Managing HIV treatment in resource-limited and dynamic environments

Containing the HIV epidemic is one of the most pressing global health care issues.\ud Antiretroviral therapy, the only treatment option for chronic HIV, inhibits the progression of the disease.\ud However, there is a severe shortage of treatment in the developing world, particularly in Sub-Saharan Africa, the area hit the hardest by the epidemic.\ud The current guidelines recommend treating HIV patients until death, known as a nonabandonment policy.\ud HIV-infected patients develop resistant mutations and they benefit marginally from treatment. Therefore, there is an opportunity cost for treating patients until they die.\ud We estimate the price of nonabandonment policies in HIV treatment where resources are limited.\ud We develop a mathematical framework to optimize the allocation of scarce HIV treatment for a broad class of admissible policies.\ud Pursuant to this goal, we develop a Markov model of the progression of a population of susceptible and infected individuals.\ud Then, we restrict our attention to two classes of admissible policies: (i) nonabandonment policies, and (ii) abandonment-permitted policies.\ud The price of nonabandonment policies is estimated by the difference between the optimal solution of these two classes of admissible policies.\ud Since the state spaces of the models are unbounded, solving the allocation problems is intractable. Therefore, we approximate the price of nonabandonment policies by the difference between a lower bound on the best performance of allocation policies in abandonment-permitted settings and an upper bound of that in nonabandonment settings.\ud We show that the price of following the nonabandonment policies in HIV treatment is as much as 41%.\ud Moreover, they shed light on the key role allocation policies play in containing the epidemic.\ud In resource-rich environments, when to start HIV treatment is a fundamental question.\ud Current models do not consider the rate of new antiretroviral development in their analysis.\ud We model the arrival of HIV pipeline drugs in resource-rich environments as a split Poisson process and incorporate it in a validated simulation model to measure the effect of HIV pipeline drugs on HIV treatment.\ud The model with the inclusion of pipeline drugs recommends earlier treatment