Modern Methods in Precision Medicine

As the concept of precision medicine spreads, there is a growing need for developing better algorithms that a) are sample efficient (i.e., require fewer samples to achieve the same accuracy level), b) think beyond association (to identify the causation hidden in the data), and c) provide insights to medical practice. In this dissertation, we investigate various problems in precision medicine, the topics ranging from opioid use disorder (OUD) and cancer treatment, to sickle cell disease (SCD). We leverage tools from stochastic learning, causal inference, and machine learning, with the objective of reducing healthcare expenditure and improving the quality of care.