My Health Record and AI: Current State and Future Possibilities

Australia’s My Health Record (MHR) system contains health information for most of the population. As AI applications in healthcare expand, the question of how MHR data might support AI becomes increasingly relevant.

The current answer: very limited. But the future could be different. Here’s my assessment of where things stand and what needs to change.

What My Health Record Actually Contains

MHR aggregates health information from multiple sources:

  • Medicare and PBS claims data
  • Hospital discharge summaries
  • Pathology and radiology results (where uploaded)
  • Specialist letters
  • GP shared health summaries
  • Immunisation records
  • Prescribed and dispensed medications

This represents a longitudinal view of patient health that no single healthcare provider has. For AI applications that benefit from comprehensive patient history, this data is potentially valuable.

Current AI Use of My Health Record Data

Today, AI use of MHR data is minimal:

Clinical use within MHR. Limited AI functionality within the MHR system itself. Some smart features exist (medication matching, duplicate detection) but these are relatively simple.

Research use. Secondary use of de-identified MHR data for research, including AI development, is possible through approved pathways. But the process is complex and not widely used for AI specifically.

Clinical decision support. Healthcare providers accessing MHR in their practice don’t have AI tools that synthesise or analyse MHR data in sophisticated ways.

Primary care integration. Practice management systems that connect to MHR don’t generally incorporate AI that processes MHR data.

In practice, MHR provides data access but not AI-augmented insights.

Why AI and MHR Haven’t Connected

Several factors limit AI use of MHR data:

Privacy Framework Constraints

The My Health Records Act establishes strict controls on data use:

  • Data access is limited to healthcare providers involved in an individual’s care
  • Secondary use for research requires de-identification and approval
  • There’s no general authority for AI training on identifiable MHR data

These constraints are intentional and appropriate—they protect patient privacy. But they also create barriers to certain AI applications.

Data Quality Limitations

MHR data has significant quality issues:

  • Inconsistent upload practices across healthcare providers
  • Variable completeness of records
  • Free-text formats that are hard to process
  • Duplicate and conflicting information
  • Historical gaps from before MHR participation

AI trained on or inferring from poor quality data produces poor results. Data quality improvement is prerequisite to many AI applications.

Infrastructure and Integration

Current MHR infrastructure wasn’t designed for AI applications:

  • Access is oriented toward individual patient record retrieval, not population analytics
  • APIs don’t support AI-friendly data extraction
  • Integration with clinical AI systems isn’t straightforward

Technical infrastructure would need enhancement to support sophisticated AI use.

Governance and Trust

Public trust in MHR has been fragile. Early opt-out controversies, concerns about data security, and privacy anxiety have made ADHA appropriately cautious about expanding data use.

New AI applications that access MHR data would face public scrutiny. Governance arrangements would need to be robust enough to maintain trust.

Potential AI Applications

Despite barriers, several AI applications could potentially use MHR data:

Longitudinal Risk Prediction

MHR provides longitudinal data that individual providers lack. AI could potentially use this to:

  • Predict chronic disease progression based on multi-year patterns
  • Identify patients at risk of adverse events based on cumulative medication history
  • Detect care gaps by analysing patterns across multiple providers

This requires population-level analytics with appropriate consent and governance.

Care Coordination Support

For patients seeing multiple providers, AI could:

  • Identify potential medication interactions across prescribers
  • Flag conflicting treatment plans
  • Highlight information that should be shared across care settings

This operates at the individual patient level during clinical care—potentially more feasible than population analytics.

Clinical Decision Support Enhancement

When clinicians access MHR, AI could:

  • Summarise relevant history from extensive records
  • Highlight clinically significant information
  • Identify patterns across data sources that individual providers might miss

This augments the clinician’s use of MHR rather than operating independently.

Research and Population Health

De-identified MHR data could support:

  • AI development and validation
  • Population health analytics
  • Health system planning and optimisation

This requires robust de-identification and governance but doesn’t face consent barriers for identified data.

What Would Need to Change

For significant AI use of MHR data, several things would need to change:

Clarity about:

  • Whether current privacy frameworks permit various AI use cases
  • What consent arrangements (if any) are needed for different applications
  • How AI-derived insights can be used and shared
  • Liability when AI using MHR data contributes to clinical decisions

Some of this might require legislative change; some might be achievable through policy interpretation and guidance.

Technical Infrastructure

Development of:

  • APIs designed for AI applications
  • Secure computing environments for population-level analytics
  • Integration pathways with clinical AI systems
  • Data quality improvement mechanisms

ADHA would need to prioritise and fund this development.

Data Quality Improvement

Investment in:

  • Standardising data formats and coding
  • Improving upload consistency across providers
  • Cleaning historical data
  • Incentivising quality contributions

This is a long-term challenge requiring sustained effort.

Trust and Engagement

Building public confidence through:

  • Transparent governance of AI use
  • Consumer involvement in AI policy development
  • Clear communication about benefits and risks
  • Strong security and privacy protections

Trust is earned over time and can be lost quickly. ADHA would need to manage this carefully.

My Assessment

My honest assessment of MHR-AI timeline:

Short term (1-2 years). Minimal change. Some pilot projects exploring potential. Policy discussions but limited action.

Medium term (3-5 years). Enhanced clinical decision support within MHR. Research use of de-identified data for AI development increases. Some integration with clinical AI systems.

Longer term (5-10 years). Depending on policy direction, either significant AI integration with MHR or continued fragmentation where AI develops separately from MHR data.

The variable is political and policy will. The technical challenges are solvable; the governance and trust challenges are harder.

What Health Services Should Do

For healthcare organisations:

Engage with ADHA consultations. When ADHA seeks input on MHR development, provide clinical and technical perspectives on AI potential.

Develop parallel AI capabilities. Don’t wait for MHR-AI integration. Build AI capabilities using your own data. If MHR integration comes, you’ll be ready to use it. Firms like AI consultants Brisbane are helping health services develop these foundational capabilities now.

Contribute quality data to MHR. Better data in MHR creates better foundations for eventual AI applications. Improve your upload practices.

Participate in research. Support research using de-identified MHR data. This builds evidence base and develops governance patterns.

Working with partners who understand both MHR and AI can help. AI consultants Melbourne and health informatics specialists can advise on how to position for future integration while building capability today.

The Broader Question

The MHR-AI question is really about what kind of health system we want:

  • Do we want AI that can see the full picture of patient health?
  • How do we balance privacy protection with AI potential?
  • Who should benefit from insights derived from population health data?
  • How do we maintain public trust while enabling innovation?

These aren’t technical questions. They’re social and political questions that will be answered through policy processes over coming years.

Whatever the answers, the intersection of Australia’s national health record and AI will be an important space to watch. The potential is significant. The path to realising it is uncertain.

Dr. Rebecca Liu is a health informatics specialist and former Chief Clinical Information Officer. She advises healthcare organisations on clinical AI strategy and implementation.