Towards Intelligent Clinical Decision Support Systems for Prehospital Trauma Triage
Doktorsavhandling, 2025
For severely injured patients, minimizing time to definitive care is crucial to increase the chance of survival and to reduce the risk of lifelong disabilities. The ambulance team is therefore key to providing optimal care to injured patients, as the remaining flow of care depends on the accuracy in the patient assessment and the decision on transport destination. However, current tools used for patient assessment of injured patients function as checklists with limited accuracy. It is hypothesized that Artificial Intelligence (AI) is needed to find more valuable patterns in patient data and thereby improve the support for the ambulance team. This work therefore explores a prehospital Clinical Decision Support System (CDSS) that utilizes AI to predict the risk of a patient being severely injured at the incident scene.
First, an AI model named On Scene Injury Severity Prediction (OSISP) was developed and evaluated on Swedish trauma data. Next, the OSISP model was applied and evaluated on Norwegian trauma data to estimate performance and triage accuracy on future patients. Following model development, packaging OSISP as a CDSS was studied by exploring workflow integration in a workshop together with clinical and industrial representatives. Scientific literature was reviewed to find inspiration for how to communicate OSISP’s predictions to the ambulance team. Lastly, a tablet prototype was built and tested together with ambulance teams. The findings of this work demonstrate that OSISP has a theoretical value as a CDSS for ambulance teams during patient assessment of injured patients, both during model development and CDSS building, and offers a promising solution for future work to continue design refinement and initiate prospective evaluations.
First, an AI model named On Scene Injury Severity Prediction (OSISP) was developed and evaluated on Swedish trauma data. Next, the OSISP model was applied and evaluated on Norwegian trauma data to estimate performance and triage accuracy on future patients. Following model development, packaging OSISP as a CDSS was studied by exploring workflow integration in a workshop together with clinical and industrial representatives. Scientific literature was reviewed to find inspiration for how to communicate OSISP’s predictions to the ambulance team. Lastly, a tablet prototype was built and tested together with ambulance teams. The findings of this work demonstrate that OSISP has a theoretical value as a CDSS for ambulance teams during patient assessment of injured patients, both during model development and CDSS building, and offers a promising solution for future work to continue design refinement and initiate prospective evaluations.
Artificial Intelligence (AI)
On Scene Injury Severity Prediction (OSISP)
User Interface (UI)
Prehospital Care
Clinical Decision Support System (CDSS)
eXplainable Artificial Intelligence (XAI)
Field Triage
Trauma
Digital Health
KB-salen, Kemigården 4, Chalmers.
Opponent: Senior research fellow Mark van Heijl, the University Medical Center Utrecht, Netherlands
Författare
Anna Bakidou
Chalmers, Elektroteknik, Signalbehandling och medicinsk teknik
On Scene Injury Severity Prediction (OSISP) model for trauma developed using the Swedish Trauma Registry
BMC Medical Informatics and Decision Making,;Vol. 23(2023)
Artikel i vetenskaplig tidskrift
Bakidou, A, Caragounis, E.C, Andersson Hagiwara, M, Røise, O, Jonsson, A, Sjöqvist, B.A, Candefjord, S. Evaluating Performance and Potential Clinical Benefit of the Swedish On-Scene Injury Severity Prediction (OSISP) Model for Prehospital Field Triage on Norwegian Trauma Data
Human-AI collaboration for prehospital trauma triage: Designing the On Scene Injury Severity Prediction (OSISP) model as a clinical decision support system
Digital Health,;Vol. 11(2025)
Artikel i vetenskaplig tidskrift
Bakidou, A, Seth, M, Andersson Hagiwara, M, Jalo, H, Caragounis, E.C, Sjöqvist, B.A, Jonsson, A, Candefjord, S. Usability of a Prehospital Clinical Decision Support System Based on Artificial Intelligence: A case study of EMS personnel’s experience of using On-Scene Injury Severity Prediction (OSISP)
OSISP, a new digital colleague in the ambulance team
Have you had any injuries? Perhaps you know others that have been injured?
Injuries can happen due to many reasons, for instance sport activities, traffic incidents or violent encounters. It may happen at any time and place, both at night and day, as well as in urban and rural locations. The broad spectrum of situations makes injuries a shared experience by many, and it is a common cause of death and lifelong disabilities.
For severely injured patients, time to care should be minimized. As injuries are typically sudden incidents, ambulance teams often represent the first healthcare contact. Their task is to assess the patient and decide on a transport destination where the injuries can be treated – which may not be the closest hospital. To enable rapid and accurate selection of destination, the ambulance teams need access to good supporting tools during patient assessment. However, current tools function as checklists with limited accuracy.
This thesis has explored the development of the tool OSISP that uses artificial intelligence to analyse patients’ conditions. We developed and evaluated OSISP with Swedish and Norwegian data. We investigated when and how OSISP can be used, and what its interface should look like. Lastly, we tested a prototype of OSISP with ambulance teams to receive feedback on its usability. The results indicate that OSISP can function as a valuable digital colleague in the ambulance, providing a more accurate support compared to current checklists.
Have you had any injuries? Perhaps you know others that have been injured?
Injuries can happen due to many reasons, for instance sport activities, traffic incidents or violent encounters. It may happen at any time and place, both at night and day, as well as in urban and rural locations. The broad spectrum of situations makes injuries a shared experience by many, and it is a common cause of death and lifelong disabilities.
For severely injured patients, time to care should be minimized. As injuries are typically sudden incidents, ambulance teams often represent the first healthcare contact. Their task is to assess the patient and decide on a transport destination where the injuries can be treated – which may not be the closest hospital. To enable rapid and accurate selection of destination, the ambulance teams need access to good supporting tools during patient assessment. However, current tools function as checklists with limited accuracy.
This thesis has explored the development of the tool OSISP that uses artificial intelligence to analyse patients’ conditions. We developed and evaluated OSISP with Swedish and Norwegian data. We investigated when and how OSISP can be used, and what its interface should look like. Lastly, we tested a prototype of OSISP with ambulance teams to receive feedback on its usability. The results indicate that OSISP can function as a valuable digital colleague in the ambulance, providing a more accurate support compared to current checklists.
Ämneskategorier (SSIF 2025)
Medicinteknik
Människa-datorinteraktion (interaktionsdesign)
Artificiell intelligens
Datorsystem
DOI
10.63959/chalmers.dt/5795
ISBN
978-91-8103-338-0
Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 5795
Utgivare
Chalmers
KB-salen, Kemigården 4, Chalmers.
Opponent: Senior research fellow Mark van Heijl, the University Medical Center Utrecht, Netherlands