A Decision Analytics Model to Optimize Investment in Interventions Targeting the HIV PrEP Cascade of Care
This repository holds the source to code to reproduce the analysis featured in our HIV transmission model evaluating the impact of three clinical HIV PrEP interventions on HIV incidence among men who have sex with men.
Jenness SM, Knowlton G, Smith DK, Marcus JL, Anderson EJ, Siegler AJ, Jones J, Sullivan PS, Enns E. A decision analytics model to optimize investment in interventions targeting the HIV preexposure prophylaxis cascade of care. AIDS. 2021; 35(9): 1479-1489. doi:10.1097/QAD.0000000000002909
The EpiModelHIV engine for this paper is release Optim-AIDS of EpiModel/EpiModelHIV. Install it with:
remotes::install_github("EpiModel/EpiModelHIV@Optim-AIDS")ARTnetData is a private package. It is needed only to re-estimate the networks. The paper's networks were estimated with ARTnet 1.0.0, which is no longer public, so renv.lock installs ARTnet 2.9.3, the first version in the public ARTnet repository. Networks re-estimated with it may not match the paper's exactly.
Gaps between recommended and actual levels of HIV preexposure prophylaxis (PrEP) use remain among men who have sex with men (MSM). Interventions can address these gaps, but it is unknown how public health initiatives should invest prevention funds into these interventions to maximize their population impact.
We used a stochastic network-based HIV transmission model for MSM in the Atlanta area paired with an economic budget optimization model. The model simulated MSM participating in up to three real-world PrEP cascade interventions designed to improve initiation, adherence, or persistence. The primary outcome was infections averted over 10 years. The budget optimization model identified the investment combination under different budgets that maximized this outcome given intervention costs from a payer perspective.
From the base 15% PrEP coverage level, the three interventions could increase coverage to 27%, resulting in 12.3% of infections averted over 10 years. Uptake of each intervention was interdependent: maximal use of the adherence and persistence interventions depended on new PrEP users generated by the initiation intervention. As the budget increased, optimal investment involved a mixture of the initiation and persistence interventions, but not the adherence intervention. If adherence intervention costs were halved, the optimal investment was roughly equal across interventions.
Investments into the PrEP cascade through initiatives should account for the interactions of the interventions as they are collectively deployed. Given current intervention efficacy estimates, the total population impact of each intervention may be improved with greater total budgets or reduced intervention costs.
These models are written and executed in the R statistical software language. To run these files, it is necessary to use the correct version of our epidemic modeling software, EpiModel, and our extension package specifically for modeling HIV/STI transmission dynamics among MSM, EpiModelHIV.
In R, load the necessary packages with the following command:
install.packages("renv")
renv::init()
Once renv has finished initializing, restart R.
To restore the full package set from renv.lock, you will need a GitHub personal access token with access to the private ARTnetData repository. It should be set either in "~/.bashrc":
export GITHUB_PAT="<your github private access token>"
or in "~/.Renviron":
GITHUB_PAT="<your github private access token>"
