Visualization of COVID-19 spread based on spread and extinction indexes

Song-Hai Zhang, Yun Cai, Jian LiScience China Information Sciences 2020 · DOI · OpenReview

Introduction

During the initial COVID-19 outbreak in early 2020, decision-makers and the public were inundated with raw data that was difficult to interpret. Traditional epidemiological models are often too complex for non-experts, and standard statistical charts struggle to represent data spanning multiple orders of magnitude—such as the massive disparity between infection rates in Hubei Province versus the rest of China. This paper introduces a specialized visualization system designed to communicate these complex dynamics to the general public through intuitive metrics and interaction-less visualizations.

Mathematical Indices

To provide immediate insight without requiring data analysis expertise, the researchers defined two primary mathematical indices:

  • Spread Index ($F_s$): The ratio of new confirmed cases to the number of patients currently under treatment from the previous day, measuring the "speed" of the spread.
  • Extinction Index ($F_e$): The ratio of the sum of cured patients and fatalities to the number of patients under treatment from the previous day, measuring the "recovery" or "clearance" rate.
  • Infection Rate ($F_r$): Total confirmed cases divided by the resident population of an area.
Overview of the mathematical indices and visualization components, including the relationship between Spread and Extinction indexes (a, b) and the resulting temporal and spatial charts (c-f).

The relationship between $F_s$ and $F_e$ is critical: when the Extinction Index exceeds the Spread Index ($F_e > F_s$), the total number of active cases begins to decline, marking a pandemic turning point. In China, this "balancing point" occurred around February 7, 2020.

Visualization Techniques

The system employs modified visualization techniques to handle the high dynamic range and data imbalance of the early outbreak:

  • Bidirectional ThemeRiver: Standard ThemeRivers were redesigned to plot Hubei Province (which accounted for over 80% of cases) and all other provinces in opposite directions along a horizontal axis. This maintains legibility for both Hubei and the rest of the country.
  • Logarithmic Bubble Charts: To represent cities with vastly different case counts (from dozens to tens of thousands), the system uses logarithmic scales for axes and bubble areas.
  • Animation: Time-series animations show the evolution of the pandemic without requiring user input, making the data accessible to "unskilled individuals."
Detailed view of the bidirectional ThemeRiver (d) and the logarithmic bubble chart (f) showing the state of the pandemic in late February 2020.

Key Insights and Results

The visualization system successfully identified several key trends:

  1. Leading Indicators: The Spread Index began decreasing well before the daily number of new cases peaked, providing an early signal that containment measures were effective even while raw numbers were still climbing.
  2. Identifying Hidden Risks: By using the Infection Rate ($F_r$) as bubble size, the system identified E’zhou as a high-risk city on February 11, despite its lower absolute case count compared to other major hubs.
  3. Public Impact: The visualization reports were published daily on Tencent UR, garnering over 20 million page views by the end of February 2020, demonstrating the effectiveness of the communication strategy.

Conclusion

While the system is empirical rather than predictive and relies on the accuracy of official reporting, it proved to be a powerful tool for public communication. By simplifying complex epidemiological data into intuitive indices and clear visual structures, the authors helped the public rationally understand the pandemic's progression and the impact of government policies.

Summary updated

BibTeX
@article{zhang2020visualization,
  author = {Song-Hai Zhang and Yun Cai and Jian Li},
  title = {{Visualization of COVID-19 spread based on spread and extinction indexes}},
  journal = {Science China Information Sciences},
  year = {2020},
  doi = {10.1007/s11432-020-2828-1},
}
Figure