Learning Treatment Allocations with Risk Control Under Partial Identifiability
This addresses the challenge of minimizing unnecessary harm from treatments with adverse side effects in patient populations, offering a solution for partially identifiable scenarios.
The paper tackles the problem of learning beneficial treatment allocations in precision medicine while controlling treatment risk, which is not generally identifiable from data, and proposes a certifiable learning method that achieves risk control with finite samples in partially identified settings, as demonstrated on simulated and real data.
Learning beneficial treatment allocations for a patient population is an important problem in precision medicine. Many treatments come with adverse side effects that are not commensurable with their potential benefits. Patients who do not receive benefits after such treatments are thereby subjected to unnecessary harm. This is a `treatment risk' that we aim to control when learning beneficial allocations. The constrained learning problem is challenged by the fact that the treatment risk is not in general identifiable using either randomized trial or observational data. We propose a certifiable learning method that controls the treatment risk with finite samples in the partially identified setting. The method is illustrated using both simulated and real data.