Fellowship Grant: What Does It Take to Bring AI Models into the Exam Room? 

Published on: November 16, 2026

Things are going great, they’re going well—in fact, they’re actually getting better and better. And then, all of a sudden, they’re not. This pattern occurs regularly with Crohn’s disease. A patient may have no symptoms for months and then suddenly experience a flare-up, with blood in the stool, diarrhea, and abdominal pain. Sometimes it’s so severe that hospitalization is necessary. What if you could see such a flare-up coming? 

AI can help predict this. The Fellowship Grant research group “Navigating the Impact of AI in Healthcare” developed and tested an AI model for this purpose. Here’s how it works: patients periodically fill out a questionnaire about symptoms, fatigue, stress, and significant life events. The model makes a prediction based on that data. If the indicator turns red, it indicates an increased risk and an earlier consultation is needed; the doctor can then intervene before the flare-up begins. If it remains green, the patient does not need to come in, which saves time.  

These types of applications offer opportunities for better care. Yet, in practice , they are still rarely used. Why is that, and what can we do about it? Rianne Fijten of Maastricht University is conducting research on this topic, with support from an AiNed Fellowship Grant. 

Only three percent of AI projects in healthcare scale up to everyday practice. These are missed opportunities, and they mean we’re providing poorer care than we could. Fijten sees this happening with his own eyes. “We can build great AI models, but as soon as they enter the clinical setting, they often go unused. That’s a waste of money, a waste of investment, and a waste of time. And ultimately, it’s a waste for the patient, too. We can simply offer a better quality of life with these kinds of AI models.” 

"Navigating the Impact of AI in Healthcare " is a research project funded by the National Growth Fund as part of the Fellowship Grant program and conducted at Maastricht University. It investigates how AI can be used responsibly in treatment decisions, focusing on four areas: trust, communication with patients, organizational obstacles, and the development and evaluation of AI for chronic intestinal diseases. The project runs from 2024 to 2029 and funds five Ph.D. candidates in medicine, computer science, and philosophy. 

Gut feeling 

So why isn’t a model that works well being used? Fijten: “Many factors play a role; one of them is a lack of trust. Doctors wonder: Is the data accurate? Is the prediction reliable? Is this really better than what we’re doing now? When they have that gut feeling, it’s very easy to say, ‘I’m not going to use it; I’ll just keep doing what I’ve always done.’” 

But where does that distrust come from? Fijten hired a doctoral candidate with a background in philosophy to investigate this. She reviewed eighty-two academic publications on trust in medical AI and identified five ways in which trust is conceptualized. These range from building trust by following a framework that can be designed into a system to a relationship that must be maintained. 

In that literature, distrust is primarily viewed as an obstacle that must be removed, not as a phenomenon that must first be understood. According to the researchers, however, building trust in a new, innovative technology such as AI is not a roadmap or step-by-step plan that you can simply check off. It is more like an endless maze in which you must find your way anew each time. The study therefore recommends looking and listening in clinical settings—through observations, interviews, and case studies—rather than systematically addressing trust. 

Who will take responsibility? 

The second obstacle for medical AI lies within the exam room itself. Doctors are required to inform patients that they have used an AI model, but they aren’t sure how to do so. “It’s easy to say that something was generated using AI. But what does that mean? What are the risks? What should patients be aware of? If you explicitly state that you, as a doctor, are using an AI model and communicate that to your patient, then you’re also taking on a certain responsibility for the model’s accuracy.” According to Fijten, this makes some doctors hesitant “to present themselves as someone who relies on AI.”  

The use of AI also carries risks. Doctors tell Fijten that patients are increasingly coming into the exam room with a mountain of information they’ve looked up themselves. “What I’m afraid of is that it’s mainly the highly educated patients who know how to use this information. They also know how to view information critically.” In addition, there is a group with limited health literacy: about one-third of the Dutch population. “This group has to process an overwhelming amount of information about their illness; if information about the risks or impact of an AI model is added to that, it becomes too much. What I’m afraid of is that AI models will widen the health gap. And we absolutely don’t want that.” 

