AI internship for students on reducing administrative burden in healthcare
Health and well-being staff spend at least 25 per cent of their working hours on admin.* Time that could be used for care. Much of this work could be automated with artificial intelligence. The LUMC is the first in the Netherlands to offer medical students AI language model programming internships.
Hospitals have umpteen examples of manual processes that take a lot of time. Doctors, for example, have to summarise information from referrals. This can take a few hours every week. And some researchers have to extract data from patient files for their research – at 200 pages per file and times 2,000, 20,000 or even 200,000 patients. That takes a vast amount of time. Many departments therefore want to use large language models (AI models that can recognise and generate text) to automate such tasks. But doctors do not usually have programming expertise.
Doctors of the future
To meet these future challenges, something must change. Julius Heemelaar (postdoc researcher and cardiology trainee) and Marieke van Buchem (innovation manager at Cairelab, the expertise centre for artificial intelligence at the LUMC) took action. ‘We know that we will have more and more patients coming to us for treatment but we won’t have more doctors or nurses – we will actually have fewer’, says Heemelaar. ‘We know that if we carry on like this, things are bound to wrong. If more doctors have programming experience, they will be able to use that knowledge to reduce the bulk of their admin. You don’t necessarily have to program yourself, but knowing how programming work can already have a big impact.’
‘At present there is no way for future doctors or medical students to achieve that expertise’, Heemelaar adds. ‘So nothing will change despite there being a need for this at the hospital. That’s why we’ve started large language model programming internships for not only Medicine and Technical Medicine students but also students who would like to do this out of interest alongside their regular programme.’
Programming basics
Marit Hamer is one of the first students to do this internship alongside her regular degree programme. ‘My programme had no artificial intelligence training, so I’m learning the basics of programme in my internship. It’s a lot of trial and error but you always get a step further in the end.’
‘The aim is for students to be able to program’, says Heemelaar. ‘Most have no programming experience, so they are given an introduction programme where they learn the basic machine-learning principles. Then they soon set to work on their own projects. They obviously work in a secure environment to ensure the data does not end up elsewhere. We now have five interns, four at the LUMC and one that I’m supervising at OLVG in Amsterdam. We hope departments will eventually take on the tools but the projects are still too young for that.’
Collaboration
Hamer likes working on her own project best – and working with other experts at the LUMC. ‘I like how it’s easy for us to connect. We talk to the people behind the supercomputer network, who are always eager to help, and to people from the departments and among us interns. We do and share a lot together, which helps.’
‘Everyone is also itching to do something with new technologies like generative artificial intelligence (comparable with GPT-4)’, says Heemelaar. ‘That’s what you keep hearing but it doesn’t get done because no single person is working on it. Our IT staff are pleased to see that students want to work with their methods. And the doctors are pleased because their work is suddenly much faster.’
Future of AI in healthcare
Hamer and Heemelaar both see the opportunities AI presents for the healthcare of the future. ‘I hope people will be more open to the idea that AI in healthcare can help’, says Hamer. ‘And I think that in a few years the degree programmes will focus more on using it. Now there is an AI course for medical students at the LUMC but that didn’t exist in my year.’
‘I’m 100 per cent convinced that this innovation will soon help the hospital tackle the challenges it faces’, says Heemelaar. ‘Obviously we will all have to take extra care that it is done responsibly and that data remains protected. My biggest hope is that this technology will mean more time for patients and less time doing admin. That is not just because I always think we should have more time for patients but also because we need to.’
*From a study commissioned by the FNV in May 2024.
To find out more about AI at the LUMC, see the Cairelab website.