Abstract
There has been an outbreak of coronavirus disease (COVID-19) in Wuhan city, Hubei province, China since December 2019. Cases have been exported to other parts of China and more than 20 countries. We provide estimates of the daily trend in the size of the epidemic in Wuhan based on detailed information of 10,940 confirmed cases outside Hubei province.
Introduction
In the early weeks of the COVID-19 pandemic, Wuhan was the epicenter of a rapidly growing crisis. However, official case counts were widely suspected to be underestimations due to overburdened hospitals, testing kit shortages, and asymptomatic or mild cases remaining undetected. While previous studies attempted to estimate the epidemic's size using international travel data, small sample sizes led to high statistical variability. This research addresses the need for a precise, high-resolution daily trend of the infected population by leveraging a much larger dataset: confirmed cases found in all Chinese provinces outside of Hubei, as well as international cases.
Methodology
The study employs a statistical framework based on a simplified SIR (Susceptible-Infectious-Recovered) model, specifically using an Ordinary Differential Equation (ODE) to describe logistic growth. The core approach involves back-calculation via exportation: by knowing the probability of an individual leaving Wuhan ($p$) and the time from infection to detection ($d$), the researchers used the number of cases discovered elsewhere to estimate the total infected population in Wuhan.
The number of imported cases on a given day was modeled using a binomial distribution. For cases lacking travel history, an imputation method was applied, assuming the proportion of imported cases matched the observed proportion for that day. Parameters such as the growth rate ($r$) and the total susceptible population ($K$) were estimated by maximizing a likelihood function based on daily imported case counts.
Data Analysis
The analysis utilized data from 10,940 confirmed cases reported outside Hubei province between mid-January and February 13, 2020. These records were sourced from provincial and municipal health commissions in China and health ministries in over 20 countries. The dataset included region, gender, age, symptom onset date, and confirmation date. Of these, 3,039 cases were explicitly identified as "imported" from Wuhan.
Results and Key Insights
The model reveals a significant discrepancy between estimated and reported cases. By January 11, 2020, the model suggests there were already 4,094 cases in Wuhan, while official counts remained in the dozens. By February 13, 2020, the estimated total cases reached 58,153 (95% CI: 56,532 – 59,811), compared to the officially reported 35,991.
A critical insight from the study is the evolution of the reporting rate. On January 20, 2020, the estimated reporting rate was a mere 1.41%. This rate improved significantly over time as public health infrastructure and testing capacity expanded, reaching 61.89% by February 13.
Furthermore, by solving the growth equation for the point of the first infection ($N_{t_0} = 1$), the researchers estimated the start of the outbreak to be November 30, 2019. The mean time from symptom onset to confirmation was calculated at 5.54 days.
Limitations and Conclusion
While the model provides a more robust estimate by using domestic travel data, it has limitations. It assumes travel probability is independent of infection status; however, severely ill individuals are less likely to travel, potentially leading to an underestimation. Additionally, the model relies on travel data prior to the Wuhan lockdown on January 23. Future work should focus on more thorough screening of mild cases to bridge the reporting gap and better understand the true spread of the virus.
Summary updated