143 AI-based risk predictions: Evaluation of pilot DRAAI 5 the implementation process. Without this ongoing evaluation, DRAAI would probably have failed. Therefore, our key recommendations include performing continuous reexamination of the implementation process, ethical aspects and quality control and involving patients. Continuous re-examination of the implementation process ensured us to detect that nurses felt that DRAAI’s risk predictions were unreliable. However, we were able to address this issue by distinguishing between PU risk predictions and PU detection. This experience underscores the critical importance of nurses’ trust in AI. As this is the first AI risk prediction model implemented in practice for nurses, it raises the question whether they will maintain this level of scrutiny towards future AI models. On the other hand, it remains essential to consider ethical aspects, rigorously perform quality control checks of AI technologies in healthcare [55, 56]. Especially when relying on AI becomes more relevant. As AI systems become more integrated into healthcare, one potential strategy to enhance trust is to involve patients. By sharing AI-generated predictions with patients, we might be able to foster greater patient engagement and participation in taking measures to prevent the development of PU [57]. This approach may not only empower patients but also promotes a collaborative environment where both healthcare providers and patients work together towards better health outcomes. Recommendations for future research should incorporate the effectiveness of AI DRAAI in reducing PU incidence. A hybrid type 2 study, which simultaneously examines the effect of the intervention and the implementation strategies, could provide more comprehensive insights [58]. Additionally considering using AI for designing educational materials and designing and selecting implementation methods and strategies should be explored [59]. Conclusion This pilot implementation study gave us with confidence that implementing an AI-based prediction model to identify patients at risk for developing PUs is feasible. Continuous evaluation, intensive facilitation, daily reminders, and formally committing to this implementation were key success factors. While DRAAI shows promise in improving PU prevention, future research should focus on its clinical impact and effectiveness in reducing PU prevention. Ultimately, integrating clinical decision tools into nursing workflows could enhance nurses’ ability to deliver more effective PU prevention care.
RkJQdWJsaXNoZXIy MTk4NDMw