Click here to watch this seminar
Presentation 1: A Scalable Open-Source Interactive Cancer Analytics Web Application for Routinely-collected Data
By: Millie Wagstaff
Abstract: Cancer data such as survival outcomes and waiting times are routinely collected across hospitals. Whilst organisations such as NHS Digital and Public Health England analyse these data, their outputs can take years to produce and provide limited information at a local or regional level.
The aim of this project is to develop software that allows hospitals across Yorkshire and Humber to obtain immediate insights from their own cancer data. This involves developing an interactive web application and automated reports using open source R packages.
The easy to use app will allow health care professionals and hospital managers to interact with, analyse and export waiting time and outcome data, soon after they are collected. Alongside this, automated data reports can be emailed to key decision makers within hospitals at regular time intervals. Together, their outputs can aid managerial oversight, improve understanding of clinical outcomes and be used to help inform patients.
Overall, the app aims to enhance the utilisation of routinely collected cancer data within hospitals across The Yorkshire and Humber Region. The app being based on nationally collected datasets and using open source tools, add the potential for scaling deployment nationally, ultimately leading to the improvement of patient care nationwide.
Presentation 2: Modelling Uncertainty in the Risk of Intensive Care Unit Readmission
By: Ben Cooper
Abstract: Intensive Care Units (ICU) provide the highest level of care possible to patients in the most dire clinical need, with the intention of progressing patients towards a stable condition for discharge to lower-priority wards. However, as many as 1 in 10 patients discharged from ICU are subsequently readmitted within the same hospital stay, as their condition deteriorates again. Not only does this present a significant strain on hospital resources, but patients readmitted to ICU have poorer prognoses and face a much greater risk of in-hospital mortality. Several quantitative models exist to predict the risk of ICU readmission, but none have yet been shown consistently produce well-calibrated and highly accurate predictions, nor are any in regular use by clinicians.
All of these models share an underlying concern – how should they deal with variables which are not available at the point a prediction is made? Many models suggest that missing data should be assumed to be within a normal range, but this is a potentially dangerous assumption to make, and may contribute to the relatively poor performance of these models. In this project, we aim to demonstrate two separate but related methods for dealing with missing clinical variables. First, we will use Bayesian analysis to carry forward the uncertainty introduced by missing data into the final predictions. Second, we will use Gaussian process regression modelling to provide a better method for imputing missing variables from both the variables which are known, and their previous values. We will use these methods to produce a framework for handling missing data that can then be applied to any model for predicting ICU readmission risk.
Presentation 3: Isolation and Inclusion in a Post-social Distancing COVID world
By: Rosalind Martin
Abstract: The disparate impacts of COVID-19 and the associated lockdown have been much discussed recently, particularly in terms of age, deprivation, or employment sector. As countries slowly relax their lockdown regulations, it is equally apparent that the after-effects are likely to be long-lasting, whether through continued mitigation efforts such as social distancing or the economic impacts of economic shutdown, and that these after-effects are likely to further unevenly impact some groups over others. Although there are many dashboards reporting information on COVID cases, deaths and now vaccine rollout, detailed information on the socio-economic impacts on general population and businesses is largely missing. This work presents initial definitions of those who are expected to experience increased isolation under a variety of Covid-easing regulations, with the aim of presenting outputs via an interactive dashboard.
If you wish to ask the presenters any questions or offer feedback, please email LIDA