Understanding the effectiveness of infection control methods in reducing and preventing SARS-CoV-2 transmission in healthcare settings is of high importance. We sequenced SARS-CoV-2 genomes for patients and healthcare workers (HCWs) across multiple geographically distinct UK hospitals, obtaining 173 high-quality SARS-CoV-2 genomes. We integrated patient movement and staff location data into the analysis of viral genome data to understand spatial and temporal dynamics of SARS-CoV-2 transmission. We identified eight patient contact clusters (PCC) with significantly increased similarity in genomic variants compared to non-clustered samples. Incorporation of HCW location further increased the number of individuals within PCCs and identified additional links in SARS-CoV-2 transmission pathways. Patients within PCCs carried viruses more genetically identical to HCWs in the same ward location. SARS-CoV-2 genome sequencing integrated with patient and HCW movement data increases identification of outbreak clusters. This dynamic approach can support infection control management strategies within the healthcare setting.

Article Figures and data Abstract Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Understanding the effectiveness of infection control methods in reducing and preventing SARS-CoV-2 transmission in healthcare settings is of high importance. We sequenced SARS-CoV-2 genomes for patients and healthcare workers (HCWs) across multiple geographically distinct UK hospitals, obtaining 173 high-quality SARS-CoV-2 genomes. We integrated patient movement and staff location data into the analysis of viral genome data to understand spatial and temporal dynamics of SARS-CoV-2 transmission. We identified eight patient contact clusters (PCC) with significantly increased similarity in genomic variants compared to non-clustered samples. Incorporation of HCW location further increased the number of individuals within PCCs and identified additional links in SARS-CoV-2 transmission pathways. Patients within PCCs carried viruses more genetically identical to HCWs in the same ward location. SARS-CoV-2 genome sequencing integrated with patient and HCW movement data increases identification of outbreak clusters. This dynamic approach can support infection control management strategies within the healthcare setting. Introduction The severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) pandemic continues to place a significant burden on healthcare services worldwide (Miller et al., 2020; Propper et al., 2020; Maringe et al., 2020). Reducing the spread and outbreak of SARS-CoV-2 infections is particularly important in hospitals and care homes where individuals at high risk of developing severe responses to infection are vulnerable to transmission due to close and regular contact between patients and healthcare workers (HCWs) (Clark et al., 2020; Nguyen et al., 2020; Rivett et al., 2020). Whilst the recent development of promising SARS-CoV-2 vaccines may reduce risk to individuals in hospitals who receive the vaccine (Walsh et al., 2020; Anderson et al., 2020; Keech et al., 2020; Lodge, 2020), risk of infection will not be completely mitigated. Standard methods of infection control will continue to be required to ensure patient and HCW safety. Regular and iterative testing for SARS-CoV-2, in both patients and HCWs, underpins approaches that quantify and control nosocomial transmission (Black et al., 2020) but will not provide insights into how the virus may have spread throughout an institution, alone. Since the degree to which different groups (patients, HCW) propagate SARS-CoV-2 transmission remains uncertain, the utility of screening approaches to prospectively prevent nosocomial spread is difficult to evaluate. It is well established that HCWs are an important component in pathogen outbreak investigations (Peacock et al., 2018; Wenger et al., 1995; de Swart et al., 2000). Here, we have integrated viral genome sequencing with patient admissions records and staff workplace information to investigate SARS-CoV-2 nosocomial outbreaks. We demonstrate that such an approach can be used to identify highly likely nosocomial transmission events of SARS-CoV-2 between HCWs and the patients in their care. Results Sample demographics We generated high-quality sequencing datasets for 173 samples that had been collected from inpatient wards, accident and emergency departments (A and E) and from HCWs across geographically distinct hospitals. All samples were collected between calendar week 11 (commencing 8 March 2020) and calendar week 23 (ending 6 June 2020). Of the 173 high-quality sequenced samples, 39 (23%) were HCW samples. The remaining 134 samples were collected from patients admitted to a total of 31 wards and units situated across the five hospitals. Forty-four (25%) of the 173 high-quality samples were from three hospital locations, which had seen sudden rises in SARS-CoV-2-positive cases; an additional 35 samples from these locations failed to meet our quality criteria for sequencing quality (Figure 1—figure supplement 1). Forty-seven (35%) of 134 patient samples were from A and E departments. The median age of the 134 patients with sequenced samples was 81 years (mean = 75 years; range = 6 weeks–100 years). Sequencing metrics Comparison of SARS-CoV-2 amplicon yield versus target coverage found that samples consistently met 10× minimum coverage thresholds when a total yield of >400 ng was obtained (median = 846; range = 86–3667; Figure 1—figure supplement 1). Where RT-qPCR cycle threshold (Ct) values were provided, we found samples with Ct values of 31 or more resulted in lower amplicon yield and frequently failed to meet sequencing coverage thresholds (Figure 1—figure supplement 1). In the 173 high-quality samples, we identified 268 genomic variants in comparison to MN908947.3, of which 86 were recurrent variants across samples (range = 2–126 samples) and 182 were unique to single samples. Global lineage assignment We utilised Pangolin for placement of the 173 high-quality viral genome sequences within the global SARS-CoV-2 phylogenetic tree. Eight-seven percent (151/173) of sequenced genomes were confidently assigned to an existing lineage (SH-alrt > 80%, UFbootstrap > 90%). We identified 11 distinct lineages in our cohort, with a bias towards viral genomes assigned to lineage B.1.1 (71%, 122/173; Figure 1—figure supplement 2). There were no samples assigned to lineage A. Incorporating the calendar week of sample collection into this analysis suggested a constant relative frequency of viral lineage B.1.1 over time (Figure 1—figure supplement 2). Viral lineage B.1.1 was present in samples collected from 29 different wards, including A and E, and in HCWs, and was present in our cohort between calendar weeks 12–23 (Figure 1—figure supplement 2). Local phylogenetic networks implicate nosocomial transmission within hospital wards To understand how the viral genomes collected from different areas across hospital sites related to one another we created a local phylogenetic tree rooted to a SARS-CoV-2 genome originally sequenced in Wuhan, China (MN908947.3) (Minh et al., 2020; Wu et al., 2020). We overlaid locational origins of the samples (i.e. ward and units from which samples were collected) onto the phylogenetic tree (Figure 1—figure supplement 3). We observed clusters within the phylogenetic tree that were formed predominantly from viral genome sequences taken in individual wards (Figure 1—figure supplement 3). For example, 19/31 of the samples in one of these identified clusters were collected from a single hospital ward (H2_W7, Figure 1—figure supplement 3). The phylogenetic clusters were supported by maximum-likelihood and consensus (10,000 ultrafast bootstraps) approaches for tree creation. Incorporating staff and patient movement enhances identification of nosocomial transmissions Patient contact clusters Utilising patient admissions data over the period of the pandemic, we first created a network of potential direct or indirect patient–patient contacts, inferred through the presence of two individuals on the same ward on the same calendar day. We adopted national guidelines for definition of nosocomial infection to identify contacts between individuals relative to the date of sample collection for a positive SARS-CoV-2 test. Using a contact window of 3–7 days prior to a positive SARS-CoV-2 test (termed herein as likely period of infection), we developed a network of patient–patient contacts (Figure 1). We identified eight significant clusters of individuals (patient contact clusters [PCCs]), defined by multiple potential contacts between two or more individuals within the likely period of infection for each individual (A–H, Figure 1). We assessed the pairwise similarity of viral genomes in identified clusters, demonstrating significantly higher viral genetic similarity within clusters compared to non-clustered samples (Figure 2, p<0.001). This trend was further supported by overlaying the PCCs onto the local phylogenetic tree (Figure 1—figure supplement 4). We identified areas of the phylogeny where there was a 20-fold increase, than expected by chance, in potential contacts between a patient and six or more patients with the most closest genetically related viral genome samples. Consecutive windows of increased (20-fold) patient–patient contacts were merged to identify seven distinct clusters of individuals defined by high level of genetic relatedness of the viral genome sequence and a high degree of potential patient–patient contacts during the likely period of infection (Figure 1—figure supplement 4). Figure 1 with 4 supplements see all Download asset Open asset Incorporating healthcare worker (HCW) and patient admissions data into the analysis of viral genetic relatedness improves certainty of nosocomial outbreaks. (a) Network of direct and indirect potential patient–patient contacts within the window of likely infection (3–7 days prior to positive SARS-CoV-2 test) defines eight significant patient contact clusters (PCCs, overlaid boxes); (b) network including HCW interactions one week prior to positive SARS-CoV-2 test and patient infection classification. Nodes represent individual patients or HCWs, with ordinal numbers representing their position in the constructed local phylogenetic tree. Edges indicate presence on the same hospital ward on the same calendar day. Inclusion of HCWs brings together originally disparate PCCs (b) and (c) increases the number of individuals within viral clusters (VCs) – defined as clusters of identical viral samples or derived viral samples which differ by a single genomic variant. We identified 44 individuals within VCs in the newly defined HCW contact clusters (HCW_A, HCW_B, HCW_C), 21 of whom were not identified within VCs using PCCs alone. The shape of symbols within the enlarged boxes displays the classification of SARS-CoV-2 infection in patients: community, community-acquired infection (positive test within 2 days of hospital admission); possible, possible hospital-acquired infection (positive test 3–7 days after hospital admission); probable, probable hospital-acquired infection (positive test 8–14 days after hospital admission); definite, definite hospital-acquired infection (positive test >14 days after hospital admission). The presence of several patients with definite and probable hospital-acquired infections within the PCC and HCW interaction clusters further reinforces the risk of SARS-CoV-2 transmission events between patients and HCWs on the same hospital wards. Figure 2 Download asset Open asset SARS-CoV-2 viral genomes are more similar in groups of patients and healthcare workers (HCWs) who have been in contact prior to a positive SARS-CoV-2 test. (a) Pairwise genomic variant similarity comparisons of SARS-CoV-2 genomes by the status within patient contact clusters (PCCs) demonstrates increased genetic similarity when patients have been in direct or indirect contact with one another 3–7 days prior to positive SARS-CoV-2 test. Pairwise comparisons within PCCs (n = 544) were tested against all pairwise comparisons that were not defined within the PCCs (n = 29,212). (b) Pairwise genomic variant similarity comparisons of SARS-CoV-2 genomes by patient–HCW interactions in the week prior to positive SARS-CoV-2 test (n, yes = 98, no = 11,836). (c) Pairwise genomic variant similarity comparisons of SARS-CoV-2 