Multidrug-resistant organisms (MDROs) are increasing globally, and they disproportionately affect hospital patients.Reference Cassini, Högberg and Plachouras1,Reference Health Organization2 Infections with these pathogens may be acquired in healthcare settings or in the community, associated with increased morbidity, mortality, length of hospital stay, and healthcare costs.Reference Health Organization2–Reference de Kraker, Davey and Grundmann4 Although many healthcare systems, including in Australia, have successfully implemented surveillance programs for low-burden, high-impact pathogens such as carbapenemase-producing Enterobacterales (CPE),Reference Sherry, Lane and Kwong5–Reference Pfeiffer, Cunningham and Poissant10 these surveillance systems do not always comprehensively address more common multidrug-resistant organisms (MDROs). This results in incomplete data regarding some MDROs frequently affecting patients, such as extended-spectrum β-lactamase–producing E. coli (ESBL-Ec) and methicillin-resistant Staphylococcus aureus (MRSA).
Although some MDROs are clearly healthcare-associated and are rarely isolated in the community (eg, vancomycin-resistant Enterococcus [VRE]), others have more complicated patterns of transmission including both community and healthcare acquisition, such as MRSA and ESBL-Ec. In Australia, the successful implementation of programs to minimize healthcare-associated MRSA, such as Hand Hygiene Australia,Reference Grayson, Stewardson and Russo11 has led to the belief that most MRSA is now community acquired. Similarly, ESBL-Ec are thought to be predominantly community-acquired, yet no data exist to support these beliefs because routine microbiological and surveillance methods have insufficient resolution to address this question. In the absence of evidence, infection control teams have assumed that these MDROs are usually not acquired in a hospital and, therefore, may not be managed using additional infection prevention precautions.
In this pilot genomics implementation study, we performed comprehensive surveillance of the clinical and genomic epidemiology of MDROs across multiple hospitals over a 2-month period to answer these questions: (1) What is the local burden of MDRO infection and colonization? And (2) can genomics feasibly be used to predict in-hospital MDRO transmission, and estimate a transmission rate to allow comparisons between sites and over time?
Methods
Study design
We conducted a prospective multicenter study of 8 hospital sites from 4 hospital networks (Table 1) covering ~2,800 acute and subacute patient beds. Isolates were collected during an 8-week pilot study (April 24 to June 18, 2017) conducted as part of a larger study for the Melbourne Genomics Health Alliance, using genomics for MDRO surveillance in hospitals. Clinical and screening isolates of 6 MDROs were collected from hospital inpatients: (1) vanA vancomycin-resistant Enterococcus faecium (vanA VRE, confirmed by PCR), (2) methicillin-resistant Staphylococcus aureus (positive cefoxitin screen or oxacillin MIC >2 mg/L), (3–4) extended-spectrum β-lactamase (ESBL)-phenotype Escherichia coli and Klebsiella pneumoniae (ESBL-Ec and ESBL-Kp, defined by ceftriaxone resistance with MIC ≥4 mg/L and AmpC phenotypes included), (5) carbapenem-resistant Acinetobacter baumannii complex (CRAb, meropenem MIC ≥8 mg/L), and (6) carbapenem-resistant Pseudomonas aeruginosa (CRPa, meropenem MIC ≥8 mg/L and resistant to piperacillin-tazobactam and ceftazidime). Carbapenem-resistant Enterobacterales (CPEs) were excluded because they were already collected for state-wide CPE surveillance.Reference Sherry, Lane and Kwong5,12 Although vanB VRE are dominant in Australia, we focused on vanA VRE because it emerged more recently in Victoria, has greater associated antimicrobial resistance and costs, and may be more amenable to infection control interventions than vanB VRE. Additional details on inclusion and exclusion criteria are available in the Supplementary Data (online).
Table 1. Hospital Sites and Characteristics
Note. MDRO, multidrug-resistant organism; ICU, intensive care unit; MRGN, multidrug-resistant gram negatives (includes ESBL and carbapenem-resistant phenotypes); BMT, bone marrow transplant (allogeneic). ESBL-Ec, extended-spectrum β-lactamase phenotype E. coli; MRSA, methicillin-resistant S. aureus; vanA VRE, vanA-producing vancomycin-resistant E. faecium.
