Estimation of Near-kink Reproduction Numbers During the Emergent Variants of the COVID-19 Pandemic: Log-quadratic and Forward-imputation Approach

Abstract

AbstractBackgroundSketching the major portraits of the COVID-19 epidemic when variants of the pathogen emerge is critical to inform the dynamics of disease transmission, reproduction (i.e., the average counts of individuals of secondary infections generated by an index individual infected by the virus) strength of the pathogen, and countermeasure strategies. Multiple approaches, including log-linear, EpiEstim (an R package generally utilized to estimate the evolution traits of epidemics), and near-log-linear techniques, have been exploited to evaluate the principal parameters such as basic and effective reproduction numbers of an epidemic outbreak.ObjectiveThis study focuses on the kink corner (i.e., sharp alternation of direction of the transmission curve) presenting differentiated log-quadratic traits where more infectious variants of viruses emerge at the diminishing transmission phase of an infectious disease.MethodsA novel log-quadratic trending framework was proposed to project potentially unidentified cases (i.e., forward imputing approximately one week ahead) of COVID-19 around the kink, where the transmission of the pandemic initially lowered and accelerated subsequently, and exercised with the updated framework of classic EpiEstim and Log-linear model. I first compared the performance near the kink using the proposed technique versus the two traditional models taking into account a variety of levels of transmissibility, data distribution (Weibull, Gamma, and Lognormal distributions), and reporting rates (0.2, 0.4, 0.6, 0.8 and 1.0 respectively). Thereafter I utilized the revised framework on the outbreak data of four settings including Bulgaria, Japan, Poland, and South Korea from June to August 2022.ResultsThe proposed framework reduced the estimation bias versus traditional EpiEstim and log-linear methods near the kink. The coverage estimates of 95% confidence intervals improved. The proposed forward-imputation method implied generally a consistent ascending trend of effective reproduction number estimation applying to a precipitous transition from diminishing to diverging scenarios versus the irregular zigzagging outcomes in classic methods when more contagious variants of the virus were present in the absence of effective vaccines.ConclusionsThe log-quadratic correction accounting for transmissibility, data distribution, reporting rates, sliding windows, and generation intervals improved the basic and effective reproduction numbers estimation at the kink corner versus the classic EpiEstim and log-linear models by refined amendment of curve fitting. This is of concern when essentially the fundamental transmission traits of a pandemic alter expeditiously and countermeasures are needed at the earlier variant phases of the transiting climax with the advancement of the pandemic.