Causal Transformer for Estimating Counterfactual Outcomes
This work addresses the challenge of handling complex, long-range dependencies in counterfactual outcome estimation for applications like personalized medicine, representing an incremental advance by adapting transformer architectures to this domain.
The paper tackles the problem of estimating counterfactual outcomes over time from observational data, such as in personalized medicine, by developing a Causal Transformer that captures complex, long-range dependencies and achieves superior performance over current baselines on synthetic and real-world datasets.
Estimating counterfactual outcomes over time from observational data is relevant for many applications (e.g., personalized medicine). Yet, state-of-the-art methods build upon simple long short-term memory (LSTM) networks, thus rendering inferences for complex, long-range dependencies challenging. In this paper, we develop a novel Causal Transformer for estimating counterfactual outcomes over time. Our model is specifically designed to capture complex, long-range dependencies among time-varying confounders. For this, we combine three transformer subnetworks with separate inputs for time-varying covariates, previous treatments, and previous outcomes into a joint network with in-between cross-attentions. We further develop a custom, end-to-end training procedure for our Causal Transformer. Specifically, we propose a novel counterfactual domain confusion loss to address confounding bias: it aims to learn adversarial balanced representations, so that they are predictive of the next outcome but non-predictive of the current treatment assignment. We evaluate our Causal Transformer based on synthetic and real-world datasets, where it achieves superior performance over current baselines. To the best of our knowledge, this is the first work proposing transformer-based architecture for estimating counterfactual outcomes from longitudinal data.