A new model for accurate short-term prediction of COVID-19 cases

New research by researchers at Texas A & M University PLOS ONE Learn more about a new model for short-term prediction of daily COVID-19 cases that is accurate, reliable, and easy to use by public health authorities and other organizations.

Researchers led by Hongwei Zhao, a professor of biostatistics at the Texas A & M School of Public Health, used a method based on the SEIR (susceptibility, exposure, infection, recovery) framework to COVID-19 in the next 2-3 weeks. Predicted the incidence of. Only for the observed incidence. This model assumes constant or slight changes in the prevalence of the virus that causes COVID-19 in a short period of time.

This model uses public data on newly reported cases of COVID-19 in Texas from the COVID-19 data repository by the Center for Systems Science and Engineering, Johns Hopkins University. Texas A & M researchers used this data on incidence rates in several counties, including Texas and the Texas A & M campus, to estimate COVID-19 infection rates.

The results show that this model can be used to reasonably predict COVID-19 cases 2-3 weeks in advance using only the current incidence. The simplicity of this model is one of its greatest strengths because it is easy to implement even in resource-poor organizations. Predictions from this model help healthcare organizations prepare for spikes and help public health authorities determine if mask obligations or other policies are needed. “

Hongwei Zhao, Professor of Biostatistics, Faculty of Public Health, Texas A & M

They predicted future infections under three possible scenarios. Infection rates are 5 percent higher than current levels, reflecting reduced practices to prevent infection or increased conditions that promote infection. The transmittance is 5% lower.

Estimating the current effective transmission rate can be difficult because day-to-day fluctuations in both infection and reporting can dramatically affect this estimate. Therefore, researchers used a 3-day weighted average to smooth out daily report variability and perform additional smoothing to account for data anomalies such as counties reporting months of cases at once. did.

Researchers compared the incidence and projections reported in Texas over the four periods of 2020 (April 15, June 15, August 15, and October 15). After the gradual resumption began in Texas, it shut down and began to increase in early May. Since Memorial Day, the numbers have increased sharply and have been declining since the state-wide mask order was enacted in the summer. After Labor Day, infections increased again, but then remained flat until mid-October, when a dramatic increase in infection rates was observed again.

The state-wide application of the model deviates from the actual recorded incidence during the second period, probably due to the number that changed dramatically when the COVID-19 outbreak occurred around Memorial Day holidays. Predictions alone have shown that it works reasonably well. The model worked just as well at the county level, but population decline and population changes affected the reporting of new cases, such as students moving in and out of the area during the school year.

However, the model is limited by the data used. Local test and reporting policies and resources can affect the accuracy of your data, and assumptions about transmission rates based on current incidence are less likely to be more accurate in the future. And as more people become infected with COVID-19 and recover or are vaccinated, the susceptible population can change and affect the infection.

Despite these limitations, researchers said the model could be a valuable tool for healthcare and public health authorities, especially when combined with other sources. The COVID-19 pandemic isn’t over yet, so it’s important to have a tool that can help you determine when and where another surge will occur. Similarly, researchers want these new tools to be at their disposal to meet the needs of future infectious diseases.

In addition, this model is being used to create dashboards that provide real-time data on the COVID-19 epidemic across the state. It is used locally by university administrators and public health authorities.