Laboratory diagnosis

Nasopharyngeal swabs in viral PCR solution (1:1 ratio of EasyMag Nuclisens Extraction Buffer (BioMerieux, France) and EMEM) were extracted and tested according to the availability of assays at the diagnostic laboratory.ⁱ Surplus RNA extract was collected for sequencing with ethical approval from the NHSGGC biorepository (16/WS/0207NHS).

Sequencing

Sequencing was performed on either the ONT MinION/GridION or the Illumina MiSeq platform, as previously described.16 Briefly, libraries were prepared in accordance with the ARTIC network protocols (v1 and v2) (https://artic.network/ncov-2019). For nanopore reads the ARTIC-nCov-2019 bioinformatics protocol was used, reads were size filtered, demultiplexed, trimmed with Porechop (https://github.com/rrwick/Porechop) and mapped against the reference strain Wuhan-Hu-1 (GenBank accession number MN908947), followed by clipping of primer regions. Variants were called using Nanopolish 0.11.3 (https://github.com/jts/nanopolish). For Illumina, reads were trimmed with trim_galore (bioinformatics.babraham.ac.uk/projects/trim_galore/) and mapped with BWA20 to the Wuhan-Hu-1 reference sequence, followed by primer trimming and consensus calling with iVar.21 A read coverage of least 10 was used for the consensus.

Phylogenetic and probabilistic analysis

Retrospective phylogenetic analysis of whole-genome sequences was performed as follows. All full genomes of SARS-CoV-2 from Scotland sequenced as part of the COG-UK consortium were included and aligned using MAFFT v7.310 and with the HKY+I+G4 nucleotide substitution model determined by modeltest. The global lineage and UK lineage assignments for the dialysis samples were determined using civet (https://github.com/artic-network/civet).

We applied a novel algorithm, the Sequence Reporting Tool (SRT) to estimate the probability of healthcare- vs community-acquired infection in each case, based on the statistical approach developed for the COG-UK hospital-onset COVID-19 infection (HOCI) study (https://clinicaltrials.gov/ct2/show/NCT04405934).22 The approach is based on Bayesian principles and involves comparison of the proportion of similar viral sequences (with maximum pairwise SNP difference of two, with no difference where there is an overlap in ambiguous nucleotide codes or an ‘N’ in either sequence) observed within potential locations of infection for the case of interest: i) patients’ RDU and elsewhere in this hospital, ii) inpatient ward and hospital (if the patient was admitted) all within the prior three weeks; along with a weighted proportion of similar sequences in the local community of the patient within the prior six weeks based on the outer postcode of their home address. There are 61 districts based on this outcode (49 in Glasgow and 12 in Lanarkshire). We assumed 0.5 prior probability of infection within the RDU before consideration of sequence data, and the prior probability of admission-related infection among the inpatients was based on the interval from admission to diagnosis and the incubation distribution of COVID-19.23 The SRT algorithm outputs two posterior probabilities in all cases: that of acquiring the virus directly from the RDU (p_RDU) and that of acquiring the infection through use of facilities within the hospital but not in the RDU (e.g. toilets, cafes, lobbies, or shared transport, etc) (p_hRDU). A posterior probability (p_RDU) of 1 indicates that the transmission occurred within the RDU; if the p_RDU stays at 0.5 then it remains unclear where the transmission occurred; if p_RDU is 0 then the infection was most likely community acquired. If the haemodialysis patient was an inpatient and continued to attend the RDU at the time of COVID-19 diagnosis, the SRT algorithm also gives the probability of acquiring the infection from the ward of admission (p_wADM) and that of infection elsewhere in that hospital (p_hADM). The SRT algorithm was coded in R version 3.6.0, using the ape v5.3 package for calculation of pairwise SNP differences and PostcodesioR v0.1.1 and gmt v2.0-1 packages to calculate distances between postcodes.22