Fellowship Grant: What Does It Take to Bring AI Models into the Exam Room?  AIC4NL | AI for the Netherlands

I think that’s one of the most important things: that we can actually quantify it. That will also help us demonstrate this to doctors and organizations. This isn’t just a fad—it actually delivers real benefits. I’d like to be able to prove that AI brings positive results and doesn’t just cost money.

Rianne Fijten, researcher at Maastricht University

The entire organization needs to get on board 

The third obstacle is the organization. To implement digital innovations, for example, care pathways must change, and IT systems must accommodate them or make room for them. That doesn’t happen on its own. “If you want to integrate an AI model into a hospital where senior management isn’t interested in AI, you can try as hard as you want as a doctor, but nothing will happen. The same goes for the IT department.” 

There are also cases where it does work. “When AI is successfully applied in healthcare, it’s striking that everyone in the organization in question is fully behind it, from the top down to the front lines.” Fijtens’ team is now identifying the bottlenecks for each organization. This should result in an implementation plan over the next six months. 

Multiple disciplines in a single group 

The team consists of researchers from the fields of medicine, computer science, and philosophy. According to Fijten, the fact that this issue is viewed from different perspectives is a benefit. “The multidisciplinary approach is truly one of the most important aspects of the Fellowship. There’s a lot of cross-pollination among the Ph.D. candidates. They really learn from each other, and I learn a great deal from them. And while I’m not an ethicist myself, it’s very valuable that we can also examine these kinds of issues from a philosophical perspective.” 

Despite these differing perspectives, a clear common thread runs through the research: the patient’s perspective. It is still too easily assumed that patients are negative or hesitant about AI. “We often think we know what patients want. But when we ask patients themselves, we often discover that we have the wrong impression,” Fijten notes. This is also evident from her own work. In 2023, Fijten participated in an interview study on AI in the doctor’s office involving twelve women who had had breast cancer. The women were largely open to an AI-based prediction of side effects, but when it came to survival rates, most of them became hesitant. A second round of interviews is now underway, and the initial signs point in the opposite direction. “There’s a shift taking place in people’s perceptions of how they view the use of AI.” 

Go ahead and prove it 

Fijten wants to use the coming period to introduce a practical model—such as the Crohn’s model—into the clinic, including an implementation plan and incorporating her recent insights on communication and trust. She then wants to measure the results to generate evidence. Will decision-making regarding the disease change? Will people choose different treatments? Will there be fewer side effects? Will there be fewer appointments? 

Once the patient is on board, the doctor is the only one left to convince. Doctors want to see concrete evidence that AI provides added value. Fijten: “I often hear doctors say, ‘First, prove what such an AI solution delivers.’ There’s some truth to that, because studies demonstrating that often don’t exist.” 

About AiNed Fellowship Grants

AiNed Fellowship Grants help Dutch research institutions attract and retain top international AI talent. The grant funds the appointment of an AI researcher as an assistant professor, associate professor, or full professor. The Fellowship Grants are part of the AiNed National Growth Fund program and are administered by NWO.

Fellowship Grant: What Does It Take to Bring AI Models into the Exam Room?  AIC4NL | AI for the Netherlands

Related posts

Share via:

Category

Instrument

With an AiNed Fellowship Grant, Dutch universities and research institutions attract top international AI talent to the Netherlands. The grant funds the appointment of an AI researcher as a (tenure-track) assistant professor, associate professor, or

Scope of work

The Dutch healthcare system is under pressure. Demand for care is rising, costs are increasing, and the staffing shortage is growing. AI can help alleviate these issues. Think of reduced administrative work, faster and more accurate diagnoses,

Want to contribute to the development of AI in the Netherlands?

Join AIC4NL

Do you have a question or want to get in touch?

Ask your question at info@aic4nl.nl

Follow us on LinkedIn and stay up-to-date with the latest news about AIC4NL

Join AIC4NL

Are you interested in contributing to AI innovation and becoming part of the largest AI community in the Netherlands?

2 persons AI NED