Preventing adverse patient outcomes in home care nursing through predictive process monitoring
The global shortage of nursing staff poses a major threat to health care. Indeed, nursing, especially in the home setting, offers a model of person-centered, preventive and coordinated care that can reduce hospitalizations and help people stay in their own homes. The application of AI and data science can fill some of this gap in the area of early detection of adverse events, such as falls. However, little research in the field of AI and data science is particularly focused on nursing care. In this project a multidisciplinary team of experts will explore how nursing knowledge can be incorporated into a high-quality prediction model.
In the Netherlands, 550,000 patients receive home care. This is a lifeline for many people, especially older people, and plays an important role in maintaining independence, managing long-term conditions and treating acute illnesses. However, patients receiving nursing care are at high risk for complications or adverse events.
Patient complications such as infections, delirium, pain and pressure ulcers are so-called nurse-sensitive outcomes. Nurses are trained to prevent these complications, but research shows that screening and intervention are still suboptimal due to lack of time due to a high workload. Moreover, nurses record a lot of data in patient records that are not used to improve nurses’ work process and clinical decision-making. Modern AI technologies provide a means to detect negative patient outcomes before they arise.
Predictive Process Monitoring
In this study, a multidisciplinary team with experts from University Medical Center Utrecht, Utrecht University and Eindhoven Technical University, will use and develop Predictive Process Monitoring techniques. This branch of data science allows them to predict the outcomes of processes in real-time. The techniques draw on historical data to discover the process and predict the outcome in a particular scenario. This allows organisations to prevent an undesirable scenario by taking countermeasures beforehand.
High quality prediction model
In order to make accurate and reliable predictions, a high-quality prediction model is required. In order to realize such a model, there are three challenges that remain unsolved. The research questions therefore are:
- How to incorporate nursing knowledge into the model? The team will use observations and document analysis (including nursing guidelines) to understand the home care processes and incorporate this into the prediction model.
- How to record the patient-nurse interactions in an accurate way in home care setting? The team will collect data through microphones and transform conversations between patients and nurses to text. They use semantic technology and computational linguistic tooling based on ontologies to transform the information from the conversations to input for the predictive model.
- How to combine the patient and nurse perspective in the model? The team will develop and/or evaluate object-centric process mining techniques to take into account the complexity and interconnectedness of patient-nurse interactions and other aspects that potentially influence home care.
The results of the study will have an impact on different levels. The predictive process monitoring techniques developed in this project will improve the care for older patients who receive home care. It also contributes to decreasing the registration burden of home care nurses by using speech recognition as the for recording patient-nurse interactions. Furthermore, nurses will be supported with information to enhance clinical decision-making. And last, but not least, the project also contributes to innovative development and application of AI in preventive health.
|In 2023, this project team received seed funding from the Institute for Preventive Health to get this research started. During the course of the project, the team will share its findings, a.o. through i4PH’s communication channels.|