Advanced Analytics in Healthcare

Creating cutting-edge national-scale analytics infrastructure for health research and research translation
Four medical researchers gathering around a computer in a lab
Who will benefit
Health researchers using contemporary analytical methods such as AI to derive insights from data for better diagnostics and clinical care

The Challenge

The sensitive nature of much health information results in silos of data in local or secure environments, but modern data science depends on data aggregations of considerable scale and integration. Applying techniques like machine learning to sensitive data across multiple secure data repositories for national-scale health insights creates challenges that require national coordination and capability.

The Approach

The ARDC is uniquely positioned to coordinate and deliver national infrastructure capability to support advanced analytics in health research through the People Research Data Commons (People RDC).

Framework Development Phase

The Framework Development Phase took place in the first half of 2024. It involved co-designing a national framework, the specification and reference architecture for national infrastructure to support AI and other analytics in health research.

Two activities were carried out:

Framework Project

This project methodically captured researchers’ perspectives to help prioritise the ARDC’s infrastructure support for advanced health analytics.

This is a partnership between the ARDC and the Australian Data Science Network, a national network of data science research groups.

Pathfinder Project

As a companion to the Framework Project, this project was created to delve deeper into federated learning, , a well known challenge around learning from sensitive data. The project provides a test bed for constructing a federated learning infrastructure for healthcare data.

This project brought together collaborators from:

The project focused on developing federated learning implementation, which includes: 

  • comparing existing federated learning tools
  • selecting best-of-breed federated modelling options for healthcare data
  • considering national infrastructure provision. 

Infrastructure Development Phase

UBased on these 2 projects, we are developing a reference architecture and roadmap for the next 4 years for future research infrastructure in healthcare analytics. These will prepare us for infrastructure development on multiple fronts simultaneously, affording flexibility and minimising critical paths.

Coordinated projects and services will commence in 2025 in collaboration with:

  • universities, medical research institutes and national research collaborations
  • government data custodians and health systems players
  • NCRIS facilities.

The implementation of the roadmap over the next 2 years will create a long-lasting, safe, and flexible AI-enabled research infrastructure with:

  • open labs and ’playgrounds’
  • the ability for secure data from different jurisdictions to be shared and analysed
  • new ways to analyse data (e.g. AI/ML, federated and foundational models) using combined datasets.

In health analytics, patient privacy is fundamental. The future program includes secure underpinning cloud and privacy preserving technology. These will be supported by social and technical resources like training pathways, governance, co-developed standards, frameworks, guidelines, facilitations and communities of practice. The system will support analysis, modelling and decision-support tools in a variety of disease areas with measurable diagnostic improvements, as well as an ecosystem of engaged key partners for knowledge sharing to influence policy and systemic behavioural change.

Collaboration

At the Framework Development Phase, we worked with selected national groups and conducted broad consultation. 

The Infrastructure Development Phase will offer further opportunities to participate for researchers, policymakers and innovators. There will also be more opportunities for partnerships with institutions, data custodians and technology service providers.

We invite you to register your interest in the People RDC and keep in touch with us through public consultations.

Target Outcomes

At the Framework Development Phase (the first half of 2024), a Framework Project and a Pathfinder Project were carried out with the following outcomes:

Framework Project

Through surveys, workshops and interviews, the ARDC and the ADSN consulted a wide range of stakeholders to identify the needs, aspirations and challenges of advanced analytics in healthcare. 

Based on our findings, we have developed an infrastructure framework to create a comprehensive and coordinated approach to health analytics infrastructure, leveraging national and international collaborations to enhance research capabilities, data governance and technological advancements. We have made the following key recommendations, which are aimed at creating an Australian Harmonized Approach (AHA) to health data analytics research infrastructure:

  • enhance computation resources and data environments
  • standardise data governance and curation
  • promote collaborative and ethical research initiatives
  • support workforce development and practical implementation.

Read the report. Also watch a recording of our launch event for the framework, where we delved into the framework’s development, discussed its key recommendations and explored how it can support advanced health analytics in the coming years.

Pathfinder Project

While the framework project was about capturing the breadth of issues, a companion pathfinder project was carried out to delve deeper and:

  • explore the uses, needs and challenges of federated learning in the context of sensitive health-related data, while ensuring the maintenance of privacy and confidentiality
  • identify and establish a collaborative network among similar research groups
  • develop suitable demonstrator artefacts to centre the dialogues around them. 

This project provided an overview of requirements and current experiences with federated learning, covering:

  • a comparison of key federated learning tools available and infrastructure requirements to support federated learning with the goal of establishing a suitable blueprint for a federated learning architecture that can be effectively implemented
  • consideration of use cases that could become cardinal edge cases for the development of a national infrastructure, including discussion of case study of designs and deployments that are available to or informing national infrastructure
  • healthcare-focused conclusions and recommendations to the ARDC on infrastructure and other support required to enable and encourage use of federated learning by Australian research groups.

Read the report. Also watch a recording of our launch event for the pathfinder.

Based on the 2 projects, we are developing a reference architecture and roadmap for advanced analytics in healthcare. It will facilitate meta-specifications for scheduled ARDC services, capabilities and co-investments.

The Infrastructure Development Phase (2025–28) will provide national coordination, capability, partnerships and services supporting national scale analytics of health data.