JMIR Publications

ABSTRACT

Background:

Early warning systems are widely used to detect acute clinical deterioration, which may be defined as a significant worsening in health over a few hours that may lead to adverse outcomes such as code blue activation, unplanned intensive care unit admission, or death. These systems rely on regular measurement of physiological parameters, such as heart rate and blood pressure, which are converted into warning scores using deterioration prediction algorithms (DPAs). A range of DPAs are currently in use, most commonly simple track-and-trigger tools or summative scoring systems. More complex machine learning approaches have been proposed that may improve prediction accuracy. However, heterogeneity in outcome definitions and reported model performance metrics hinders evidence synthesis needed to support deployment of proposed models in clinical contexts.

Objective:

This scoping review aims to identify the range of DPAs developed for use in pediatric inpatient early warning systems, as well as operational definitions of deterioration and reported performance metrics.

Methods:

The review will follow the Joanna Briggs Institute methodology for scoping reviews and the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) reporting guidelines. The population of interest is hospitalized children. The concept under review is deterioration prediction algorithms, defined as decision-support tools that use routinely monitored physiological parameters to alert clinicians to worsening clinical status. The context will be inpatient ward settings, excluding emergency departments, neonatal units, and intensive care environments. Studies will be identified from searches on the MEDLINE (Ovid), Scopus, Web of Science, Cochrane, and ACM DL databases. Studies will be screened by two independent reviewers against the inclusion and exclusion criteria. A broad range of study types, including prospective and retrospective analyses, will be eligible for inclusion. Data on the choice of algorithmic approach, definition of deterioration, and reported performance metrics will be collated and analyzed. The results will be presented descriptively in tabular and narrative formats.

Results:

At the time of submission, the protocol has been registered and the search strategy finalized. A formal database search has been carried out in April 2026. Screening and data extraction are expected to occur over the following 6 months, after which the findings will be published.

Conclusions:

This protocol describes the planned scoping review of deterioration prediction algorithms for pediatric inpatient care. The completed review will summarize the types of algorithms evaluated, the outcomes used to define deterioration, and the performance metrics reported. These findings will support further evidence synthesis in this emerging field. Clinical Trial: Registered on Open Science Framework at https://osf.io/eg9cs(DOI: 10.17605/OSF.IO/EG9CS)