In 2023, several interdisciplinary projects received seed funding from the institute for Preventive Health (i4PH). Annemieke Groenenboom has interviewed researchers from four seed funding projects: what are they researching, and how did they start up their interdisciplinary and transdisciplinary research?
Featured in the last episode of 2024, Nienke Bleijenberg (UMC Utrecht) and Iris Beerepoot (UU) about their interdisciplinary project: AI@HomeCare.
Infections or delirium following a fall are common complications for elderly people living at home. Community nurses can detect these issues early and help prevent complications, but they are under significant pressure. To support them, TU/e, UU, and UMC Utrecht are developing a predictive model that will lead to improved care. This was demonstrated in the first year of research, funded by seed money from the Institute for Preventive Health, part of the EWUU alliance.
In the Netherlands, more than 500,000 people receive community nursing care, 80% of whom are over the age of 66. This group often faces multiple complex health issues, putting them at greater risk of acute events, such as falls or confusion caused by infections. “Community nurses are trained to detect these issues early, but in practice, they don’t visit patients’ homes every day,” explains Nienke Bleijenberg of UMC Utrecht.
“At the same time, we’re not yet utilising the nursing documentation they record during home visits in the electronic health record. By developing a predictive model that incorporates this data, we can support community nurses and, in the future, predict complications even before they occur.”
The Research
“For this model, we analyse data from electronic health records using AI and data science techniques, such as Predictive Process Monitoring. This focuses on real-time predictions of process outcomes. The model then provides community nurses with alerts, enabling them to intervene earlier and prevent visits to the emergency department,” Bleijenberg explains.
Developing such a model requires interdisciplinary collaboration. At the start of 2023, Nienke Bleijenberg, Lisette Schoonhoven (Nursing Science, UMC Utrecht), Iris Beerepoot (Data Science, including process mining, UU), Boudewijn van Dongen (Artificial Intelligence, TU/e) and Renata Medeiros de Carvalho (Artificial Intelligence, TU/e) decided to join forces. With seed funding from the institute for Preventive Health (i4PH), part of the EWUU alliance, they launched the research and began preparing a larger grant application.
The Approach
Bleijenberg explains: “Given the central role of data in this project, we first examined what community nurses do during home visits and how well this aligns with the information recorded in patient records. With the help of master’s students, we made 27 audio recordings of conversations between community nurses and patients. Using specialised software, we compared these recordings to the patient records and, with the help of AI, linked the interventions to a nursing classification system. This provided valuable insights into the interventions nurses use for patients with specific care needs.”
”One of the follow-up questions was: what steps are needed to actually develop a predictive model, and how can we optimise it using data from patient records? Each team member contributed their expertise to address this.”
“One of the follow-up questions was: what steps are needed to actually develop a predictive model, and how can we optimise it using data from patient records? Each team member contributed their expertise to address this question. For example, UU explored the possibilities for automating documentation using Care2Report software, which could save community nurses time in the future. While the output wasn’t a perfect report, the foundation has been laid. We also learned valuable lessons about working with audio recordings. In one recording, for instance, a community nurse and a patient discussed wound care without specifying the location of the wound—a crucial detail for the patient record.”
Based on the audio recordings, UMC Utrecht also investigated the extent to which community nurses follow guidelines. Bleijenberg explains: “We integrated the guidelines into the software, using loneliness as an example, and analysed how often the terms and interventions from the guideline appeared in the recordings. This analysis is still ongoing, but we can already see that providing this feedback to community nurses has the potential to improve the quality of care.”
”This analysis is still ongoing, but we can already see that providing this feedback to community nurses has the potential to improve the quality of care.”
“These intermediate steps provide input for improving the software and refining the predictions in the predictive model,” Bleijenberg continues. “We are building step by step, incorporating the perspectives of community nurses themselves through focus groups. For example, one nurse suggested recording only a summary at the end of the home visit instead of the entire visit. This approach is less privacy-sensitive and allows for the use of professional terminology that the software can recognise. By continuously evaluating what works best and aligns with practice, we aim to develop practical and effective solutions.”
Interdisciplinary Collaboration
“The whole is greater than the sum of its parts in this project,” concludes Bleijenberg. “Each team member identified opportunities from their own disciplines to contribute to better care for older adults at home and to make better use of routine healthcare data. A key success factor was the significant time we invested at the start of the project to get to know one another and each other’s expertise. This not only fostered a positive atmosphere but was also essential for translating technological processes into nursing practice—and vice versa. At every step, we check whether the desired outcome has been achieved, and that’s only possible when you truly understand each other.”