To date, HIV has been responsible for over 35 million deaths. As of today, over 37.9 million people around the world live with HIV (PLHIV). Sub-Saharan Africa accounts for 25.7 million PLHIV, over 65% of current worldwide estimates. There exist different intervention strategies used to combat HIV. In this paper, we propose a framework to optimally allocate capital to those intervention strategies subject to budget constraints. The framework uses the Epidemic MODel (EMOD) model to evaluate the effectiveness of any suggested intervention policy and uses the second-order gradient-based Newton’s Method algorithm for the optimization. To improve the computational efficiency of our framework, we approximate the complex EMOD model using tensor trains. Our framework is adaptive and iterative. It suggests points in the parameter space to run the EMOD model to improve the accuracy of the surrogate model. As such we are guaranteed to converge to an optimum with respect to the full EMOD model and not with respect to the approximate surrogate model. That allowed us to optimize resources in both South Africa and Kenya. The results of our work conclude that, under a constrained budget, AntiRetroviral Therapy (ART) medication is the most effective investment in reducing HIV-related death. This is a more effective investment than other popular intervention strategies such as Pre-Exposure Prophylaxis (PrEP), Voluntary Male Medical Circumcision (VMMC), or Home Counseling and Testing (HCT) programs. However, our framework manages to effectively allocate funds at any appropriate budget to all the above intervention policies, and some more.

To date, HIV has been responsible for over 35 million deaths. As of today, over 37.9 million people around the world live with HIV (PLHIV). Sub-Saharan Africa accounts for 25.7 million PLHIV, over 65% of current worldwide estimates. There exist different intervention strategies used to combat HIV. In this paper, we propose a framework to optimally allocate capital to those intervention strategies subject to budget constraints. The framework uses the Epidemic MODel (EMOD) model to evaluate the effectiveness of any suggested intervention policy and uses the second-order gradient-based Newton’s Method algorithm for the optimization. To improve the computational efficiency of our framework, we approximate the complex EMOD model using tensor trains. Our framework is adaptive and iterative. It suggests points in the parameter space to run the EMOD model to improve the accuracy of the surrogate model. As such we are guaranteed to converge to an optimum with respect to the full EMOD model and not with respect to the approximate surrogate model. That allowed us to optimize resources in both South Africa and Kenya. The results of our work conclude that, under a constrained budget, AntiRetroviral Therapy (ART) medication is the most effective investment in reducing HIV-related death. This is a more effective investment than other popular intervention strategies such as Pre-Exposure Prophylaxis (PrEP), Voluntary Male Medical Circumcision (VMMC), or Home Counseling and Testing (HCT) programs. However, our framework manages to effectively allocate funds at any appropriate budget to all the above intervention policies, and some more.