genomes by presence within PCCs including interactions with HCWs. Pairwise comparisons within HCW clusters (n = 846) were tested against pairwise comparisons that were not defined within the PCCs including HCW interactions (n = 28,168). Colours of boxplots reflect the PCCs identified in Figure 1, and asterisks indicate significance level determined through a two-sided Wilcoxon rank-sum test (*<0.05; **<0.01; ***<0.001 after Bonferroni correction for multiple testing). Contact clusters including patients and HCWs Next, we created networks of potential interactions between HCWs and patients (Figure 1). This was inferred through the presence of patients in the direct workplace of staff members for at least one calendar day in the 7 days prior to a patient testing positive for SARS-CoV-2. We identified 49 potential contacts between HCWs and patients, including 10 HCWs and 18 patients. Incorporating this information onto the local phylogeny expanded the density of contacts within hotspots (Figure 1—figure supplement 4) and altered the structure of the PCCs within the likely period of infection (Figure 1). Overall, we identified significantly increased genetic similarity in viral samples between patients and HCWs in the same ward locations in comparison to patients and HCWs from different wards (Figure 2, p<0.001). Moreover, we observed greater genetic similarity within each of the PCCs with HCW interactions incorporated (Figure 2, p<0.001). For example, HCW_A illustrates that the addition of HCWs created a previously hidden link between PCC_A and PCC_C (Figure 1); these newly identified connections increased the number of individuals within viral clusters (VCs), defined as clusters of identical viral samples or viral samples that differed by just a single genomic variant (Figure 1). Within the newly defined contact clusters including HCWs and patients, we identified six genetically identical VCs including 18 individuals, and 24 individuals with viral sequences differing by a single genomic variant. The number of individuals in VCs was expanded by including patient–HCW contact networks, over and above those identified using patients alone (Figure 1). Temporal patterns within identified patient and HCW contact clusters In order to establish the most likely SARS-CoV-2 transmission pathways, we examined temporal patterns within proposed nosocomial outbreaks. We created a median joining network for each of the patient and HCW contact networks and incorporated time of sample collection. As multiple entry points into the outbreaks may complicate inference, we cleaned the data to create median joining networks for the 10 most genetically related viral samples within each cluster (inferred from position in local phylogeny, Figure 1—figure supplement 3). This identified trends in the datasets that further reinforced the likelihood of nosocomial infection (Figure 3). We observed that all identified contact clusters could be rooted back to potential 'founder' viral samples that occurred early during the suspected outbreak. For example, in PCC_B and PCC_D (Figure 3), which occurred on the same hospital ward at different time points, the original 'founder' viral samples contained the novel (at time of sample collection) genomic variant MN908947.3–3228-T-G. Eighteen (95%) of 19 viral samples subsequently collected from this hospital ward also contained the same novel variant, at least 16 of these cases were probable or definite hospital-acquired infections. Reducing the quality threshold for sequencing datasets (≥50% coverage at ≥10× coverage) identified the MN908947.3–3228-T-G variant in an additional five samples collected from this hospital location (83%, n = 6 additional samples included with modified criteria). These data highlight that samples excluded from our analyses due to sequencing quality criteria may be missing links within SARS-CoV-2 transmission pathways. Figure 3 Download asset Open asset Temporal patterns in SARS-CoV-2 genomic similarity identify potential viral transmission pathways within patient contact clusters (PCCs) including healthcare worker (HCW) interactions. For each of the highlighted contact clusters, a median joining network is presented with size of nodes representing number of samples and numbers indicating day of nasal or throat swab collection. The presented network suggests a possible path of viral transmission within each contact cluster, hatches represent single genomic variants that differ between viral clusters. The top scatterplot shows that the number of genomic variants identified against the MN908947.3 reference genome increases over time. The bottom scatterplot shows the number of other samples within the contact cluster that are identical or expected to be derived from samples collected at specific calendar days – these are defined as other samples that are identical but with the presence of additional genomic variants. The observed trends show that samples collected early during the suspected outbreaks have a greater number of derived or identical samples than those collected at a later day. These data support that the samples collected early during the highlighted contact clusters are early founder events during a nosocomial outbreak. We observed further trends indicative of nosocomial outbreaks, including the number of genomic variants identified against the MN908947.3 reference genome increasing over time within each of the contact networks, and consistently observed that samples collected early during the suspected outbreaks had a greater number of derived or identical samples within the outbreak than those collected at a later day (Figure 3). Both of these trends would be expected if single ancestral SARS-CoV-2 sequences were original founders of an outbreak. Discussion In this study, we obtained high-quality SARS-CoV-2 genomic sequences for 173 individuals across five hospitals in the North West of the UK, including both patients and HCWs. We incorporated potential contacts between the two groups into a phylogenetic analysis and comparison of pairwise genetic similarity. Overall, this demonstrated that inclusion of contact data increased confidence in the characterisation of nosocomial