Treatment-Response Models for Counterfactual Reasoning with Continuous-time, Continuous-valued Interventions
This work addresses a critical problem in causal inference for healthcare and policy, enabling more precise counterfactual reasoning with continuous-time interventions, though it is incremental as it builds on existing Gaussian Process frameworks.
The paper tackles the challenge of estimating potential outcomes from observational data when treatment doses vary continuously over time and outcomes are irregularly measured, by modeling treatment response curves using linear time-invariant dynamical systems and multiple-output Gaussian Processes. It demonstrates significant accuracy gains over state-of-the-art models on simulated and clinical datasets.
Treatment effects can be estimated from observational data as the difference in potential outcomes. In this paper, we address the challenge of estimating the potential outcome when treatment-dose levels can vary continuously over time. Further, the outcome variable may not be measured at a regular frequency. Our proposed solution represents the treatment response curves using linear time-invariant dynamical systems---this provides a flexible means for modeling response over time to highly variable dose curves. Moreover, for multivariate data, the proposed method: uncovers shared structure in treatment response and the baseline across multiple markers; and, flexibly models challenging correlation structure both across and within signals over time. For this, we build upon the framework of multiple-output Gaussian Processes. On simulated and a challenging clinical dataset, we show significant gains in accuracy over state-of-the-art models.