The Digital Futures team are hosting talks by The University of Manchester's new Turing Fellows for 2021-22.
Presenting at this event are Glen Martin and Matthew Sperrin.
To register for this event, please visit the Eventbrite link opposite.
Glen Martin. Glen is a Senior Lecturer in Health Data Sciences. As a health data scientist, Glen undertakes multidisciplinary research at the intersection of mathematics, statistics, epidemiology and data science to conduct clinical investigations. Specifically, he uses data-driven methods on routinely collected health data to improve healthcare quality and clinical outcomes.
Glen's talk is titled 'Multi-Outcome Risk Prediction Modelling: current state-of-play and future research'.
Clinical prediction models (CPMs) are used to predict the risk of clinically relevant outcomes or events (e.g. the risk of developing cardiovascular disease within ten years). They are increasingly used to support clinical decisions, yet they seldom reflect the interplay between developing multiple comorbidities in terms of both pathophysiologic and treatment interactions. Specifically, the typical situation is to develop a CPM for one particular outcome, but this fails to capture multi-outcome patterns evolving over time. With rising emphasis on the prediction of multi-morbidity, there is a growing need for CPMs to simultaneously predict risks for each of multiple future outcomes. A common approach to multi-outcome risk prediction is to derive a CPM for each outcome separately, then multiply the predicted risks. This approach is only valid if the outcomes are conditionally independent given the covariates, and it fails to exploit the potential relationships between the outcomes. In this talk, I will introduce the notion of a multivariate (multi-outcome) CPM, present an overview of our current research on this topic and explore avenues for future work.
Matthew Sperrin. Matt is a senior lecturer in the Division of Informatics, Imaging & Data Sciences in Decision Sciences. Matt researches new statistical methodology to make inference with observational health data, collaborating closely with clinicians, epidemiologists, health informaticians, software engineers and statisticians.
Matt's talk is titled 'Role of causality in prediction'
?Causality and prediction are traditionally viewed as two separate activities. In this talk I will argue that causal inference can be useful when building prediction models in a range of scenarios. Primarily, being able to make predictions under hypothetical interventions - which is often done wrongly in models that do not have causal awareness. For example, a model that predicts mortality in hospitalized covid-19 patients might be used to triage patients – with patients at low risk being considered for discharge. However, a patient may be predicted as low risk because of the care that similar patients received in hospital in the past – which would not be equivalent to the care that they would receive post-discharge. Incorporating causality can also help improve generalisability of a prediction model, and answer questions about fairness in protected groups.