Low-Barrier Access and Monitoring for HIV PrEP

This repository holds the model code for:

Le Guillou AV, Marcus JL, Krakower DS, Violette LR, Jenness SM. Potential risks and benefits of low-barrier access and monitoring for HIV pre-exposure prophylaxis: a modeling study. J Infect Dis. 2026; 234(4): 519-529. https://doi.org/10.1093/infdis/jiag369

It contains the R scripts for network estimation, model calibration, the over-the-counter PrEP scenarios, and the processing that produced the paper's tables and figures, along with the model parameter file (data/input/model_parameters.csv) and the lockfiles that record package versions (renv.lock for local runs, renv.lock.hpc for the cluster). Simulation output is not included. The work was funded by NIH/NIMH R21 MH136855 (OFFSCRIPT).

Engine

The model runs on EpiModelHIV. The version used for the paper is this release: https://github.com/EpiModel/EpiModelHIV/releases/tag/LowBarrier-JID

remotes::install_github("EpiModel/EpiModelHIV@LowBarrier-JID")

Setup

Open the project in R 4.5 and restore the packages with renv:

renv::restore(exclude = "ARTnetData")

ARTnetData is a private package and is needed only for network estimation (step A below). Every later step reads the fitted network objects that step A saves. The paper's networks were estimated with ARTnet 2.9.0, which is no longer public, so the lockfiles install ARTnet 2.9.3, the first version in the public ARTnet repository. Networks re-estimated with it may not match the paper's exactly.

Running the pipeline

Scripts are run from the project root. Each step folder has a z-context.R that switches between a small local model and the full model on a SLURM cluster (set hpc_context <- TRUE). For the cluster, edit mail_user and the node setup in R/hpc_configs.R.

1. Networks (R/A-networks/). 1-estimation.R fits the partnership network models from ARTnet (needs ARTnetData), and 2-diagnostics.R and 3-assess.R check them. workflow-networks.R runs both on the cluster.
2. Calibration (R/Z-calibration/). workflow-swfcalib.R calibrates the model to the Atlanta targets with swfcalib (swfcalib_config.R defines the waves), and workflow-restart_point.R with choose_restart.R saves the restart point. finalize_params.R writes the calibrated values to data/input/model_parameters.csv. R/z-gfr_optim.R fits the starting distribution for the renal function module.
3. Intervention scenarios (R/F-intervention_scenarios/). The scenarios are built in code (utils-scenarios.R and the 0*-make_scenarios*.R scripts) and run from the restart point by the workflow-intervention*.R scripts: workflow-intervention.R for the main scenarios, workflow-intervention_plots.R for the coverage grid behind the contour plots, workflow-intervention_find.R for the search for target coverage levels, and workflow-intervention_indics.R for the sensitivity analyses. 1-scenarios.R runs a small version locally.
4. Tables and figures (R/F-intervention_scenarios/). 2-process_tables.R and 2-process_plot_data.R summarize the merged simulations, and 3-contour_plot.R, 4-filter_csv_tables.R, and 6-tables.R build the figures and tables in data/output/.

Source

This is a snapshot of the private project repository, taken July 9, 2026, at the revision submitted to the journal. Its code matches the private repository at commit 30dc8a0, apart from two added lines in 3-contour_plot.R that save the contour figure as TIFF and PDF.