131 AI-based risk predictions: Evaluation of pilot DRAAI 5 Process evaluation domain Defined in this study as Outcome Data Source Implementation The process through which interventions are delivered, and what is delivered in practice. Fidelity – % of admissions with at least one at-(high)risk prediction that were followed-up with creating a nursing care plan Routine monitoring data Timeliness – % of admissions that receive preventive measures <24h Routine monitoring data Feasibility of the implementation strategies Field notes Adaptations - alterations made to DRAAI and the implementation in order to achieve better contextual fit Field notes Updates DRAAI Dose – how many implementation strategies were delivered, e.g. in number of visits by the research team Field notes Mechanism of impact The intermediate mechanisms through which DRAAI results in intended (or unintended) effects. Nurses’ responses – how participants interact with DRAAI Acceptability – to assess if nurses had any concerns towards accepting DRAAI (trust, easy to consult, added value) Questionnaire Field notes Feasibility – to assess potential challenges impacting the usability of DRAAI in practice (time, skills, fit with current work processes) Questionnaire Field notes Abbreviations: NA - Not Applicable, PU - Pressure ulcer, DRAAI - Decubitus Risk Alert Artificial Intelligence, EPR - Electronic patient record Finally, the mechanisms of impact: Investigating how the intervention and implementation strategies interact with the participants, and trigger changes [24]. The outcomes are integrated within the previously mentioned aspects, all aspects are elaborated in the following paragraphs and summarized in table 1. Outcomes The study outcomes were based on the Implementation Outcomes Framework by Proctor, Silmere [38]. The selected outcomes included acceptability, feasibility, and fidelity related to the implementation strategies. The evaluated implementation outcomes were adaptations, dose, and feasibility. Additionally, the service outcome of timeliness was assessed for the follow-up of the predictions. Detailed definitions of these outcomes are provided in Table 1.
RkJQdWJsaXNoZXIy MTk4NDMw