Looking forward >
DAGs Forum: 12pm – 1pm Thursday 13 April. Tonsley Campus – Room 5.29, and online via Teams (Lunch will be provided).
Presenter: A/Prof Murthy Mittinty
Title: What must we aim for: Is it clinical utility or predictive accuracy?
There are two aspects to data modelling, variance and bias. Variance can be minimized by having a large sample, but bias must be handled using additional assumptions and techniques. Identifying a predictive model helps in understanding: 1) model uncertainty and 2) how the outcome is generated. Even then, this information does not provide insights into the treatment’s comparative effectiveness and handling of costs associated with treatments. Two important aspects of clinical utility.
On the other hand, the counterfactual approach is helpful in estimating the potential outcomes estimates for each individual respondent using the prediction model thus addressing the confounding bias. Additionally, the field of causal inference has given us a new outlook on how we can emulate trials using observational data. In this presentation, I will combine these fields of science; machine learning and causal inference, and show how one can aim for attaining clinical utility than just predictive accuracy.
< Looking back
Network Science; a transformative science for healthcare research and technology? (12pm Thursday 24th November 2022)
Presenters: Prof David Ben-Tovim
Professor David Ben-Tovim presented work on Graph Theory in the context of healthcare research. David has been a member of a collaborative group from the College of Medicine and Mathematical and Computing Science with a long history of healthcare research and practice. The group has been exploring the application of Network Science to understanding how hospitals really work. Following a brief introduction to Network Science and explained how the problem of organising hospitals can be thought of as being a Mesostate problem, being one of organisational complexity related to uncertainties in factors such as admissions, transfers, available care, and length-of-stay that administrators must somehow deal with and ensure that the hospitals remain organised. David then gave some examples from their recent research and how network theory can be used to represent major components of existing hospital services, and how novel hypotheses can then be generated.