Clinical treatment evaluation based on real-world data often requires adjusting for population differences in order to draw meaningful inference. This problem is considered in the context of estimating mean outcomes and treatment effects in a well-defined target population using clinical data from a study population that differs from but overlaps with the target population in terms of patient characteristics. The current literature includes a variety of statistical models which generally require the correct specification of at least one parametric regression model. In this work, we propose the use of machine learning methods to estimate nuisance functions, incorporating these methods into existing doubly robust estimators. The resulting nonparametric estimators are $\sqrt n$-consistent, asymptotically normal, and asymptotically efficient under general conditions. Simulation results demonstrate that the proposed methods perform well in reasonable settings. These methods are also illustrated with a concrete cardiology example concerning standard of care for aortic stenosis. Finally, the ignorability assumption is examined through the development of global sensitivity analysis methods for two of the commonly used parametric approaches.