Behavior-Aware COVID-19 Forecasting Using Markov SIR Models on Dynamic Contact Networks: An Observational Modeling Study

Background: Traditional compartmental models for infectious diseases, such as COVID-19, often assume static contact patterns and overlook behavioral responses, limiting their ability to predict epidemic dynamics accurately. Methods: We developed a novel framework integrating time-varying contact networks, Markov chain-based Susceptible-Infectious-Recovered (SIR) dynamics, and adaptive behavioral responses. Populations are modeled as dynamic networks, with nodes representing individuals and edges reflecting social interactions that evolve based on mobility data and random updates. Infection rates adjust dynamically via an exponential feedback function tied to epidemic intensity, capturing behaviors like social distancing. The model was validated using non-military COVID-19 case data (2020-2021) from Ghana, India, and the United States, with random (Erdős-Renyi) and scale-free network structures. Results: Behavioral feedback reduced peak infection rates by 10-15%. Scale-free networks, which mimic real-world hubs like urban centers, predicted larger epidemics due to super-spreader effects. Interventions, such as a 50% reduction in transmission probability, decreased epidemic size by up to 45%. Simulations closely matched observed case data in timing and magnitude of epidemic peaks across datasets. Conclusion: Integrating dynamic contact networks and behavioral feedback enhances epidemic forecasting accuracy. This computationally efficient framework provides actionable insights for targeted public health interventions, such as vaccination prioritization and mobility restrictions.