outbreaks. This work was undertaken across multiple, geographically separate hospitals in the UK with responsibility for the care of large numbers of SARS-CoV-2-positive patients. At the time of analysis, there was no routine SARS-CoV-2 screening of asymptomatic HCWs. Similar to other UK hospitals, the different locations within the hospitals were assigned as either green (SARS-CoV-2-negative) or red (SARS-CoV-2-positive) zones. This strategy, in combination with additional infection control measures such as staff bubbles, is widespread as a method to reduce nosocomial infection. However, as patients tested positive after spending prolonged periods of time in green areas, it became apparent that there were unrecognised transmission events between the two areas or from the community into green zones. Viral genome sequencing offers a realistic possibility to track and identify root causes of nosocomial transmissions (Lucey et al., 2020; Meredith et al., 2020). Here, we applied genome sequencing to throat and nasal swabs obtained from HCWs and patients. We identified 268 unique genomic variants in the 173 high-quality samples and placed our samples within recognised global lineages (Figure 1—figure supplement 2). The predominance of a single SARS-CoV-2 lineage meant that precise differentiation of viral samples from individual hospital wards was not possible at this level (Figure 1—figure supplement 2). We therefore created a local phylogeny for our sequenced genomes rooted to MN908947.3 (Figure 1—figure supplement 3) and calculated pairwise similarity in the genomic variants identified in each of the SARS-CoV-2 genomes. It has been previously noted that the low genetic diversity of SARS-CoV-2 causes complexity in the identification of nosocomial outbreaks, as samples may be genetically identical by chance rather than through transmission between individuals (Meredith et al., 2020; Gudbjartsson et al., 2020). Our data reemphasise this low genetic diversity of SARS-CoV-2, with a median number of 11 (range = 2–16) variants identified per sample, and an average pairwise similarity of 61.5% (Figure 2). Integrating our analyses with patient admissions record (available for 104/134 patients for at least one day prior to SARS-CoV-2-positive test) and location of staff workplaces (available for 31/39 HCWs) identified clusters of individuals who had interacted during their most likely periods of infection. These analyses confirmed that individuals who had been in contact during this period were more likely to have genetically more similar viral samples than individuals who had not been in contact (Figure 2). While we cannot exclude that viral samples are genetically identical or similar by chance due to the low genetic diversity of SARS-CoV-2, the spatial and temporal patterns of viral genetic relatedness that we observe provide strong evidence for nosocomial transmission amongst both patients and HCWs. These trends are observed in at least five distinct clusters, across three geographically distinct hospitals. These are further reinforced by the presence of novel genomic variants (at the time of analysis) transmitting through identified clusters (Figure 1—figure supplement 3). We suggest these data support the widespread adoption of iterative screening strategies for HCWs who may be pre-symptomatic or asymptomatic shedders of SARS-CoV-2 (Black et al., 2020; Arons et al., 2020; Buitrago-Garcia et al., 2020). Pre-symptomatic or asymptomatic individuals have been demonstrated to be important contributors to SARS-CoV-2 outbreaks (Rivett et al., 2020; Kasper et al., 2020; Letizia et al., 2020). Others have shown that comprehensive characterisation of outbreak clusters can be hindered by the existence of hidden links between individuals, even where all individuals within clusters are known and completely isolated from external contacts (Sekizuka et al., 2020). In our data, we observed the presence of cohort-specific genomic variants shared between individuals but without a known connection between the sampled individuals. Here, it is likely that missing individuals or connections between surveyed individuals are adding complexity to our analyses. While the use of digital contact tracing is difficult in hospital environments, track and tracing smartphone software could be useful to extend the characterisation of contacts between individuals and to understand the accuracy of the assumptions enforced in this study (Firth et al., 2020; Ferretti et al., 2020). Here, we infer contacts between individuals through their presence on the same hospital ward on the same calendar day. This approach identifies individuals within our cohort that are likely to have been in face-to-face contact. We note that this approach has imperfect assumptions but is likely to dilute rather than inflate the statistical significance of the investigations reported in this study (Figures 1–3). Future work may enable additional co-factors to be considered in models for network creation such as infection control measures in place on hospital wards (e.g. personal protective equipment utilised), symptomatic status of individuals, and the length, type, and the proximity of physical contacts between individuals. These approaches are supported by recent data demonstrating that over half of SARS-CoV-2 transmissions occur when individuals are pre-symptomatic and that transmission likelihood increases with the duration and proximity of contact (Sun et al., 2021). Collecting data to incorporate these factors into network models in the healthcare setting may enable the generation of more precise binary contact clusters according to specified parameters or the development of weighted networks biased by the relative importance placed on co-factors (Firth et al., 2020). Understanding the concordance of empirical datasets of SARS-CoV-2 transmission, as reported here, and computational models of transmission is an important avenue for future work to identify the most influential factors to decrease the likelihood of SARS-CoV-2 transmission in both healthcare and community settings. Our data demonstrate that SARS-CoV-2 genome sequencing alongside patient admission and staff workplace information