Enabling adoption of AI in The Australian Healthcare System

Recently, I posted here in LinkedIn about the "5 steps to have AI adopted in healthcare in Australia". What I shared wouldn't be new or surprising to many of those involved in developing and implementing AI models or medical technology here in Australia, but I was asked to clarify or explain the steps by some. I acknowledge what I posted previously was at a high-level and just an outline, so I will provide more detail to the steps and add some context in this article. But first, some background. While Australia is a leader in developing and implementing world-class technology, including digital technologies ; when it comes specifically to implementing and adopting AI in our healthcare system, we could do much better.

It is understandable that some individuals and entities, here in Australia, express concerns about AI being a relatively new technology and emphasise the need for caution when integrating it into existing platforms and workflows. While this perspective is valid and warrants consideration, it is equally important to recognize that an overly cautious approach, combined with excessive barriers and the influence of stakeholders with vested interests, may put Australia at risk of falling behind in the adoption of AI compared to other nations.

Countries such as the United States, the United Kingdom, Singapore, and others are making significant strides in the development and deployment of AI in healthcare models. To ensure that Australia remains competitive and at the forefront of this technological advancement, it is crucial to strike a balance between necessary caution and proactive engagement with AI technologies. If it needs us reminding how behind Australian healthcare is in terms of adoption of AI, this excellent review by Anna Janssen et al (2023) identified only 11 eligible AI applications that were implemented in routine practice. Yes, for a nation of 700 public hospitals, 630 private hospitals, 7000 general practices, 2700 aged care services and 150 ACH services, we have only 11 AI applications in routine practice! In a supporting review by van der Vegt and colleagues (2024), it was found there was only one Australian hospital that had an AI related clinical trial underway. They also found numerous barriers for adoption of AI in the Australian healthcare system. In my own preliminary research, I found very few native Australian AI applications listed on the Australian Register of Therapeutic Goods.

Which brings us to the question? What is required to enable adoption and integration of AI in the Australian healthcare system? In a recent article , published in an eminent implementation science journal, I discuss in general terms how AI, specifically generative AI, should be implemented in the healthcare context, but here I want to outline the specific steps we should take to adopt AI in The Australian healthcare system.

Step 1: Create accessible and vetted medical datasets for AI model training

High-quality, diverse datasets are essential for developing accurate and unbiased AI models. For AI to be reliably adopted in Australian healthcare, we must establish large-scale, accessible datasets. Australia should prioritize curating medical datasets that are easily accessible to researchers and companies building healthcare AI solutions. Importantly, these datasets must be carefully vetted to protect patient privacy and are representative of Australia's population. By doing so, we'll be creating a valuable resource for the development and training of AI models that can accurately reflect and serve Australia's diverse population.

Step 2: Create affordable and defined regulatory pathways for AI SaMD

While TGA provides a pathway and classfication system for AI-based Software to be approved for clinical use here in Australia, this is under generic medical software regulations. AI Software as a Medical Device (SaMD) has unique considerations compared to traditional medical devices. Further, the current costs and steps involved in gaining TGA approval make it prohibitive for SMEs/startups to undertake regulatory approval here in Australia. Considering the market size and economies of scale here in Australia, many developers are considering overseas markets for regulatory approval and release of their AI applications. Establishing clear, streamlined and affordable regulatory pathways tailored for AI SaMD will provide innovators much needed guidance and reduce barriers to bringing safe and effective AI tools to market.

Step 3: Create AI training programs for healthcare professionals

As AI becomes more prevalent in healthcare settings, it's critical that medical professionals know how to effectively use these tools. These programs should focus not only on how to use AI tools but also on understanding their limitations. Developing such comprehensive training programs will give clinicians the knowledge and skills to work alongside AI, understand its strengths and limitations, and maintain meaningful human oversight. This upskilling will be essential for the Australian healthcare workforce to adapt to and embrace AI integration. In a previous article , myself and my colleague, provide a template as to how this training should occur.

Step 4: Fund clinical trials to validate multi-modal AI models in healthcare

Rigorous studies are needed to demonstrate the real-world safety and efficacy of AI multi-modal models that analyze multiple data types like imaging, lab results, clinical notes, etc. Securing government and industry funding for these resource-intensive clinical trials will accelerate the responsible translation of cutting-edge AI research into validated healthcare applications.?Also, this step is pivotal in demonstrating the practical value of AI in improving healthcare outcomes, thus facilitating its adoption among clinicians and patients alike.

Step 5: Strengthen industry-academia-provider collaborations to hasten AI adoption

Deploying AI in healthcare requires close coordination between researchers developing the models, companies commercializing them, and hospitals implementing them. Australia should establish initiatives to foster these multidisciplinary partnerships, enabling a robust ecosystem for healthcare AI innovation. These partnerships will facilitate knowledge sharing, innovation, and the translation of research into practice. Such a collaborative ecosystem can accelerate AI adoption by aligning the objectives and resources of all stakeholders involved.

AI has the power to revolutionize healthcare in Australia. By following this 5-step plan - from foundational elements like curated datasets and regulatory clarity, to translational efforts like clinical validation and multistakeholder collaboration - Australia can become a leader in the responsible adoption of AI in healthcare. As we move forward, it will be the collective effort of government bodies, private organizations, healthcare professionals, and the wider community that will ensure these technologies are implemented effectively, ethically, and to the benefit of all Australians. The provided roadmap balances innovation with safety considerations to ultimately leverage AI to improve the lives of patients and support the vital work of healthcare professionals.

?

?

?