Abstract: 

Reliably estimating the individualized effects of treatments is crucial for decision making in the healthcare setting. Machine learning methods for causal inference can be used to leverage the increasingly available patient observational data, such as electronic health records (EHRs), in order to estimate heterogeneous treatment effects. In this context, several causal inference methods have been developed to estimate the effects of binary treatments in the static, cross-sectional setting. Nevertheless, estimating the effects of treatments over time poses unique opportunities such as understanding how diseases evolve under different treatment plans, which are optimal timings for assigning treatments, but also how individual patients respond to medication over time.

This talk will highlight the challenges that arise when using longitudinal patient observational data for causal inference. Then, it will introduce the Counterfactual Recurrent Network [1], a novel sequence-to-sequence model that employs the recent advances in representation learning and domain adversarial training to overcome the problems of existing methods for causal inference over time. In addition, the talk will describe the Time Series Deconfounder [2], a method that enables the estimation of treatment effects over time in the presence of hidden confounders. Finally, the talk will outline future research directions that could lead to achieving the full potential of utilizing electronic health records and machine learning methods for causal inference to make individualized treatment recommendations.

[1] Bica, I., Alaa, A.M., Jordon, J. & van der Schaar, M. ""Estimating counterfactual treatment outcomes over time through
adversarially balanced representations"" International Conference on Learning Representations (ICLR) 2020
[2] Bica, I., Alaa, A.M. & van der Schaar, M. ""Time Series Deconfounder: Estimating Treatment Effects over Time in
the Presence of Hidden Confounders”, International Conference on Machine Learning (ICML) 2020

Bio: 

Ioana Bica is a PhD student at the University of Oxford and at the Alan Turing Institute. Her PhD research focuses on building machine learning methods for causal inference and individualized treatment effect estimation from observational data. In particular, she has developed methods capable of estimating the heterogeneous effects of time-dependent treatments, thus enabling us to determine when to give treatments to patients and how to select among multiple treatments over time. Recently, Ioana has started working on methods for understanding and modelling clinical decision making through causality, inverse reinforcement learning, and imitation learning.

She has previously completed a Bachelor’s degree and a Master’s degree in Computer Science at the University of Cambridge where she has worked on multi-modal data integration and unsupervised learning for genomics data.