While the Hack Aotearoa 2020 – Artificial Intelligence for Health conference and datathon may be over for another year, we’re still fizzing with energy and enthusiasm for what we’ve learnt.
Adding a health datathon to the event for the first time, Hack Aotearoa 2020 was a great opportunity for inter-disciplinary collaboration between data scientists, academics and healthcare professionals. The two-day datathon saw delegates grouped into teams to work on a problem using the Massachusetts Institute of Technology (MIT) eICU Collaborative Research Database (eICU-CRD) and Medical Information Mart for Intensive Care (MIMIC-III) datasets.
Conference organiser Dr Mataroria Lyndon said about the datasets available:
“Often there’s limited access to large healthcare datasets, so it was a fantastic opportunity for clinicians and data scientists to be able to work with the MIMIC and eICU data – large-scale datasets that are well curated, high resolution and have many variables. This sort of data allows much more sophisticated analyses to be performed, compared with smaller or less well-curated datasets,” says Dr Lyndon.
Precision Driven Health (PDH) returned as an event sponsor for the second year running, and several of our researchers participated in the weekend-long health datathon. PDH-funded masters student Hamish Huggard valued the chance to experience doing data science using Google services.
“As well as myself and PDH intern Harshil Gulati, Rima Rōpū (Team Five) included a medical student, a clinician, and a health consumer. We worked on investigating the level at which hypokalemia (low blood potassium) poses an increased risk of arrhythmia (irregular heart beat).
“We spent most of Saturday writing SQL queries in order to access the data, so unfortunately we only had a few hours on Sunday to produce some graphs of our results. We’d have loved to have more time to get into some deep analysis.
“For me, a highlight was learning how to coordinate a team with diverse skills and levels of data science competence in a tightly constrained and high uncertainty environment.”
Hamish Huggard, PDH-funded masters student
PDH data scientist Dr Ning Hua was really looking forward to the cross-disciplinary collaboration, and she wasn’t disappointed.
“Our team, eMAKO, included Chad Paraone, PDH’s Māori Health Director; Sanja Mirkov, Professional Teaching Fellow at the University of Auckland’s School of Pharmacy; and Marek Lipert, data scientist and Chief Technology Officer of Spinney Mobile Development.
“We worked on answering the question: ‘Is ICU admission during the weekend more detrimental to kidney function compared to admissions during weekdays?’. By narrowing down our study population we were able to establish that if a patient has poor kidney function, it deteriorates further if they are admitted to ICU on the weekend.
“Using Google BigQuery was also a challenge for me. I’m used to using PostgreSQL, and it’s quite different to that. It was really fun to try out a new database management system though!”
Dr Ning Hua, PDH data scientist
“The cross-collaboration was great. Many people attended the datathon as ‘observers’, curious to find out how to use data science tools to solve a healthcare problem.
“Interestingly, we found it a challenge to define the ‘problem’: we spent a long time deciding on the specific question we were hoping to answer, and what data and statistical test would be required to answer it – both things that are really important parts of the process of doing data science. I think we would have had many more interesting outcomes if we had more time to work with the data.
“Using Google BigQuery was also a challenge for me. I’m used to using PostgreSQL, and it’s quite different to that. It was really fun to try out a new database management system though!”
Congratulations to the Hack Aotearoa datathon leadership team for organising such a fascinating and valuable event – we’re already looking forward to the next one!
Learn more about Hack Aotearoa – Artificial Intelligence for Health at www.hackaotearoa.co.nz