MinterEllison continues to advise market leading healthcare providers and government about how to appropriately guardrail use of Artificial Intelligence (AI). We assist clients to maximise the opportunities presented by AI, while appropriately managing the risks, including the significant legal, reputational and privacy risks presented by this technology.

One question we are commonly asked is 'what are you observing in the market?'. This article summarises our observations.

Key health AI deployment focus areas

As briefly outlined in our article 'Early Adopter's Playbook: Deploying AI in healthcare', the use cases for AI in healthcare are vast, and developing rapidly.

Based on our experience, the primary use cases for AI in healthcare broadly fall in these seven categories:

1. Healthcare administration and workflow optimisation: AI that automates administrative tasks like scheduling, patient flow management and billing, to improve efficiency and reduce costs.
2. Clinical Decision Support Systems (CDSS): AI systems that provide healthcare professionals with evidence-based clinical knowledge and patient data, aiding clinical decision-making.
3. Medical imaging analysis: AI systems used to analyse X-rays, CT scans and other imaging modalities, to support detection of abnormalities more quickly and accurately.
4. Telemedicine and virtual health assistants: AI that powers virtual health assistants and telemedicine platforms, allowing for remote monitoring, diagnosis and consultation, thus improving healthcare accessibility.
5. Mental health and therapy: AI applications that provide mental health support through chatbots, that can analyse speech and text patterns to identify mental health issues.
6. Personalised medicine: AI that can analyse data, including genetic information, to tailor treatment plans to individual patients.
7. Health monitoring and wearables: AI that integrates with wearable technologies to monitor patients' health in real-time, providing alerts for potential health issues and enabling proactive management of chronic diseases.

Deploying AI in healthcare

There are now a number of AI tools available in the market that fall into these seven use case categories. We have briefly summarised some of these tools below. (The information outlined below has been compiled from a scan of publicly available information and does not constitute endorsement of any of the products.)

Elsevier Health

Elsevier Health has partnered with OpenEvidence to create the AI model 'ClinicalKey AI'. ClinicalKey AI was designed and developed specifically for use by treating clinicians, as a clinical decision support tool, providing secure access to medical content. Much like other AI platforms such as ChatGPT, ClinicalKey AI can be accessed through a conversational search function. When and if prompted, ClinicalKey AI can account for patient context, such as any underlying health conditions or comorbidities. Then, ClinicalKey AI works by gathering information from reliable and authoritative sources such as journal articles, government publications and medical reference tests to provide a summarised response to the clinician's query.

Epic Systems

Privately owned American company, Epic Systems, has integrated AI in its software to improve its service offering. Epic Systems is an Electronic Health Records (EHR) system, and has incorporated Generative AI into its EHR. According to Epic Systems, the tool works by auto-drafting patient summaries and messages, along with handoff notes, reducing a clinician's time spent on administrative or repetitive tasks, ultimately improving efficiency.

Philips and AWS

Philips has partnered with Amazon Web Service (AWS) to create a tool that aims to provide clinical support, enable more accurate diagnoses and automate administrative tasks. Philips will use Amazon HealthLake Imaging, which forms part of AWS, to store, access and analyse medical imaging data. The idea is to optimise image access speeds while simultaneously reducing costs associated with medical imaging.

Google Cloud

Google Cloud has launched Vertex AI Search (Vertex AI). Vertex AI is an information retrieval system. It works by searching across a wide variety of data sources, which may include public websites as well as an organisation's enterprise data. The ability to search private data means that Vertex AI can search and summarise, for example, a patient's health records to the practitioner. Vertex AI cites the information it includes in its responses in order to improve transparency.

Salesforce

Salesforce has launched its AI model the 'Einstein Copilot'. The Einstein Copilot can provide answers that consider an organisation's unique data. Einstein Copilot can be used by healthcare providers to summarise clinical data and automate processes.

PathAI

United States based company PathAI has created the digital pathology tool AISight Image Management System for AP Laboratories. It works by ordering data and assisting users to manage their digital pathology workflow. One particular feature of AISight is the product TumorDetect. TumorDetect works by arranging the images as a matter of priority through assessing and quantifying the number of detectable tumours. AISight, combined with TumorDetect, aims to expedite diagnoses and therefore accelerate treatment plans.

Key AI in healthcare issues to consider

As these use cases demonstrate, the application of AI in healthcare is vast and promising. It presents the opportunity for significant increased efficiency, accuracy and cost effectiveness. Leading healthcare organisations are focused as much on managing risks presented by this technology as they are on maximising the opportunities it presents.

1. We have noticed that our market-leading clients are taking a Governance-First Approach to AI. By this, we mean, they are carefully putting in place governance structures to mitigate key risks posed by AI, including privacy, data management, operational and workplace risks. This includes putting in place lighthouse ethical principles for the business that align with Australia’s 8 Ethics Principles;
2. Preparing a written policy about how AI can and cannot be used within the organisation (including specific instructions to the workforce); and
3. Preparing User Terms and Conditions that specifically set out how the AI tool can and cannot be used, and requirements for ensuring there is a 'human in the loop'.

Australia launched its National Policy Roadmap for Artificial Intelligence in Healthcare (Policy) in November 2023. As businesses design their AI approach and governance framework, they should keep in mind the following key issues:

• Governance and Regulation: Use of AI in Australia is regulated by a number of different legislative instruments. To compound this issue, the sheer volume of data from different sources can make it challenging to comply with basic obligations, such as ensuring individuals are informed about their data being collected and provide consent. We recommend clients seek advice about how use of AI in their business is regulated.
• Data privacy and security: AI applications often involve storing and analysing sensitive patient data. This requires robust data protection measures, strict compliance with privacy regulations, and comprehensive systems to operationalise relevant data protection practices. We recommend businesses seek advice about how AI will use personal information, and ensure there is a robust data privacy and security framework in place.
• Bias and Fairness: Australian healthcare data is an evolving area, meaning that it may inaccurately represent the population. These discrepancies can reflect inherent biases in our society and create complications for organisations utilising AI. If AI models are trained on biased data, they could perpetuate or amplify discrimination against certain patient populations. In turn, this could create liability issues for users of these systems. Businesses should carefully assess whether the AI tool has imbedded biases, and consider how it will ensure use of that AI is fair.
• Accountability: As AI becomes increasingly embedded in clinical decision-making, it becomes increasingly important to understand liability for mistakes. Clear governance frameworks (including policies regulating use of AI within an organisation) are needed to manage risks and ensure responsible AI. Healthcare organisations must establish clear lines of responsibility and controls to ensure accountability. This can be achieved through written policies, written agreements (including with AI providers) and User Terms and Conditions.

With the right safeguards, these AI use cases present tremendous upside for Australian healthcare organisations – and the healthcare system itself. However the upside requires upfront investment in governance and responsible guardrails to mitigate risks and embed compliance.

If you are interested in understanding how your organisation can appropriately manage the legal and operational risks associated with implementation and use of AI in healthcare, please contact us for a confidential discussion.

MinterEllison specialises in helping healthcare clients to implement governance, regulatory and workforce risks associated with digital transformations and adoptions of new technology.