A Shiny app visualizer for agent-based model in R simulating rural breast cancer screening uptake, incorporating race, poverty, insurance, and access-to-care dynamics.
This repository contains code for an app that visualizes an agent-based model (ABM) implemented in R to simulate breast cancer screening behaviors in rural populations. The model explores how race, poverty, insurance status, and geographic access interact to influence screening intentions and uptake over time.
- Generates synthetic rural social networks using
igraph - Assigns sociodemographic attributes to agents:
- Race (White vs. Racialized Minority)
- Poverty status
- Insurance status
- Distance from care (>25 miles vs. <25 miles)
- Models initial screening intentions using a logistic regression–style scoring system
- Simulates screening uptake dynamics over 24 time steps
- Screening intention rises when ≥3 connected neighbors are screened
- Agents with screening intention ≥0.9 may get screened (with stochastic noise)
- Supports multiple model runs and scenario-based experimentation via user-supplied parameter data frames
- Outputs:
- Simulation logs (aggregate counts over time, by race)
- Node histories (agent-level screening status and timing)
Full credit for the ABM goes to Dr. Jennifer Cruz who led and owns this project, and who hired me as a collaborator (https://service.harvard.edu/people/jen-cruz). Due to the publication being in progress, Dr. Cruz's code has been witheld from this repository, and only the app code remains.
The Shiny app provides a user-friendly interface with sliders and presets to configure simulations. Example outputs include:
- Cumulative screening uptake by time step
- Proportion of newly screened agents by run
