EpidemicKabu a new method to identify epidemic waves and their peaks and valleys

Abstract

INTRODUCTION: The dynamical behavior of epidemic curves is an oscillation between a very low and very high number of incident cases throughout the time. These oscillations are commonly called waves of the epidemic curve. The concept of epidemic waves lacks a consensual definition and a simple methodology that can be used for many diseases. OBJECTIVE: We describe in this study the EpidemicKabu method to identify the start and the end of past epidemic waves but also their peaks and valleys. METHOD: The methodology is divided into processing of the curve, waves detection, and peaks and valleys delimitation. For processing the curve, a Gaussian kernel was used to diminish the noise and to smooth the curve. The first and second derivatives of the curve were used for the detection of waves, delimitation of peaks and valleys. The methodology was derived into the open access library. The method was tested using COVID-19 daily cases reported between 2020 and 2022 for different countries. After detection of waves, we made some measures related to the size of the waves for those countries. RESULTS: The results of the method were the dates of start and end of waves, peaks, and valleys. The dates are displayed on graphs and added as a new column in a dataset. We found that Belgium was the country recording the highest ratio of incident cases per 100 people by day in a wave. CONCLUSION: The EpidemicKabu method is simple, easy to use, and very useful in estimating epidemic waves. The methodology requires expert judgment in order to set a parameter that could only have three possible values.