Evaluating transmission heterogeneity and super-spreading event of COVID-19 in a metropolis of China

Yunjun Zhang, Yuying Li, Lu Wang, Mingyuan Li, Xiao‐Hua Zhou

Abstract

Background: COVID-19 caused rapid mass infection worldwide. Understanding its transmission characteristics including heterogeneity is of vital importance for prediction and intervention of future epidemics. In addition, transmission heterogeneity usually evokes super spreading events (SSEs) where certain individuals infect large numbers of secondary cases. Till now, studies of transmission heterogeneity of COVID-19 and its underlying reason are far from reaching an agreement.

Methods: We collected information of all infected cases between January 21 and February 26, 2020 from official public sources in Tianjin, a metropolis of China. Utilizing a heterogeneous transmission model based on branching process along with a negative binomial offspring distribution, we estimated the reproductive number $R$ and the dispersion parameter $k$ which characterized the transmission potential and heterogeneity, respectively. Furthermore, we studied the SSE in Tianjin outbreak and evaluated the effect of control measures undertaken by local government based on the heterogeneous model.

Results: A total of 135 confirmed cases (including 34 imported cases and 101 local infections) in Tianjin by February 26th 2020 entered the study. We grouped them into 43 transmission chains with the largest chain of 45 cases and the longest chain of 4 generations. The estimated reproduction number $R$ was at 0.67 (95%CI: 0.54~0.84), and the dispersion parameter $k$ was at 0.25 (95% CI: 0.13~0.88). A super spreader causing six infections in Tianjin, was identified. In addition, our simulation results showed that the outbreak in Tianjin would have caused 165 infections and sustained for 7.56 generations on average if no control measures had been taken by local government since January 28th.

Conclusions: Our analysis suggested that the transmission of COVID-19 was subcritical but with significant heterogeneity and may incur SSE. More efforts are needed to verify the transmission heterogeneity of COVID-19 in other populations and its contributing factors, which is important for developing targeted measures to curb the pandemic.

Introduction

In the early stages of the COVID-19 pandemic, public health modeling primarily focused on the Basic Reproduction Number ($R_0$), representing the average number of secondary cases generated by a single infected individual. However, averages can obscure critical dynamics. If one individual infects ten people while nine others infect none, the average is one, yet the epidemiological risk and necessary response differ vastly from a scenario where ten people each infect one person.

This study addresses the "missing piece" of early pandemic modeling: transmission heterogeneity. By quantifying the variation in how individuals spread the virus (characterized by the dispersion parameter $k$), researchers can better predict the likelihood of explosive outbreaks and evaluate the efficacy of targeted versus broad interventions.

Methods

The researchers employed a stochastic branching process model to reconstruct the transmission history of 135 confirmed cases in Tianjin. Key technical components included:

  • Negative Binomial Offspring Distribution: Unlike a Poisson distribution, which assumes uniform transmission risk, the negative binomial distribution accounts for "overdispersion." A smaller $k$ value indicates higher heterogeneity and a greater propensity for super-spreading events (SSEs).
  • Chain Reconstruction: Transmission was categorized into three types: simple (clear history), ordinary (clear primary case, unclear secondary links), and complex (multiple potential primary cases).
  • EM Algorithm: For complex chains with ambiguous primary cases, an Expectation-Maximization (EM) algorithm was used to estimate the most probable partition of those chains.
  • Monte-Carlo Simulations: 5,000 simulations were conducted to model the outbreak's trajectory in the absence of the control measures implemented by the Tianjin government on January 28, 2020.

Data and Transmission Dynamics

The study analyzed 135 confirmed cases reported between January 21 and February 26, 2020, sourced from the Tianjin Municipal Health Commission. The dataset comprised 34 imported cases and 101 local infections, organized into 43 distinct transmission chains.

Proportion and temporal distribution of COVID-19 cases in Tianjin by infection source.

As shown in the data, household transmission was the most prevalent source of infection, followed by public places and imported cases. The temporal distribution indicates a peak in cases around early February, followed by a decline as interventions took effect.

