Presentation description

The focus of this project is to model and analyze how infections spread in hospital settings using a combination of agent-based simulation and empirical dynamic modeling (EDM). The main goal is to explore whether transmission events, colonization prevalence, and clinical detections generated from simulations exhibit nonlinear dynamical behavior, which has important implications for forecasting infectious disease outbreaks.

I used Repast Simphony, a Java-based simulation platform, to create an agent-based model of a hospital. Patients are admitted (either colonized or susceptible) and discharged based on defined parameters like admission rate, importation probability (probability of being colonized at time of admission) and discharge rate. Inside the hospital, patients can become colonized through transmission, become detected through clinical or surveillance testing, and be isolated. Key parameters such as importation probability, mean time to detection, and transmission rate influence how infections spread.

The simulation produces time series data (e.g., daily/ weekly/ monthly colonization counts, clinical detections, and transmissions), which we analyze using rEDM, a toolkit for empirical dynamic modeling. We apply Simplex and S-Map to assess nonlinear behavior, test forecast skill (ρ), and examine accuracy using different embedding dimensions (E) and Time offset of embedding (tau). We also systematically vary library (lib) and prediction (pred) ranges to assess how they affect predictability and dynamic structure.

This approach allows us to:

Examine how changes in key parameters affects transmission, clinical detection and colonization patterns.

Assess whether the simulation-generated time series can support nonlinear forecasting methods.

Explore how varying time settings, such as library/prediction ranges and time delays (tau)-influence prediction accuracy

Simulation and EDM together provide a strong approach for understanding and improving infection control in real-world hospital settings.