Application of Causal Inference Methods to Improve Treatment of HIV Infection

Advances in causal inference have improved the ability of statistical methods to answer practical questions in epidemiology and clinical research. In particular, the counterfactual framework and marginal structural models have provided the basis for practical and theoretical advances in the statistical estimation of causal parameters. The dissertation focuses on the development and application of statistical methods based on the counterfactual framework. Three methods are presented and each is applied to answer a practical research question. The presentation of each method is aimed at a distinct audience: clinicians, epidemiologists, and statisticians, respectively. The unifying theme is the use of state-of-the-art causal inference methods to answer questions surrounding the treatment of antiretroviral resistant HIV infection. Specific topics include use of point treatment marginal structural models to investigate antiretroviral salvage strategies, use of history-adjusted marginal structural models to study the effects of non-suppressive antiretroviral therapy, and application of a new methodology for the estimation of direct causal effects to compare antiretroviral regimens.