MDRO screening protocols
Existing MDRO screening protocols varied between hospitals (Table 1; details in Supplementary Data online). We assessed hospital infection control practices, including patient isolation and terminal cleaning, at baseline and at study conclusion; no changes were made during the study period. The results of genomic analyses were not available to hospitals during the study period.
Clinical data collection
Demographic and clinical data were collected for each patient, including whether the patient was likely to be colonized (ie, no symptoms of infection and no antibiotic treatment required) or infected with the MDRO (ie, symptoms of infection requiring targeted antibiotic treatment). History of infection or colonization with the same MDRO in the previous 12 months, from clinical or screening samples based on infection control alerts, laboratory data, or medical history, was also collected. We attempted to collect data regarding history of overseas travel or hospitalization; however, the travel history for most patients was unclear from the electronic medical record, and could not be clarified because most patients were discharged at the time of data collection. Hence, these data were excluded from the analysis. Antimicrobial exposure history was not collected.
Hospital laboratory workflow
MDROs were identified, analyzed, and reported by the hospital laboratory according to their usual protocols. For patients meeting study inclusion criteria, a pure subculture was sent to the central laboratory for sequencing and isolate storage. Results from automated susceptibility testing (Vitek 2 platforms, bioMérieux, Marcy-l’Étoile, France) were collected from each laboratory. Hospital laboratory records were audited at the end of the study period to assess completeness of isolate capture for the study.
Sequencing laboratory workflow
At the central sequencing laboratory, a single colony from the subculture received from the hospital laboratory was selected for subculture on horse blood agar; on day 2, 1–2 colonies were selected and placed in lysis buffer for sequencing. DNA was extracted on the JANUS Chemagic Workstation using Chemagic Viral DNA/RNA kit (CMG-1033, PerkinElmer, Waltham, MA). Whole-genome sequencing was performed on the Illumina NextSeq platform using Nextera XT libraries and protocols (Illumina, San Diego, CA), as previously described.Reference Carter, Buultjens and Ballard13 Sequences not meeting predefined quality metrics (minimum average quality score 30, target sequencing depth ≥40×) were resequenced.
Bioinformatic analysis
De novo assembly of all isolates was conducted as part of standard laboratory quality control workflows, using Shovill (v1.0.4; https://github.com/tseemann/shovill) for use in multilocus sequence typing and screening for acquired antimicrobial resistance genes.
In silico multilocus sequence typing (MLST) was conducted using the mlst tool (https://github.com/tseemann/mlst), and either the pubMLST sequence type (ST) database (for Escherichia coli, Enterococcus faecium, and Staphylococcus aureus) or the BIGSdb Institut Pasteur ST database (for Klebsiella pneumoniae).Reference Jolley, Bray and Maiden14 The BLAST-based Abricate tool (v0.9.5, https://github.com/tseemann/abricate, minimum coverage and identity 100%) was used to detect a subset of acquired antimicrobial resistance genes (Supplementary Table S1 online) from the NCBI Bacterial Antimicrobial Reference Gene database (version 2019-02-08).15
Transmission analysis
Fine-scale transmission analysis was conducted within each ST, except ST131 E. coli (2 subclades). Sequence reads for all isolates within an ST were aligned to a reference genome and variants called using Snippy (v4.4.3, https://github.com/tseemann/snippy) with default settings. Where possible, publicly available complete chromosomes of the same ST were chosen as the reference genome. In the absence of a complete, publicly available, ST-matched reference genome, de novo assemblies of the earliest isolate from the ST group were created using SPAdes (v3.13.0)Reference Bankevich, Nurk and Antipov16; these completed genome assemblies (used as reference genomes for read mapping) have been uploaded to GenBank (BioProject no. PRJNA565795). Reference genome data (including pairwise SNP distances to isolates for within-ST analyses) is available in Supplementary Table S2 (online).
Following read mapping and variant calling with Snippy, Gubbins was used to detect and mask recombination in the alignments for ST transmission analyses of E. coli, K. pneumoniae, and E. faecium. Reference Croucher, Page and Connor17 Recombination was not screened and masked for S. aureus because it has not been shown to have significant impact on this type of analysis.Reference Didelot and Wilson18 The resulting core genome alignments were used to calculate pairwise SNP differences for all pairs of isolates within each ST transmission analysis.
