Abstract
WHAT IS ALREADY KNOWN ABOUT THIS TOPIC? The exact number of incident cases of emerging infectious diseases on a daily basis is of great importance to the disease control and prevention, but it is not directly available from the current surveillance system in time. WHAT IS ADDED BY THIS REPORT? In this study, a Bayesian statistical method was proposed to estimate the posterior parameters of the gamma probability distribution of the lag time between the onset date and the reporting time based on the surveillance data. And then the posterior parameters and corresponding cumulative gamma probability distribution were used to predict the actual number of new incident cases and the number of unreported cases per day. The proposed method was used for predicting COVID-19 incident cases from February 5 to February 26, 2020. The final results show that Bayesian probability model predictions based on data reported by February 28, 2020 are very close to those actually reported a month later. WHAT ARE THE IMPLICATIONS FOR PUBLIC HEALTH PRACTICE? This research provides a Bayesian statistical approach for early estimation of the actual number of cases of incidence based on surveillance data, which is of great value in the prevention and control practice of epidemics.