Impact of artificial intelligence on hospital admission prediction and flow optimization in health services: a systematic review.

Abstract

Artificial Intelligence (AI)-assisted prediction of hospital admission is an innovative tool that optimizes resource allocation and improves patient flow within emergency departments. Health institutions need to decongest these departments to maintain sustainability and become efficient. Increasing demand and excessive competition for limited resources directly contribute to these challenges. This systematic review aims to evaluate the use of artificial intelligence in the prediction of hospital admissions, evaluating the accuracy of machine learning models, their impact on clinical decision- making, and their role in the optimization and allocation of resources. A systematic review of studies published between 2019 and 2024 followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Studies were selected based on a pre-defined Population, Intervention, Comparison, and Outcome (PICO) framework, and methodological quality was assessed using the Quality In Prognosis Studies (QUIPS) and the Checklist for Artificial Intelligence in Medical Imaging (ChAMAI) tools. A total of 20 studies were included; most of the studies evaluated were retrospective. AI-based models demonstrated superior accuracy (85 % to 95 %) compared to traditional methods, with Random Forest (RF) and Neural Networks outperforming classical statistical models. Studies incorporating unstructured data through Natural Language Processing (NLP) have significantly improved patient flow and resource allocation. The integration of predictive analytics resulted in a reduction in avoidable hospitalizations, optimized bed occupancy, and a decrease in emergency room overcrowding. AI-driven admission prediction is promising in the hospital setting, as it improves efficiency and allows for proactive and rapid decision-making, optimizing available resources. Future research is promising and should focus on prospective studies to validate practical applicability.