Data-driven decision support for the modern radiation therapy process - Technical solutions and computational insights to improve daily working strategies

Abstract

Numbers of cancer patients are continuously growing, as is the demand for radiotherapy (RT). Globally, RT departments are facing many operational challenges due to the increased demand. Additionally, patient waiting times is also noted as a recurring problem in RT. While optimization and scheduling models are well studied, decision support approaches to assist in solving operational/organizational issues within RT are often overlooked. This thesis aimed to enable secondary use of RT data from the Oncological Information Systems (OIS) and explore strategies that can improve different aspects of the RT workflow. In Paper I, A data extraction tool was created to easily extract cleaned and processed RT datasets for historical analysis or as input to external applications. Overall, this reduced the manual cleaning task from approximately 5 to 6 weeks to a few minutes. In Paper II, data from the OIS was utilized to identify characteristics that affect patient queues and to simplify and automate the patient scheduling task. Results suggested that the algorithm may reduce the average delay from 3-4 weeks to 1-2 weeks per month with rare diagnosis groups scheduled promptly within 1-2 weeks compared to an average delay of 3 weeks at the time of the study. This work also identified that definition of ‘wait time’ was ambiguous and referral-to-treatment intervals needed to be further explored to understand if a specific interval time was beneficial for overall workflow efficiency. The impact of various definitions of waiting time used in RT departments was explored in Paper III. Data from 14 Swedish RT departments showed no consistent definition used nationally making standardization difficult. Overall, both local and national data suggested using definitions according to the diagnoses’ timelines to accurately depict departmental delays. In Paper IV, a total of 13 RT professionals were interviewed to understand how transformations over the decade were realized in the clinic and their impact on daily RT practices. Thematic qualitative analysis was used with an aim to find certain recurrent themes between different professions and to what extent reported major events coincided with finding from quantitative OIS data. Operational issues overall were reported as the most repeated concern followed by treatment related changes. Overall, qualitative aspects increased the understanding of effects by changes in patient volumes and department capacities by providing a holistic perspective on the past and current workflows. Overall, various data-driven operational improvement strategies were explored aiming to deepen the understanding of work practices at modern RT departments. Results and integration of the findings will hopefully assist in streamlining the overall RT workflow in the future to allow for a higher patient throughput and reduced patient waiting time.