Artificial intelligence (AI) in healthcare has the ability to revolutionise healthcare delivery but has yet to live up to its hype due to structural obstacles. That could change in 2019 with the promotion of AI through policies that encourage innovation. With more than half of the world’s population lacking access to the healthcare services they need, according to the World Health Organisation (WHO), governments and industry can leverage AI to provide tailored, cost-effective, high quality and more accessible healthcare.
AI-based solutions are already reducing the cost of healthcare and freeing up time for patient interaction and treatment by automating administrative tasks such as writing chart notes, prescriptions, and voice-to-text transcriptions. Virtual nursing assistants — apps which provide basic medical information based on the patient’s symptoms — can ‘filter out’ less serious conditions to local doctors and pharmacies, reducing hospital visits and the risk of exposure to illnesses at hospitals.
These efficiency savings, while necessary and helpful, are not the healthcare revolution anticipated by many in the medical world. In the future, AI could play an important role in early detection and diagnosis of illnesses, by using cognitive technology (such as machine learning and advanced analytics) to store and combine scientific data with personalised medical information. For example, AI can suggest the most optimal surgical approaches based on individual patient records, as well as medical research and data on past surgeries.
The technology is developing but several obstacles remain. Privacy regulations on medical data limit what information can be used, while the low case numbers of many conditions prevent bulk analysis. For example, Google’s acquisition and subsequent relocation of DeepMind Health raised concerns over the transfer of UK residents’ data to the US. Also, accusations of bias in AI algorithms, differing resource levels, and different medical cultures limit the extent of usability of AI-based solutions.
Therefore, to use AI to help achieve the WHO’s triple billion targets — an additional one billion people receiving universal healthcare, one billion people better protected from health emergencies, and one billion people experiencing better health — the WHO, in partnership with the International Telecommunication Union (ITU), has launched a Focus Group on Artificial Intelligence in Health (FG-AI4H). The group’s objective is to establish a standardised assessment framework for the evaluation of AI-based methods for health, diagnosis, triage or treatment decisions.
Over the coming year, the Focus Group hopes to develop international standards for collecting, aggregating, analysing, and developing AI models for various stakeholders in the health system including patients, healthcare providers, insurers, and health authorities. While the Focus Group has noted that it does not intend to specify the algorithms themselves, it does represent an opportunity for those developing AI systems to have them evaluated and benchmarked. If proven successful, these AI systems could be adopted in countries around the world.
The widespread adoption of AI could benefit health systems in developing countries that face structural challenges, such as shortages of staff and supplies, accessibility barriers, and lack of awareness on certain health issues. A standardised framework of AI-based solutions can facilitate the uptake of AI and help bridge these gaps.
Some governments are already investing in AI in the health sector. The UK government, for example, is encouraging AI uptake in its National Health Service (NHS), and local trials are being held in several London suburbs for the ‘C the Signs’ tool, which helps doctors recommend cancer investigations or referrals based on the presented symptoms. Governments can continue to promote the development of AI in healthcare by creating thoughtful policies that foster innovation, including:
- Increase investment in ICT infrastructure to enable stakeholders to develop AI solutions.
- Encourage the reduction of algorithmic bias to ensure discrimination and prejudice are removed from AI solutions.
- Develop data privacy regulation that protects data without hindering cross-border data transfer.
- Expand datasets through government investment in statistical agencies that can publish accessible data.
- Encourage data sharing platforms by funding academic and scientific research institutions in order to increase the data pool used by stakeholders.
- Ensure intellectual property regulation properly protects against infringement of algorithms.
- Develop international standards to ensure data portability and interoperability across borders.
Author: Anne Shannon Baxter, Policy Analyst, Access Partnership