Big Data/Machine Learning/AI
Using Natural Language Processing On Clinical Text For Health Outcomes Research: Challenges And Future Advances Abayomi Adegunlehin* Abayomi Adegunlehin
As clinical data become increasingly available in healthcare, machine learning (ML) has revolutionized many areas, like early intervention for disease progression and comorbidities, clinical predictions, clinical natural language processing, and so on. Despite the immense feats and possibilities of applying ML techniques in the medical field, its impact remains limited. This gap stems from persistent challenges hindering the possibilities of clinical natural language processing (NLP) in healthcare research and patient health outcomes.
Therefore, this annotated bibliography seeks to provide future researchers with an understanding of these common challenges to guide their research. When these challenges are known beforehand and handled, clinical NLP’s possibilities in improving health outcomes will significantly increase.
To achieve the goal of this research, this annotated bibliography systematically identified and analyzed critical challenges within the past decade’s research, drawing from the PubMed Medline Database. The systematic review yielded 15 articles using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The range of clinical topics identified in these articles are radiology, cancer, Joint Arthroplasty, Drug Overdose, Opioid, ischemia, Disorder normalization, HIV Risk Assessment, Pneumonia, Aortic Stenosis, Axial Spondyloarthritis, Atrial Fibrillation, and Liver Disease.
This review study identified critical challenges in patient healthcare outcomes research and provided valuable recommendations for future research endeavors. Among the challenges identified are data quality limitations, model validation issues, data standardization, addressing model bias, and ethical concerns surrounding bias and patient privacy. Adopting these recommendations in future patient healthcare outcomes research will lead to more robust, accurate, and equitable research outcomes. This, in turn, will contribute to enhanced healthcare decision-making, improved patient care, and, ultimately, better health outcomes for individuals and communities.
