Superspreaders drive the largest outbreaks of hospital onset COVID-19 infections

  1. Christopher JR Illingworth  Is a corresponding author
  2. William L Hamilton
  3. Ben Warne
  4. Matthew Routledge
  5. Ashley Popay
  6. Chris Jackson
  7. Tom Fieldman
  8. Luke W Meredith
  9. Charlotte J Houldcroft
  10. Myra Hosmillo
  11. Aminu S Jahun
  12. Laura G Caller
  13. Sarah L Caddy
  14. Anna Yakovleva
  15. Grant Hall
  16. Fahad A Khokhar
  17. Theresa Feltwell
  18. Malte L Pinckert
  19. Iliana Georgana
  20. Yasmin Chaudhry
  21. Martin D Curran
  22. Surendra Parmar
  23. Dominic Sparkes
  24. Lucy Rivett
  25. Nick K Jones
  26. Sushmita Sridhar
  27. Sally Forrest
  28. Tom Dymond
  29. Kayleigh Grainger
  30. Chris Workman
  31. Mark Ferris
  32. Effrossyni Gkrania-Klotsas
  33. Nicholas M Brown
  34. Michael P Weekes
  35. Stephen Baker
  36. Sharon J Peacock
  37. Ian G Goodfellow
  38. Theodore Gouliouris
  39. Daniela de Angelis
  40. M Estée Török
  1. MRC Biostatistics Unit, University of Cambridge, East Forvie Building, Forvie Site, Robinson Way, United Kingdom
  2. Institut für Biologische Physik, Universität zu Köln, Germany
  3. Department of Applied Mathematics and Theoretical Physics, Centre for Mathematical Sciences, United States
  4. University of Cambridge, Department of Medicine, Cambridge Biomedical Campus, United Kingdom
  5. Cambridge University Hospitals NHS Foundation Trust, Cambridge Biomedical Campus, United Kingdom
  6. Public Health England Clinical Microbiology and Public Health Laboratory, Cambridge Biomedical Campus, United Kingdom
  7. Public Health England Field Epidemiology Unit, Cambridge Institute of Public Health, Forvie Site, Cambridge Biomedical Campus, United Kingdom
  8. University of Cambridge, Department of Pathology, Division of Virology, Cambridge Biomedical Campus, United Kingdom
  9. Cambridge Institute for Therapeutic Immunology and Infectious Disease, Jeffrey Cheah Biomedical Centre, United Kingdom
  10. Wellcome Sanger Institute, Wellcome Trust Genome Campus, United Kingdom
  11. MRC Epidemiology Unit, University of Cambridge, Level 3 Institute of Metabolic Science, United Kingdom
  12. University of Cambridge, School of Clinical Medicine, Cambridge Biomedical Campus, United Kingdom
  13. Public Health England, National Infection Service, United Kingdom
5 figures, 1 table and 3 additional files

Figures

Preliminary analysis of the data with A2B-Covid.

Squares indicate the extent to which an individual-to-individual transmission event is consistent with the data collected, when considered on a pairwise level. Our analysis highlighted multiple …

Figure 2 with 2 supplements
Maximum likelihood transmission networks for wards A to D.

Circles represent individuals and arrows show transmission events. White circles represent patients while grey circles represent health care workers. Individuals for which no transmission events …

Figure 2—figure supplement 1
Maximum likelihood sources of patient and HCW infections.

Statistics were calculated across maximum likelihood network reconstructions. The great majority of patient infections were inferred to arise from other patients, while HCWs were infected roughly …

Figure 2—figure supplement 2
Ensemble transmission networks for wards A to D.

Data were compiled over sets of plausible reconstructions, weighted by likelihood. The width of each arrow is proportional to the probability that a specific transmission event occurred. Arrows are …

Figure 2—figure supplement 2—source data 1

Posterior probabilities of transmission between individuals on each ward.

https://cdn.elifesciences.org/articles/67308/elife-67308-fig2-figsupp2-data1-v2.xlsx
Maximum likelihood transmission network for ward E.

Circles represent individuals and arrows show transmission events. White circles represent patients while grey circles represent health care workers. Individuals for which no transmission events …

Figure 4 with 5 supplements
Models of viral transmission.

(A) Fit of the output of the Poisson model (black dots) to the ensemble data (yellow bars). The weighted number of transmissions per individual reflects the uncertainty in the network reconstruction …

Figure 4—source data 1

Distributions of number of individuals infected by each individual and fits to these data using Poisson and Negative Binomial models.

https://cdn.elifesciences.org/articles/67308/elife-67308-fig4-data1-v2.xlsx
Figure 4—figure supplement 1
Modelling of viral transmission on the green wards.

