UTS Faculty of Health Higher Degree Research student and nurse, Julia Pilowsky alongside her team of clinicians and data scientists showcased the potential of Natural Language Processing (NLP) in healthcare, pitching the winning idea at the recent National Healthcare Datathon.
The National Healthcare Datathon (opens external site) is run by IntelliHQ and the Australia and New Zealand Intensive Care Society (ANZICS). The event brings clinicians and data scientists together to use healthcare data to solve real world clinical problems.
During the datathon, each team is given access to real patient data from a variety of sources, including NSW Health and the ANZICS Adult Patient Database. Teams must come up with data-driven solutions to a chosen problem, and pitch their ideas to a panel of judges at the regional hub – with winners going on to the national pitch.
Julia’s team which included Nhi Nguyen and Lizzie Barrett, clinicians from the Agency for Clinical Innovation (ACI) and data scientists Jae-Won Choi, eHealth NSW and Thomas Beltrame, Flinders University moved through the regional hub, and were awarded winners of the national pitch.
Julia’s team had initially planned to develop an algorithm to better predict which critically ill patients were more likely to develop a pressure injury during their stay in the intensive care unit (ICU).
However, after interrogating the data from the NSW Health ICU electronic medical record (eRIC) they realised only 12% of patients had their pressure injury status filled at any point during their admission.
It was Julia’s previous work using a Natural Language Processing (NLP) algorithm that offered the solution. NLP is what enables computers to understand human’s natural language, whether spoken or written. It uses artificial intelligence to take real-world input (text or speech) and decode it in a way a computer understands.
The team presented the idea of an NLP algorithm capable of determining a patient’s pressure injury status based on progress note documentation.
Julia says, “We imagine the algorithm could be embedded locally within each ICU’s electronic medical record and provide high quality data on pressure injuries in the unit.
“The algorithm could also be generalised to other under-researched topics that are difficult to obtain complete data on, for example delirium or agitation.
“We also anticipate the algorithm could be used to automatically complete the vast array of tick-box fields present in the clinical information system by using information from progress notes, thereby reducing the documentation burden for clinicians.
She says, “The datathon was a great opportunity to meet and work with other professionals who are passionate about using big data to come up with innovative solutions to improve patient care.
“Ever since discovering the power of NLP in my PhD, I have been looking for ways to implement it in real-world healthcare computer systems to enable researchers and clinicians to access the wealth of information locked behind text-based data.
“With the support our team has received from winning the datathon we hope to develop and implement an algorithm to do just that”.
Find out more about Faculty of Health Higher Degrees by Research