Introduction by Croakey: A forum was held in Sydney today to examine issues around the adoption of artificial intelligence (AI) in healthcare, bringing together clinicians, digital health leaders, clinical governance experts and a range of other stakeholders (notably, the importance of First Nations and consumer health representation was not highlighted on the conference website).
In a keynote address to the forum, Australian Medical Association President Professor Steve Robson argued that AI can be a “game changer” for Australia’s healthcare sector, but said proper regulation is needed to ensure it doesn’t worsen health inequities across the system.
According to an AMA media release, Robson called for measures to ensure all Australians have equal access to future AI diagnostic technology and treatment options. He also wants regulation to address issues relating to patient safety, privacy and professional autonomy.
Meanwhile, Croakey’s managing editor Alison Barrett has been reviewing recent publications on AI in health, and provides an overview below.
Alison Barrett writes:
The World Health Organization has emphasised the importance of safe and effective systems and regulations to manage artificial intelligence in health, especially given the rapid growth and development of the technology.
In a report published last week, the WHO outlined considerations for the regulation and management of AI in health, including the need for transparency and commitment to data quality.
“Artificial intelligence holds great promise for health, but also comes with serious challenges, including unethical data collection, cybersecurity threats and amplifying biases or misinformation,” WHO’s Director-General Dr Tedros Adhanom Ghebreyesus said in a statement.
Similar sentiments were made by PLOS Digital Health’s Editor-in-Chief Dr Leo Anthony Celi in the journal’s recent collection of articles on the digital determinants of health.
Celi warned that while digital health technologies, including AI, may transform access to and delivery of healthcare, they also have the potential to widen already existing health inequities.
Shining a spotlight on the digital determinants of health, articles in the collection discuss digital literacy, bias in AI algorithms and the potential “pitfalls to access and equality in technology-enabled healthcare”.
While some articles acknowledged the intersections between digital determinants of health and the social determinants – for example, digital health literacy is linked to education level – they do not discuss the intersections with the commercial determinants.
Digital determinants had previously been studied as part of the social determinants of health, but according to Healthcare Consultant Casey Holmes Fee and colleagues, the escalating growth in digital health technology warranted “independent exploration”.
Paradox
While digital health technologies could bridge some gaps in healthcare access, groups who are most likely to be underserved by healthcare are also more likely to be “excluded from the digital world through their sociodemographic characteristics,” according to Dr Maria del Pilar Arias Lopez from the Intermediate Care Unit in Hospital de Ninos Ricardo Gutierrez in Argentina, and colleagues.
“As healthcare becomes more reliant on technology-based tools, the digital divide stands to further exacerbate existing healthcare access disparities,” the authors wrote.
Accessibility, affordability, usability and digital literacy are some of the factors that influence digital determinants of health and, if not addressed sufficiently, could result in widening inequities.
The mass adoption of virtual/telehealth during the pandemic is a good example of the paradox of digital health. While it was beneficial in facilitating healthcare access during COVID-19 lockdowns, inadequacies in digital infrastructure in some rural and remote areas of Australia meant that virtual/telehealth was not accessible for all.
A recent article in The Conversation reported that almost half of remote First Nations communities and homelands in Australia do not have mobile phone coverage, and therefore are “highly excluded from increasingly important digital services and tools”.
Authors Dr Daniel Featherstone and Associate Professor Lyndon Ormond-Parker from RMIT University said that the digital gap “widens substantially with remoteness”.
Featherstone and Ormond-Parker highlight the importance of First Nations leadership – via the First Nations Digital Inclusion Advisory Group – on closing the digital gap, by providing “practical, appropriate and evidence-based input on key policy areas that affect First Nations people and communities”.
Digital health literacy
An individual’s level of digital health literacy is associated with age, gender and level of education, one of the reviews in the DDOH collection found.
It found that people with higher levels of education had a greater level of digital health literacy, as did people with a higher level of digital experience and skill.
Older age – no specific age range provided – was found to be a barrier to learning and navigating technology. Cognitive, memory loss, hearing and visual impairments also contributed to the digital divide in older people.
The review also found that people with better digital health literacy experience better health outcomes, including higher quality of life, sense of purpose and optimism.
