User-Centered Design of A Novel Risk Prediction Behavior Change Tool Augmented With an Artificial Intelligence Engine (MyDiabetesIQ): A Sociotechnical Systems Approach

Abstract

Background Diabetes and its complications account for 10% of annual health care spending in the United Kingdom. Digital health care interventions (DHIs) can provide scalable care, fostering diabetes self-management and reducing the risk of complications. Tailorability (providing personalized interventions) and usability are key to DHI engagement/effectiveness. User-centered design of DHIs (aligning features to end users’ needs) can generate more usable interventions, avoiding unintended consequences and improving user engagement. Objective MyDiabetesIQ (MDIQ) is an artificial intelligence engine intended to predict users’ diabetes complications risk. It will underpin a user interface in which users will alter lifestyle parameters to see the impact on their future risks. MDIQ will link to an existing DHI, My Diabetes My Way (MDMW). We describe the user-centered design of the user interface of MDIQ as informed by human factors engineering. Methods Current users of MDMW were invited to take part in focus groups to gather their insights about users being shown their likelihood of developing diabetes-related complications and any risks they perceived from using MDIQ. Findings from focus groups informed the development of a prototype MDIQ interface, which was then user-tested through the “think aloud” method, in which users speak aloud about their thoughts/impressions while performing prescribed tasks. Focus group and think aloud transcripts were analyzed thematically, using a combination of inductive and deductive analysis. For think aloud data, a sociotechnical model was used as a framework for thematic analysis. Results Focus group participants (n=8) felt that some users could become anxious when shown their future complications risks. They highlighted the importance of easy navigation, jargon avoidance, and the use of positive/encouraging language. User testing of the prototype site through think aloud sessions (n=7) highlighted several usability issues. Issues included confusing visual cues and confusion over whether user-updated information fed back to health care teams. Some issues could be compounded for users with limited digital skills. Results from the focus groups and think aloud workshops were used in the development of a live MDIQ platform. Conclusions Acting on the input of end users at each iterative stage of a digital tool’s development can help to prioritize users throughout the design process, ensuring the alignment of DHI features with user needs. The use of the sociotechnical framework encouraged the consideration of interactions between different sociotechnical dimensions in finding solutions to issues, for example, avoiding the exclusion of users with limited digital skills. Based on user feedback, the tool could scaffold good goal setting, allowing users to balance their palatable future complications risk against acceptable lifestyle changes. Optimal control of diabetes relies heavily on self-management. Tools such as MDMW/ MDIQ can offer personalized support for self-management alongside access to users’ electronic health records, potentially helping to delay or reduce long-term complications, thereby providing significant reductions in health care costs.

Usability and Preliminary Efficacy of an Artificial Intelligence–Driven Platform Supporting Dietary Management in Diabetes: Mixed Methods Study

“…This may need to be considered when the platform is offered to people with chronic health conditions to increase their engagement and motivation to use features for recipes, meal planning, and creating web-based shopping lists. Apart from personalized recipe suggestions, using AI could improve engagement with such platforms and diabetes care [ 48 ]. Although participants were experienced in terms of computer use, and the majority (42/73, 58%) owned a smartphone or laptop computer as demonstrated by the survey, it seemed from the semistructured interviews that participants did not embrace technology fully in daily life, as illustrated by some of the interview participants not taking their mobile phone with them while shopping or their preference to write down their shopping list using paper and pencil as well as to write down their favorite recipes in a notebook.…”

Impact of Artificial Intelligence (AI) on Psychological and Mental Health Promotion: An Opinion Piece

“…Research has shown that involving end users through usability studies in designing AIgenerated applications has been instrumental in ensuring context-based messaging, curbing language limitations and addressing cultural disparities in delivering and engaging health services (Cordova et al, 2015;Shields et al, 2022). Therefore, to address stigma related to mental health, collaborative efforts in the design of interventions are crucial.…”

Planning an artificial intelligence diabetic retinopathy screening program: a human-centered design approach

“…HCD offers a methodology that allows researchers to explore why variation exists in order to create solutions that are integrated within the needs of various stakeholders (40). The iterative nature of HCD enhances usability and avoids unintended consequences by involving intended users throughout the entirety of the design process, gathering feedback to prioritize user needs using qualitative assessments and interviews (41). Beede and Baylor et al, the first human-centered observational study of a deep learning system deployed in a clinical setting, models the benefit of using HCD to evaluate AI systems, as this methodology helps understand end user needs and their environment, demonstrating how contextualizing these factors is key to the successful implementation of AI technology and acknowledging how success in a clinical environment cannot just depend on the accuracy of the technology alone (42).…”