Abstract
In recent years, “real-world data (RWD)” has attracted much attention in the medical science. These data are expected to be used in a wide range of applications, such as personalized medicine and precision medicine. In the midst of this trend, the development of statistical methods for applying RWD is also progressing rapidly. The interest is in statistical inference on the effect of treatment on patients with arbitrary background information (covariates) (i.e., the difference between the outcome of the active and the standard regimen), i.e., the heterogeneous treatment effect (HTE). The statistical model for estimating HTE is the treatment effect model. Treatment effect models are being actively developed in the fields of statistical science and machine learning.
In this paper, we organized treatment effect models based on a typology of subgroup identification methods, and evaluated their performance through numerical simulation and case studies. In addition, variable importance and partial dependence were examined as graphical representations for the treatment effect model.
