127 AI-based risk predictions: Evaluation of pilot DRAAI 5 DRAAI provides daily predictions, categorizing patients into low risk, at-risk, or highrisk for developing PU during hospitalization. Nurses access these predictions through a stand-alone web application, with defaults in displaying only at-risk and high-risk patients, though it can be configured to show all patients. A feedback loop is integrated to provide real-time updates on actions taken in response to the predictions, aligning with the hospital protocol that involves creating nursing care plans for at-risk patients. Figure 1 shows the DRAAI dashboard as it appeared at the start of the pilot study. DRAAI operates as a clinical decision support system (CDSS), designed “to enhance decision making in the clinical workflow” by providing tools such as alerts, guidelines, order sets, templates, and summaries [29]. Utilizing person-specific data and reasoning mechanisms, a CDDS generates and presents valuable information to nurses [30]. The intervention is founded on three causal assumptions daily risk predictions will: 1) reduce the perceived registration burden among nurses 2) increase focus on PU prevention, and 3) increase timely application of preventive measures leading to a reduction in PU incidence. Context - At baseline The context at baseline was assessed through observation, reviewing current protocols and informal conversations with nurse champions and managers from each ward. Nurse champions are dedicated to PU prevention and wound care. They serve as primary contacts for ward nurses, provide education, address questions, identify problems, and monitor quality of care related to PU prevention and management. The context is described into detail in table 1. Implementation The Implementation Research Logic Model (IRLM), as outlined by Smith, Li and Rafferty [31], provides a structured approach for examining the relationships between an intervention, here DRAAI, barriers and facilitators, implementation strategies, mechanisms of change, and implementation outcomes. Each of these components is informed by implementation frameworks, models, and theories, as illustrated in Figure 2. In the left column of Figure 2, barriers and facilitators –referred to as determinants – are categorized using the NASSS (non-adoption, abandonment, scale-up, spread, sustainability) framework [32]. Determinants were identified by DS through a review of literature, particularly studies on implementing prediction models [33] and clinical decision support systems [34]. Research on the implementation of (AI-assisted) clinical decision support systems implementation of (AI-assisted) clinical decision support
RkJQdWJsaXNoZXIy MTk4NDMw