Authors
Weijia Xiong, Liping Peng, Xiaotong Huang, Tim K Tsang
Published in
Royal Society open science. Volume 13. Issue 11. Pages 260996. Nov 04, 2026. Epub Nov 18, 2026.
Abstract
Timely monitoring of epidemics such as SARS-CoV-2 transmission is essential for guiding public health and social measures, yet real-time estimation of the time-varying effective reproductive number (Rt ) from case data suffers from right-censoring bias, systematically underestimating recent Rt values and requiring 7-14 days of additional data to stabilize. To address this limitation, we developed and evaluated a three-step framework that decouples case forecasting from Rt estimation. Daily PCR-confirmed case counts are first forecast up to 14 days ahead using an ensemble of 10 statistical, machine learning and deep learning models. Forecasted cases are then deconvolved to reconstruct infection timelines, from which Rt is estimated for up to 7 days ahead. This approach mitigates right-censoring artefacts by generating Rt estimates from projected rather than truncated case data. We applied this framework to Hong Kong's Omicron outbreaks (February to December 2022), incorporating 15 epidemiological, meteorological, air pollution and policy-related predictors. An unweighted ensemble of top-performing models substantially improved forecast accuracy, reducing the root mean square error and the mean absolute percentage error by 10 and 9% for case numbers and by 38 and 34% for Rt . By separating predictive modelling from epidemiological inference, this framework enables accurate real-time monitoring of epidemic burden and transmission dynamics to support timely public health decision-making.
PMID:
43048363
Bibliographic data and abstract were imported from PubMed on 18 Nov 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 1
- Comments 0