Predicting Patient Deterioration with AI: Systematic Review Protocol

This document is a systematic review protocol aimed at identifying the most effective artificial intelligence (AI) and machine learning (ML) algorithms for analyzing physiological data sets to predict patient deterioration in hospital settings. The review will follow the PRISMA and PICOS frameworks and will include studies published from 2007 onwards. Eight databases will be searched to gather relevant literature. The systematic review aims to improve the predictive accuracy of the National Early Warning Score (NEWS2), which is widely used in the UK and internationally for assessing acute illness. The review will also explore the integration of additional variables, such as age and comorbidities, into the predictive models. Ethical approval is not required as the data will be sourced from published studies. The findings are expected to contribute to the literature by providing guidance on model selection for predicting clinical deterioration, thereby enhancing early detection in healthcare systems.