JMIR Publications

ABSTRACT

Patient messaging technologies offer treatment information and recommendations through web-based platforms, text messages, patient portals, and mobile apps. Many of these technologies have started to incorporate messages that are crafted by artificial intelligence (AI). Such tools are most effective when constructed with theoretical grounding and iterative input from end users. Thus, we outline a theory-based and human-centered design approach for developing patient-facing technology interventions that prioritize diverse patient feedback. We illustrate this approach through a case in which we applied it to the development of an automated messaging intervention that promotes treatment opportunities for patients with chronic pain. Messages were initially developed by subject matter experts and refined using self-determination theory and motivation and behavior change techniques. Using a rapid prototyping approach, we sequentially met with three patient engagement boards to elicit feedback on message prototypes and enhance their content. We synthesized and aligned disparate feedback across boards with motivation and behavior change techniques. Drawing upon the input from the engagement boards, existing co-design approaches, and the field of human-centered AI, we recommend strategies to collaborate with patient partners to enhance health technology designs. Recommended strategies include: 1) involve engagement boards early in the design process, 2) target end users who represent diverse perspectives to reduce biases and promote adoption, 3) acknowledge and set expectations to integrate unique experiences and views, 4) prioritize tailoring of communication for the population of interest, 5) incorporate continual feedback mechanisms, and 6) keep the “human” in human-centered design. Our approach and recommendations can be used to guide multidisciplinary research and design teams to build patient-centered health technologies, with special consideration for AI-guided messaging interventions. As technologies and interventions are adapted over time, linking continual end user input with theoretical grounding is valuable to develop effective tools that can improve health outcomes.