Identifying real-time surveillance indicators that can be used to estimate COVID-19 hospital admissions

Abstract

AbstractNumerous approaches have been used to track COVID-19 trends, from wastewater surveillance to laboratory reporting of diagnostic test results. However, questions remain about how best to focus surveillance efforts during and after public health emergencies. Using an archive of SARS-CoV-2 surveillance data, we reconstructed seven real-time surveillance indicators and assessed their performance as predictors of 7-day moving average COVID-19 hospital admissions in Colorado from October 2020 to March 2024. Models were constructed using neural network models and Ordinary Least Squares regression. We found that hospital census data, emergency-department based syndromic surveillance, and daily COVID-19 hospital admissions were the best indicators of COVID-19 hospital demand in Colorado during the public health emergency (PHE) (October 2020 – May 2023) and after (May 2023 – March 2024). The removal of wastewater from our multi-indicator models resulted in a decrease in model performance, indicating that wastewater provides important and unique information. By contrast, capacity to predict COVID-19 hospital admissions was not meaningfully reduced when sentinel test positivity, statewide test positivity, and/or case report data were dropped from our prediction models during and after the PHE. These findings underscore the importance of hospital-based reporting for monitoring COVID-19 hospital admissions, and, conversely, suggest that case reporting and percent positivity are not essential to monitoring COVID-19 hospitalizations in Colorado.