To screen for potential MDRO transmission, isolates with pairwise SNP distances of ≤25 SNPs to another study isolate (including same patient), or ≤15 SNPs for MRSA, were selected, based on results from previous studies (K. pneumoniae and E. coli,Reference Sherry, Lane and Kwong5,Reference David, Reuter and Harris19–Reference Pightling, Pettengill, Luo, Baugher, Rand and Strain21 MRSA,Reference Young, Golubchik and Batty22–Reference Harris, Cartwright and Török24 and E. faecium 25,Reference Health Organization26). Further epidemiologic data on these patients were collected by infection control teams at the participating sites, detailing admission history (dates, hospitals, and wards) from 12 months before the patient’s earliest study isolate, through to the end of the study period. Epidemiologic data were used to construct Gantt charts (not shown) for each species and ST with potential transmission, and overlapping admissions were then identified. ‘Probable transmission’ was defined as admission to the same ward at the same time; ‘possible transmission’ was defined as admission to same ward at different time (within 60 days), or admission to different ward in same hospital at same time; and all other patients were classified as ‘unlikely transmission’ (modified after Voor In‘t Holt et alReference Health Organization27). To investigate the relationship between genomic relatedness and epidemiology, pairwise SNP distances were plotted by species (and ST for major clones) and likelihood of transmission was plotted by epidemiologic classification.
Ethics approval
This study was approved by the Melbourne Health Human Research Ethics Committee (HREC) and endorsed by the corresponding HREC at each participating site.
Data availability
Raw sequence data were uploaded to the Sequence Read Archive (BioProject no. PRJNA565795).
Results
Isolate numbers, patient demographics, and specimen types
During the 8-week study period, 408 MDRO isolates from 358 patients were collected; most patients (88.3%) only had a single isolate included: 10.1% had 2 isolates and 1.7% had ≥3 isolates. The median age of patients was 67 years (Table 2). Overall, 47.5% of isolates were collected for screening purposes, although this varied between species (Fig. 1a and Supplementary Table S3 online). Of the clinical samples, urine specimens were most common (45.8% of clinical isolates), followed by nonsterile sites (25.2%), blood cultures (10.7%), respiratory specimens (9.8%), and other sterile sites (8.4%) (Fig. 1b). Of the clinical isolates, 28.5% were thought to represent colonization rather than infection (Table 2).
Table 2. Characteristics of Patients and Isolates
Note. ESBL-Ec, extended-spectrum β-lactamase phenotype E. coli; MRSA, methicillin-resistant S. aureus; vanA VRE, vanA-producing vancomycin-resistant E. faecium; ESBL-Kp, extended-spectrum β-lactamase phenotype K. pneumoniae; CRPa, carbapenem-resistant P. aeruginosa (also resistant to piperacillin-tazobactam and ceftazidime); CRAb, carbapenem-resistant A. baumannii; MDRO, multidrug-resistant organism.
a 28 patients had >1 species isolated; hence, percentages add to >100%.
b Infection defined as symptoms of infection and receiving targeted antibiotic treatment for the MDRO.
c Colonization was defined as no symptoms of infection, and the patient did not receive targeted antibiotic treatment for the MDRO.
d Includes any isolation of the same MDRO in the 12 months prior, either from clinical or screening samples.
Fig. 1. Characteristics of isolates, reasons for sample collection and specimen type. (a) Reason for sample collection. (b) Sample type. Note. ESBL-Ec, extended-spectrum β-lactamase phenotype E. coli; MRSA, methicillin-resistant S. aureus; vanA VRE, vanA-producing vancomycin-resistant E. faecium; ESBL-Kp, extended-spectrum β-lactamase phenotype K. pneumoniae. See Supplementary Table S3 (online) for further details.
High rates of MDRO isolation, especially ESBL-Ec and MRSA
To define the incidence of each MDRO, we calculated rates per 100,000 occupied bed days (OBDs). MDRO infections occurred at a rate of 107.1 patients per 100,000 OBDs, whereas the overall MDRO burden (infection and colonization) was 294.5 patients per 100,000 OBDs. Considering infection only (not affected by different screening practices), rates were much higher in patients in high-risk wards or ICU (infection rate, 151.1; total burden, 900.4 patients per 100,000 OBDs) (Fig. 2). ESBL-Ec infections were most common, followed by MRSA and ESBL-Kp. CRPa and CRAb were uncommon in all participating sites.
