Data rectification to account for delays in reporting disease incidence with an application to forecasting COVID-19 cases

Effective monitoring of infectious disease incidence remains a major challenge to public health. Difficulties in estimating the trends in disease incidence arise mainly from the time delay between case diagnosis and the reporting of cases to public health databases. However, predictive models usually assume that public data sets faithfully reflect the state of disease transmission. In this paper, we study the effect of delayed case reporting by comparing data reported by the Johns Hopkins Coronavirus Resource Center (CRC) with that of the raw clinical data collected from the San Antonio Metro Health District (SAMHD), San Antonio, Texas. An insight on the subtle effect that such reporting errors potentially have on predictive modeling is presented. We use an exponential distribution model for the regression analysis of the reporting delay. The proposed model for correcting reporting delays was applied to our recently developed SEYAR (Susceptible, Exposed, Symptomatic, Asymptomatic, Recovered) dynamical model for COVID-19 transmission dynamics. Employing data from SAMHD, we demonstrate that the forecasting ability of the SEYAR model is substantially improved when the rectified reporting obtained from our proposed model is utilized. The methods and findings demonstrated in this work have ample applicability in the forecasting of infectious disease outbreaks. Our findings suggest that failure to consider reporting delays in surveillance data can significantly alter forecasts.