Seminar 2024-15-03

Network Analysis of Multimorbidity Patterns Across Multiple EHR Systems

Yaomin Xu

Assistant Professor

Department of Biostatistics

Vanderbilt University

Date: Friday, March 15, 2024

Time: 11:00 A.M. – 12:00 P.M. Eastern Time

Location: Nguyen Engineering Building, Room 1109

The phenomenon of multimorbidity, where multiple health conditions coexist nonrandomly within an individual, presents significant challenges to healthcare systems and society. Insights into consistent multimorbidity patterns and the corresponding underlying mechanisms can shed light on developing novel preventative strategies, interventions, and personalized treatments. Large-scale electronic health record (EHR) systems have emerged as vital resources for data-driven biomedical research, offering extensive datasets on real-world patient health dynamics. However, the utility of these datasets is often concerned by EHRs being primarily designed for billing and administration, raising questions about the consistency and reproducibility of EHR-based research discoveries. In this work, we employed multivariate analysis and network modeling to investigate disease multimorbidity patterns and compare them across two major EHR systems. Using International Classification of Diseases (ICD) codes as disease phenotypes, our analyses revealed robust multimorbidity patterns that extend beyond individual EHR systems. Through network models, we investigated the graph-theoretic properties of the multimorbidity network at the local-scale of nodes and edges, the global-scale of network statistics and graph invariants, and the meso-scale of neighboring connection patterns and clustering structures. This provided valuable insights for developing efficient framework to analyze and compare complex structures, offering new perspectives on the inherent complexity of multimorbidities. Our case study demonstrated that the multimorbidity networks replicate known disease condition clusters and showcased a complete online network clustering approach is effective in identifying those clusters, inspired by graph spectral characteristics of the multimorbidity networks. Using hydronephrosis as an example, we illustrated the representation of known clinical disease relationships and highlighted the potential of using the multimorbidity network to deduce causal relationships among disease phenotypes. To facilitate access to these complex datasets and promote further discovery research and hypothesis generation, we have developed a suite of interactive visualization tools for complex online data analysis leveraging data from multiple EHR/Biobank data sources. These tools are designed to enable clinical researchers to intuitively explore the multifaceted relationships within the multimorbidity networks, thereby enhancing our collective understanding and fostering the development of novel medical interventions in the context of multimorbidities. Lastly, our study represents the intersection of data science, statistics, and medical research, offering a novel lens through which to explore the complex patterns of disease multimorbidity as a specific case study. By integrating diverse EHR systems and providing advanced tools for data analysis, we seek to catalyze research dedicated to advancing medicine with real-world big data. We encourage our colleagues to consider the potential of such cross-disciplinary approaches in deciphering the complexities of human health and disease.

Dr. Yaomin Xu, an Assistant Professor of Biostatistics and Biomedical Informatics at Vanderbilt University Medical Center, specializes in leveraging machine learning approaches to extract novel insights from large-scale, real-world health system data, including Biobanks and Electronic Health Records (EHRs). His technical expertise encompasses multivariate data analysis, data visualization, and unsupervised learning, all with a concentrated focus on the application and problem-solving within the realms of statistics, bioinformatics, and health informatics

Nicholas Rios

David Kepplinger