Bayesian reconstruction of SARS-CoV-2 transmissions highlights substantial proportion of negative serial intervals
Cyril Geismar,
Vincent Nguyen,
Ellen Fragaszy,
Madhumita Shrotri,
Annalan M.D. Navaratnam,
Sarah Beale,
Thomas E. Byrne,
Wing Lam Erica Fong,
Alexei Yavlinsky,
Jana Kovar,
Susan Hoskins,
Isobel Braithwaite,
Robert W. Aldridge,
Andrew C. Hayward,
Peter J. White,
Thibaut Jombart,
Anne Cori
Affiliations
Cyril Geismar
MRC Centre for Global Infectious Disease Analysis and NIHR Health Protection Research Unit in Modelling and Health Economics, Department of Infectious Disease Epidemiology, School of Public Health, Imperial College London, London, UK; Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK; Corresponding author at: MRC Centre for Global Infectious Disease Analysis and NIHR Health Protection Research Unit in Modelling and Health Economics, Department of Infectious Disease Epidemiology, School of Public Health, Imperial College London, London, UK.
Vincent Nguyen
Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK
Ellen Fragaszy
Institute of Epidemiology and Health Care, University College London, London, UK; Department of Infectious Disease Epidemiology, London School of Hygiene & Tropical Medicine, London, UK
Madhumita Shrotri
Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK
Annalan M.D. Navaratnam
Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK
Sarah Beale
Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK; Institute of Epidemiology and Health Care, University College London, London, UK
Thomas E. Byrne
Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK
Wing Lam Erica Fong
Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK
Alexei Yavlinsky
Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK
Jana Kovar
Institute of Epidemiology and Health Care, University College London, London, UK
Susan Hoskins
Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK
Isobel Braithwaite
Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK
Robert W. Aldridge
Centre for Public Health Data Science, Institute of Health Informatics, University College London, London, UK
Andrew C. Hayward
Institute of Epidemiology and Health Care, University College London, London, UK
Peter J. White
MRC Centre for Global Infectious Disease Analysis and NIHR Health Protection Research Unit in Modelling and Health Economics, Department of Infectious Disease Epidemiology, School of Public Health, Imperial College London, London, UK
Thibaut Jombart
MRC Centre for Global Infectious Disease Analysis and NIHR Health Protection Research Unit in Modelling and Health Economics, Department of Infectious Disease Epidemiology, School of Public Health, Imperial College London, London, UK
Anne Cori
MRC Centre for Global Infectious Disease Analysis and NIHR Health Protection Research Unit in Modelling and Health Economics, Department of Infectious Disease Epidemiology, School of Public Health, Imperial College London, London, UK
Background: The serial interval is a key epidemiological measure that quantifies the time between the onset of symptoms in an infector-infectee pair. It indicates how quickly new generations of cases appear, thus informing on the speed of an epidemic. Estimating the serial interval requires to identify pairs of infectors and infectees. Yet, most studies fail to assess the direction of transmission between cases and assume that the order of infections - and thus transmissions - strictly follows the order of symptom onsets, thereby imposing serial intervals to be positive. Because of the long and highly variable incubation period of SARS-CoV-2, this may not always be true (i.e an infectee may show symptoms before their infector) and negative serial intervals may occur. This study aims to estimate the serial interval of different SARS-CoV-2 variants whilst accounting for negative serial intervals. Methods: This analysis included 5 842 symptomatic individuals with confirmed SARS-CoV-2 infection amongst 2 579 households from September 2020 to August 2022 across England & Wales. We used a Bayesian framework to infer who infected whom by exploring all transmission trees compatible with the observed dates of symptoms, based on a wide range of incubation period and generation time distributions compatible with estimates reported in the literature. Serial intervals were derived from the reconstructed transmission pairs, stratified by variants. Results: We estimated that 22% (95% credible interval (CrI) 8–32%) of serial interval values are negative across all VOC. The mean serial interval was shortest for Omicron BA5 (2.02 days, 1.26–2.84) and longest for Alpha (3.37 days, 2.52–4.04). Conclusions: This study highlights the large proportion of negative serial intervals across SARS-CoV-2 variants. Because the serial interval is widely used to estimate transmissibility and forecast cases, these results may have critical implications for epidemic control.
