During the 6th Amsterdam Medical Data Science meetup on 15 January 2019, the focus was on the challenges in applying artificial intelligence and data science in a clinical setting to improve patient care
Applying data science in a clinical setting.
For Dr Lucas Fleurens – one of the organisers of this monthly event at Amsterdam UMC and an intensive care specialist – practical uses of AI and machine learning in patient care is a subject dear to his heart. “There are all these ingenious models being developed – some that can even outperform doctors when it comes to certain tasks, diagnoses and treatments. But how can we bring these health innovations into the underlying workflow of a hospital – and to the patient?”
The greatest challenge, according to Fleurens, is convincing hospital staff that these new technologies will truly benefit the patient. “But that’s true for any new technology you want to implement,” he added. “So that’s our job: to show through robust research and clinical trials how machine learning can really work to improve the quality of care.”
Big data is like teenage sex
Dr Ronald Petru, one the evening’s speakers, needs no convincing. He is the chief medical information officer at the Radboud University Medical Centre in Nijmegen – and still spends half his time working as a general practitioner.
In 2012, Dr Petru helped implement the electronic patient record system EMRAM to his hospital – which today rates as the only hospital in Europe that enjoys EMRAM’s top Stage 7 award. They have applied the technology widely, including in automating bed planning and patient nutrition requirements.
While Petru regards “data as the new oil”, he also cites Professor Dan Ariely of North Carolina’s Duke University on the current state of the technology: “Big data is like teenage sex: everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it so everyone claims they are doing it.”
Indeed, he sees many challenges ahead. How do you convince all these different individuals, departments, hospitals and institutions to start collecting – and sharing – data in the same way? How do you turn information into data and then back into applicable information again without losing any potentially important meaning? Can we maintain the ‘human touch’? And what about US liability laws if a machine makes the wrong diagnosis? After all, Netflix and Google, who use the same technologies, have it easy: no one dies from getting the wrong advert or recommendation.
Improving electronic patient records
But first things first: how do you make data collection and management part of an organisation’s primary processes? “It’s essential that everyone can see the big picture. It’s not about bookkeeping,” says Petru. “It’s a system that provides a platform for improvement – and not an improvement in itself. It’s only as good as your configuration, processes and user input. So continual training and feedback is essential.”
Petru believes the key motivator is reward: to give the data back to the gatherers for use in their own research – and thereby completing the circle so that everyone benefits. Already, Radboud has developed a Digital Research Environment where you can request data. Once a privacy and ethics committee has approved your request, you can then use the platform for your research and to help publish your fully traceable results.
Dr Martijn Schut, the evening’s second speaker, is a university professor and researcher at the Medical Informatics department at Amsterdam UMC. “I’m not a doctor, I’m a techie,” he says. However, his 25-odd year career in AI has seen him apply machine learning to an assortment of different problems, including helping to decipher the famously incoherent scribbles of some doctors.
Currently, Schut is applying sequence- and process-mining to medical data. In short, he writes software to look for patterns in data and then to make future predictions based on these patterns. For example, the program may recognise that the emergency room gets busier after an important local football match or during certain weather conditions – so the hospital’s staff can be better prepared during these scenarios.
Schut sees a lot of potential in using AI to improve the care pathway of patients. By running raw process data through Disco algorithms, for example, he can generate animated maps that immediately show the locations of potential bottlenecks. “In this way, hospitals can work to enhance their processes. But it can also be used to ensure compliance – that the right processes are being followed. But, perhaps most importantly, it can be used for discovery. “We’re now really entering an age of discovery. As one of my colleagues says: ‘We’ve been feeding data to these machines for 20, 30, 40 years. Now it’s payback time.’”
For more reports from previous Medical Data plus Pizza Meet-ups, click here.
About Amsterdam Medical Data Science
The Amsterdam Medical Data Science Group meetup takes place on the third Tuesday of every month in the Delta Room at the VU University Medical Center Amsterdam’s Intensive Care Unit. The next meeting will take place on 19 february.
The Amsterdam Medical Data Science Group meetings are supported by The Right Data Right Now consortium, which includes Amsterdam UMC, OLVG, Vrije Universiteit, Pacmed, and the Amsterdam Economic Board.