Asthma in children is unpredictable. One moment, a child is running carefree across the football field, the next moment, they are too short of breath to participate. What if a smart AI model could predict when things will go wrong? At Medisch Spectrum Twente (MST), researchers are working on PREVENT: a dashboard that helps parents and doctors stay ahead of asthma attacks.
About 7% of Dutch children have asthma, and in some of them the disease remains unstable despite medication. When does it change? What factors play a role? "That is sometimes difficult for us as doctors to fathom," says Dr Mattiènne van der Kamp, technical physician at MST, Reggeborgh Research Fellow and principal investigator of the PREVENT project.
PREVENT is a collaboration between MST, Ziekenhuisgroep Twente, Deventer Hospital, the University of Twente (TechMed Centre) and the company Evidencio. The Pioneers in Health Care voucher made it possible to start
"Current asthma care does not always match the erratic nature of the disease," explains Van der Kamp. "Children visit the clinic a few times a year, where we mainly look back: how have things been over the past few months? But asthma comes in waves, and anything can happen between two visits." In addition, many children have difficulty recognising shortness of breath, which means that a deterioration is sometimes noticed too late. By then, lung function may have declined to such an extent that hospitalisation is necessary.
Djamael (8) from Enschede loves to play football all day long. As a midfielder, he has to run a lot, but he hardly notices when his asthma flares up. "I hardly ever feel it," he says. His mother Bianca notices it sooner. "You see him lagging slightly behind on the pitch, while he himself says that nothing is wrong. But sometimes his lung function has already dropped to around 50%. That’s what makes asthma in children so treacherous."
In order to be able to intervene more quickly, Van der Kamp’s team developed the PufferApp: an eHealth care pathway that allows children to measure their lung function at home and keep in touch with the care team via an app. "This enables us to give much better and faster advice," says Van der Kamp. "As a result, the number of hospital admissions has halved compared to a few years ago."
Bianca sees the same effect at home: "The PufferApp takes away a lot of the anxiety. If I’m unsure, we do a lung function test. If I don’t trust it, I send a message via the chat." She taps on her phone screen: "It feels like the care team is always there, right here in this little box. That gives me peace of mind."
PREVENT (Personalized Risk EValuation of Asthma Exacerbation using AI based on multimodal data) builds on this. The project combines artificial intelligence (AI) with data from various sources: electronic patient records, home measurements via the PufferApp, questionnaires, but also external data such as pollen counts, air quality and the number of viruses in circulation.
"What is important for one child may hardly matter for another child," explains Tamara Arendshorst-Ruuls. She is a technical physician and PhD candidate on the project. "With AI, we can recognise these patterns for each child and predict when things are likely to go wrong. What is special is that the model not only predicts, but can also explain why. Using so-called explainability techniques, it provides insight into which risk factors contribute to the likelihood of deterioration."
Instead of reacting to deterioration, the team wants to look ahead. "If we know in advance when and why things are likely to go wrong, we can take corrective action sooner," says Arendshorst-Ruuls. "This could involve adjusting medication, avoiding personal asthma triggers or simply being extra alert."
All these insights will soon be consolidated into a personalised dashboard. And the nice thing is that the children themselves thought about the design during so-called co-design sessions. They made dashboards with stickers and colours to indicate what they want to see. "Children wanted lots of colour, playful elements and information about the weather," says Arendshorst-Ruuls. "By designing the dashboard together with them, it will really match how families want to use it at home."
PREVENT is a collaboration between MST, Ziekenhuisgroep Twente, Deventer Hospital, the University of Twente (TechMed Centre) and the company Evidencio. The Pioneers in Health Care voucher made it possible to start
"The University of Twente contributes expertise in the field of AI and computer science," says Van der Kamp. "The hospitals provide the clinical knowledge and data. And Evidencio is helping to ensure that the model can actually be used as a medical device, with all the laws and regulations that this entails."
The fact that the data comes from multiple hospitals is important for reliability. "We don’t just want to train a model locally, we want to ensure that it works for all children with asthma," explains Arendshorst-Ruuls. "Ironically, we are already doing so well here that we have few admissions to train the model," she says. "That’s why we are looking at the bigger picture: not only predicting admissions, but also when extra medication is needed."
The project is still in full swing. In the coming period, the model will be further optimised and tested. The next step is to deploy the dashboard alongside existing care, allowing doctors to test whether the predictions are accurate. Arendshorst-Ruuls is optimistic: "Technical physicians have one foot in healthcare and one foot in technology. We see the problems, know the new possibilities, and can connect them. This project is a good example of that."
Van der Kamp concludes: "My wish is that every child with asthma will soon be able to receive this care. That we no longer have to chase the symptoms, but can stay ahead of them. And that children can simply play football, play and participate with their peers."
The Pioneers in Health Care (PIHC) Innovation Fund is a collaboration between the University of Twente (TechMed Centre), Saxion University of Applied Sciences, and the hospitals MST, ZGT, and Deventer hospital. Each year, the Fund provides ¤600,000 for 10 innovative projects that make smart use of technology for the healthcare of tomorrow. PIHC brings doctors and researchers together to develop new technology for better patient care, or to use existing technology for new medical applications.
Text: Susanne Bosscha, Photocredits: Rikkert Harink fotografie