Machine learning for causal inference: developing modern tools for personalised optimal treatments

Scientists evaluating the effects of medicines or social policies (collectively referred to as treatments) are interested in identifying treatment combinations that maximise benefits as well as determine the best time for treatment initiation, changes or discontinuation. Such tailored interventions depend on genetic, demographic, environmental and lifestyle factors, treatment history and co-morbidities. Machine-learning - automated algorithms that learn from the data- bring a novel opportunity to answer these questions. How to handle the large number of variables involved, make predictions and take data-driven decisions, is a complex task. By successfully addressing these challenges, I will deliver methods that effectively harness machine-learning to obtain treatment effects and individualised predictions from a "what-if" (so-called causal) perspective, allowing us to obtain information about each individual's likely outcome if they follow a particular treatment. Methods for finding optimal treatments, targeted to maximise health benefits and minimise exposure to unnecessary treatments will also be provided.