In a recent study published on medRxiv* preprint server, Pfizer Inc scientists determined whether the rates of coronavirus disease 2019 (COVID-19) cases in the United States (US) and Europe (EU) follow a seasonal pattern at the using a time series model.
Study: Is COVID-19 seasonal? A time series modeling approach. Image Credit: CKA/Shutterstock
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) likely follows a pattern of seasonality like other viral respiratory infections. An interaction between the viral pathogen, host, and environmental factors, including seasonal temperature, humidity, and increased indoor activity during winters, may impact viral stability outside the host. ‘host.
Additionally, vaccination strategies to date have struggled to keep pace with the emergence of novel variants of concern (COVs) of SARS-CoV-2. In the absence of data, public health agencies have no idea when to deploy booster doses that could help maintain optimal levels of protection against serious illness throughout the pandemic.
Based on these observations, identifying the seasonal pattern of SARS-CoV-2 infections is crucial for public health planning as well as determining the optimal timing for booster doses.
About the study
In the current study, researchers retrieved COVID-19 data from the Our World in Data (OWID) database for the United States and five EU (EU5) countries, including Spain, France, Germany, Italy and the United Kingdom, from March 2020 to April 2022. They identified anomalies for each country using from Twitter decomposition method and the Generalized Extreme Studentized Deviate (GESD) tests.
Twitter time series decomposition with generalized extreme student gap (GESD) anomaly detection of COVID-19 rates, March 7, 2020 – April 9, 2022. Panel A: United States, Our World in Data; Panel B: EU5: Italy, Germany, France, Spain and United Kingdom, Our World in Data. Shaded areas represent the normal range of data points.
Decomposition of Twitter time series with generalized extreme student gap (GESD) anomaly detection of influenza cases, March 1, 2014 through April 10, 2020. Shaded areas represent normal range of data points. Data Source: US Centers for Disease Control and Prevention FluView
Additionally, the team used Meta’s Prophet approach to decompose US time series rates by adjusting the age-specific proportion of fully vaccinated individuals over time, US holidays, predominant variant in circulation, and seasonality. Additionally, they used Markov chain Monte Carlo simulations to calculate uncertainty intervals for the decomposed components of the Prophet time series.
The researchers performed sensitivity analyzes to determine the impact of the data sources. For example, they compared data from the US Centers for Disease Control and Prevention [CDC] to OWID data. Finally, they ran seasonal flu trends using FluView data from the US Centers for Disease Control and Prevention (CDC) to determine the accuracy of their time series models.
Census Bureau regions of the northern, midwestern, and western United States saw similar seasonal spikes in SARS-CoV-2 cases between November and February. However, the authors also noted a second seasonal peak in the southern region of the United States in late summer 2021. Anomaly detection analyzes reliably predicted seasonal influenza peaks over six U.S. influenza seasons . Similarly, anomaly plots detected higher than expected COVID-19 rates in the United States between November and March each year, which returned to normal in October 2021.
Sensitivity analyzes confirmed that the threshold for a proportion of anomalous observations had little impact on seasonal trends. Additionally, the data source did not impact seasonal patterns of COVID-19 in the United States. As a result, they were similar for US CDC and OWID data.
Sensitivity analyzes based on Meta’s Prophet model also showed an annual seasonal component for U.S. COVID-19 case rates between December and February, with holiday effects near Christmas and New Year’s Day, independence and spring break. The same analyzes showed remarkably similar trends for influenza but with slightly more residual variation.
The results of the study supported the deployment of protective measures against SARS-CoV-2 in the same time frame as for annual influenza prevention. In the United States and the EU5, cases of seasonal respiratory viral infections increase each year between December and March. Therefore, according to the authors, administering booster shots for COVID-19 before the winter months will most significantly reduce the burden of COVID-19 disease.
The methodology used in the study was superior to anomaly detection methods typically used to identify seasonal trends in common respiratory viruses. It showed that the flu season was longer than the COVID-19 season, possibly due to additional non-pharmaceutical measures taken for COVID-19, including masking and social distancing. The authors highlighted the need for further confirmatory studies in the Southern Hemisphere and other regions of the US and EU to inform public health strategies and stay ahead of upcoming seasonal waves of COVID. -19.
Studies have shown that giving more than one booster dose of COVID-19 vaccines each year is a programmatic challenge and can also lead to a phenomenon called “booster fatigue.” Thus, timing of annual COVID-19 vaccine administration is the most conservative approach to confer vaccine protection at the time of increasing COVID-19 cases. In addition, careful epidemiological, benefit-risk, and programmatic considerations will be needed to administer additional boosters to certain high-risk groups in the future. Thus, although vaccination alone is unlikely to lead to the elimination of SARS-CoV-2, timely deployment of vaccines can still mitigate future waves of SARS-CoV-2 infection.
medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be considered conclusive, guide clinical practice/health-related behaviors, or treated as established information.