Modeling Gonorrhea Transmission in Baltimore and San Francisco

The gcRegional R package implemented here defines a model and analyzes outcomes from this model for the purpose of providing simulation model evidence for the forthcoming research manuscript The potential population-level impact of different gonorrhea screening strategies in Baltimore and San Francisco: an exploratory mathematical modeling analysis.

Authorship

The author list includes:

Minttu M. Rönn^#*1, Christian Testa^#1, Ashleigh R. Tuite¹, Harrell W. Chesson², Thomas L. Gift², Christina Schumacher^3,4, Sarah L. Williford^3,4, Lin Zhu¹, Meghan Bellerose¹, Rebecca Earnest¹, Yelena Malyuta¹, Katherine K. Hsu⁵, Joshua A. Salomon⁶, Nicolas A. Menzies¹

^# Contributed equally
^* Corresponding author

Affiliations:

1. Department of Global Health, Harvard T. H. Chan School of Public Health
2. Centers for Disease Control and Prevention, Division of STD Prevention
3. Johns Hopkins University School of Medicine
4. Baltimore City Health Department
5. Division of STD Prevention & HIV/AIDS Surveillance, Massachusetts Department of Public Health
6. Center for Health Policy / Center for Primary Care and Outcomes Research, Stanford University

Funding Source

This work was funded by the U.S. Centers for Disease Control and Prevention, National Center for HIV, Viral Hepatitis, STD, and TB Prevention Epidemiologic and Economic Modeling Agreement (NEEMA, #5U38PS004644)

Data Sources

In this project we used data from the NHANES, NSFG, and SSuN data surveys.

Methodology

The gcRegional R package implements a deterministic compartmental model of gonorrhea transmission.

The approach taken here is to write the simulation module (i.e. a differential equations solver) in C++ accessible through an Rcpp interface in this package. This simulation model is calibrated to the observed data, and using calibrated models we are able to estimated the impact of changes in screening patterns on the fit models' future outcomes.

We take a Bayesian modeling approach to calibrate our model; We formally construct the dLogPosterior function and estimate the parametric uncertainty of our model using an adaptive MCMC algorithm.

Installing the Package

Installation Prerequisites

One must have Rcpp installed and configured correctly in order to install and use gcRegional.

Install gcRegional

Installing this package will allow the user to run the model and to produce plots such as those we have included in the manuscript and in this package.

Install the package and its dependencies using library(devtools); install_github("PPML/gc-regional.git", dependencies=T).

Cloning the Project for Calibration

Fitting the model and storing the outcomes of the calibration process is functionality that we advice one does by cloning this git repository into a local directory.

This is because the code to store calibrated results uses the here package in the context of the gcRegional package to store results in the inst/ directory to make them available in the package.

Running the Model

library(gcRegional) # load our package

# Use the load_start function to load the model configurations for either
# Baltimore or San Francisco.
load_start('BA') # or 'SF'

# The parametrization for the model in the selected location is given by 
# parameters in the environment `gc_env`.

# > str(gc_env$theta)
#  Named num [1:106] -0.0306 -0.7014 -0.0688 -0.9072 -0.1881 ...
#  - attr(*, "names")= chr [1:106] "logit.epsilon.1" "logit.epsilon.2" "logit.epsilon.3" "logit.epsilon.4" ...

e <- create_gcSim_param_env(theta=gc_env$theta)
run_gcSim_with_environment(e)