can identify transmission events within the healthcare setting. Looking forward, we expect that the adoption of genomic approaches in real time, for example within 48 hr, alongside consideration of patient movement datasets will enable rapid identification of linked hospital-acquired SARS-CoV-2 infections. Such approaches could optimise infection control management strategies, lead to targeted interventions, reduce nosocomial transmission, and ultimately prevent avoidable harm to vulnerable individuals who acquire COVID-19 whilst in the healthcare setting. Materials and methods Sample selection Request a detailed protocol Throat and nasal swab samples were collected from patients and healthcare professionals based at MFT hospital sites. Diagnostic SARS-CoV-2 RT-qPCR assays were performed by the Clinical Virology Department of the Manchester Medical Microbiology Partnership (MMMP; Manchester, UK). RT-qPCR-positive samples were selected for SARS-CoV-2 whole-genome analyses at the Manchester Centre for Genomic Medicine (MCGM; Manchester, UK). We attempted to sequence all available SARS-CoV-2-positive samples from hospital wards highlighted by infection control surveillance officers as potential outbreaks within our sample collection period due to sudden rises in positive cases. Demographic, hospital location, and laboratory data were included with each referral. All ward names have been anonymised for publication. Sample and NGS library preparation Request a detailed protocol Nucleic acid re-extraction was performed using the chemagic Viral DNA/RNA 300 Kit on the chemagic 360 instrument (PerkinElmer Inc, Waltham, MA). All extracted RNA samples underwent cDNA synthesis using either LunaScript RT SuperMix kit (New England Biolabs, Ipswich, MA) or SuperScriptIV (Thermo Fisher Scientific, Waltham, MA), in accordance with manufacturers protocols. SARS-CoV-2 whole-genome libraries were prepared using SureSelectXT Low Input kit CoVHuman6X enrichment capture-based method (Agilent Technologies, Santa Clara, CA) or the ARTIC tiled amplicon multiplex PCR protocol (version three primer set) with NEBNext Ultra II DNA Library Prep Kit (New England Biolabs). PCR and library preparation quality validations were obtained using TapeStation D1000 and HSD1000 (Agilent Technologies). Final libraries were sequenced using MinION flow cells version 9.4.1 (Oxford Nanopore Technologies, Oxford, UK) or MiSeq (Illumina, San Diego, CA) using reagent kits for 600 cycles (for tiled PCR SARS-CoV-2 amplification) or 300 cycles (for Agilent SureSelectXT enrichments) for paired end sequencing. Bioinformatics and analysis Request a detailed protocol Sequencing reads were deduplicated on instrument for Illumina MiSeq datasets or using Guppy for Oxford Nanopore datasets. Reads were aligned to the SARS-CoV-2 reference genome (MN908947.3) using BWA-MEM (Li, 2013) for Illumina MiSeq datasets and using Minimap2 (Li, 2018) for Oxford Nanopore MinIon datasets. Reads were filtered and variants identified using iVar v1.2.2 (Grubaugh et al., 2019). Samples with ≥75% of the MN908947.3 reference genome covered by ≥10 high-quality reads with at least 50 aligning nucleotides were included for downstream analysis. Variants with an allele fraction of at least 0.6 in high-quality mapped reads were identified in comparison to MN908947.3, and a consensus FASTA built using iVar. Multi-way alignments were performed using MAFFT v7.407 (Katoh et al., 2002), and maximum-likelihood trees rooted to MN908947.3 using 1000 bootstraps were generated with IQ-TREE v1.6.12 (Minh et al., 2020). Trees were visualised in Geneious Prime software v2020.1.2 (https://www.geneious.com). Pangolin v2.0 (Rambaut et al., 2020) was utilised for positioning of sequences within the global phylogenetic tree (lineages v2020-05-19). Median joining networks were created in Pop-ART (http://popart.otago.ac.nz/). Pairwise similarity analyses were performed using a bespoke script and calculated the number of exact matches in genomic variants between samples after adjusting for regions masked by low coverage. All high-quality genome sequences were shared with COG-UK (COVID-19 Genomics UK (COG-UK), 2020). Patient admissions and movement Request a detailed protocol We collected hospital admission data for all patients with high-quality sequenced genomes. For each patient, we identified other individuals within the cohort who were present on the same hospital wards on the same calendar day, leading to potential indirect or direct contacts between patients. This method for defining contacts assumes that close face-to-face contact is the most likely method for SARS-CoV-2 transmission between individuals and aims to identify individuals within our cohort who are most likely to have had such interactions. This assumption is supported through recent meta-analyses concluding that physical distancing of less than 1 m increases likelihood of SARS-CoV-2 transmission between individuals (Chu et al., 2020). We assessed all potential contacts for each patient in relation to the calendar day that the positive SARS-CoV-2 nasal or throat sample was collected from the patient. The windows of contacts are defined in accordance with national guidelines for SARS-CoV-2 nosocomial outbreak: community-acquired infection (positive test within 2 days of hospital admission); possible hospital-acquired infection (positive test 3–7 days after hospital admission); probable hospital-acquired infection (positive test 8–14 days after hospital admission); and definite hospital-acquired infection (positive test >14 days after hospital admission). For patient–patient contacts, we identified any potential contacts within the possible hospital-acquired infection period (3–7 days) and developed a binary matrix. We made additional assumptions for including HCWs in the contact networks, s

Introduction to key concepts: education policy, economic necessity and public service reform Class, comprehensives and continuities: a short history of English education policy Current policy models and The UK government's approach to public service reform Current key issues: forms of policy and forms of equity A sociology of education policy: past, present and future.

