Source-linked AI summary
Predicting the cumulative number of cases for the COVID-19 epidemic in China from early data
Zhihua Liu, Pierre Magal, Ousmane Seydi, Glenn Webb
TL;DR
The paper asks how early reported data can predict China’s COVID-19 epidemic while accounting for interventions, unreported cases, and asymptomatic infectiousness. It models transmission with a time-dependent rate and simulates epidemic trajectories. The results show that intervention timing and restriction level, unreported-case identification, and infectious-period assumptions substantially affect projected epidemic size.
Problem
The paper addresses the need to predict China’s cumulative COVID-19 cases while accounting for public policies, unreported cases, and asymptomatic infectious cases.
Method
The authors fit an exponentially decreasing transmission rate to reported data after government restrictions and numerically simulate reported and unreported epidemic trajectories.
Results
Earlier or more restrictive interventions produce substantially fewer projected cases, while higher unreported transmission produces a larger epidemic.
Takeaways & Limitations
Control depends strongly on early, restrictive public measures and identifying and isolating symptomatic unreported infectious cases.
Takeaways & Limitations
The unreported-case estimate is uncertain because reported data contain a gap after February 15 and the reporting fraction is assumed.
Abstract
from arXiv · showhide
We model the COVID-19 coronavirus epidemic in China. We use early reported case data to predict the cumulative number of reported cases to a final size. The key features of our model are the timing of implementation of major public policies restricting social movement, the identification and isolation of unreported cases, and the impact of asymptomatic infectious cases.
1 Introduction
The paper develops a China-focused COVID-19 epidemic model that examines how public restrictions and reported versus unreported cases shape epidemic progression. Using reported data after January 29, it projects the epidemic timeline toward a final size.
- The model focuses on government containment policies and reported and unreported COVID-19 cases in China.
- After January 23 restrictions, the model represents transmission as an exponentially decreasing time-dependent rate rather than a constant rate.
- Reported case data after January 29 are used to identify the decreasing transmission rate and project the epidemic forward toward a final size.
2 Model
The model represents susceptible, asymptomatic infectious, reported symptomatic, and unreported symptomatic populations with a system of ordinary differential equations. It assumes reported symptomatic individuals are immediately isolated, while transmission comes from asymptomatic and unreported symptomatic individuals.
- The epidemic dynamics are specified by four ordinary differential equations governing flows among susceptibility, infection, reporting, recovery, and unreported infection.
- The model tracks susceptible S(t), asymptomatic infectious I(t), reported symptomatic R(t), and unreported symptomatic U(t) individuals.
- Reported symptomatic individuals are assumed to be isolated immediately and cause no further infections.
- All infections are acquired from asymptomatic or unreported symptomatic infectious individuals.
3 Data
The study uses mainland-China case data from the National Health Commission and Chinese CDC through February 15, 2020. It presents daily reported cases alongside cumulative reported cases.
- The dataset covers mainland China reported cases through February 15, 2020, from the National Health Commission and Chinese CDC.
- Table 2 contains cumulative daily reported case data from January 20 through February 15, 2020.
- Figure 2 plots daily reported cases and cumulative reported cases.
- The reported-case epidemic turning point is approximately February 4, 2020, day 35 when January 1 is day 1.
4 Model parameters
The parameters combine early exponential case growth with assumptions about reporting and infectious periods. Government interventions are modeled through an exponentially decreasing transmission rate fitted to ongoing reported-case data.
- The model assumes f = 0.8, so 20% of symptomatic infectious cases go unreported.
- The assumed average infectious period is 7 days for reported, unreported symptomatic, and asymptomatic infectious individuals.
- The assumed parameter values can be modified as further epidemiological information becomes known.
- Early reported-case growth is fitted with χ1 = 0.16, χ2 = 0.38, and χ3 = 1.1, using Wuhan’s population as S0 = 11,000,000.
- After January 24, transmission is modeled as τ(t) = τ0 exp(−µ(t −24)) with µ = 0.12 fitted from ongoing reported-case data.
5 Model simulation
Numerical simulations project epidemic trajectories under the baseline model and vary intervention timing, reporting completeness, asymptomatic infectious duration, and restriction strength. The simulations show substantial changes in projected epidemic size across these scenarios.
- Asymptomatic infections: The asymptomatic infectious-case turning point is approximately day 35, preceding the day-41 turning points for reported and unreported cases.The day-by-day weekly reported data turn at approximately day 38.
- Intervention timing: Earlier intervention produces approximately 5,750 total cases, whereas intervention one week later produces approximately 1,234,000 total cases.The earlier scenario turns at day 34, while the later scenario turns at day 47.
- Unreported cases: When the reported fraction f is reduced from 0.8, the projected final size is approximately 164,700 cases for f = 0.4 and approximately 110,700 cases for f = 0.6.The model varies the transmission-rate parameters τ0 and µ for each reporting-fraction scenario.
- Sensitivity analyses: Using ν = 1/3 instead of ν = 1/7 produces a small decrease in projected final epidemic size, while decreasing µ from 0.12 to 0.09 produces a significant increase.Here ν represents the asymptomatic infectious-period parameter, and µ controls the modeled decline in transmission under public restrictions.
6 Discussion
The model incorporates policy timing and restrictiveness, reported and unreported symptomatic infections, and asymptomatic transmission to simulate China’s COVID-19 epidemic. It fits Chinese CDC data through February 15, 2020 and indicates that earlier, stronger restrictions reduce epidemic size.
- Model formulation: The model represents asymptomatic infectious individuals separately and divides symptomatic infections into reported and unreported cases.This structure supports interpreting reported cases while accounting for infections that may remain unreported.
- Policy effects: A time-dependent exponentially decreasing transmission rate models the effects of government restrictions beginning after January 23.The model uses the previously identified constant transmission rate during the early exponential-growth phase, then fits the declining rate to national data.
- Model fit: The simulations fit Chinese CDC reported case data for all of China with increasing accuracy through February 15, 2020.The fitting uses the early-phase constant transmission rate and the later time-dependent transmission rate.
- Policy effects: One week earlier implementation of major restrictions would significantly reduce total cases, whereas one week later implementation would significantly increase them.The model also predicts a significant increase in epidemic size if restrictions on public movement are less restrictive.
- Unreported and asymptomatic infections: A higher fraction of unreported cases significantly increases total cases, while shorter asymptomatic or unreported symptomatic infectious periods reduce them.The results emphasize identifying and isolating unreported symptomatic infectious cases and understanding infectiousness periods.
- Scope: Although specified for China, the model is applicable to COVID-19 outbreaks in other locations.Its formulation is presented as location-general beyond the China outbreak.