Year: 2026
Type: preprint
Abstract: Background Artificial intelligence (AI) is increasingly promising to support decisions in acute cardiovascular care, including whether patients presenting to the emergency department can be safely discharged. Ethical and trustworthy-AI frameworks call for involving patients in designing such tools, yet patient involvement in AI has largely been limited to gathering views, acceptability, and interface preferences. This consultative role leaves largely untested what patients contribute when involved strategically, shaping what a tool is for and how it works. What patients contribute when they help design a high-stakes cardiovascular risk prediction tool has rarely been examined. Methods We conducted a co-design workshop as a pilot in which patients with lived experience of heart failure (n = 11) participated, within a European project developing an AI-enabled cardiovascular risk prediction tool. The study used a two-stage process: preparatory capacity-building through e-learning and an online webinar, followed by a two-day in-person workshop of 5 facilitated sessions. The workshop comprised five facilitated sessions using participatory design activities structured around trustworthy-AI (FUTURE-AI) design-phase principles. Recordings, co-design artefacts, and surveys were analysed using reflexive thematic analysis. Results Patients shifted from a familiar consultative role into an active design role. Working in two independent groups, participants converged on similar designs. Both reframed the tool from clinician-facing to patient-facing and specified: whole-person inputs beyond acute clinical variables; qualitative, and actionable risk rather than a numerical output; the withholding of predictions the model cannot make reliably; and layered, patient-controlled explanation. Because the prediction concerned their own risk of death, participants built emotional safeguards into the design. Conclusion Patients involved as strategic partners produce coherent, concrete design specifications that a purely technical process would miss. Realising this contribution requires preparing patients for co-design methods, treating the tool's intended beneficiary as an open question, and building routes by which patient input reaches development. Trial Registration Not applicable.
Source: Research Square
Language: English
Subfield: Health Informatics
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7204452300