(A) Fit of the output of a Poisson model (black dots) to the ensemble data (yellow bars). The weighted number of transmissions per individual reflects the uncertainty in the network reconstruction …

Figure 4—figure supplement 1—source data 1

Distributions of number of individuals on green wards infected by each individual and fits to these data using Poisson and Negative Binomial models.

https://cdn.elifesciences.org/articles/67308/elife-67308-fig4-figsupp1-data1-v2.xlsx
Figure 4—figure supplement 2
Ct values of viral samples.

Distributions of known Ct values collected from all samples from the wards studied (yellow) and from individuals identified as superspreaders (grey). Samples from superspreaders were not …

Figure 4—figure supplement 3
Inferred timings of transmission events in ward A.

(A) Maximum likelihood network of transmission events. (B) Maximum likelihood spread of infection given this network. (C) Distributions of the times at when transmission occurred were calculated …

Figure 4—figure supplement 3—source data 1

Distributions of timings of transmission events on ward A.

https://cdn.elifesciences.org/articles/67308/elife-67308-fig4-figsupp3-data1-v2.xlsx
Figure 4—figure supplement 4
Inferred timings of transmission events in ward A.

(A) Maximum likelihood network of transmission events. (B) Maximum likelihood spread of infection given this network. Timings are shown relative to the first transmission event. Infection of the …

Figure 4—figure supplement 4—source data 1

Distributions of timings of transmission events on ward B.

https://cdn.elifesciences.org/articles/67308/elife-67308-fig4-figsupp4-data1-v2.xlsx
Figure 4—figure supplement 5
Inferred timings of transmission events in ward D.

(A) Maximum likelihood network of transmission events. (B) Maximum likelihood spread of infection given this network. Timings are shown relative to the first transmission event. Infection of the …

Figure 4—figure supplement 5—source data 1

Distributions of timings of transmission events on ward D.

https://cdn.elifesciences.org/articles/67308/elife-67308-fig4-figsupp5-data1-v2.xlsx
Figure 5 with 5 supplements
Overview of events on different wards.

Blue squares show days on which individuals became symptomatic, while green squares show inferred days of individuals becoming symptomatic when these dates were unknown or not applicable. Red …

Figure 5—figure supplement 1
Ensemble transmission networks for wards B, D, and E generated without extending the times at which HCWs were present beyond the direct observations made.

Data were compiled over sets of plausible reconstructions, weighted by likelihood. The width of each arrow is proportional to the probability that a specific transmission event occurred. Arrows are …

Figure 5—figure supplement 1—source data 1

Posterior probabilities of transmission between individuals on each ward inferred under a model in which no padding for HCW locations was included.

Data are shown only for wards in which the inference was different from the original.

https://cdn.elifesciences.org/articles/67308/elife-67308-fig5-figsupp1-data1-v2.xlsx
Figure 5—figure supplement 2
Modelling of viral transmission in the absence of an extension to HCW locations.

(A) Fit of the output of a Poisson model (black dots) to the ensemble data (yellow bars). (B) Fit of the output of the negative binomial model (black dots) to the ensemble data (yellow bars). (C) …

Figure 5—figure supplement 2—source data 1

Distributions of number of individuals on all wards infected by each individual, as inferred under a model in which no padding for HCW locations was included, alongside fits to these data using Poisson and Negative Binomial models.

https://cdn.elifesciences.org/articles/67308/elife-67308-fig5-figsupp2-data1-v2.xlsx
Figure 5—figure supplement 3
Assigning mutations to the transmission tree.

(A) Case of the last transmission event. Transmission occurs from A to B at time tAB. Viral sequence data is collected from A at time DA and from B at time DB. Grey markers show points which may be …

Figure 5—figure supplement 4
Restrictions placed on the network by sequence variants.

Here sequences collected from the individuals B and C (i.e. in the set Ia) have the variant a, but no other sequences have this variant. Data from individual i was collected at time Di. We assume …

Figure 5—figure supplement 5
Convergence of the statistical ensemble of networks for ward A.

Comparisons of the number of infections per individual and the probabilities of specific edges being found in the transmission network for a ‘partial’ set of networks and for a more complete, ‘full’ …

Figure 5—figure supplement 5—source data 1

Probabilities of transmission events between individuals inferred from data describing ward A calculated across partial and fuller sets of networks.

The values reported here show convergence in the model with the addition of further networks.

https://cdn.elifesciences.org/articles/67308/elife-67308-fig5-figsupp5-data1-v2.xlsx

Tables

Table 1
Case numbers in the five major ward clusters.

Total cases before network analysis’ were derived by adding patients with potential hospital-acquired COVID-19 infections and HCW cases from each ward. The five wards with the largest combined …

Ward nameWard typeTotal cases before network analysisHAI cases before network analysisHCW cases before network analysisCases after network analysisCases after network analysis with genomic data
AGreen141221615
BGreen11291512
CGreen12572019
DGreen144101616
ERed133104739

Additional files

Download links