Natural Language Processing of Clinical Notes on Chronic Diseases: Systematic Review (1908.05780v1)

Abstract: Of the 2652 articles considered, 106 met the inclusion criteria. Review of the included papers resulted in identification of 43 chronic diseases, which were then further classified into 10 disease categories using ICD-10. The majority of studies focused on diseases of the circulatory system (n=38) while endocrine and metabolic diseases were fewest (n=14). This was due to the structure of clinical records related to metabolic diseases, which typically contain much more structured data, compared with medical records for diseases of the circulatory system, which focus more on unstructured data and consequently have seen a stronger focus of NLP. The review has shown that there is a significant increase in the use of machine learning methods compared to rule-based approaches; however, deep learning methods remain emergent (n=3). Consequently, the majority of works focus on classification of disease phenotype with only a handful of papers addressing extraction of comorbidities from the free text or integration of clinical notes with structured data. There is a notable use of relatively simple methods, such as shallow classifiers (or combination with rule-based methods), due to the interpretability of predictions, which still represents a significant issue for more complex methods. Finally, scarcity of publicly available data may also have contributed to insufficient development of more advanced methods, such as extraction of word embeddings from clinical notes. Further efforts are still required to improve (1) progression of clinical NLP methods from extraction toward understanding; (2) recognition of relations among entities rather than entities in isolation; (3) temporal extraction to understand past, current, and future clinical events; (4) exploitation of alternative sources of clinical knowledge; and (5) availability of large-scale, de-identified clinical corpora.

Systematic Review on NLP of Clinical Notes for Chronic Diseases

This review paper presented by Sheikhalishahi et al., explores the intersection of clinical narratives and NLP with a focus on chronic disease management. Highlighting the necessity to translate the rich, yet unstructured data within electronic health records (EHRs) into actionable insights, the authors conduct a holistic evaluation of existing NLP applications targeting chronic disease-related notes.

Summary of Methodology

The authors employ the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline to ensure a comprehensive collection of literature. This paper spanned over a decade from 2007 to 2018 and involved screening and analysis of over two thousand articles from multiple databases, eventually distilling the findings to 106 relevant works. These studies were methodically classified into categories of chronic diseases per the International Classification of Diseases, 10th Revision (ICD-10), where circulatory system diseases dominated with 38 studies, and endocrine/metabolic diseases had notably fewer studies, likely due to inherent data structuring differences.

Key Findings and Observations

A notable trend identified is the preference shift from rudimentary, rule-based NLP methods to more sophisticated machine learning algorithms. Despite this shift, the emergent role of deep learning in clinical NLP remains nascent with a mere three studies utilizing deep learning models. This oversight is partly attributed to both a lag in journal publications' coverage of deep learning advancements and insufficient availability of large-scale, annotated corpora for training robust models. The discussion notes the inequity in data availability across disease types, which may impact the deployment of advanced methods.

The majority of studies emphasize phenotyping and risk factor identification over more complex tasks like extraction of comorbidities or integration with structured data. The latter remains an underexplored area that could vastly enhance clinical decision-making processes. There's an ongoing reliance on interpretable yet limited machine learning models such as Support Vector Machines (SVMs) and Naïve Bayes due to the demand for transparent decision-making tools in medical settings.

Implications for Future Research

Despite the compelling utility of NLP in unearthing insights, the paper suggests several pathway improvements. These include advancing beyond mere entity recognition to relational and temporal understanding, which would allow for more dynamic patient trajectory modeling and decision-making support. A significant hurdle remains the dearth of publicly available, de-identified data, which hinders progress in creating generalizable models. The authors advocate for initiatives such as patient data donation frameworks or developing in-situ algorithms to bolster dataset availability.

Additionally, the exploration of transfer learning approaches that leverage existing embeddings could potentially overcome data paucity challenges. There's also advocacy for models that balance performance with interpretability, as this will foster trust and adoption within the clinical community.

Conclusion

The review underscores the burgeoning potential of clinical NLP while candidly addressing existing challenges. With ongoing methodological advancements, especially in areas like deep learning, NLP's role in chronic disease management can be significantly enhanced, paving the way for more informed, precise healthcare outcomes. The paper serves as a call to action for the research community to address these gaps and drive forward clinical informatics capabilities.