HARNESSING ARTIFICIAL INTELLIGENCE FOR PREDICTING AND MANAGING CARDIAC ARREST RISK: INNOVATIONS IN EARLY DETECTION, PREVENTION, AND PERSONALIZED CARE

Abstract

The sudden breakdown of heart activity has the potential to become fatal. The prediction of cardiac arrest at an early stage enables proper intervention to stop such occurrences or protect patients from experiencing an arrest. Artificial intelligence (AI) and big data technologies have gained popularity for enhancing the prediction power and emergency planning for vulnerable patients. The research evaluated the methods described in published literature to forecast cardiac arrest events. The study followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) extended guidelines throughout the scoping review procedure. The researchers utilized Scopus together with ScienceDirect along with Embase and the Institute of Electrical and Electronics Engineers and Google Scholar databases for their study search. The researchers checked reference lists from studied publications. Independently two reviewers executed the selection of studies and data extraction tasks. The researchers conducted a narrative synthesis of gathered data from all selected studies. Citation reviews enabled researchers to retain 41 of 697 identified studies while another 6 studies were added after this review process. Artificial intelligence tools used for forecasting cardiac arrest appeared in the investigated studies. Research methods within the 47 reviewed studies consisted of patient-specific analysis for predicting cardiac arrest in 26 (55%) studies as well as AI-based alarm system development in 16 (34%) studies. Fifteen percent (5/47) of the research examined how to recognize patients who were likely to experience cardiac arrest versus those without cardiac arrest risks. Studies that analyzed pediatric patients accounted for only two cases among the total of 47 (4%), while all other studies examined adult populations (45/47; 96%). The majority of studies analyzed data samples which contained less than 10,000 records (32 out of 47 or 68 percent). The usage of machine learning models emerged as the predominant AI technique in 38 out of 47 studies (81%) which investigated cardiac arrest prediction while neural networks became the algorithm most frequently applied across the research (23 out of 47 studies, 49%). Among the evaluation tools researchers applied to their studies K-fold cross-validation stood as the most frequently used (24/47, 51%). The healthcare field uses artificial intelligence technology as a predictive tool in multiple cardiac arrest scenarios. The role of technology will be essential in enhancing cardiology medicine. Additional review is necessary to determine why barriers exist for implementing AI technologies in clinical practice. Research must determine effective methods of supporting healthcare professionals to learn about and implement this technology in their everyday work.