Genomic and epidemiological investigation

Residual RNA extract from 53 of 60 patients with SARS-CoV-2 positive samples were obtained for virus genome sequencing. 39 of these sequences plus one from a healthcare worker were of sufficient quality and coverage for further analysis. The samples belonged to 13 different UK lineages (Fig. 2 ). Whilst a number of patients had indistinguishable sequences from the same UK lineage and shared dialysis sessions, some fell outwith phylogenetic lineages, providing evidence of community-transmission. The recent introduction of SARS-CoV-2 into the human population and its relatively low mutation rate, mean sequences in the same UK lineage and phylotype cannot be interpreted as direct transmission events, with further temporal and epidemiological data required to quantify the probability of transmission. Conversely, sequences from different UK lineages would disprove direct transmission. Epidemiological investigation identified clusters of SARS-CoV-2 positive patients with shared dialysis sessions and sometimes transport and this was analysed with the phylogenetic data (Figs. 2 and ​and3 ).3 ). We found five of the six RDUs spanning two health boards in the West of Scotland had evidence of unit-linked transmission events.

Open in a separate window

Fig. 2

Phylogenetic tree showing the relationship of 39 sequences from RDU patients and additional SARS-CoV-2 genomes from Scotland. Sequences are colour-coded by RDU location. Dashed boxes highlight the UK lineage and are shown in more detail in Fig. 3. The numerical suffixes of the CVR identifier indicate the posterior probability (as a percentage) of the patient acquiring SARS-CoV-2 from the RDU (p_RDU) or from the wider hospital where dialysis takes place (p_hRDU). The scale bar indicates substitutions per nucleotide site.

Open in a separate window

Fig. 3

Timeline of detection of first SARS-CoV-2 positive results in haemodialysis patients in RDUs with details of dialysis sessions and shared patient transport in relation to the UK lineage. The phylogenetic trees are derived from the dashed boxes in Fig. 2. Circled numbers in the phylogenetic tree represent the number of indistinguishable sequences from Scotland for the given node on the phylogeny. The numerical suffixes of the CVR identifier indicate the posterior probability (as a percentage) of the patient acquiring SARS-CoV2 from the RDU or from another healthcare-related infection (i.e., hospital where dialysis takes place and ward and/or hospital they have been admitted to), respectively. The scale bar indicates substitutions per nucleotide site.

RDU1

In RDU1, viral sequences from seven haemodialysis patients and one HCW from the same unit clustered within the UK40 lineage (Fig. 3). Five of these sequences were indistinguishable to each other (CVR248, CVR284, CVR495, CVR987 and CVR1404), suggestive of within-unit transmission. Applying the SRT, we found the probability of within-unit transmission in RDU1 was indeterminate based on sequence data alone ranging from 0.53 to 0.68. Further epidemiological analysis suggested transmission in some cases – for example, CVR248 and CVR1404 had indistinguishable sequences and shared dialysis sessions (Fig. 3). However, CVR284 and CVR495, also with indistinguishable sequences, did not overlap with each other or anyone else on the unit, suggesting community-acquisition. HCW, CVR987, having been in direct contact with CVR248, self-isolated on 22nd March, prior to testing positive seven days later. However, this viral sublineage of UK40, was widespread in the community, with 63 other indistinguishable sequences detected within the geographical location of RDU1 and patient communities; so this transmission could not reliably be inferred. Two of the seven patients whose sequences derived from the wider UK40 lineage (CVR780 and CVR1404) (Fig. 3) were linked epidemiologically, sharing both dialysis sessions and transport from home to the unit. However, the estimated probabilities of within unit transmission for CVR780 and CVR1404 were 0.63 and 0.54, respectively. This result was supported by close inspection of the data. CVR780 was found to have tested positive for SARS-CoV-2 six days prior to CVR1404 and had a single nucleotide polymorphism relative to the Wuhan reference not found in CVR1404, making transmission from CVR780 to CVR1404 less likely. The widespread distribution of UK40 lineages in the community, the early preventative measures implemented by RDU1 and the lack of definitive epidemiological evidence for transmission suggest that individual transmissions were not due to infection prevention and control (IPC) challenges in this unit. The final case occurred nine days after “lockdown” and the implementation of enhanced PPE measures.