Background: Understanding the effectiveness of infection control methods in reducing and preventing SARS-CoV-2 transmission in healthcare settings is of high importance. Infection control is challenging in these environments due to regular contact between healthcare workers (HCWs) and patients. This is amplified by increased frequency of severe adverse responses to SARS-CoV-2 in patients with underlying health conditions. Methods: We sequenced SARS-CoV-2 genomes for patients and HCWs across multiple geographically distinct UK hospitals. All hospitals were actively enforcing zoning approaches (SARS-CoV-2 negative and SARS-CoV-2 positive areas) as an infection control measure. We integrated patient movement and staff location data into the analysis of viral genome data in order to understand geographical and temporal dynamics of SARS-CoV-2 transmission. Findings: We obtained 173 high-quality SARS-CoV-2 genomes from patients ( n =134) and HCWs ( n =39). The median number of genomic variants per sample of 11 (range=0-16), with a 61.5% average pairwise similarity in the variants (range=0-100%). Integration of patient movement identified eight patient contact clusters (PCC) with significantly increased similarity in genomic variants compared to non-clustered samples ( p <0.001). Incorporation of HCW location further increased the number of individuals within PCCs. Patients within PCCs carried viruses more genetically identical to HCWs in the same ward location ( p <0.001). Interpretation: SARS-CoV-2 genome sequencing integrated with patient and HCW movement data increases identification of outbreak clusters and improved understanding of the role of patient-HCW interactions. This dynamic approach to SARS-CoV-2 outbreak monitoring in a healthcare setting is able to support infection control management strategies within the healthcare setting. Funding: JME is funded by a postdoctoral research fellowship from Health Education England. WGN is supported by the Manchester NIHR BRC (IS-BRC-1215-20007).Declaration of Interests: The authors declare no conflicts of interest.Ethics Approval Statement: The study was conducted to investigate hospital outbreak investigation/surveillance. Ethical approval was obtained from the Manchester Biomedical Research Centre COVID-19 rapid response group for viral genome analysis.

Objective: This study investigates the prognostic significance of pre-operative symptom status and type of symptom in outcomes after carotid endarterectomy (CEA). Methods: This review was conducted and reported in accordance with the Preferred Reporting Items for Systematic reviews and Meta-analysis (PRISMA) to identify studies reporting peri-operative outcomes of CEA in symptomatic and asymptomatic patients. The last search was conducted in August 2019 and a methodological assessment was performed using the Newcastle Ottawa Scale. A meta-analysis of outcome data using the odds ratio (OR) as the summary statistic was conducted, and the precision of the effect was reported as 95% confidence interval (CI). Fixed effect or random effects models were used to calculate the pooled estimates. Results: Eighteen studies reporting a total of 91 895 patients were included in the meta-analysis. Asymptomatic patients had a lower peri-operative risk of stroke (OR 0.5, 95% CI 0.45-0.54; p < .001) and death (OR 0.66, 95% CI 0.57-0.77; p < .001) than symptomatic patients, but the risk of myocardial infarction was not significantly different (OR 0.98, 95% CI 0.84-1.15; p = .82). Those suffering a pre-procedural stroke had an increased peri-operative risk of stroke and death vs. patients suffering a pre-procedural transient ischaemic attack or amaurosis fugax. Conclusion: Patients undergoing CEA after a stroke have worse peri-operative outcomes in terms of stroke and death. Further research needs to be performed to ascertain the value of this finding in risk stratification systems and to investigate potential aetiological associations between pre-operative symptom status and peri-operative risk following a CEA.

