Dean Russell MP
Member of Parliament for Watford and Chair, Digital Health APPG
The rise of artificial intelligence (AI) in the health arena creates potentially unbounded opportunities to improve patient experience and health outcomes.
Since the turn of the century, AI has gone from a futuristic possibility to a modern-day reality. The technology is being used across industries to replicate – and even outdo – complex human analysis at an unprecedented scale. Therefore, it is no surprise that healthcare would be an obvious choice for its application.
Today, the scale of healthcare AI research is mind-blowing. Global brands such as Google and IBM are dedicating resources to this fast-emerging arena along with a plethora of start-ups. In the UK, NHSX is also working at the cutting edge with the NHS AI Lab, exploring a range of opportunities from early detection of common diseases to finding cures for (currently incurable) diseases.
Big opportunities bring big challenges though, in healthcare AI these challenges are not just technological but also ethical.
As we embark on the journey towards a brand-new future of healthcare, we must make sure no one is left behind.
Tackling healthcare bias
Unfortunately, with AI increasingly unleashed in the real world, one of these ethical issues is the evidence that bias exists even in the minds of machines. In a Harvard article entitled: ‘Algorithmic Bias in Health Care Exacerbates Social Inequities — How to Prevent It’, the article states that “Bias can creep into the process anywhere in creating algorithms: from the very beginning with study design and data collection, data entry and cleaning, algorithm and model choice, and implementation and dissemination of the results.”
The root causes of bias in AI can be complex. However, a commonly referenced issue is the essential need to ensure access to diverse data. A November opinion piece in Scientific American entitled ‘Health Care AI Systems Are Biased’ by Amit Kaushal, M.D., PhD, explored many routes to reducing bias. A critical point was the need to “Ensure the diversity of data used to train algorithms is central to our ability to understand and mitigate biases of AI.”
Improving diversity of data
Any exploration of the potential issues indicates that data diversity is not the only solution, but it is an integral part. Most importantly, knowing the risk of bias is essential; otherwise, we could use systems for decades that have inherited flaws which could unwittingly discriminate across a range of areas from race to disability.
There is no doubt the digital healthcare revolution has begun and AI could be the key to unlocking some of our biggest health challenges. As we embark on the journey towards a brand-new future of healthcare, we must make sure no one is left behind.