Nowcasting epidemic trends using hospital- and community-based virologic test data

Area of Science:

• Epidemiology
• Virology
• Public Health Surveillance

Background:

• Traditional COVID-19 surveillance relies on case counts and hospitalizations.
• Reverse transcription quantitative polymerase chain reaction (RT-qPCR) cycle threshold (Ct) values offer a potential alternative for monitoring epidemic trends.
• The strengths, limitations, and statistical power of Ct values for real-world epidemic surveillance remain underexplored.

Purpose of the Study:

• To explore the utility of SARS-CoV-2 viral load Ct values for epidemic nowcasting.
• To quantify biological and logistical factors influencing Ct-based epidemic surveillance accuracy.
• To assess the predictive power of Ct values compared to traditional metrics.

Main Methods:

• Utilized SARS-CoV-2 RT-qPCR data from hospital and municipal testing.
• Employed a combination of theoretical analysis and simulation studies.
• Fitted generalized additive models to predict epidemic growth rates and direction using Ct value distributions.
• Assessed nowcasting accuracy over two-week windows using metrics like RMSE and AUC.

Main Results:

• Changes in peak viral load, growth/clearance rates, and sampling logistics significantly impact the Ct value-growth rate relationship.
• Models accurately predicted epidemic trends in synthetic data (RMSE: 0.0192, AUC: 0.910).
• Modest accuracy was observed with real-world data (RMSE: 0.039-0.052, AUC: 0.72-0.80), with outlier trimming improving performance.

Conclusions:

• Ct values can complement traditional incidence metrics for SARS-CoV-2 surveillance.
• The accuracy of Ct-based nowcasting is influenced by various biological and logistical factors.
• While real-world performance is modest, Ct values provide valuable insights into epidemic dynamics.