Is AI in shape to transform healthcare?

AI and robotics certainly have the potential of transforming healthcare: prevention, be it through technology and apps encouraging a healthier lifestyle, early disease detection, faster diagnosis and decision-making, better disease management, end-of-life care supported by humanoid robots, advanced drug research processes, etc. The opportunities for engaging emerging technologies in this field are tremendous.

All of these applications hold the promise of improving health outcomes, boosting patient experience and creating easier access to health services. They either increase productivity or provide more and better care to people. Some require highly complex AI systems, while others are less complicated. But in general, the adoption of AI applications in healthcare takes time and we must avoid the trap of overhyping the potential.

So where are we today? What are the most promising use cases? And can we accelerate the scaling of current AI-driven healthcare applications? 

AI as the engine for growth

By 2050, one in four people in Europe and North-America will be over the age of 65, so healthcare systems must prepare for dealing with a greater number of patients with more complex needs.

Yet today, healthcare is among the least digitised sectors in Europe, lagging behind in digitisation of work and processes. In order to fully grasp the benefits of AI-powered advances in healthcare, the sector needs to get the digital basics in place first.

According to Accenture analysis, key clinical health AI applications can potentially create 150 billion dollars in annual savings for the US healthcare economy by 2026.

A 2018 survey conducted by NHS England and the AHSN Network AI Initiative came to similar conclusions: 94% of the UK’s AI thought-leaders cite AI as being extremely important or very important for diagnostics, 89% support this view for operational and administrative goals, and 79% have this opinion in regard to the benefits for health promotion and preventative health solutions.

Solving real problems

Taking into account that AI-applications in healthcare range from very complex (e.g. diagnostic decision support systems) to less complex (e.g. image processing), it is safe to say that the most likely ones to be adopted at scale are the ones that solve frequent frustrations or problems encountered by care professionals. According to the NHS report, these game-changing use cases can be found in the domains of diagnostics, non-clinical processes (operational and administrative efficiency) and health promotion and preventative health.

Concrete examples in the areas cited above include: assisted reporting and screening in radiology and dermatology, prioritisation of scans to be brought to human attention, increased efficiency in paperwork and scheduling, workflow management and predictive modelling for preventative healthcare. But also many more applications facilitating the shift from reactive care to a more preventative health model in which people are more empowered to take care of their own health.

Transforming the workforce

None of these applications aim at replacing human care professionals by machines.

On the contrary, a recent McKinsey report states that only 35% of time spent in healthcare is potentially automatable. And we must add that ‘potentially automatable’ is not the same as ‘likely to be automated’.

Most AI uses cases will rather increase the ability of care professionals to focus on jobs that draw on uniquely human skills, like empathy and big-picture integration. On the other hand, there is this harsh reality of a significant workforce gap. The overall demand for healthcare workers is expected to grow to 18.2 million across Europe by 2030. The current supply of 8.6 million nurses, midwives and healthcare assistants will not meet that future need. The sector as a whole could thus greatly benefit from the adoption of AI to remove or minimise time spent on routine administrative tasks, which can take up to 70% of a healthcare professional’s time.

Preparing healthcare for AI

So what is holding us back? It is not the technology readiness as such, but rather everything required to ensure adoption in daily practice. AI systems must be approved by regulating bodies, integrated in health systems, standardised to a sufficient degree, taught to clinicians, paid for by public or private partners and updated on a regular basis. And then there some crucial ethical questions to be answered as well. The use of smart machines inevitably raises issues of accountability, transparency, permission and privacy. Let’s be honest: it will not be easy to accept a diagnosis by a machine that will not necessarily be able to explain why and how it came to a specific diagnosis. Practitioners too will want to understand how it all works and may be very reluctant to trust the concept of AI as a black box.

These hurdles won’t be overcome overnight, and there is no simple answer on how to accelerate AI adoption. But it does make very much sense to start addressing the low-hanging fruit in order to gain knowledge and trust to then start moving towards more complex AI-powered solutions.

In the meantime both healthcare organisations, health systems and governments must start preparing for change and improving the availability and quality of health data as a prerequisite.

So yes, AI is definitely in shape to transform healthcare. Now the healthcare sector itself must start building up speed to benefit from its value.