JMIR Publications

ABSTRACT

Background:

Interpretability and visualization of data analysis play a vital role in medical knowledge generation through ‘big data’. An understandable and visually appealing representation of statistical findings improves the communication of research results with clinicians and patients.

Objective:

We aim to capture the results of a thorough statistical analysis based on a complex data source in an intuitive yet expressive way. For this purpose, we introduce HCHSGraphXplore, a novel graph-based data representation of the HCHS dataset, a large cohort study conducted in the city of Hamburg, Germany.

Methods:

The association of subject properties with cardiovascular diseases as well as conditional relationships between properties are measured using robust statistical metrics. The results are captured in a knowledge graph. The generated graph structure is accompanied by a dashboard showcasing distributions of metric properties. All data structures can be accessed and visually explored by researchers, clinicians, and patients.

Results:

An exemplary data analysis is conducted based on the generated knowledge graph and dashboard. This way, we aim to showcase the applicability of the derived data structures. The findings are compared with associations known from literature.

Conclusions:

The exemplary analysis shows the intuitive character and usability of the generated knowledge graph and dashboard. Several well-established biomarkers for cardiovascular diseases are confirmed. Additionally, relationships between laboratory measurements known from literature are found. HCHSGraphXplore balances complex analysis and explainable visual presentation for exploring and communicating clinical research.