Fig. 2. Rates of MDRO infection and colonization per 100,000 OBDs. High-risk wards include hematology, oncology, renal ward (including renal transplant), liver transplant ward, and ICU (intensive care unit). Occupied bed day is defined as number of beds occupied by patients at midnight, excluding day cases, mental health and hospital-in-the-home. Note. OBD, occupied bed days; ESBL-Ec, extended-spectrum β-lactamase phenotype E. coli; MRSA, methicillin-resistant S. aureus; vanA VRE, vanA-producing vancomycin-resistant E. faecium; ESBL-Kp, extended-spectrum β-lactamase phenotype K. pneumoniae. See Supplementary Table S5 for more detailed data.
Very few MDROs were isolated from patients without healthcare contact
Of MDROs, 25.7% were isolated from patients with a known history of infection or colonization with the same MDRO in the previous 12 months; 60.7% of these current episodes were thought to represent infection rather than colonization (mostly ESBL-Ec and MRSA, data not shown) (Table 2).
Very few patients (42 patients, 11.7%) had MDROs isolated without a history of healthcare exposure: admitted from home, no known admissions in previous 12 months, not known to be colonized in last 12 months or unknown colonization status (Fig. 3). Most of these had ESBL-Ec (32 patients, 25.0% were clinical isolates), and 18 of 32 patients had ESBL-Ec isolates within the first 2 days of admission. In contrast, only 4 patients with MRSA (4.6% of MRSA isolates), and 6 patients with vanA VRE (10.0% of VRE isolates) had a similar lack of healthcare exposure. Further data regarding wards and medical units where MDROs were isolated are detailed in Supplementary Table S4 (online).
Fig. 3. Patient admission source and history. (a) Admission source (ie, where patient was admitted from). (b) Admission history in previous 12 months. Note. ESBL-Ec, extended-spectrum β-lactamase phenotype E. coli; MRSA, methicillin-resistant S. aureus; vanA VRE, vanA-producing vancomycin-resistant E. faecium; ESBL-Kp, extended-spectrum β-lactamase phenotype K. pneumoniae. See Supplementary Table S3 for further details.
High rates of transmission detected for some MDROs
To investigate potential transmission events for major MDROs in this study, genomic comparisons were performed, and potential genomic links to other study patients (pairwise SNPs at or below transmission screening threshold) were determined. Overall, 113 of 358 patients (31.6%) had potential genomic links to other study patients: 95.0% of vanA VRE, 23.3% of ESBL-Kp, 20.2% of ESBL-Ec, and 11.6% of MRSA. Of these potential genomic links, 78 of 113 patients (69.0% under genomic link screening threshold) had probable transmission confirmed by epidemiology, and a further 19 patients had possible transmission by epidemiology (Table 3 and Fig. 4). Conversely, in the absence of genomic links (ie, considering matches to patients with different species, different ST, or same species and ST but above the screening threshold), only 13.9% of isolate pairs were classified as probable or possible transmission by epidemiology alone, compared with 34.6% of isolate pairs with genomic links (χ2, P < .0001). Bed movement data were only available for patients with at least 1 isolate with genomic links, which likely led to overestimation (Supplementary Fig. S1 online). Notably, 18 of 113 patients (15.9%) with genomic links were only identified from the point-prevalence survey at network A.
Table 3. Likelihood of MDRO Transmission by Epidemiology by Speciesa
Note. ESBL-Ec, extended-spectrum β-lactamase phenotype E. coli; MRSA, methicillin-resistant S. aureus; vanA VRE, vanA-producing vancomycin-resistant E. faecium; ESBL-Kp, extended-spectrum β-lactamase phenotype K. pneumoniae; MDRO, multidrug-resistant organism; OBDs, occupied bed days, ie number of patients admitted overnight (excluding mental health and hospital-in-the-home services); ST, sequence type; SNP, single nucleotide polymorphism.
a Definitions of likelihood of transmission by epidemiology: Probable, patients admitted to same ward at the same time; possible, patients admitted to same hospital at same time, or same ward within 60 days (but without overlapping stays); unlikely, all other patients outside these definitions; same patient, isolates from same patient at different times. Modified after Voor In’T Holt et al.Reference Voor In ‘t Holt, Wattel and Boers27
b Some patients represented under >1 species; hence, totals may add to more than the overall number of patients.
c Potential genomic links: isolates analyzed for core genome SNPs by species and ST; isolate pairs with SNP distances below the transmission screening threshold [≤15 SNPs (MRSA) or ≤25 SNPs (other species), excluding same patient pairs] were designated as ‘potential genomic links’ for further epidemiologic investigation.
d No. of patients with both genomic and epidemiologic links to other patients in the study. eFor some patient pairs, admissions overlapped in multiple wards.
f High-risk wards, includes hematology, oncology, renal ward (including renal transplant), and liver transplant wards.
g Subacute care includes aged care, rehabilitation, palliative care, and spinal wards.
