ESTRO 2023 - Abstract Book

S475

Sunday 14 May 2023

ESTRO 2023

visualization. The aim of this work is to perform a process mining - machine learning based analysis of the major events involving patients’ path of cure in a high-flow Radiation Oncology department. Materials and Methods All patients treated from 2017 to 2021 in our department have been enrolled in the study. An analysis of the main events representative of the care path has been performed. Main events included are: first consultation, dose prescription, CT simulation, plan contouring, start and end of the treatment. Treatment suspensions/cancellation have been analyzed by categorizing underlying causes. A custom script has been developed to refine and convert data coming from our institutional database to event log template. Process mining analysis is expected to be performed by pMineR v.045b. Results More than 10,000 patients, that gave informed consent, were included and more than 100,000 events were considered in the study. Our data refining framework involved data collection and a preliminary analysis for event logs consolidation. The former step involved patient’s age, site of tumor, prescribed dose, dates of the main events, linear accelerator, priority of the treatment. Cause of suspension/cancellation of the treatment were collected for patients undergoing these events. Later, the events and the attributes have been converted in an event log template suitable for machine learning analysis. The results of the preliminary analysis have showed that the 1 out of 5 radiotherapy treatment candidates has undergone treatment suspension/cancellation. Among the most common categories we found logistic issues and clinical impairments, both accounting for more than 20% of the involved cases. Among logistic issues, the most common causes were the rescheduling of the treatment due to machine ordinary and extraordinary maintenance, while among the clinical causes, 1 out of 3 patients had tumor-related complications (e.g. chemotherapy schedule not ended, etc). Conclusion Our work clearly demonstrates the importance of the establishment of a process mining methodology that would allow clinicians to gain more awareness on the actual workflow to apply mitigation strategies and optimize patients’ care path. 1 Maastricht University Medical Centre+, Department of Radiation Oncology (Maastro), Maastricht, The Netherlands; 2 Maastricht University, Department of Health Services Research, Maastricht, The Netherlands; 3 Maastro Clinic, Research Affairs department, Maastricht, The Netherlands; 4 Maastricht University Medical Centre+,, Department of Radiation Oncology (Maastro), Maastricht, The Netherlands Purpose or Objective Artificial intelligence (AI) has the potential to revolutionize cancer care. In spite of the efforts made in developing AI tools, adoption in clinical practice is lagging behind. As healthcare providers strive to improve patient satisfaction and quality of care, patients' attitudes toward AI are important in facilitating this clinical adoption, but few studies have explored this. The purpose of this study is to investigate how cancer patients perceive AI-based decision aids in medical decision-making. Materials and Methods To explore different aspects of patients' opinions of AI systems, semi-structured interviews were conducted covering different scenarios including treatment recommendations, prediction of side effects, survival, and recurrence. Twelve former breast cancer patients were recruited via the Dutch Breast Cancer Association (BVN) for our study and were interviewed. Afterward, all audio-recorded interviews were transcribed verbatim and analyzed using open, axial, and selective coding after which recurring themes and associated quotes were extracted. Results Our results revealed that two factors affect Patients' willingness to use AI-based decision aids: (i) the scenario under study and (ii) the type of information the AI system is intended to provide. In general, patients had a positive attitude toward using AI for less impactful scenarios such as side effects and treatment recommendations but were hesitant toward using AI in situations of life and death such as the prediction of survival and recurrence after treatment. Additionally, patients (n=11) were willing to use AI-based decision aids when used together with their medical doctor, rather than if they use it on their own (n=9). Negative impact on mental health, anxiety, disappointing previous experiences, and the lack of trust in AI were listed among the reasons for patients' reluctance to use AI systems. Figure 1 demonstrates the summary of our findings. OC-0594 Patients' attitudes on the use of AI-based decision aid: A qualitative study H. Hasannejadasl 1 , E. Essink 1 , C. Roumen 2 , C. Offermann 3 , A. Dekker 4 , R. Fijten 1