Estimating COVID-19 Incidence During the Pre-Vaccination Era in North Carolina Counties by Leveraging Spatial Characteristics (2023)
Undergraduate: Cindy Pang
Faculty Advisor: Paul Delamater
Department: Biostatistics, Carolina Population Center, Geography
During the COVID-19 pandemic, variable testing availability across space and time has posed challenges to understanding the true magnitude of COVID-19 infections, resulting in an underestimate of the true number of infections due to SARS-CoV-2. To address these issues, we developed a model to estimate COVID-19 infection incidence in North Carolina counties from March 2020 to January 2021, using a back-casting method which accounts for time lags in reporting and age-specific disease parameters. Ensemble models were used as comparisons, and the Root Mean Squared Error (RMSE) and Spearman's Rank Correlation Test were used to measure the magnitude of error and temporal similarity between the models. Model results were mixed. Our model exhibited strong temporal similarity with the Death Infection Estimates model but was not robust in accounting for the magnitude of error, with high RMSE values across all ensemble models. Some North Carolina counties had more robust models. The results highlight the challenges of estimating COVID-19 infections, particularly when data sources are limited.
Link to Poster