When someone is first diagnosed with cancer, patient and doctor jointly need to decide which treatment to opt for from the range of possible treatments. Decision tools have been developed to support this difficult process, but, unfortunately, these tend to be generic and population-based, focus exclusively on long-term survival and lack personalized explanations. Moreover, they are ‘doctor driven’ and hence not easily understandable and accessible for patients. As a result their usefulness is limited.
In this project, we use data from millions of Dutch cancer patients (from the Netherlands Cancer Registry and the PROFILES registry) to empower newly diagnosed patients during treatment decision making. We do this by building new statistical models, determining the advantages and disadvantages of the relevant treatment options for individual patients. Based on the person and tumor characteristics, these models not only include predictions for (long term) survival, but also factors like side-effects of treatment and quality of life after treatment. Moreover, we develop a data-to-text system which automatically generates personalised explanations of the outcomes, using non-technical language and visualisations. This system is enhanced with personalised explanations of uncertainties and risks associated with different treatments. End users (doctors and patients) are involved in all stages of the project, from the beginning, to gauge their wishes and needs, to the end, to evaluate the systems that were developed.
We aim to show that our data-driven, personalised approach makes patients more knowledgeable about different treatment options and empowers them during shared decision making about treatments. The project is a collaboration between Tilburg University (department of Communication and Cognition and the department of Methodology and Statistics) and the Netherlands Comprehensive Cancer Organisation, IKNL (department of Research and Innovation)