Key Results and Insights

Transmission Parameters

The estimated reproduction number $R$ was 0.67 (95% CI: 0.54–0.84). In epidemiology, $R < 1$ indicates a "subcritical" state where the outbreak will eventually terminate. However, the dispersion parameter $k$ was 0.25 (95% CI: 0.13–0.88), signifying significant heterogeneity. This low $k$ value confirms that COVID-19 is highly prone to super-spreading events, where a small minority of cases drive the majority of transmission.

Super-Spreading Events

Using a 99th-percentile criterion from a homogeneous model, a super-spreader was defined as an individual infecting six or more people. One such individual was identified in the Tianjin data, notably within the complex transmission chains.

Reconstructed transmission chains in Tianjin, illustrating simple, ordinary, and complex structures. Chain D represents a significant cluster.

Impact of Interventions

The study found that interventions significantly altered transmission dynamics. Before February 1st, $R$ was 0.74 with a $k$ of 0.14. Following the implementation of control measures, $R$ dropped to 0.53 and $k$ increased to 0.77, indicating more uniform and less "spiky" transmission. Simulations suggested that without these measures, the outbreak would have resulted in approximately 165 infections and lasted nearly twice as many generations.

Discussion and Limitations

The findings suggest that while the average spread was contained, the inherent heterogeneity of COVID-19 transmission poses a constant risk of explosive clusters. The authors hypothesize that this heterogeneity may be driven by asymptomatic shedding or pre-symptomatic transmission, which complicates traditional contact tracing.

Limitations of the study include a relatively small sample size (135 cases), which leads to wider confidence intervals for $k$, and potential observation bias inherent in early-pandemic data collection. Future research should investigate the biological and environmental factors—such as "super-shedders" or high-density environments—that contribute to this transmission variance.

Summary updated

Abstract

Abstract Background COVID-19 caused rapid mass infection worldwide. Understanding its transmission characteristics including heterogeneity is of vital importance for prediction and intervention of future epidemics. In addition, transmission heterogeneity usually envokes super spreading events (SSEs) where certain individuals infect large numbers of secondary cases. Till now, studies of transmission heterogeneity of COVID-19 and its underlying reason are far from reaching an agreement. Methods We collected information of all infected cases between January 21 and February 26, 2020 from official public sources in Tianjin, a metropolis of China. . Utilizing a heterogeneous transmission model based on branching process along with a negative binomial offspring distribution, we estimated the reproductive number R and the dispersion parameter k which characterized the transmission potential and heterogeneity, respectively. Furthermore, we studied the SSE in Tianjin outbreak and evaluated the effect of control measures undertaken by local government based on the heterogeneous model. Results A total of 135 confirmed cases (including 34 imported cases and 101 local infections) in Tianjin by February 26th 2020 entered the study. We grouped them into 43 transmission chains with the largest chain of 45 cases and the longest chain of 4 generations. The estimated reproduction number R was at 0.67 (95%CI: 0.54~0.84), and the dispersion parameter k was at 0.25 (95% CI: 0.13~0.88). A super spreader causing six infections in Tianjin, was identified. In addition, our simulation results showed that the outbreak in Tianjin would have caused 165 infections and sustained for 7.56 generations on average if no control measures had been taken by local government since January 28th. Conclusions Our analysis suggested that the transmission of COVID-19 was subcritical but with significant heterogeneity and may incur SSE. More efforts are needed to verify the transmission heterogeneity of COVID-19 in other populations and its contributing factors, which is important for developing targeted measures to curb the pandemic.
BibTeX
@article{zhang2020evaluating,
  author = {Yunjun Zhang and Yuying Li and Lu Wang and Mingyuan Li and Xiao‐Hua Zhou},
  title = {{Evaluating transmission heterogeneity and super-spreading event of COVID-19 in a metropolis of China}},
  journal = {International Journal of Environmental Research and Public Health},
  year = {2020},
  doi = {10.3390/ijerph17103705},
}
Figure