MERLIN – MachinE leaRning et Lésions de pressIon Nosocomiales

Background

Hospital-acquired pressure injuries (HAPI) are among the most frequent hospital-acquired complications and are among the most common nursing-sensitive outcomes. HAPI are associated with negative outcomes for patients and the healthcare system. The nurse’s role is essential in HAPI prevention and the first step in preventing HAPI is early risk detection. The development of advanced data-driven methods based on artificial intelligence (AI) could improve HAPI risk assessment using routinely collected electronic health record (EHR) data. Several predictive models based on ML techniques have been developed. These predictive models are often based on retrospective data and have not been validated in real-world clinical conditions. Enhancing the early identification of HAPI risk requires the integration of predictive models into clinical decision support system (CDSS) and their practical implementation in healthcare settings.

Overall aim

The overall aim of the project is the development, implementation and testing of MERLIN (MachinE leaRning et Lésions de pressIon Nosocomiales), an AI-based CDSS for the early detection of HAPI combining nursing, implementation, and data-science methods.

Methods

The Exploration, Preparation, Implementation, Sustainment (EPIS) framework and the Consolidated Framework for Implementation Research 2.0 (CFIR 2.0) will be used to guide different steps of this implementation research. For this project, the first three phases of EPIS will be performed.

Exploration Phase

Development and validation of the predictive model (between 2021 and 2023):

This work builds on the implementation of an AI-based, time-aware predictive model for the early detection of inpatients at risk of HAPI developed between 2021 and 2023 (Pouzols et al., 2023). This retrospective observational study was conducted in the Lausanne University Hospital and approved by the Cantonal Commission for Ethics in Human Research (CER-VD/Project ID: 2021-02136).

Design and development of MERLIN (between 2023 and 2025):

To design and to develop the CDSS integrating the predictive model for the real-world and real-time use, design sessions using the nominal group technique (NGT) was conducted.  This qualitative method will allow the participants’ involvement in the design process and achieve a consensus about MERLIN design and features.

Preparation Phase

Context Analysis and implementation strategies (2024):

A multimethod context analysis was conducted in six units of a Swiss university hospital. The CFIR card game method was used to identify implementation determinants, and the NGT to prioritize the implementation determinants identified. A quantitative descriptive study was done to complete these qualitative findings. Based on the results of the context analysis, implementation strategies were identified and proposed to the implementation teams. Using the NGT, the implementation strategies were prioritized, then developed.

Implementation Phase

Implementation of MERLIN (2025)

MERLIN will be implemented in six units, three units with implementation strategies and three without implementation strategies. During this phase MERLIN will be adapted according to the user’s feedback.

Hybrid effectiveness-implementation study (between 2025 and 2026)

A type 2 hybrid quasi-experimental study will be conducted in nine units (three units with implementation strategies, three units without implementation strategies, and three control units). A pre-post comparison of quantitative and qualitative data collected according to the CFIR 2.0 will be used to evaluate the implementation strategies and outcomes. To assess the impact of MERLIN, a descriptive analysis and inter-unit comparison will be carried out.