Personalized AI-driven decision support tool for Shared Decision Making about adjuvant systemic treatment for breast cancer- integrating prediction modeling with user centred research

This project was a collaborative effort between industry and science to develop a personalized AI-driven decision support tool for Shared Decision Making about adjuvant systemic therapy for breast cancer. With the tool, the consortium - Pacmed, IKNL, PATIENT+, Amsterdam UMC- aimed to provide the best possible personalised decision support for clinicians and patients, using diagnosis, therapy, and outcome data of all representative Dutch cancer patients (NKR-data). The tool was integrated in Zuyderland hospital, where patients and healthcare professionals responded positively.

Through use of the tool that includes the algorithm’s predictions of 5/10-years survival, patients and professionals are optimally supported in their conversations about adjuvant systematic treatment. This will enhance Shared Decision Making, one of the cornerstones of Passende Zorg. It empowers patients in decisionmaking and stimulates health literacy. As a result, patients may now choose more frequently conservative treatment, and – in case of deciding for more invasive treatment- show more compliance. Although it is too soon to conclude this for the Zuyderland pilot, patients included as research participants throughout our project stated consistently that they wanted to know the no-treatment option and specifically if adjuvant systematic treatment benefit is worth the negative impact. The personalized decision support tool provides this information.

The project was unique in that the tool was co-designed and tested with patients themselves, including those health literate patients This contributes to health literacy-proof tools and ultimately to tackling healthcare inequalities.

The key deliverable is the newly developed personalized AI-driven decision support tool, that includes integration with the algorithm to predict 5 and 10- years survival. The consortium sees opportunities to further standardize and implement this approach for other diseases and clinical decisions, after CE marking. The project has also yielded a knowledge-base about translation of complex medical information from a prediction model to understandable information for patients.