicrobial multidrug-r esistan t outb reak in critical car e. Mic rob G enom 2021 ; 7(3). 17. Munckhof WJ, Schoon eveldt J, Coombs GW, Ho are J, Nimmo GR . Emergenc e of community- acquired me thicillin-resist ant S taphyloco ccus aureus (MRSA) infection in Que ensl and, Aust ralia . Int J I nfect Dis 2003 ; 7(4): 259-64. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 20 18. Chua KY, Seemann T, Harrison PF, e t al. T he dominant Austr alian community-acqu ired methicillin-resis tant S taphylococcus aur e us clone ST93-IV [2B] is highl y virulent and genetically distinct. PLoS One 2011 ; 6(10): e25887. 19. Lebret on F, van Schaik W, McGui re A M, et al. Eme rgence of epid emic multidrug-r esistan t Enterococcus faecium from animal and c ommensal strains . mBio 2013 ; 4(4). 20. Rober ts LW, Harris PN A, Ford e BM, e t al. Integr ating multipl e genomic tech nologi es to investigate a n outb reak of carbap enema se-producing Enter obact er ho rmaechei . Nat Commun 2020 ; 11(1): 466. 21. Rober ts LW, Catchpool e E, Jen nison AV, et al. Genomic an alysis of carbapenemas e-producing Enterob acte riace ae in Que ensland r eveal s widespread t ransmission of bla IMP-4 on an IncHI2 plasmid. Microb Genom 2020 ; 6(1). 22. Gorri e C, Higgs C, Carter G, St inea r TP, Howden B. G enomics of vancomycin-resistant Enterococcus faecium. Mic rob G enom 2 019 ; 5(7). 23. Slayton RB, To th D, Lee BY, et al . Vital Sig ns: Estimated Effects of a Coordina ted A pproach for Action t o Reduce Antibio tic-Resist ant Inf ections in Heal th Care F acilities - Unit ed Stat es. MM WR Morb Mor tal Wkly Rep 2015 ; 64(30): 826-31. 24. Kwong JC, Lane CR, Romanes F, et al . Tra nslating genomics into pr actice for r eal-ti me surveillance and r esponse to carba pen e mase-producing Ente robact eriac eae : evi dence from a complex multi-insti tuti onal KPC outbreak . PeerJ 2018 ; 6 : e4210. 25. Elliott TM, Ha re N, H ajkowicz K, et al. Evaluating th e economic effects of genomic sequencing of pathogens to prio ritise h ospital pa tien ts competing for isolati on beds . Aust He alt h Rev 2021 ; 45(1): 59-65. 26. Elliott TM, Lee XJ , Foeglei n A, Har ris PN, Gordon L G. A hybrid simula tion model a pproach to examine b acte rial genome s equencing d uring a hospital ou tbr eak. BMC Infec t Dis 2020 ; 20(1): 72. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 21 27. Gordon L G, Elliot t TM, Ford e B, e t al. Bud get impact analysis of rout inely using whole-genomic sequencing of six multidrug-r esistan t bac terial p athog ens in Queensl and, Aus trali a. BMJ Open 2021 ; 11(2): e041968. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 22 TABLES Table 1 : Lineage diversity among sequenced isola tes Species Total # isolates # ST identified Prominent STs^ # Isolates with uncharact erised STs* A. baumannii 35 12 ST1050 5 E. coli 879 122 ST131, ST1193, ST38, ST69, ST10, ST648 26 Enterobacter spp. 168 23 ST90, ST830 6 K. pneumoniae 330 122 ST13, ST15, ST17 12 E. faecium 433 26 ST78, ST80, ST203, ST1421, ST1424 8 S. aureus 612 42 ST1, ST5, ST22, ST30, ST93, ST97 21 ^ Lineages with > = 20 identified isol at es (excep t for A. baumannii ST1050 [n=15]) *Isola tes may belong to the same unch ar acteris ed ST Table 2: Clustering of predominant sp ecies Species Clustered isolates #Clusters STs with clusters Maximum Cluster Size (Cluster ID) A. baumannii 18 3 ST1050, ST451, ST229 14 isolates (Ab1050- A1) E. coli 249 27 ST10, ST38, ST69, ST73, ST88, ST127, ST131, ST141, ST185, ST349, ST410, ST457, ST1193, ST4553 76 isolates (Ec131-A8) Enterobacter spp.^ 60 4 ST90, ST513, ST599, ST830 23 Isolates (Eh415-B1) K. pneumoniae 67* 7 ST13, ST22, ST133, ST348, ST429, ST711, ST1412 17 isolates (Kp13-A1) E. faecium 289 19 ST17, ST78, ST80, ST203, ST555, ST780, ST1421, ST1424 74 isolates (Ef1421- A2) S. aureus 149 17 ST5, ST30, ST72, ST78, ST88, ST93, ST97, ST239 51 isolates (Sa93-A2) * Includes 19 K. michiganensis isolates fr om a neonat al SCU outbr eak (KmNA-A1). ^ Includes: E. bugandensis, E. cloacae, E. hormaechei . