Decoding machine learning in nursing research: A scoping review of effective algorithms
Jeeyae Choi,Hanjoo Lee,Yeounsoo Kim‐Godwin
DOI: https://doi.org/10.1111/jnu.13026
2024-09-19
Journal of Nursing Scholarship
Abstract:Introduction The rapid evolution of artificial intelligence (AI) technology has revolutionized healthcare, particularly through the integration of AI into health information systems. This transformation has significantly impacted the roles of nurses and nurse practitioners, prompting extensive research to assess the effectiveness of AI‐integrated systems. This scoping review focuses on machine learning (ML) used in nursing, specifically investigating ML algorithms, model evaluation methods, areas of focus related to nursing, and the most effective ML algorithms. Design The scoping review followed the Preferred Reporting Items for Systematic Review and Meta‐Analysis Extension for Scoping Reviews (PRISMA‐ScR) guidelines. Methods A structured search was performed across seven databases according to PRISMA‐ScR: PubMed, EMBASE, CINAHL, Web of Science, OVID, PsycINFO, and ProQuest. The quality of the final reviewed studies was assessed using the Medical Education Research Study Quality Instrument (MERSQI). Results Twenty‐six articles published between 2019 and 2023 met the inclusion and exclusion criteria, and 46% of studies were conducted in the US. The average MERSQI score was 12.2, indicative of moderate‐ to high‐quality studies. The most used ML algorithm was Random Forest. The four second‐most used were logistic regression, least absolute shrinkage and selection operator, decision tree, and support vector machine. Most ML models were evaluated by calculating sensitivity (recall)/specificity, accuracy, receiver operating characteristic (ROC), area under the ROC (AUROC), and positive/negative prediction value (precision). Half of the studies focused on nursing staff or students and hospital readmission or emergency department visits. Only 11 articles reported the most effective ML algorithm(s). Conclusion The scoping review provides insights into the current status of ML research in nursing and recognition of its significance in nursing research, confirming the benefits of ML in healthcare. Recommendations include incorporating experimental designs in research studies to optimize the use of ML models across various nursing domains. Clinical Relevance The scoping review demonstrates substantial clinical relevance of ML applications for nurses, nurse practitioners, administrators, and researchers. The integration of ML into healthcare systems and its impact on nursing practices have important implications for patient care, resource management, and the evolution of nursing research.
nursing