The University of Sydney’s Westmead Applied Research Centre has been awarded a $1 million grant from technology giant Google to research and develop ways artificial intelligence could be used in digital tools to reduce the risk of heart attacks.
Westmead Applied Research Centre will utilise the Google AI Impact challenge grant to create customised digital health tools that enable clinicians and health services to support more people to prevent cardiovascular disease.
By combining clinical and consumer-derived data, such as from mobile phone apps and wearables, the program will offer tailored advice using machine learning to assess participants who have been to the hospital with chest pain, harnessing their digital footprint to reduce the risk of a heart attack.
Dr. Harry Klimis, a cardiologist and researcher at WARC, explained to HealthcareITNews that modifiable risk factors account for more than 90 percent of the risk of heart attack worldwide.
These include abnormal cholesterol, hypertension, smoking, excess alcohol consumption, physical inactivity, diabetes, obesity, psychosocial factors and diet.
"Existing prevention programs are underutilised and thus ineffective," he said. "Therefore, we need improved methods of identifying high-risk individuals prior to getting a heart attack, and high quality, accessible, cost-effective and customisable prevention programs that we can target them with."
Dr. Klimis pointed out a cloud of data surrounds individuals – clinical data, consumer derived data, social and health system data: Harnessing this data with AI could mean both greater accuracy in risk assessment and a more personalised digital solution adaptable to change.
"Adoption of such a digital solution by health services could improve health service efficiency and effectiveness," he said. "This will be achieved through greater accuracy in identification, better access prioritisation based on risk, and customisation of management and monitoring intensity based on individual risk."
He noted AI-driven adaptive digital solutions connecting our health services could transform care for patients and populations, and potentially reduce hospitalisations and healthcare costs.
Dr. Klimis said the most challenging part of training a machine learning (ML) model is ensuring the delivery of a program based on unbiased data representative of the general population.
"To address this, we have already delivered standard text-message programs, without AI or machine learning, to over 3000 consumers from different socio-demographic classes. which has allowed us to collect data on consumer characteristics, responses to different message types, and how this affects health outcomes," he said. "This dataset will be crucial in leveraging for an AI model for the current project."
He explained a large number of people already track their health statistics on a daily basis through various apps and wearables, and this data supplemented with health information gives greater insights into individual and population health.
This provides health professionals with insight into the day-to-day patterns, needs and preferences of patients and enables them to provide better guidance, as well as feedback and support to their patients.
The project’s initial plan is to link data from existing secondary sources such as hospital and clinic presentations, to deliver customised programs tailored to the individual through the digital platform they are most comfortable with, such as an SMS and/or app.
"By doing this we believe AI can improve engagement, health behavioural change, and thus health outcome measures – such as future cardiac events and hospitalisations – with our mobile digital health cardiovascular disease prevention program," Dr. Klimis said.