People with higher digital health literacy are better able to self-manage and engage in their own medical decisions, according to the review.
However, the authors highlighted limitations – not many experimental studies have been undertaken on the topic of digital health literacy, and different definitions of digital health literacy were used across studies.
Interventions to improve digital health literacy include education, training and social support from tech-savvy family, professionals and peers. Dr Maria del Pilar Arias Lopez and colleagues suggest that digital literacy interventions should be designed with patients, and monitored and evaluated to determine effectiveness and sustainability.
Bias in data
AI and machine learning (ML) are limited by the quality of data used to program the technology, which can have the effect of amplifying health inequities.
For example, if health data going into AI/ML models is not inclusive of underrepresented populations, the outputs can lead to generalisations and worse outcomes, according to Kenneth Eugene Paik from MIT University in Massachusetts and colleagues.
“Despite technological advancements, communities with lower health outcomes often continue to have poorer outcomes regardless of the improvements in technology,” they wrote.
Bias in the data and also development of AI algorithms can be introduced in multiple aspects of a project including during data collection and the selection process for datasets chosen for algorithms.
To generate equitable data, Paik and colleagues recommend using accurate, inclusive data, representative of the needs of diverse populations, and continuous monitoring and evaluation to measure impacts of biases in the AI/ML design.
Another paper in the DDOH collection recommends following AI frameworks or checklists to mitigate bias in AI-based health models, as well as strong collaboration among stakeholders to “enhance representation in the AI models”.
Transformation
The rapid development of digital health is seen as a means of delivering efficient healthcare in Australia in an environment that is becoming increasingly challenging with a growing and ageing population and complex care needs, according to an article in the latest Australian Prescriber, by Jodie A Austin, Associate Professor Michael A Barras and Associate Professor Clair M Sullivan.
Digitisation of Australia’s health system could enable data sharing, development of new models of care and integration of data with other services, research facilities and hardware, according to Austin and colleagues.
“Prescribers will have the ability to ‘foretell the future’ as Hippocrates prophesised and make informed decisions based on population data but tailored to meet the individual’s needs,” they wrote.
They describe the digital health transformation across three horizons:
- Implementing digital systems to collect and use data during routine care, for example, dose-range checking and alerts for drug interactions
- Leveraging real-time patient data to create analytics and optimise future decisions – for example, monitoring for adverse drug events
- Integration of data and digital technology to embed new models of care into clinical workflows.
However, the authors acknowledged some challenges with digital health, including the potential introduction of “new types of medication error,” and the need to address complexities in Australia’s complex health system, which includes different digital health platforms across and within the sector, as well as across states and territories.
Privacy of health data is another challenge to consider – “secure methods for data collection, transfer, storage and access need rigorous enforcement through government policy and regulations”, Austin and colleagues wrote.
They recommend the Australian National Digital Health Strategy’s Framework for Action to address system-wide issues and overcome barriers to digital health transformation.
Regulation
The WHO’s six regulatory considerations for AI in health are:
- The importance of transparency and documentation to foster trust. They recommend documenting the entire product lifecycle and development process
- Cybersecurity threats, training models, human interventions and other issues must be comprehensively addressed for risk management. Models should be made as simple as possible
- Externally validating data and being clear about the intended use of AI, which helps assure safety and facilitates regulation
- A commitment to data quality, through rigorously evaluating systems pre-release, is vital to ensuring systems do not amplify biases and errors
- Understanding scope of jurisdiction and consent requirements – such as the General Data Protection Regulation (GDPR) in Europe – for privacy and data protection
- Fostering collaboration between regulatory bodies, patients, healthcare professionals, industry representatives, and government partners, can help ensure products and services stay compliant with regulation.
Further reading
- Regulatory considerations on artificial intelligence for health, World Health Organization
- Digital Determinants of Health collection, PLOS Digital Health
- ‘Digital inclusion’ and closing the gap: how First Nations leadership is key to getting remote communities online, Daniel Featherstone and Lyndon Ormond-Parker, The Conversation
From Twitter
See Croakey’s archives of articles on AI, digital technology and digital platforms