Fig. 4. Transmission analysis results. (a) Flow chart describing transmission analysis. (b) Percentage of isolates in each transmission category by species. aGenomic links denotes an isolate that is genomically closely related to another isolate in the study, defined as below the screening threshold for pairwise single nucleotide polymorphisms (SNPs) on core genome alignment (≤15 SNPs for MRSA, ≤25 SNPs for other species). No genomic links denotes an isolate that is not genomically closely related to any other isolate in the study. Transmission categories refer to categorization of patient pairs by epidemiologic data. If pairwise SNP distance was under the screening threshold; hospital admission, ward and bed data were collected for 12 months prior to the patient’s earliest study isolate. ‘Probable’ patients were admitted to same ward at the same time; ‘possible’ patients were admitted to same hospital at same time, or to the same ward within 60 days but without overlapping stays. ‘Unlikely’ patients were all other patients outside these definitions. ‘Same patient’ isolates were from same patient at different times. ‘Above screening threshold’ indicates pairwise distances between isolates exceeded the transmission screening threshold (>15 SNPs for MRSA, >25 SNPs for other species); therefore, bed movement data were not collected. ‘No transmission analysis’ indicates isolates that did not meet the criteria for transmission analysis (ie, ST only contained a single isolate or only isolated from a single patient). Note. SNP, single-nucleotide polymorphism; ST, sequence type; ESBL-Ec, extended-spectrum β-lactamase phenotype E. coli; MRSA, methicillin-resistant S. aureus; vanA VRE, vanA-producing vancomycin-resistant E. faecium; ESBL-Kp, extended-spectrum β-lactamase phenotype K. pneumoniae.
Genomic data have not previously been used to define transmission rates in the hospital setting (based on patient throughput), and we have attempted to use these data to estimate transmission rates. The highest rate of transmissions occurred in vanA VRE (36.5 probable transmissions per 100,000 OBDs), followed by ESBL-Ec (15.9 per 100,000 OBDs), ESBL-Kp (5.6 per 100,000 OBDs), and MRSA (4.0 per 100,000 OBDs) (Fig. 4). No transmission was found for Pseudomonas and Acinetobacter isolates or between hospital networks. Probable transmission occurred mostly in intensive care units and acute-care wards (Table 3). There was no clear genomic threshold separating the pairwise SNP distributions for pairs designated as ‘probable,’ ‘possible,’ or ‘unlikely’ transmission (Supplementary Fig. S2 online).
Discussion
In this pilot study of combined prospective genomic and epidemiologic MDRO surveillance, we have collected comprehensive clinical and genomic data on 4 high-prevalence MDROs (ESBL-Ec, ESBL-Kp, MRSA, and vanA VRE) as well as 2 low-prevalence but high-impact MDROs (CRPa and CRAb). In doing so, (1) we have demonstrated the feasibility of this workflow, using a centralized genomic sequencing and analysis laboratory; (2) we have established the local burden of these MDROs; (3) we have identified historical patient factors potentially leading to increased MDRO risk; and (4) we have used genomic data to infer putative MDRO transmission in hospitals, validated by epidemiologic data.
Hospital-based surveillance for common MDROs with mixed community and healthcare-associated acquisition can be difficult to interpret. Here, we have demonstrated that, although most MDROs (57.2%) are isolated in the first 2 days after hospital admission, very few patients (11.7%) had an MDRO isolated without a history of healthcare exposure. This finding may suggest either that there is more transmission of these MDROs in hospitals than is currently recognized (particularly for MRSA) or that there is a co-occurrence of risk factors between patients who are likely to acquire MDROs and are also likely to be frequent consumers of healthcare. It is not possible to resolve this important question without using the high-resolution typing capabilities of genomics to investigate potential MDRO transmission on a patient-to-patient scale.
