[Paper Review] PatientEG Dataset: Bringing Event Graph Model with Temporal Relations to Electronic Medical Records

Xuli Liu, Jihao Jin+5 more|arXiv (Cornell University)|Dec 24, 2018

Biomedical Text Mining and Ontologies3 references4 citations

TL;DR

This paper proposes the PatientEG model, a patient event graph representation that explicitly encodes medical entities, events, and temporal relations in electronic medical records (EMRs) using a semantic graph model. By constructing a dataset of 191,294 events, 3,429 entities, and 545,993 temporal relations linked to a Chinese biomedical knowledge graph, the authors enable efficient, complex clinical queries via SPARQL, significantly improving query expressiveness and efficiency over traditional relational databases.

ABSTRACT

Medical activities, such as diagnoses, medicine treatments, and laboratory tests, as well as temporal relations between these activities are the basic concepts in clinical research. However, existing relational data model on electronic medical records (EMRs) lacks explicit and accurate semantic definitions of these concepts. It leads to the inconvenience of query construction and the inefficiency of query execution where multi-table join queries are frequently required. In this paper, we propose a patient event graph (PatientEG) model to capture the characteristics of EMRs. We respectively define five types of medical entities, five types of medical events and five types of temporal relations. Based on the proposed model, we also construct a PatientEG dataset with 191,294 events, 3,429 distinct entities, and 545,993 temporal relations using EMRs from Shanghai Shuguang hospital. To help to normalize entity values which contain synonyms, hyponymies, and abbreviations, we link them with the Chinese biomedical knowledge graph. With the help of PatientEG dataset, we are able to conveniently perform complex queries for clinical research such as auxiliary diagnosis and therapeutic effectiveness analysis. In addition, we provide a SPARQL endpoint to access PatientEG dataset and the dataset is also publicly available online. Also, we list several illustrative SPARQL queries on our website.

Motivation & Objective

• To address the limitations of relational databases in representing medical events and temporal relations in EMRs.
• To develop a semantic, graph-based model that explicitly defines medical entities, events, and temporal relations for clinical research.
• To construct a publicly available, linked data PatientEG dataset from real EMRs with normalized entity values using a Chinese biomedical knowledge graph.
• To enable complex clinical queries—especially those involving temporal sequences—through a SPARQL endpoint and illustrative query examples.
• To support clinical applications such as auxiliary diagnosis, therapeutic effectiveness analysis, and mortality risk prediction through structured, machine-processable EMR data.

Proposed method

• Propose a patient event graph (PatientEG) model based on the Simple Event Model (SEM), defining five types each of medical entities, medical events, and temporal relations.
• Extract and structure EMR data from Shanghai Shuguang Hospital’s hospital information system into a graph format using event-centric representation.
• Normalize entity values (e.g., synonyms, abbreviations, hyponymies) by linking them to the Chinese Biomedical Knowledge Graph (CBioMedKG).
• Publish the PatientEG dataset as linked data using RDF format and provide a public SPARQL endpoint for programmatic access.
• Design and implement a set of illustrative SPARQL queries for clinical research tasks, including single-event, multi-event, and constrained multi-event queries.
• Adopt best practices in linked data publishing to ensure reusability, extensibility, and interoperability with other biomedical knowledge bases.

Experimental results

Research questions

• RQ1 How can medical events and temporal relations in EMRs be effectively modeled to support complex clinical queries?
• RQ2 What is the impact of entity normalization using a biomedical knowledge graph on query accuracy and expressiveness in EMR analysis?
• RQ3 Can a graph-based representation of EMRs outperform traditional relational databases in terms of query efficiency and expressiveness for clinical research?
• RQ4 How can temporal relations between medical events be systematically captured and utilized in clinical decision support systems?
• RQ5 To what extent can a public, SPARQL-accessible dataset of EMRs with semantic enrichment improve clinical research workflows?

Key findings

• The PatientEG dataset contains 191,294 events, 3,429 distinct medical entities, and 545,993 temporal relations extracted from real EMRs in a Chinese hospital.
• Linking entities to the Chinese Biomedical Knowledge Graph (CBioMedKG) enables normalization of synonyms, abbreviations, and hyponymies, improving query robustness.
• The dataset is published as linked data with a public SPARQL endpoint, enabling programmatic access and integration with other knowledge bases.
• Illustrative SPARQL queries demonstrate the model’s capability to express complex clinical queries involving multiple events and temporal constraints, such as tracking treatment sequences and outcomes.
• The PatientEG model supports efficient execution of multi-table join-like queries by replacing them with graph traversal operations, significantly improving performance for clinical research workloads.
• The model is extensible and can be enhanced with additional relations (e.g., causal links between drugs and diseases) for broader clinical applications.