When COVID-19 Approaches Again: Early Detection of an Incoming New Wave of COVID-19 Positive Cases Inspired by Linear Breakpoint Models

From time to time, breakouts of COVID-19 income as waves of rapidly increasing numbers of positive cases alternated by time periods, varying in length, of low numbers of positive cases. Early detection of an incoming COVID-19 wave may enable pro-actively and quick activating all epidemiological measures that belong, together with vaccination, between a few effective weapons against unlimited COVID-19 transmission in a population. In this work, we applied an approach for the early detection of an incoming COVID-19 wave inspired by linear breakpoint models. Moreover, we refined the task of early detection as an optimization. Fitting a linear model on increasing numbers of positives from the present day to the past requires a sufficient number of daily records of positives from the past to get a significant positive regression slope. However, the more daily records of positives, usually stagnating in the past, we need to consider, the flatter the regression line becomes, which lowers the chance the regression slope is significant. Thus, finding a breakpoint, i. e. the beginning of the new wave of positives, is tricky. To address this issue, we propose an iterative algorithm searching for a breakpoint followed by a significantly positive regression slope. Finally, considering that a new wave is fitted by increasing the regression line, we also discuss an average number of days after the beginning of the incoming wave, when the early detection is, on average, firstly possible on a given statistical confidence level.