Rectification of delays in reporting disease incidence with an application to forecasting COVID-19 cases

Abstract: Real-time tracking of the trends in an infectious disease incidence is of major public health concern. Systemic errors in estimating a disease case counts are mostly introduced by the delay between onsets of symptoms and reporting to central databases. Whilst reporting error in epidemic data remains inevitable, its potential to undermine the predictability of disease transmission has often been overlooked by many predictive models. In this article, we analyze the delays in disease incidence data reported by the John Hopkins Coronavirus Resource Center (CRC). We show that when there are reporting errors in the data used for predictive modelling, its forecasting ability is significantly biased. A method for adjusting reporting delays in COVID-19 case counts was developed, following the analysis of the raw clinical data collected from the San Antonio Metro Health District (SAMHD), San Antonio, TX, and that of the CRC. The proposed method uses an exponential distribution model for the regression analysis of the reporting delay. Our method also minimized the time-lag between the CRC and SAMHD time-series, by estimating the optimal data time for the CRC data by taking the difference between the CRC and SAMHD data. The proposed model for adjusting reporting delays was applied to our recently developed SEYAR (Susceptible, Exposed, symptomatic, Asymptomatic, Recovered) dynamical model for COVID-19 transmission dynamics. A measure of the agreement between our model and data is presented. We illustrate that the predictability of the SEYAR model with rectified reporting is significantly improved. The methods and results shown in this study are useful for adjustment of reporting errors in disease incidence data, as well as forecasting epidemics. Our findings highlight the potential and pitfalls of reporting errors in case counts when forecasting pandemic.

Back to Search Research Symposium Abstracts