Data scientists develop decision-making tool to improve treatment of Covid patients.
Coventry University and Milton Keynes University Hospital (MKUH) joined forces during the pandemic to develop a tool that hopes to ease the Covid burden on the NHS.
Data scientists from Coventry University’s Centre for Computational Science and Mathematical Modelling (CSM) and MKUH used machine learning methods (ML) to support clinicians in predicting the effects of COVID-19 on each diagnosed patient involved in the study, alongside the associated risks of further ill health.
Predictive factors include the duration of a hospital stay, the risk of developing blood clots in the lungs, the likely need for ventilator support, or the probability of possible death.
Dr Alireza Daneshkhah, associate professor and curriculum lead in data science and artificial intelligence at Coventry University, stated: ‘After entering one of the greatest global catastrophes of the century, clinicians all around the world were puzzled as to what puts a person at risk from COVID-19, and whether this risk could be quantified in order to compare between individuals. The shortage of hospital beds, oxygen as well as ventilators prompted the need for this clinical decision-making tool which could assist clinicians in resource allocation during the pandemic, including the allocation of staff, beds and ventilation equipment.’
During the first wave of the pandemic, Dr Daneshkhah together with Dr Abhinav Vepa, senior house officer in general medicine at MKUH, reviewed and assessed 44 risk factor variables in more than 355 COVID-19 inpatients using the Bayesian Network ML method – a model used to represent knowledge about an uncertain domain.
The method included collating data on pre-existing health conditions, blood tests and importantly considered patient demographics including age, gender, and ethnicity.
Dr Vepa said: ‘The methodology showcased in this research has the potential to be applied to all diseases and all outcomes in order to improve clinical care. In order to prove that our model is robust to the extent at which it can influence lives, it would be good practice to test the model on a second data set, which is why collaboration with other research teams with independent data sets would be highly beneficial.
‘After external validation, and with a larger amount of data, this methodology could be applied to predict a solution to many clinical problems which could assist clinical decision-making and thus ease pressures in the NHS.’
This research seeks further external input and analysis through a wider data set to ensure it is fit-for-purpose, it could then be implemented within hospitals to support a multitude of clinical requirements.