Ethical Design of Data-Driven Decision Support Tools for Improving Cancer Care: Embedded Ethics Review of the 4D PICTURE Project

Abstract

Oncology patients often face complex choices between treatment regimens with different risk-benefit ratios. The 4D PICTURE (Producing Improved Cancer Outcomes Through User-Centered Research) project aims to support patients, their families, and clinicians with these complex decisions by developing data-driven decision support tools (DSTs) for patients with breast cancer, prostate cancer, and melanoma as part of care path redesign using a methodology called MetroMapping. There are myriad ethical issues to consider as the project will create data-driven prognostic models and develop conversation tools using artificial intelligence while including patient perspectives by setting up boards of experiential experts in 8 different countries. This paper aims to review the key ethical challenges related to the design and development of DSTs in oncology. To explore the ethics of DSTs in cancer care, the project adopted the Embedded Ethics approach-embedding ethicists into research teams to sensitize team members to ethical aspects and assist in reflecting on those aspects throughout the project. We conducted what we call an embedded review of the project drawing from key literature on topics related to the different work packages of the 4D PICTURE project, whereas the analysis was an iterative process involving discussions with researchers in the project. Our review identified 13 key ethical challenges related to the development of DSTs and the redesigning of care paths for more personalized cancer care. Several ethical aspects were related to general potential issues of data bias and privacy but prompted specific research questions, for instance, about the inclusion of certain demographic variables in models. Design methodology in the 4D PICTURE project can provide insights related to design justice, a novel consideration in health care DSTs. Ethical points of attention related to health care policy, such as cost-effectiveness, financial sustainability, and environmental impact, were also identified, along with challenges in the research process itself, emphasizing the importance of epistemic justice, the role of embedded ethicists, and psychological safety. This viewpoint highlights ethical aspects previously neglected in the digital health ethics literature and zooms in on real-world challenges in an ongoing project. It underscores the need for researchers and leaders in data-driven medical research projects to address ethical challenges beyond the scientific core of the project. More generally, our tailored review approach provides a model for embedding ethics into large data-driven oncology research projects from the start, which helps ensure that technological innovations are designed and developed in an appropriate and patient-centered manner.

Keywords: IT; artificial intelligence; big data; decision support tools; ethics; medical decision-making; oncology; shared decision-making.

Figure 1

Schematic illustration of the MetroMap that forms the core of the 4D PICTURE (Producing Improved Cancer Outcomes Through User-Centered Research) project. The MetroMap is a comprehensive visualization of the general care trajectory. Feeding into the MetroMap will be the results of two types of models developed in the project: (1) a treatment outcome prediction tool for each cancer type and (2) a conversation tool developed by analyzing patient experiences through text mining. The result of integrating these data-driven tools into the MetroMap will be a personal care path navigator for each patient that serves as a decision aid in shared decision-making.

Figure 2

Method of conducting an embedded review in the 4D PICTURE (Producing Improved Cancer Outcomes Through User-Centered Research) project—flowchart of the process of identifying and refining ethical challenges based on the literature and discussions with researchers in the project.

Figure 3

Choice architecture—visualization of the default and deviated decision paths.