UNDERREPORTING IN INFECTIOUS DISEASE CASES: A TIME SERIES REGRESSION-SIR APPROACH

Failure to account for the underreporting of infectious disease cases distorts the understanding of infectious disease dynamics. Underreporting creates a false sense of security, allowing the disease to persist or resurge and undermining the effectiveness of public health interventions. This study aims to address underreporting and identify the underlying distribution that best describes the Coronavirus disease 2019 (COVID-19) cases in Nigeria. A Time Series Regression Susceptible-Infected-Recovered (TSIR) model, incorporating Poisson, Gaussian, and Quasi-Poisson distributions with various link functions, was applied to weekly cumulative COVID-19 case data. This dataset spans from February 28, 2020, to July 3, 2022, and includes 110 weekly records. It was sourced from the Nigerian Centre for Disease Control (NCDC) through publicly available weekly epidemiological reports. Microsoft Office Excel 2016 was utilized to collate the database, and the NCDC’s online platform served as the primary data source. The data were divided into two sets: training data from February 28, 2020, to March 20, 2022, comprising 100 cases for modeling TSIR, and testing data from March 27, 2022, to July 3, 2022, encompassing 10 weekly cases for model performance evaluation. These research findings revealed that the reporting rate of COVID-19 data under study is about 35%, indicating underreporting. When accounting for underreporting, the transmission rate was reduced by approximately 0.15. The quasi-Poisson distribution with the log function was the best at describing the distribution of the incidence cases. The study established that the COVID-19 incidence cases in Nigeria are underreported and follow a quasi-Poisson distribution.