Bayesian Estimation of COVID-19 Under-Reporting with Time-Varying Rates: A Case Study of Morocco

Abstract

Accurate estimation of COVID-19 cases remains challenging in many low and middle-income countries due to substantial under-reporting driven by limited testing capacity, asymptomatic infections, and surveillance system constraints. This study develops an advanced Bayesian hierarchical framework to estimate the true scale of the COVID-19 pandemic in Morocco while explicitly modeling time-varying case detection and seasonal effects. Using national surveillance data from March 2020 to June 2022, we estimate a time-varying reporting rate increasing from 10.7% to 11.5% throughout the pandemic, indicating significant improvements in surveillance capacity. The model reveals that the actual epidemic was substantially larger than official reports suggested, with a final attack rate of 30.6% (95% HDI: 24.9--37.3%) compared to the reported 3.2%.. Our methodological innovations include temporal reporting rate estimation and seasonal component integration, providing a more realistic framework for epidemic assessment in settings with evolving surveillance systems.