Bayesian imputation of COVID-19 positive test counts for nowcasting under reporting lag

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Reporting delays in COVID-19 test results hinder real-time infection monitoring in the UK.
• Delays of up to five days for 'Pillar 2' swab tests complicate timely data collation for symptomatic individuals.

Purpose of the Study:

• To develop a statistical temporal model for inferring total COVID-19 infection counts from partial, incoming data.
• To provide a smoothed, nowcasted time-series of expected daily positive tests with uncertainty estimates.

Main Methods:

• A Bayesian statistical temporal model was employed, utilizing the stability of under-reporting over time.
• The model incorporates subjective priors and a hierarchical structure for latent infection intensity.
• Sequential Monte Carlo methods were used for inference.

Main Results:

• The model generates a nowcasted time-series of daily COVID-19 positive test counts.
• Uncertainty bands are provided, enhancing the utility of the nowcasted data for decision support.
• The approach effectively addresses the challenge of reporting lags in infection data.

Conclusions:

• The developed statistical model offers a reliable method for nowcasting UK COVID-19 infections despite reporting lags.
• Smoothed time-series data with uncertainty estimates can significantly aid public health decision-making.
• This approach improves the timeliness and accuracy of epidemiological surveillance for infectious diseases.