Providing sufficient testing capacities and accurate results in a time-efficient way are essential to prevent the spread and lower the curve of a health crisis, such as the COVID-19 pandemic. In line with recent research investigating how simulation-based models and tools could contribute to mitigating the impact of COVID-19, a discrete event simulation model is developed to design optimal saliva-based COVID-19 testing stations performing sensitive, non-invasive, and rapid-result RT-qPCR tests processing. This model aims to determine the adequate number of machines and operators required, as well as their allocation at different workstations, according to the resources available and the rate of samples to be tested per day. The model has been built and experienced using actual data and processes implemented on-campus at the University of Illinois at Urbana-Champaign, where an average of around 10,000 samples needed to be processed on a daily basis, representing at the end of August 2020 more than 2% of all the COVID-19 tests performed per day in the USA. It helped identify specific bottlenecks and associated areas of improvement in the process to save human resources and time. Practically, the overall approach, including the proposed modular discrete event simulation model, can easily be reused or modified to fit other contexts where local COVID-19 testing stations have to be implemented or optimized. It could notably support on-site managers and decision-makers in dimensioning testing stations by allocating the appropriate type and quantity of resources.