Multi-Source Agent-Based Modeling to Optimize Influenza Mitigation Strategies in Hong Kong
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Seasonal influenza control faces significant challenges from variable vaccine effectiveness and uncertain non-pharmaceutical intervention (NPI) performance across diverse contexts. While vaccination remains the primary strategy, effectiveness varies substantially between vaccine-matched and mismatched seasons. The COVID-19 pandemic has increased public acceptance of NPIs such as staying home when sick and mask use, enhancing feasibility for influenza control, yet optimal combination approaches remain poorly understood. We used an agent-based model to analyze influenza transmission across six seasons in Hong Kong (2009-2013) with varying vaccine effectiveness and epidemic characteristics. We integrated surveillance, serological, and school absenteeism data for calibration, enabling accurate estimation of reported and unreported infections. We evaluated age-targeted vaccination, staying home when sick, mask use, and school-based interventions across diverse real-world scenarios. Comparing with actual coverage levels, child vaccination consistently outperformed other strategies, with targeting those under 12 yielding the greatest population-level attack rate reduction (up to 2.17 percentage points absolute reduction and 8.5% relative reduction per 100,000 vaccinated individuals with coverage increases up to 50 percentage points). Among NPIs, 40% mask coverage reduced attack rates by 18-45%, comparable to 25% of symptomatic individuals staying home (17-49% reduction), though mask effectiveness remained stable while staying home declined with higher asymptomatic proportions. During vaccine-mismatched seasons, combining high-coverage mask use and staying home can reduce attack rates by 77-84%. School-based vaccination at high coverage was more effective than closures, reducing student attack rates by up to 86% versus 31% for 14-day closures. Our multi-source calibration approach provides robust evidence for prioritizing child vaccination and strategic NPI combinations.