JOIN HINZ   |   Print Page   |   Contact Us   |   Report Abuse   |   Sign In News Articles

NZ’s taonga – linked data – underused, says expert

Tuesday, 22 January 2019  
Share |

Return to home page

Picture: University of Michigan assistant professor of computer science and engineering Jenna Wiens speaking at Hack Aotearoa. editor Rebecca McBeth


New Zealand leads the world in linked data, but it is being underused, a data expert told attendees at a global artificial intelligence conference in Auckland.


Nicholson Consulting general manager Kylie Reiri was speaking at Hack Aotearoa, held at Auckland Business School on January 18–19.


The conference explored the use of predictive data, robotics and new smart technologies to develop better health and wellbeing outcomes for New Zealanders, with a strong focus on Māori.


Reiri says the Integrated Data Infrastructure is a taonga/treasure that links more than 60 de-identified datasets from across the health and social sector as well as areas such as education, justice and police.


It includes a range of datasets like ACC claims, dispensed medications, hospital admissions, and births, deaths and marriages records, all linked at an individual level and available for research. The oldest record goes back to 1840.


“About two thirds of the money our government spends on people is linked in this research database,” Reiri says.


“Just because the data we have today is not perfect, is not an excuse to not use it.


“We should be doing more analysis on it and the value delivered by that analysis will drive demand for more and better data.”


Reiri advocates putting people at the centre of data analysis, bringing together data to understand their lived experience.


University of Michigan assistant professor of computer science and engineering Jenna Wiens spoke at Hack Aotearoa about the need for AI to augment the role of clinicians, rather than replace them.


Her team has used machine learning to develop a risk model for patients being infected with clostridium difficile while in hospital, by using the structured contents of a hospital’s electronic health record. Data from more than 200,000 hospital admissions was analysed.


She explained that rather than limiting themselves to looking at a range of known risk factors, her team looked at all available data and also developed a model that tracks how a patient’s risk evolves over time.


The model is able to predict that a patient is at high risk five days before any clinical suspicion.


“This is enough time to do something about it and change a patient’s risk and outcomes,” Wiens says.


However, in order for these types of algorithms to be meaningfully adopted by clinicians, they must be regarded as accurate, credible, robust and actionable, she says.


Wiens told attendees that the model is hospital specific and cannot simply be transferred from one institution to another as there are differences in patient populations, clinical protocols and EHRs.

Return to home page

HiNZ, PO Box 300125, Albany, Auckland 0752, New Zealand.

Membership Management Software Powered by YourMembership  ::  Legal