RDU2

In RDU2, there was evidence of five introductions of SARS-CoV-2 from the community, of which two lineages spread within the unit or on hospital transport to the unit. CVR3289 and CVR3290 had indistinguishable sequences, only seen in two other non-geographically linked community cases. These patients shared the same dialysis session and transport, with an estimated 100% probability of within-unit transmission (Fig. 2, lineage UK658). Likewise, CVR1003 and CVR1924 had indistinguishable sequences and shared the same dialysis sessions (Fig. 3, lineage UK51). CVR2314 had a p_RDU of 1, but this patient had no epidemiological link to other two patients, suggesting nosocomial transmission from fomites or an untested staff member (staff were not routinely tested for SARS-CoV-2, at this time).

RDU3, 4 and 5

In RDU3, three introductions and two separate transmission events were identified. Although CVR375 and CVR1511 had indistinguishable sequences, this was shared with 161 other Scottish samples (Fig. 2, lineage UK5098). The SRT identified nosocomial infection in CVR375 due to within-hospital rather than within-dialysis unit transmission (p_hADM 0.95). CVR1511 acquisition of infection from RDU3 (p_RDU 0.68) was less clear. CVR1817 is a close sequence match to CVR375 and CVR1511 based on the 2 SNP threshold, leading the SRT to estimate probable unit-based transmission (p_RDU = 0.73). However, on phylogenetic analysis CVR1817 falls into the separate UK501 lineage (Fig. 2) and appears likely to have been community-acquired on consideration of all available information. In lineage UK39, CVR937 was community-acquired while the related CVR1816 was probable within-unit transmission (p_RDU = 0.74).

In RDU4, two patients tested positive for SARS-CoV-2 but there was no linkage found on sequence analysis (Supplementary Table 1) with no evidence of within unit transmission of SARS-CoV-2 in this dialysis unit.

Within RDU5, three cases of SARS-CoV-2 infection were detected. CVR1217 (lineage UK5098) was community-acquired while CVR1843 (p_RDU of 0.9) was highly suggestive of within RDU transmission. These related sequences differed by 2 SNPs. This strongly suggests that intermediary modes of transmission should ideally have been investigated, including untested asymptomatic individuals, for example members of staff.

RDU6

SARS-CoV-2 was introduced to RDU6 on at least 5 occasions with evidence of onward transmission in two cases and hospital-transmission in one. Of the 26 (12•3% of total) SARS-CoV-2 positive patients, 16 were sequenced. Nine of these patients were within the UK429 lineage (Fig. 3). The SRT verified a high likelihood of within-unit transmission, with p_RDU ranging from 0•96 to 1 (Fig. 3). CVR3373 (p_RDU = 0, the first of this phylotype found in RDU6) and CVR3362 were within a separate lineage, UK5098 (Fig. 3). CVR3362 had been hospitalised for a month prior to the positive SARS-CoV-2 test, whilst maintaining dialysis within RDU6 and had a p_RDU of 1. Further discussion with the infection control team confirmed that CVR3379 (non-haemodialysis patient) and another unsequenced SARS-CoV-2 positive haemodialysis patient, shared the same hospital bay with CVR3362. It is possible that SARS-CoV-2 was brought onto this bay by the other haemodialysis patient from their dialysis sessions on RDU6. This highlights one of the limitations of phylogeny and the SRT algorithm in an outbreak investigation – sequencing data needs to be representative of prevalent cases for the results to be interpreted. Finally, CVR3732 (Fig. 2, UK370) had a high p_hRDU (0•96) indicating acquisition of SARS-CoV-2 from elsewhere in the hospital.

COG-UK is supported by funding from the Medical Research Council (MRC) part of UK Research & Innovation (UKRI), the National Institute of Health Research (NIHR) and Genome Research Limited, operating as the Wellcome Sanger Institute. Also MRC (MC UU 1201412), UKRI/Wellcome (COG-UK), Wellcome Trust Collaborator Award (206298/Z/17/Z – ARTIC Network; TCW Wellcome Trust Award 204802/Z/16/Z.