What will be the role of Artificial Intelligence (AI) in healthcare? An interview with three experts: Daniel Nathrath (Co-Founder and CEO, Ada Health), John Crawford (European Healthcare Industry Leader, IBM), Jama Natequi (Co-Founder, Symptoma).
What AI can offer in healthcare? In which areas will it substitute or supplement doctor’s work?
Daniel Nathrath, Ada Health: On a daily basis, healthcare professionals make tough decisions that affect the wellbeing of their patients. Yet, with increasing demand for services and a global shortage of healthcare workers, there’s less time for physicians to gather a comprehensive picture of each patient’s case. AI-powered health tools can empower patients with the information they need to better understand their condition. These personalized records can also be used to give doctors a more timely, holistic picture of the patient’s health, helping them to make more informed decisions and allowing them to spend more time on patient care and prevention.
AI-powered scanners and symptom assessments can also make a big difference when it comes to understanding and identifying rare diseases. These ‘orphan’ illnesses affect less than 5 in 10,000 people, which makes them very difficult for doctors to identify and diagnose, and patients may sometimes spend years visiting multiple specialists and undergoing tests before they can be treated. Advanced AI has the potential to cut those years down to seconds by suggesting possible rare conditions to doctors who may not otherwise think to consider them. There are many areas where AI can have a meaningful effect on healthcare and medicine by facilitating more real-time sharing of resources and knowledge, and by supporting doctors in analysing new patterns of disease and health issues on a whole new level.
John Crawford, IBM: The use of AI in healthcare is not new. For example, the 2006-8 EuResist project that IBM was involved in, to develop decision support for HIV treatment, using large databases of previous case histories and viral genome data. Predictive algorithms were developed and refined, leading to antiretroviral treatment plans as good as expert doctors could produce (HIV Medicine 12(4) pp. 211-218). Since then, AI (and in particular Machine Learning) has found its way into many areas of medicine, particularly those involving image analysis such as radiology and dermatology. It also shows potential in cardiology, using ECG and other data to predict heart problems. We also see the emergence of AI in the form of ‘chatbots’, allowing people to have conversational interactions about their symptoms or treatment options. A recent example is the virtual assistant IBM developed for Arthritis Research UK. So far, AI has been limited to relatively narrow problem domains, where it has allowed doctors to make better decisions, or helped citizens to understand health issues. There is much more scope for AI to be used as a form of triage, so that less urgent medical needs can be handled more quickly and effectively. AI may also take over some of the more routine and time-consuming diagnostic tasks, allowing doctors more time to focus on the human aspects of working with patients. It seems unlikely that AI will completely replace doctors, at least in the short term, although it will change some medical specialisms.
Jama Natequi, Symptoma: In my humble opinion, in the next five years, AI will first supplement and then probably even start substituting tasks demanding the interpretation of large amount of data within the next 10 years. This comes in handy for diagnostic decision making especially.
Advanced AI has the potential to cut those years down to seconds by suggesting possible rare conditions to doctors who may not otherwise think to consider them.
What are the biggest challenges, chances and threats of AI in healthcare?
Daniel Nathrath, Ada Health: Simply put, the biggest challenge is that healthcare is extremely complex and, of course, there may be very serious consequences if anything goes wrong. Unlike other areas of tech, you cannot launch a minimally viable product in healthcare – you have to be sure that your technology is safe and accurate.
This is why we rigorously test our technology to make sure that it is as accurate and as easy to understand as possible. We also make sure that we are very clear about how Ada should be used. For instance, Ada does not offer a diagnosis. It offers a health assessment which aims to provide the patient and their doctor with information about the possible underlying causes for their symptoms.
Of course, the potential benefits once you get this right are vast. The biggest opportunity lies in empowering people to play a much more active role in managing their own health. We need solutions that shift care to more appropriate, effective and convenient patient-centered approaches that support people through every stage of their healthcare journey.
John Crawford, IBM: The biggest challenge remains the availability of a sufficient quantity and quality of data, to train AI systems, and to allow them to continuously improve. There is also insufficient independent evidence of the effectiveness and safety of AI solutions, but this will surely develop faster over the next few years. To gain the support of the medical profession, transparency is also important. It must be possible to understand why AI systems have produced their recommendations, and to look at the evidence behind them, as we have done with the IBM Watson for Oncology solution. Finally, it is vital that AI is deployed in such a way that it does not replace human contact and judgement, especially where complex needs have to be assessed, or decisions made about treatment plans.
Jama Natequi, Symptoma: A major technological challenge to overcome are the ambiguities, lack of available information and the structure of data for any kind of interpretation. However, the biggest challenge is not technological, but philosophical. Are we, as a society ready to allow a machine to tell us, what condition we might have?
Once these challenges have been overcome, we might be able to provide everyone in the world access to free healthcare at the highest levels. This will disrupt healthcare as we know it. Increasing costs, the diverse quality of healthcare given, the lack of doctors and specialists … these will all become problems of the past (our present).
The biggest threat we are facing has probably already become a reality. We have willingly and unwillingly allowed companies and governments to collect our data. As always, this may be used for good or bad. Advancing AI into healthcare is only adding another layer to the existing situation.
The biggest challenge remains the availability of a sufficient quantity and quality of data, to train AI systems, and to allow them to continuously improve.
When and under what conditions will the use of AI in healthcare become common?
Daniel Nathrath, Ada Health: The desire from patients has been there for years now, with people commonly Googling their symptoms to get more information about their health. New AI-powered technologies have the potential to give users much more accurate, personalized information and to truly change the way people engage with and manage their own health. They are already doing so, in fact: Ada’s mobile app has completed more than 5 million assessments since it was launched, with over 22,000 now completed every day.
It will take several years before AI becomes fully integrated into healthcare systems but the move towards a more patient-centric, digitally enabled, proactive approach to health management has begun and it is really building momentum. In order for the full potential of AI to be realized and become commonly used within the healthcare ecosystem, it will be critical for emerging players, traditional healthcare companies, providers, insurers and regulators, to work together to establish clear policies that encourage patient-centric innovation and safety.
John Crawford, IBM: AI is already being used in healthcare, and its use is likely to accelerate fastest in areas where the AI solutions are at least as good as the alternatives, but provide faster access and more convenience. Certainly in areas such as improving health literacy, symptom checking, and the rapid analysis of medical images, it will become common in the next few years. It will take longer to find its way into clinical pathways, but when it does, it will transform the ability of doctors to speedily make the most accurate diagnosis and best treatment decisions – precision medicine will become a reality.
Dr Jama Natequi, Symptoma: It is apparent that a lack of doctors within an aging population will at some point threaten the collapse of the various healthcare systems worldwide. At the very least, this breaking point will provide advanced AIs an opportunity to demonstrate their value at scale.
It will take several years before AI becomes fully integrated into healthcare systems but the move towards a more patient-centric, digitally enabled, proactive approach to health management has begun and it is really building momentum.