Predicting individual treatment effects

Sign in to queue


Treatments for complicated diseases often help some patients but not all and predicting the treatment effect of new patients is important in order to make sure every patient gets the best possible treatment. We propose model-based random forests as a method to detect similarities between patients with respect to their treatment effect and on this basis compute personalized models for new patients to obtain their individual treatment effect. The whole procedure focuses on a base model which usually contains the treatment indicator as a single covariate and takes the survival time or a health or treatment success measurement as primary outcome. This base model is used to grow the model-based trees within the forest as well as to compute the personalized models, where the similarity measurements enter as weights. We show how personalized models can be set up using the cforest() and predict.cforest() functions from the "partykit" package in combination with regression models such as glm() ("stats") or survreg() ("survival"). We apply the methods to patients suffering from Amyotrophic Lateral Sclerosis (ALS). The data are publicly available from and data preprocessing can be done with the R package "". The treatment of interest is the drug Riluzole which is the only approved drug against ALS but merely shows minor benefit for patients. The personalized models suggest that some patients benefit more from the drug than others.





Download this episode

The Discussion

Add Your 2 Cents