Transmission onset distribution of COVID-19

June Young Chun, Gyuseung Baek, Yongdai KimmedRxiv 2020 · DOI

Abstract

The distribution of COVID-19 transmission onset relative to symptom onset is a critical parameter for effective infection control. Because identifying the exact moment of transmission between individuals is challenging, this study inferred transmission onset times from 72 infector-infectee pairs in South Korea. Using Bayesian inference methods and accounting for incubation periods, the study estimated the median transmission onset to be 1.31 days after symptom onset, with a peak occurring 0.72 days before symptoms appear. The pre-symptomatic transmission proportion was estimated at 37%. These findings suggest that standard symptom-based isolation measures may occur too late to prevent a significant portion of SARS-CoV-2 transmissions.

Introduction

In the early stages of the COVID-19 pandemic, public health strategies relied heavily on identifying and isolating symptomatic individuals. However, the effectiveness of this approach depends entirely on the "transmission dynamics" of the virus—specifically, the timing of when an infected person becomes contagious relative to when they begin to feel ill. If a significant portion of transmission occurs before the onset of symptoms, traditional containment measures like temperature checks and self-reporting are perpetually one step behind the virus.

This study addresses the challenge of the "silent window" by analyzing high-quality epidemiological data from South Korea to map the transmission onset distribution. By understanding this window, researchers can better evaluate the viability of symptom-based screening and the necessity of more aggressive contact tracing and universal masking.

Methodology

The researchers utilized a Bayesian inference framework to estimate the time interval ($W$) between an infector’s symptom onset ($S$) and the actual moment of transmission ($P$).

Data Selection

The study analyzed public health data collected in South Korea between January 23 and March 31, 2020. From an initial pool of over 8,000 cluster-linked cases and 560 imported cases, the authors identified 72 "clean" infector-infectee pairs. These pairs were selected based on the clarity of their causal relationship and well-documented contact histories. The sample included:

  • 22 single-point contact cases: Where the exact date of exposure was known.
  • 16 duration-of-exposure cases: Where exposure occurred within a specific, limited timeframe.
  • 34 continuous exposure cases: Primarily household contacts where the infector and infectee lived together.

Statistical Modeling

Because the exact moment of transmission is rarely observed directly, the authors used Markov Chain Monte Carlo (MCMC) algorithms to impute missing transmission dates based on known incubation periods. The distribution of $W$ was modeled using a mixture of a shifted Gamma distribution (to capture the primary trend) and a Uniform distribution (to account for potential outliers or negative values). Sensitivity analyses were performed by applying incubation period distributions from other major studies (e.g., Lauer et al., 2020) to ensure the robustness of the findings.

Key Results

The analysis revealed several critical temporal parameters regarding the spread of SARS-CoV-2:

  • The Pre-symptomatic Peak: The probability of transmission was found to peak at 0.72 days (approximately 17 hours) before the onset of symptoms. This finding is consistent with biological studies of viral shedding, which show peak viral loads in the upper respiratory tract just before or at the start of clinical illness.
  • Pre-symptomatic Proportion: The study estimated that 37% of transmissions (95% Credible Interval: 16–52%) occur during the pre-symptomatic phase.
  • Median Transmission Onset: The median time of transmission was 1.31 days after symptom onset. While the peak is pre-symptomatic, the "tail" of the distribution extends into the symptomatic period.
  • Incubation and Serial Intervals: The median incubation period was estimated at 2.87 days (95% CI: 2.33–3.50), and the median serial interval (the time between symptom onset in successive cases) was 3.56 days (95% CI: 2.72–4.44).

Discussion and Implications

The finding that over one-third of transmissions occur before symptoms manifest has profound implications for public health policy. If 37% of the spread happens "silently," even a perfect system that isolates every symptomatic individual the moment they feel ill would fail to prevent a substantial portion of new infections.

The relatively short incubation and serial intervals observed in this South Korean cohort—shorter than some early estimates from the initial outbreak in Wuhan—likely reflect the impact of intensive contact tracing. In South Korea, aggressive testing and rapid isolation of contacts likely truncated the transmission distribution, preventing the "late-stage" transmissions that might occur in unmanaged environments.

Limitations

Several factors may influence these estimates. First, epidemiological data is subject to recall bias, as patients must remember exact dates of contact and symptom onset. Second, the study excluded large-scale "super-spreader" events to maintain causal accuracy, meaning the dynamics in high-density environments might differ. Finally, the study focused on individuals who eventually developed symptoms; the role of "true" asymptomatics (those who never develop symptoms) remains a separate but related challenge for global health authorities.

Conclusion

This study confirms that SARS-CoV-2 transmission peaks prior to symptom onset. Given that a significant proportion of transmission is pre-symptomatic, reliance on symptom-based screening is insufficient for containment. Effective control requires proactive measures, including rapid contact tracing, widespread testing regardless of symptoms, and the use of personal protective equipment in public spaces to mitigate the risk of silent spread.

Summary updated

Abstract

Abstract Objectives The distribution of the transmission onset of COVID-19 relative to the symptom onset is a key parameter for infection control. It is often not easy to study the transmission onset time, as is difficult to know who infected whom exactly when. Methods We inferred transmission onset time from 72 infector-infectee pairs in South Korea, either with known or inferred contact dates by means of incubation period. Combining this data with known information of infector’s symptom onset, we could generate the transmission onset distribution of COVID-19, using Bayesian methods. Serial interval distribution could be automatically estimated from our data. Results We estimated the median transmission onset to be 1.31 days (standard deviation, 2.64 days) after symptom onset with peak at 0.72 days before symptom onset. The pre-symptomatic transmission proportion was 37% (95% credible interval [CI], 16–52%). The median incubation period was estimated to be 2.87 days (95% CI, 2.33–3.50 days) and the median serial interval to be 3.56 days (95% CI, 2.72–4.44 days). Conclusions Considering the transmission onset distribution peaked with the symptom onset and the pre-symptomatic transmission proportion is substantial, the usual preventive measure might be too late to prevent SARS-CoV-2 transmission.
BibTeX
@misc{chun2020transmission,
  author = {June Young Chun and Gyuseung Baek and Yongdai Kim},
  title = {{Transmission onset distribution of COVID-19}},
  year = {2020},
  doi = {10.1101/2020.05.13.20101246},
  howpublished = {medRxiv},
}
Figure