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 23 FIGURE LEGENDS Figure 1 : Schematic representation of the workflow, including sample processing, clinical meta-data collection, SnapperRocks analysis pipeline, cluster visualisation and clinical report generation. MRO = multi-resistant organism; Abr = Antibiotic resis tanc e; MLST = multi-locus sequence typi ng; SNP = sin gle nucleotid e polymorphism; Vir = virulence; SCC mec = staphyloco ccal chr omosomal cassette mec; CA TH AI = Cluster Analysis Tool for Healthca re- Associated Infec tions. Figure 2: Distribution of clonal lineages among the 5 most prevalent species. Lineages with <20 representa tive members are c lassif i ed as “other” and isola tes whose s equence typ e had no t previousl y been de fined at the t ime of writing were des ignated as “NF”. Figure 3: Temporal distribution and frequency of resistance-associated genotypes identified among clinical Isolates . Isolates are grouped by week o f col lect ion f rom April 2017 to July 2021. Figure 4: (A) Epidemiological curve of all isolates collected during this study. Isolates are grouped by week of c olle ction from April 2017 to July 2021. Dotte d line ind icates a change in sample collect ion with a focus on a ta rgeted strategy, prio rit ising organisms of greatest concern (e.g. C PE, CRAB, VRE , E SBL- K. pneumoniae ) or invasive clin ical is olates of ot her pathogens (e.g. ESBL- E. coli, MRSA). (B) Epidemiological curve of clustering isolates coloured by genus. Only the top 5 genera are sh own, with all other clu ster ing isolates classi fied as “Other”. Figure 5: Community versus healthcare-association by genus. Clustering isolates are colour ed by genus, as per the legend, and scal ed by number of iso lates. Dashed line indicates a sample collec tion t ime of 2 days post admission (DeltaAS=2). RBWH=Royal Br isban e & Wo men’s Hospital; QCH=Queensland Children’s Hospital; PAH=Princess Alexandra Hospital. Figure 6: Turn-around times for genomic reports from day of sample colle ction t o repor t generation with a cu t of f of 90 days. Green and blue dash ed lines repres e nted the ach ieved and possible r eport ing periods, 10 and 7 days respect ively. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 24 FIGURES Figure 1: Schematic representation of the workflow, including sample processing, clinical meta-data collection, SnapperRocks analysis pipeline, cluster visualisation and clinical report generation. MRO = multi-resistant or ganism; Abr = Antibiotic resis tance; MLST = multi-locus sequence typ ing; SNP = single nucleot ide polymorphism; Vir = vi rulenc e; SCCmec = staphyloco ccal chr omosomal cassette mec; CA TH AI = Cluster Analysis Tool f or Healthcare- Associated Infec tions . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint 2 5 Figure 2: Distribution of clonal lineages among the 5 most prevalent species. Lineag es with <2 0 representa tive m em bers are class ifi e d as “other” and iso lates whose sequ ence type had n ot previousl y been de fined at the t im e of writing were des ig nated as “NF”. 5 0 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint Figure 3: Temp oral distrib ution and f requ enc y of resis t anc e- associated ge not ype s iden tified amo n g c linical Is o lates . Isolates are g r week of c o l lection f rom A pri l 2017 to Ju l y 2021 2 6 r ouped by . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint Figure 4: (A ) Epidem i o logic al curve o f al l isolates collec t ed dur ing this s t u dy. I s olat es are gr ouped by week of c ollection f r om A pr il 2 2021. Do tted line ind icates a ch ange i n s am ple collection with a focus on a targeted str ategy, priorit is ing or g anis ms of greate st con c e CRA B, VRE, ESBL - K. pneumoniae) or i n v asive c linical i solates of oth e r path ogens (e. g. ESBL - E. coli, MR SA ). (B) Epidemiological c u rve o is o lates c olou red by gen us. Only t he top 5 genera are shown, with all othe r c luster ing isolates c la s s ified as “ Ot h er”. 2 7 2 017 to J uly e r n (e.g. CPE, o f clustering . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint Figure 5: Com munit y ver s u s h ealt hc a r e-ass ociation b y g en us. Clu ster ing is olates are coloured by genus , as per t he leg en d, a n d scal e of isolates . D as h e d line indicates a s a mple collec t ion t ime o f 2 days po st a dmis sion (D eltaA S=2). RBWH=R oyal Bris b a n e & Women’ s H QCH=Queensland Children’s Hospital; PA H =Princes s A l exandr a H ospital 2 8 e d by number H os pit a l; . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint Figure 6: Turn-around times for genomic reports from day of sample collection to report generation with a cut off of 90 days. Green and blue dashed lines represented the achieved and possible reporting periods, 10 and 7 days respectively. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 3, 2022. ; https://doi.org/10.1101/2022.05.02.22273921doi: medRxiv preprint