Developing partnerships and collaborations to advance AI

AI in Health Collaborations
,
1

Developing partnerships and collaborations to advance AI

Supported by AI in Health Research Network

Key Points:

  • Barriers to partnerships and collaborations in AI health research

  • Data sovereignty and governance challenges

  • Pathways from research to implementation

  • Funding models and sustainability

  • Building trust and relationships across sectors

  • Synthetic data as a potential solution

  • Standardisation and interoperability issues

Discussion Items:

Data Access and Governance

The group discussed significant barriers to partnerships in AI health research, particularly around data access and governance. Participants noted that while data sovereignty is crucial, current processes for accessing health data are often prohibitively slow, taking months to establish agreements. This creates a significant barrier to innovation. One participant suggested using AI to create representative synthetic datasets that preserve privacy while enabling innovation. This approach would allow researchers to develop and test solutions without compromising patient privacy, with promising innovations then progressing through more formal approval processes.

  • Synthetic data was highlighted as a potential solution that could democratise access to health data while preserving privacy.

  • Participants emphasised that data governance needs to balance protection with accessibility to foster innovation.

  • The group acknowledged that data sovereignty extends beyond Māori data to include all patient data, with questions about who truly owns health data.

From Research to Implementation

A significant challenge identified was the gap between successful research and clinical implementation. Participants discussed how research funding often ends before a prototype can be developed into a clinically usable product. This creates a “valley of death” where promising innovations fail to reach patients. The group explored various models to bridge this gap:

  • Building multidisciplinary teams from the outset that include researchers, clinicians, and industry partners

  • Creating clearer pathways for evaluating the value of AI innovations in healthcare

  • Developing frameworks to assess return on investment for health AI solutions

  • Establishing incubation processes to support the transition from research to implementation Several participants shared examples of successful approaches, including Breast Cancer New Zealand’s model of bringing researchers into their system rather than sharing data externally, and the Singapore-New Zealand collaboration on AI in healthcare.

Building Trust and Relationships

The importance of trust and relationship-building emerged as a central theme. Participants noted that effective collaborations require investment in relationships before technical work begins. This includes:

  • Taking time to understand different stakeholders’ perspectives and needs

  • Ensuring diverse representation in governance and decision-making

  • Creating opportunities for ongoing dialogue between technical and clinical teams

  • Recognising and respecting different cultural perspectives on data sharing One participant from Dementia New Zealand shared how building trust with another organisation over time had enabled them to collaborate on a national dataset, demonstrating the value of patience and relationship development.

Standardisation and Interoperability

The fragmentation of health systems and lack of standardisation was identified as a major barrier to scaling AI solutions. Participants discussed:

  • The challenges of different regions using incompatible systems

  • The need for standards that enable interoperability without stifling innovation

  • The potential for AI to help bridge gaps between different data formats

  • The tension between local innovation and national coordination The group reflected on previous attempts to standardise health data in New Zealand, noting that while standards like FHIR exist, implementation has been inconsistent, limiting the potential for AI solutions to scale nationally.

AI Applications and Value

Participants shared various applications of AI in healthcare that showed promise:

  • Using AI to analyse qualitative data and identify themes in patient stories

  • Automating administrative tasks to free up clinician time

  • Improving surgical scheduling and patient flow

  • Supporting clinical decision-making with evidence-based recommendations

  • Enhancing patient engagement and self-management The group discussed the importance of focusing AI development on tasks that healthcare professionals want to stop doing, rather than automating the aspects of care they find most meaningful.

Next Steps:

  • Explore the creation of a registry or library of AI health projects to reduce duplication and foster collaboration

  • Develop frameworks for evaluating AI health solutions that consider both clinical and economic impacts

  • Investigate models for creating and using synthetic health data to accelerate innovation

  • Establish clearer pathways from research to implementation, potentially including incubation support

  • Build stronger connections between academic research and clinical practice

  • Consider how to better involve patients in AI development and governance

Challenges:

  • Balancing data protection with the need for access to enable innovation

  • Bridging the funding gap between research and implementation

  • Overcoming siloed approaches across different health organisations

  • Building trust between different stakeholders in the health system

  • Ensuring AI solutions address real clinical needs rather than technology for its own sake

  • Managing the rapid pace of AI development against the slower pace of healthcare change

  • Addressing concerns about AI replacing human judgment in clinical settings

Additional Notes:

Participants noted the irony that while New Zealand has a good reputation internationally for digital health data, internal barriers often prevent effective use of this data. The group also discussed the potential for AI to help with qualitative data analysis, particularly for understanding cultural perspectives and patient experiences. There was recognition that AI in healthcare requires different approaches depending on the application, with clinical decision support requiring more rigorous validation than administrative applications. The meeting highlighted the tension between moving quickly to adopt new AI technologies and ensuring appropriate governance and ethical frameworks are in place.

As a PDF:

Table 16 - Developing partnerships and collaborations to advance AI - Detailed insights.pdf (126.3 KB)

Converted to markdown by @NathanK using https://www.pdftomarkdown.co/ on 2025-07-01T23:15:00Z

1 Like