Urodynamics (UDS) includes invasive and complex tests with technical challenges, and often needs rigorous quality assessment and training of care providers to avoid misdiagnosis. Applying artificial intelligence (AI) to improve the trace quality and support UDS quality assessment, urodynamic pattern recognition and noisy data signals seems promising. Currently, most attention is paid to technicalities and little attention to possible implementation processes with the expectations or desires of users. Therefore, is it unclear how a desirable and appropriate embedding of AI in UDS will look like in the testing and care practice of involved care providers. In this study, we focus on the work of nurses who perform quality assessment before and during urodynamic testing procedures.

We conducted two focus groups with twelve nurses based in two non-academic teaching hospitals in the Netherlands. The protocol contained three themes; 1) How do they imagine the future of UDS with the introduction of AI-systems in their care practices, 2) could AI help improving UDS testing quality, and 3) how are changing work dynamics envisioned? Vignettes with statements, based on previous research, were used to put AI in UDS in a concrete context, to provide hypothetic ideas, and to strengthen the discussion. For example: “In the future, AI systems have access to all patient data and know what to do with different data files. The machine gives a suggested diagnosis and treatment, and tells you 'why' it is the best for a particular patient.” Verbatim transcripts were developed and analysed with an inductive approach to observe and combine various aspects into overarching themes.

We start the results in line with the three themes from the protocol. First, in view of their care practices, the nurses can imagine AI taking over all tasks they currently perform, but they are not afraid of losing their jobs. They argue in favor of ongoing collaboration in which AI can provide insight, but the human decides. Second, quality improvements are mainly envisioned in relation to measurement analysis and catheter positioning. They see significant added value in AI when it comes to positioning catheters and measuring pressures. Third, the envisioned changing work dynamics include AI-assistance in preparing patients, answering questions, performing tests, providing useful/preliminary outcomes, and administration. 
Furthermore, we were able to subdivide these results into five themes: 1) ‘Acceptance’, which nurses do not expect to be a main issue, because AI will become embedded only when it is validated. 2) ‘Doing UDS tests’, in which great value is envisioned when AI can for example answer patient’s questions before, during, and afterwards. 3) ‘Trust and explainability’, in which trust needs to grow while using, and explainability could assist in the process. 4) ‘The role of care professionals’, including changing relationships and tasks with AI. And 5) ‘Future and futuristic visions’ from possible near future towards less realistic scenarios with robots taking over.

Nurses identified five domains in which they see potential value or challenges of AI assistance in UDS: 1. Acceptance 2. Doing UDS test 3. Trust and explainability 4. The role of care professionals 5. Future and futuristic visions. In order to insure the support of key stakeholders each of these themes should be covered in the process of implentation of AI in UDS.

Development of AI tools in functional urology is rapidly evolving. However, there is great need towards qualitative research to assess needs and expectations of users, to align with current work and care practices and achieve successful implementation of AI tools. This study may help to support implementation of AI tools in urodynamic testing, and is a step forward towards explainable AI (XAI) in functional urology.