Dynamic-spread assessment and kink detection in the early mpox outbreak

Abstract

This paper proposes a varying coefficient Susceptible-Exposed-Infected-Removed (vSEIR) model to dynamically simulate the early mpox epidemic. We incorporate a time-varying infection rate and smallpox vaccination protection to capture real-time changes in transmission influenced by non-pharmaceutical interventions, setting our work apart from studies relying on fixed rates. To this end, we apply the recursive least squares algorithm with a forgetting factor for real-time identification of time-varying infection rate. The sparse Hodrick-Prescott (HP) filter, tuned with leave-one-out cross-validation, captures mpox epidemic kinks via the effective reproduction number $R_t$ obtained from the discrete vSEIR model. This allows for accurate segmentation of epidemic phases, better evaluation of intervention effectiveness, and insights that can guide preparedness for future possible outbreaks. We analyze the mpox and COVID-19 outbreaks in four countries using the proposed kink-based framework. The results show that the mpox epidemic generally entered its decline phase earlier than COVID-19, despite weaker interventions. Additionally, the early mpox epidemic exhibited more inflection points compared to the early COVID-19 pandemic, reflecting stronger non-pharmacological controls during the latter. Sensitivity analyses further indicate that mpox infections would have increased by 12% without smallpox vaccination, and data uncertainty significantly impacts $R_t$ estimates. Finally, our proposed systematic framework can also be extended to other early outbreaks of human-to-human epidemics, especially in the absence of reliable medical data on contact rates.