Efficient reporting delay calibration in spatial metapopulation models for reconstructing cross-regional epidemic dynamics

Le résumé fourni par la source

• Scientific question. • Accurate reconstruction of early-stage cross-regional epidemic dynamics is critically hindered by reporting delays, heterogeneous surveillance systems, and limited case observations. How to efficiently calibrate reporting delays and infer key epidemiological parameters in large-scale spatial metapopulation models remains an open challenge. • Evidence before this study. • Previous spatial metapopulation models have been widely applied to reconstruct epidemic spread and estimate transmission parameters. However, existing approaches often rely on computationally intensive delay compensation schemes or highly detailed model structures, leading to scalability limitations and reduced practicality for early outbreak analysis under sparse and delayed surveillance data. Particle-based Bayesian inference methods provide accurate estimation but are frequently constrained by high computational cost in multi-regional systems. • New findings. • In this study, we developed an efficient computational framework integrating a simplified stochastic metapopulation model, a novel matrix-based reporting delay calibration algorithm, and a scalable particle Markov chain Monte Carlo parameter estimation architecture. The proposed delay calibration method replaces nested-loop operations with batch matrix computation, substantially improving computational efficiency. Validation on synthetic networks and empirical Chinese COVID-19 mobility data demonstrates robust and accurate reconstruction of early epidemic dynamics, with more than a tenfold acceleration in parameter estimation for large-population cities. • Significance of the study. • This study provides a practical and scalable solution for rapid reconstruction of early-stage epidemic dynamics under reporting delays and data scarcity. By balancing epidemiological fidelity and computational efficiency, the proposed framework enhances the operational applicability of spatial metapopulation models and offers valuable methodological support for real-time epidemic assessment and public health decision-making during emerging infectious disease outbreaks. Reconstructing the early spatiotemporal dynamics of emerging infectious diseases (EIDs) is essential for effective public health response but remains difficult due to reporting delays, heterogeneous surveillance systems, and cryptic transmission chains. This study proposes a systems-oriented computational framework that tackles these challenges through three key innovations. First, we develop a stochastic infectious disease model tailored to limited early-stage case counts, grounded in a simplified metapopulation structure that enables accurate reconstruction of initial outbreak conditions while maintaining computational efficiency comparable to existing methods. Second, we introduce a matrix-based algorithm for calibrating reporting delays in spatial metapopulation models. By leveraging matrix operations to synchronize case-report updates across multiple regions, the method eliminates the need for traditional iterative traversal, thereby achieving substantial gains in computational efficiency and improving its practical utility in engineering applications. Third, leveraging complex network theory, we develop a parameter estimation framework using open-source algorithm libraries from the Medical Research Council Centre for Global Infectious Disease Analysis (MRC GIDA), achieving more than a tenfold increase in estimation efficiency for individual cities with populations exceeding one million. Validation using both simulated networks and empirical Chinese urban mobility networks covering early coronavirus disease 2019 (COVID-19) transmission scenarios demonstrates that the proposed approach substantially improves parameter estimation efficiency while ensuring robustness and accuracy. This framework provides a powerful tool for rapid, high-fidelity reconstruction of epidemic dynamics, enabling more informed responses to future public health emergencies.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.