iraca

iraca is focused on the simulation of contagion scenarios given by the interaction between mosquitoes and humans at individual levels, while allowing to include possible interventions and their aftermath. The package

iraca is a package built to model vector-borne diseases based on human-mosquito interaction. It relies on Agent-Based models to simulate individual behavior in the Colombian context. The library can be used to:

• Simulate mosquito-human interaction and possible aftermath given different contagion parameters and scenarios.
• Calibrate and obtain current parameters for simulation and analysis.
• Integrate policies and local/individual interventions to simulate their impact on vector-borne diseases.
• Support the validation process of intervention policies.

The current model is only fit for Dengue Virus. In addition, it is very important to note that the approach given by the package will be purely theoretical and should not be used for inference or immediate decision making without proper analysis.

iraca is developed at Universidad de Los Andes as part of the Epiverse-TRACE program.

Installation

You can install the development version of iraca from GitHub with:

# install.packages("pak")
pak::pak("epiverse-trace/iraca")

Motivation

Modern tools like Agent-Based Modeling (ABM) have been lately used to create robust simulations for epidemiological analysis; however, they are usually built for specific scenarios rather than a general approach. When analyzing vector-borne diseases like dengue, zika or chikungunya, part the process can be standardized since the vector is (usually) the same mosquito. Some available tools and libraries include a general ABM approach; however, they are still limited when modelling big volumes of data or very specific simulations. iraca solves the need for an ABM focused on Human-Mosquito interaction that’s specific enough to capture the behavior of a disease, but at the same time flexible to consider multiple scenarios.

This library is aimed at epidemiologists, researchers, local public health decision-makers, and anyone interested in the dynamics of vector-borne diseases.

Workflow

The user will interact with the library using a model object with the required information. Users will input direct information from the National Geostatistical Framework provided by the National Department of Statistics (DANE). This dataset should include the socioeconomic and demographic details for all municipalities of the country. In addition, climate data from the Institute of Hydrology, Meteorology and Environmental Studies should be included to simulate weather patterns. If looking for a fitted calibration, observed cases for the period of interest are required, which will be used to calculate the set of parameters that best fit the real data by minimizing RMSE. Finally, simulation can be run with or without interventions, either to recreate past events or evaluate future scenarios.

The simulation allows the differentiation of three types of agents: humans, mosquitoes and territories. All three are created from the input data stated above, and assigned different movement and behavioral patterns. Interventions can also be included in the simulation, including use of mosquito nets, insecticides and container cleaning.

To set up the model, we can use the function setup. Users may also indicate different incubation and infection periods, as well as initial parameters regarding infected population (humans and mosquitoes).

library(iraca)

# load example data, which is from the city of Ibague, Colombia
data(demographic_data_ibague)
data(temperature_data_ibague)

# Setup model
model <- setup(
  demographic_data = demographic_data_ibague,
  temperature_data = temperature_data_ibague,
  movement_data = NULL,
  demographic_data_MGN = TRUE,
  divipola_code = NULL,
  disease = "dengue",
  incubation_period = 4,
  infection_duration = 14,
  init_infected_humans = 0.005,
  init_infected_mosquitoes = 0.03
)

After setting up the model, we can start the model to make a simulation for a given amount of days (steps).

simulation(ABM_data = model,
           time = 100)

To allow larger and faster simulations, the model implementation relies on C++ and is liked to R through Rcpp. The library uses only S3 objects to make data input and output as simple as possible for the user.

Lifecycle

This package is currently a concept, as defined by the RECON software lifecycle. This means that essential features and mechanisms are still being developed, and the package is not ready for use outside of the development team.

After further consideration, this package development is temporarily halted. Usability issues were contemplated regarding the actual use in the local context given the lack of human resources and time constraints. In addition, ethical concerns were also taken into account, since the open release of very sensitive tools which rely on fine tuning and calibration could easily lead to poor evidence production for decision-making.

Contributions

Contributions are welcome via pull requests.

Code of Conduct

Please note that the iraca project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.