We consider the problem of learning an optimal prescriptive tree (i.e., a personalized treatment assignment policy in the form of a binary tree) of moderate depth, from observational data. This problem arises in numerous socially important domains such as public health and personalized medicine, where interpretable and data-driven interventions are sought based on data gathered in deployment, through passive collection of data, rather than from randomized trials. We propose a method for learning optimal prescriptive trees using mixed-integer optimization (MIO) technology. We show that under mild conditions our method is asymptotically exact in the sense that it converges to an optimal out-of-sample treatment assignment policy as the number of historical data samples tends to infinity. This sets us apart from existing literature on the topic which either requires data to be randomized or imposes stringent assumptions on the trees. Based on extensive computational experiments on both synthetic and real data, we demonstrate that our asymptotic guarantees translate to significant out-of-sample performance improvements even in finite samples.
Technical Report, University of Southern California, August 2021