The power and limitations of AI at the bedside

On 17 January 2023, the 22nd edition of the Medical Data + Pizza meetups series took place at Amsterdam UMC. The events are organised by the Amsterdam Medical Data Science (AMDS) network – now with over 2,035 members and counting.

While previous editions focused on game-changing (and brain-melting) equations, and how to increase equality and fairness in data science, this edition featured two natural born storytellers obviously experienced at bringing the right stakeholders around a table to get to grips with the fundamentals of driving innovation.

And it’s this continuing dance between stakeholders from science, higher education, the business world and government that makes the Amsterdam region so unique, according to a recent newspaper article in the Het Parool newspaper (in Dutch). And a major reason why the area continues to attract big investment dollars for AI research.

Applying AI to healthcare: part of much longer story

However, we still have a lot of work to do, according to Joanna Klopotowska, assistant professor in Pharmacoinformatics at Amsterdam UMC.

Her presentation ‘AI will see you now: recommendations for responsible development and use of AI in healthcare’ was based on a report (Dutch-language only) she wrote during 2018-19 with a dizzying number of collaborators. “A lot has happened since then – especially COVID-19,” says Klopotowska “And a lot of related reports have been written since, including one in English that echoes many of our findings.”.

Explaining the drive behind her work, Klopotowska says, “This research was really curiosity-driven. At one point as a healthcare professional, I realised that without a thorough understanding of information technologies, I would not be able to responsibly use them.” As the study progressed, she became even more convinced that AI can be a profound source for good – “but only when the results are actionable, trustworthy and aligned with end-user needs.”

And it was the voices of doctors and former ICU patients she brought together in her report on the potential of applying machine learning to predict survival outcomes for intensive care patients. “ICU admission is characterised by high costs not only in terms of money but also emotional costs for the patients, their families and the doctors.” Currently the decision on whether intake is viable is largely made by doctors on the fly. “Hence, a multidisciplinary decision-aiding tool would be very helpful – and a great starting point to better involve the patients and their families.”

Unfortunately, at the moment we are not really ready to develop such a tool according to Klopotowska. “Patients and their families often base their decisions on knowing the resulting ‘quality of life’. But this is a very difficult term to define and there’s still a lot of discussion around it. So how can you put a number on it?”

In addition, there’s still a lot of distrust of the involvement of commercial players. As one patient told her: “Google could easily develop such a model, but I think this is not a suitable party for healthcare. Commercial parties simply have different interests and quality standards.”

In general, Klopotowska recommends boosting transparency and education to provide a better understanding for everyone involved of both the possibilities and limitations of applying AI technologies. Happily, as she notes, data science is already quickly becoming an intrinsic part of the medical curriculum.

And to highlight how applying AI to healthcare is actually part of a much longer story, she ended with a quote from Georg Wilhelm Friedrich Hegel (1770-1831): “Technology can only find its place in human life if it meets the actual needs of the user.”

And now onward to happier endings: data science is already saving lives

Rob van der Mei is Chief Liaison Officer at the Centrum Wiskunde & Informatica (CWI), the national research institute for mathematics and computer science in the Netherlands. As such, he’s spent almost two decades playing cupid between the pure science brains of CWI and the more pragmatic leanings of the business world. He’s also a Professor of Applied Mathematics at the VU, specialising in stochastic operations research which confronts the question of how to make the right decisions in uncertain circumstances.

The CWI has certainly not been shy about driving innovation over the past 75 years: building the first Dutch computer in 1952, birthing the European internet in 1988, developing the popular programming language Python in the 1990s, and now in the vanguard of writing software for quantum computers.

In his presentation, ‘Mathematics and Data Science for a Better Healthcare System’, Van der Mei outlined two recent projects, including a very successful one that received much press for radically improving ambulance response times. More recently, he has been working on applying data-driven optimisation to tackle an ever-growing problem: decreasing resources to deal with an increasingly ageing population.

The multiparty DOLCE VITA project seeks to reduce waiting times for acute elderly care in the Netherlands – and thereby lessen the number of people who literally die waiting for a care bed.

He and his team came up with a more centralised and coordinated system of allocating beds that proved dramatically effective. Currently the average waiting time until placement is 211 days (232 days until preferred placement). Now with this new approach, which includes asking patients for two preferred placements, the waiting time is 33 days (105 until preferred),

“In this way we could apply a middle path between individual preferences and the efficient use of beds. And the nice part is this approach can also be applied to not only elderly care but also to psychiatry and youth care.”

Now that’s a story worth building on.

Are you a researcher looking for inspiration or exposure? You can now sign up on the new AMDS website. 

Sometimes medical professionals and data scientists do not sufficiently understand or even know each other at all. Our Right Data Right Now consortium wants to change that!

At the Medical Data + Pizza meetups, we discuss everything related to medical data science, including interesting projects, papers and the latest developments. This is where physicians and data scientists meet. It’s a great opportunity to learn, meet like-minded healthcare professionals, researchers and data scientists, get feedback on your own projects and have fun.

Read the reports about previous Medical Data + Pizza Meet-ups.

The Amsterdam Medical Data Science Group meetings are supported by The Right Data Right Now consortium, which includes Amsterdam UMC, OLVG, Vrije Universiteit, Pacmed, the Amsterdam Economic Board and Smart Health Amsterdam.