Here, we have used genomics to demonstrate that 27% of patients are likely to have acquired their MDRO in hospital, including 88% of vanA VRE. Importantly, although a lower percentage of ESBL-Ec and MRSA were likely to be acquired in hospital (14.8% and 8.1%, respectively), the high prevalence of these MDROs in our population means that significant numbers of patients seem to have acquired these MDROs in hospital, representing a preventable complication of their admission, with associated potential morbidity and mortality. Critically, almost all of these transmissions (apart from vanA VRE) would not have been detected without prospective genomic surveillance (and supported by epidemiologic investigation), underlining the potential power of applying this new technology to infection control practice. Transmission data from prospective genomic surveillance could potentially be used to approximate the rate of transmission based on patient throughput, which could be used to compare the performance of hospitals over time (eg, after changes in infection control practices) or to allow comparisons between hospitals matched for size, case mix and, most critically, screening practices.
Additionally, genomic analysis also uncovered unexpected antimicrobial resistance genes mcr-1 (encoding colistin resistance, a reserve antibiotic not routinely tested in diagnostic laboratories) and rmtB (plasmid-borne AMR gene encoding broad-spectrum aminoglycoside resistance). Although the results of our analyses were not routinely made available during this study, these results (which would not otherwise have been detected) were made available to sites for infection control purposes, prompting additional infection control measures for affected patients, and they could potentially have modified antibiotic choices for these patients.
Our study has several limitations. MDRO screening practices varied between sites, and there may have been differences in the collection methods of clinical isolates and microbiology work-ups among different hospitals. We recognize the potential bias in recall and recording of hospitalization history in last 12 months (from patient recall and medical history review; no centralized database available) and the absence of reliable data regarding patient overseas travel. The paucity of screening for MRSA colonization almost certainly means that the vast majority of MRSA transmission would not have been detected using our study methods. Similarly, our transmission analyses may be limited by only being able to collect epidemiologic data (admission history, bed movements) for patients with isolates below a screening threshold for genomic relatedness. This method was chosen due to limited resources (ie, ward data were collected manually). SNP thresholds were selected based on best available data in the literature at the time, as well as local experience; ultimately, epidemiologic data would be collected for all patients, rather than only those below the screening threshold. Nonetheless, we demonstrated that, in the absence of genomic links, there were significantly fewer patients with epidemiologic links (overlapping admissions/wards) compared to those with genomic links. This finding is still likely to be an overestimate because bed movement data were only collected in patients with established genomic links. Hence, other factors may have increased their likelihood to transmit or become colonized with MDROs.
The bioinformatic methods used for transmission analysis are constantly evolving and have not yet been standardized (ie, multiple methods are currently being used internationally). In this study, we have attempted to apply a pragmatic approach to transmission analysis to allow implementation as part of a prospective genomics surveillance program. As a starting point, we applied a single SNP threshold across each species (≤15 SNPs for MRSA, ≤25 SNPs for other species). We observed that all STs in a species are not equal. Ideally, SNP thresholds should be tailored to each ST and local experience, otherwise they may result in both missed and misattributed transmissions. Further exploration of MDRO and ST-specific SNP thresholds should be conducted in future implementation studies. Additionally, our approach was also limited in that it only allowed us to detect clonal transmission and not plasmid-associated transmission, due to the nature of short-read sequencing technology.
Despite these limitations, our results demonstrate the value of comprehensive genomic and epidemiologic surveillance for MDROs. We also illustrate the potential for genomics to inform hospital infection control, if applied in a timely manner. We plan to explore this methodology further in a larger-scale translational study, using prospective genomics to detect transmission of hospital MDROs, to inform infection control interventions. Importantly, we need to be able to measure the potential benefits of genomics against the cost, to fully evaluate its utility in this setting.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2020.1253
Acknowledgments
We gratefully acknowledge the assistance of Courtney Lane (statistics), John Greenough, Jennifer Breen, and Penny Birchmore (infection control queries and patient movement data), and Carol Wedge (data entry).
Financial support
This work was supported by the Melbourne Genomics Health Alliance, funded by the State Government of Victoria, Department of Health and Human Services, and the 10 member organizations, and by individual grants from National Health and Medical Research Council of Australia (grant no. GNT1093468 to N.L.S.; grant no. GNT1008549 to J.C.K.; grant no. GNT1105905 to B.P.H.; and grant no. GNT1116576 to M.A.S.).
Conflicts of interest
All authors report no conflicts of interest relevant to this article.
