The convergence epidemic volatility index (cEVI) as an alternative early warning tool for identifying waves in an epidemic

Konstantinos Pateras; Eleftherios Meletis; Matthew Denwood; Paolo Eusebi; Polychronis Kostoulas

doi:10.1016/j.idm.2023.05.001

The convergence epidemic volatility index (cEVI) as an alternative early warning tool for identifying waves in an epidemic

Infect Dis Model. 2023 May 7;8(2):484-490. doi: 10.1016/j.idm.2023.05.001. eCollection 2023 Jun.

Authors

Konstantinos Pateras^{1

2}, Eleftherios Meletis¹, Matthew Denwood³, Paolo Eusebi⁴, Polychronis Kostoulas¹

Affiliations

¹ Department of Public and One Health, School of Medicine, University of Thessaly, Karditsa, Terma Mavromichali St., 43131, Greece.
² Department of Data Science and Biostatistics, University of Utrecht, Postbus 85500, 3508, GA, Utrecht, the Netherlands.
³ Department of Veterinary and Animal Sciences, University of Copenhagen, Grønnegårdsvej 8, 1870, Frederiksberg, Copenhagen, Denmark.
⁴ Department of Medicine and Surgery, University of Perugia, Via Gambuli, 1, 06132, Perugia, Italy.

Abstract

This manuscript introduces the convergence Epidemic Volatility Index (cEVI), a modification of the recently introduced Epidemic Volatility Index (EVI), as an early warning tool for emerging epidemic waves. cEVI has a similar architectural structure as EVI, but with an optimization process inspired by a Geweke diagnostic-type test. Our approach triggers an early warning based on a comparison of the most recently available window of data samples and a window based on the previous time frame. Application of cEVI to data from the COVID-19 pandemic data revealed steady performance in predicting early, intermediate epidemic waves and retaining a warning during an epidemic wave. Furthermore, we present two basic combinations of EVI and cEVI: (1) their disjunction cEVI + that respectively identifies waves earlier than the original index, (2) their conjunction cEVI- that results in higher accuracy. Combination of multiple warning systems could potentially create a surveillance umbrella that would result in early implementation of optimal outbreak interventions.

Keywords: Convergence diagnostics; Early warning; Epidemic index; Surveillance system; Time-series.