The estimation of treatment effects on medical costs and cost effectiveness measures is complicated by the need to account for non-independent censoring, skewness and the effects of confounders. In this dissertation, we develop several cost and cost-effectiveness tools that account for these issues. Since medical costs are often collected from observational claims data, we investigate propensity score methods such as covariate adjustment, stratification, inverse probability weighting and doubly robust weighting. We also propose several doubly robust estimators for common cost effectiveness measures. Lastly, we explore the role of big data tools and machine learning algorithms in cost estimation. We show how these modern techniques can be applied to big data manipulation, cost prediction and dimension reduction.