The ethnographic paradigm of classroom interaction research is now a well-established element of the sociology of education in Britain. There are now several collections of papers which represent the development and current state of work in this area (Chanan and Delamont, 1975; Stubbs and Delamont, 1976; Woods and Hammersley, 1977) as well as Delamont’s (1976) exemplary introductory text. Most of the work, from the ethnographic paradigm, included in these various contributions to the field is founded to a greater or lesser extent upon a theoretical perspective derived from symbolic interactionism, although phenomenological and ethnomethodological perspective have also made themselves felt (e.g. Torode, 1976, 1977; Cicourel, et al., 1974; Payne, 1976). However, despite the growing body of empirical work on classroom interaction and the concomitantly increasing amount of theoretical commentary, surprisingly little attention has yet been given to the evolutionary and developmental nature of teacher-pupil relationships in the classroom setting. The tendency has been (with one or two exceptions in the American literature) to treat and portray classroom relationships as fixed and static patterns of interaction within which teachers select strategies or act out the constitutive rules or procedures which serve to structure this interaction. Little attention has been given to the ways in which strategies are tested or rules established and in my view this has tended to inhibit the development of a coherent formal theory of classroom interaction. In part, I want to argue, this state of affairs is an artifact of the nature of classroom interaction research itself and the constraints upon it. The problem is that most researchers, with limited time and money available to them, are forced to organise their classroom observation into short periods of time. This usually involves moving into already established classroom situations where teachers and pupils have considerably greater experience of their interactional encounters than does the observer. Even where the researcher is available to monitor the initial encounters between a teacher and pupils, the teacher is, not unreasonably, reluctant to be observed at this stage.

Introduction - Carotid Plaque Volume (CPV) correlates more closely than severity of stenosis with symptoms of cerebral ischemia in patients with carotid disease [1]. If we could measure CPV by a minimally-invasive technique, it may replace severity of stenosis as the principle indication for carotid endarterectomy (CEA) and may even be used for population screening in the future. Methods - Standard Duplex and 3D tomographic ultrasound (tUS) imaging of the carotid bifurcations were undertaken on the day of CEA in 50 patients. CPV of the endarterectomy specimen was measured using a validated modified Archimedes suspension technique. CPV by tUS was calculated by dedicated software using the intima-plaque and plaque-blood boundaries in 1mm slices through the tUS image, corrected for the plaque length. Results - The mean endarterectomy specimen CPV was 0.92±0.51cm3 and the tUS CPV was 1.08±0.61cm3 with the mean difference between the endarterectomy specimen and tUS CPV's being only -0.16±0.24cm3 (95% LOA -0.63-0.32). There was an excellent correlation (Figure) between CPV measured by tUS and the endarterectomy specimen with r = 0.92 (95%CI 0.87 – 0.96 cm3) p<0.0001. There was no correlation between CPV and the severity of stenosis measured by peak systolic velocity; (r = 0.0052, p=0.97). Conclusion - tUS measurement of CPV strongly correlated with CPV of the endarterectomy specimen and is an accurate technique for calculating atherosclerotic burden or CPV. This technique may lead to a new indication for CEA and possibly even to population screening for carotid disease associated with enhanced stroke risk. References1.Ball, S., et al., Carotid plaque volume in patients undergoing carotid endarterectomy. Br J Surg, 2018. 105(3): p. 262-269.

Fault-related structures in onshore and nearshore basins often show signs of regional subaerial erosion. In seismically imaged growth strata, syn-deformational erosion is evidenced by angular unconformities within hangingwall strata and missing section within footwall strata. Erosion complicates correlation across faults and has significant implications for burial history. This study presents a new approach to structural forward modeling that parameterizes model surfaces by age as well as depth. By including surface age, we can define complex footwall and hangingwall burial histories that include periods of erosion. The modeled fold geometry depends on fault shape, shear angles, and horizon slip, according to established kinematic theories (specifically inclined shear fault bend folding and tri-shear fault propagation folding). Where younger model surfaces intersect older fold surfaces, the younger surfaces erode and truncate the older surfaces. The models are fully interactive, allowing us to continually modify footwall to hangingwall correlations and fault shape until the computed horizon shape and unconformity geometry match the observational data. In this presentation we apply the new modeling technique to seismic examples of extensional and contractional structures with complex burial histories indicated by hangingwall unconformities. The first example is a basin-bounding growth fault within the Bohai Bay, South China Sea where over 5 km of syn-extensional erosion has removed the entire footwall section. By interactively modeling the observed hangingwall angular unconformities, we quantitatively reconstruct both the eroded footwall and burial history for the growth fault. The second two examples are inversion structures from the Subandes in Peru and the Junggar Basin in China. Both inversion structures feature multiple types of angular unconformities that independently reflect periods of extension and contraction. Forward modeling these structures refines the timing and magnitude of each deformational phase as well as providing the burial history. A final example from the Outeniqua Basin in South Africa shows how complex hangingwall unconformities can arise solely from movement along simple faults. For each example, quantitative animations show the sequential development of the structures including periods of burial, erosion, and changes in deformation style.

