Opportunities and challenges as Artificial Intelligence adoption grows.
Artificial Intelligence as a concept has been around for decades. It is now gaining popularity because of advancements in software and hardware.
Moreover, with the reduction in hardware cost, the increase in prominence of the power of the cloud now means that anyone can share large data sets efficiently, and the technology finally has the capability to draw insights from the data.
At the Artificial Intelligence in Healthcare conference, Dr Adam Chee, Chief at Smart Health Leadership Centre, Institute of Systems Science, National University of Singapore, highlighted that while various organisations, governments and private entities have been on a rampage to collect as much data as possible, the truth is that not all of this data is actually usable.
Most of the data is collected in a haphazard format, so healthcare organisations end up with incomplete or disorganised data. “Unlike finance or logistics, where there are certain rules you follow in data collection, healthcare is more ambiguous,” he said.
Data in healthcare includes electronic medical records in hospitals and environmental health, which is overlooked a lot of the time. “I’m sure it’s out there somewhere in the cloud. But we don’t marry the two data sets together. There’s social data, which a lot of times we may not necessarily have. An individual’s family history, habits, or economic status can all affect their health, but this isn’t something healthcare professionals have access to. So, most times, we’re analysing incomplete data that focuses on physical metrics,” Dr Chee explained.
Design thinking and digital transformation are also critical areas of focus in healthcare where AI can help decision-making for optimum patient outcomes.
However, there’s still a lot to be done to get healthcare professionals genuinely interested in these technologies, he said.
While Professor Rachel Dunscombe, CEO, NHS Digital Academy, UK, said that different people have different definitions of AI. One of the simple things considered part of the AI family is automation and robotic processes. For example, it can replicate data across other systems saving precious time for doctors. Other things in the AI space include algorithms that can pick up unusual patterns, detect when something may not be correct, and assist in making decisions. “I think that we will see AI assisting doctors in spotting exceptions and bringing to light the most important information about the patient. It will almost serve as a heads-up display seen in aircraft. So, I think we are going see some of this emerging in the coming five years,” she added.
While everyone is still early in the journey of using AI at scale in health and medicine, Tom Lawry, Microsoft’s National Director for AI, Health & Life Sciences, emphasised that there will be barriers that include people not understanding what AI is or the value that’s driven when properly curated in a healthcare setting. Clinicians defining where they can get better is key, allowing teams specialising in machine learning and algorithms to arrive at the right AI solutions. According to Dunscombe, the world is facing a skills emergency when it comes to AI. “We need to educate and train 90 per cent of our workforce so that they have significant digital skills in the next 15 years. We need to start that journey quickly so that people can understand the toolset they have to deliver care differently and redesign it. “I would also encourage them to become proactive in engaging with it rather than being passive about it because AI can help improve the safety and quality of care.
Augments are wonderful doctors. So, it’s not about replacing but augmenting. I would urge doctors to become curious about how technologies can augment and assist their practice and allow them to do more of what they need to do well,” she shared. Another issue that could complicate AI applications in healthcare is that the number-crunchers and statisticians tend not to be trained. As a result, they may notice obvious errors or biases, but the data fed back to the clinicians is skewed. “It’s a dilemma. You can’t start training on stuff that you don’t believe in. But, on the other hand, prediction analysis isn’t mature enough to act on,” Dr Chee added.
According to Prof Dr Nirmal Kumar, President, ENT UK, Consultant Otolaryngologist-Head & Neck Surgeon, Director of Medical Education, Research Network Lead ENT, GM CRN, Honorary Professor, UG, Clinical Lead, Edge Hill University Medical School, Council Member, The Royal College of Surgeons of Edinburgh, UK, there is discussion around working innovatively with clinicians about how they can see patients more optimally and triage patients with AI tools. He said this would help get the right patient to the right doctor at the right time and then do the procedure and discharge the patient safely. This will involve innovation using AI technology and enhanced learning. So, for example, if you’re in the consulting room, you can get the patient’s details, not just the X-rays or the scans, on a much more real-time basis. This will significantly advance the improvement of healthcare delivery.