Semantic Analysis of SNOMED CT Concept Co-occurrences in Clinical Documentation using MIMIC-IV

2025-09-03

AI Summary

Clinical unstructured text contains rich semantic information, yet mining relationships among medical concepts remains limited by the disconnection between co-occurrence statistics and semantic representations. Method: Leveraging SNOMED CT–annotated clinical notes from MIMIC-IV, we systematically analyze the correlation between concept co-occurrence patterns (quantified via normalized pointwise mutual information, NPMI) and semantic similarity derived from pretrained embeddings (ClinicalBERT/BioBERT), revealing only weak correlation—indicating that co-occurrence fails to capture implicit clinical associations. We thus propose a dual-perspective framework integrating co-occurrence and embedding signals: (i) interpretable clinical topics are generated via embedding-based clustering; (ii) clinically meaningful concept pairs—absent in explicit co-occurrence—are identified using embedding proximity. Contribution/Results: The framework significantly improves downstream diagnostic prediction and prognostic modeling (e.g., mortality, readmission). It enhances phenotyping accuracy and annotation completeness, establishing a novel paradigm for clinical decision support.

Technology Category

• Data Mining & Knowledge Management: Semantic Web
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text

Abstract

Clinical notes contain rich clinical narratives but their unstructured format poses challenges for large-scale analysis. Standardized terminologies such as SNOMED CT improve interoperability, yet understanding how concepts relate through co-occurrence and semantic similarity remains underexplored. In this study, we leverage the MIMIC-IV database to investigate the relationship between SNOMED CT concept co-occurrence patterns and embedding-based semantic similarity. Using Normalized Pointwise Mutual Information (NPMI) and pretrained embeddings (e.g., ClinicalBERT, BioBERT), we examine whether frequently co-occurring concepts are also semantically close, whether embeddings can suggest missing concepts, and how these relationships evolve temporally and across specialties. Our analyses reveal that while co-occurrence and semantic similarity are weakly correlated, embeddings capture clinically meaningful associations not always reflected in documentation frequency. Embedding-based suggestions frequently matched concepts later documented, supporting their utility for augmenting clinical annotations. Clustering of concept embeddings yielded coherent clinical themes (symptoms, labs, diagnoses, cardiovascular conditions) that map to patient phenotypes and care patterns. Finally, co-occurrence patterns linked to outcomes such as mortality and readmission demonstrate the practical utility of this approach. Collectively, our findings highlight the complementary value of co-occurrence statistics and semantic embeddings in improving documentation completeness, uncovering latent clinical relationships, and informing decision support and phenotyping applications.

Problem

• Analyzing SNOMED CT concept co-occurrence and semantic relationships
• Investigating embedding-based semantic similarity using ClinicalBERT and BioBERT
• Examining temporal and specialty-specific evolution of clinical concepts

Innovation

• Uses MIMIC-IV database for SNOMED CT co-occurrence analysis
• Applies NPMI and ClinicalBERT/BioBERT embeddings for similarity
• Clusters embeddings to identify clinical themes and phenotypes