BACKGROUND:The main indication for carotid endarterectomy (CEA) is severity of carotid artery stenosis, even though most strokes in carotid disease are embolic. The relationship between carotid plaque volume (CPV) and symptoms of cerebral ischaemia, and the measurement of CPV by minimally invasive tomographic ultrasound imaging, were investigated.METHODS:The volume of the endarterectomy specimen was measured using a validated saline suspension technique in patients undergoing CEA. Time from last symptom and severity of stenosis measured by duplex ultrasonography were recorded. Middle cerebral artery emboli were counted using transcranial Doppler imaging (TCD) in a subset of patients.RESULTS:Some 339 patients were included, 270 with symptomatic and 69 with asymptomatic carotid stenosis. Mean(s.d.) CPV was higher in symptomatic than in asymptomatic patients (0·97(0·43) versus 0.74(0·41) cm3 ; P < 0·001). CPV did not correlate with severity of carotid stenosis (P = 0·770). Mean CPV was highest at 1·03(0·46) cm3 in the 4 weeks following cerebral symptoms, declining to 0·78(0·36) cm3 beyond 8 weeks. Among 33 patients who had TCD, mean CPV was 1·00(0·48) cm3 in the 27 patients with ipsilateral cerebral emboli compared with 0·67(0·16) cm3 in those without (P = 0·142). There was excellent correlation between CPV measured by tomographic ultrasound imaging and the endarterectomy specimen in 34 patients (r = 0·93, P < 0·001).CONCLUSION:CPV correlated with symptoms of cerebral ischaemia, but not carotid stenosis. It could be a potential indicator for CEA.

Objective: Idiopathic Parkinson's disease (IPD) is the second most common neurodegenerative disorder, often complicated by dementia. Cardiovascular risk factors and spontaneous cerebral emboli (SCE) are strongly associated with Alzheimer's (AD) and vascular dementia (VaD). We measured SCE in the middle cerebral artery and arterial wall volume in the extracranial arteries in patients with IPD and controls, and explored the relationships with structural and physiological MRI brain neurovascular measures. Patients and Methods: Arterial wall volume over 2cm of the axillary and internal carotid arteries (ICA) bilaterally was measured by 3-D tomographic ultrasound in 15 IPD patients and 16 age/gender matched controls. SCE were counted by Transcranial Doppler (TCD) using international consensus criteria. Venous to arterial circulation shunting (v-aCS), usually through a patent foramen ovale (PFO), was measured using a TCD technique with intravenous microbubble contrast. Structural and physiological MRI brain neurovascular measures, acquired separately, comprised white matter lesion volume (WMLV), cerebral blood flow (CBF) and arterial arrival time (AAT). Results: Mean (95% CI) axillary and ICA wall volume was higher in IPD patients at 523 mm(3) (446, 600) and 455 mm(3) (374, 536) respectively compared with 412 mm(3) (342, 483) and 408 mm(3) (362, 454) in controls being significant for the axillary artery (p = 0.04). Cerebral WMLV was related to mean arterial wall volume for both axillary (r = 0.555, p = 0.009) and ICA (r = 0.559, p = 0.026) in all participants. SCE were detected in four IPD patients and three controls (p = 1.00). Two IPD patients and three controls were positive for a v-aCS equivalent to PFO (p = 0.477). Conclusion: Although frequent in AD and VaD, neither SCE nor v-aCS were associated with IPD. This is the first study to demonstrate arterial wall volume is increased in IPD and relates to WMLV.

AIM: To develop a high-quality national guideline for the assessment and management of CYP presenting with iTPS and/or iCDI before their 19th birthday, as a joint endeavour by the Britsh paediatric endocrine and oncology societies (BSPED/CCLG) and meeting approved commissioning standards (RCPCH/NIHCE). The interdisciplinary guideline development group (GDG) identified 64 clinical questions. These were reviewed by stakeholders and used to direct a systematic literature search (January 1990 - March 2017). 568 articles were appraised using the GRADE system. Where there was sufficient evidence, the GDG made a guideline recommendation. Where high quality evidence was lacking, the GDG drafted recommendations based on their expert opinion and reviewed these using two rounds of Delphi consensus with international experts. In 11 case series (741 paediatric patients) the commonest individual causes of TPS/CDI were Langerhans cell hystiocytosis (16%), germ cell tumours (13%) and craniopharingiomas (12%). A range of congenital defects accounted for 19% of cases. Infectious diseases (2%), trauma (1%) and inflammatory/autoimmune conditions (1%) were rare. Twenty-nine percent remained idiopathic and some causes of TPS in adults (metastatic tumours and neurosarcoidosis) were not reported in children. The definition of TPS was not consistent across studies. A guideline and decision-making flowchart were developed. The likely aetiology of TPS/CDI in children differs from that in adults and justifies the development of age-appropriate management guidelines. This will form the basis of future audits of practice and outcomes and is intended to improve the care and service provision to CYP with apparent idiopathic TPS/CDI.