Machine learning predictive models for intraoperatively acquired pressure ulcers: a scoping review protocol

Abstract

Introduction: Intraoperatively acquired pressure ulcers (IAPUs) are a frequent, costly and largely preventable complication of surgery. Traditional tools such as the Braden Scale ignore intraoperative risk factors, and surgical-specific scales capture only a small part of the data routinely recorded during surgery. Machine learning (ML) models can integrate these data, but the evidence on ML models developed specifically for IAPUs has not yet been mapped.

Objectives: To map and analyze the literature on ML predictive models for IAPUs in adult surgical patients: algorithms, input variables, outcome definitions, validation strategies, performance and calibration metrics, explainability methods, comparisons with traditional scales and barriers to clinical implementation.

Methodology: Scoping review following the Joanna Briggs Institute methodology and the Population–Concept–Context framework, reported according to PRISMA-ScR. PubMed, CINAHL Complete, Scopus, Web of Science, the Cochrane Library, IEEE Xplore and grey literature will be searched for sources published between 2000 and 2026 in English, French, Italian, Portuguese or Spanish. Two reviewers will independently screen and select sources, with a third reviewer resolving disagreements. The protocol is registered on the Open Science Framework [blinded for peer review].

Results: Data will be extracted by two independent reviewers with a piloted charting tool and presented in tables, charts and an evidence map of predictors, accompanied by a narrative synthesis.

Conclusion: This review will map the landscape of ML-based IAPU prediction, identify gaps in the literature and inform the development of an explainable, clinically deployable risk index for the operating room.