Application of Epidemiological Geographic Information System: An Open-Source Spatial Analysis Tool Based on the OMOP Common Data Model

Jaehyeong Cho, Seng Chan You, Seongwon Lee, DongSu Park, Bumhee Park, George Hripcsak, Rae Woong ParkInternational Journal of Environmental Research and Public Health 2020 · DOI

TL;DR

This paper introduces AEGIS, an open-source software tool that integrates epidemiological data with geographic information systems (GIS) to visualize and analyze disease patterns. By leveraging standardized data models, AEGIS aims to simplify and democratize spatial epidemiology, enabling researchers to identify health disparities and disease clusters across different countries.

What Problem Does This Paper Address?

Understanding the geographical distribution and clustering of diseases is crucial for public health surveillance, hypothesis generation, and resource allocation. However, performing spatial epidemiological analysis is often a labor-intensive process. This involves acquiring and standardizing diverse geographic information system (GIS) data, preprocessing raw observational health data, and then visualizing the analytical results. This complexity can hinder the widespread adoption of spatial analysis techniques, especially for large-scale, cross-country comparisons. The paper addresses this by proposing a standardized, open-source solution that streamlines these steps.

Methods

The core of this work is the development of the Application of Epidemiological Geographic Information System (AEGIS). AEGIS is designed as an interactive, open-source spatial analytical solution. Its key methodological contributions include:

  • Standardized Data Integration: AEGIS is built upon two key standardized data sources:
  • OMOP Common Data Model (CDM): This model standardizes patient-level observational health data (diagnoses, medications, procedures, etc.) from various sources, making them interoperable. The OHDSI community has already converted billions of patient records into this format.
  • Global Administrative Areas (GADM) database: This provides high-resolution spatial information (administrative boundaries and coordinates) for nearly all countries, offering a consistent geographic framework.
  • Spatial Analysis Functions: AEGIS offers two primary spatial analysis methods:
  • Disease Mapping: Visualizes regional differences in disease incidence and outcomes using methods like choropleth maps, standardized incidence ratios (SIRs), proportions, and Bayesian disease mapping models.
  • Disease Clustering: Identifies geographic areas with statistically significant clusters of adverse health outcomes using spatial scan statistics.
  • Interactive User Interface: AEGIS provides a user-friendly web application interface built with RShiny, allowing users to connect to OMOP-CDM databases, define cohorts (target and outcome groups), set time-at-risk, and select analytical parameters without requiring extensive programming skills.
  • Bayesian Disease Mapping: For areas with small populations, AEGIS employs Bayesian disease mapping models (specifically, the Besag-York-Mollié model) to provide more robust local estimates, allowing for the incorporation of covariates.
  • Spatial Scan Statistics: This method is used to detect statistically significant disease clusters by scanning circular windows across the area of interest.

Datasets

The paper utilizes two primary observational health datasets for its proof-of-concept:

  • National Health Insurance Service-National Sample Cohort (NHIS-NSC) (South Korea): A longitudinal dataset representing 1 million randomly sampled patients from 2002 to 2013, containing individual characteristics and medical treatment data.
  • Data Entrepreneurs' Synthetic Public Use File (DE-SynPUF) (United States): A synthetic database representing 5% of the US population from 2008 to 2010, derived from Medicare claim data.

The GADM database was used as the standard geographic information database. The authors note that the conversion of these datasets into the OMOP-CDM format is publicly available.

Key Insights and Ideas

  • Democratizing Spatial Epidemiology: The most significant idea is making complex spatial epidemiological analysis accessible to a broader audience through an open-source, user-friendly platform. This lowers the barrier to entry for researchers who may not have specialized GIS or programming expertise.
  • Power of Standardization: The paper strongly emphasizes the critical role of standardized data models (OMOP-CDM) and geographic data (GADM) in enabling cross-country comparisons and ensuring the reusability and transparency of research.
  • Bridging Health and Geospatial Data: AEGIS effectively bridges the gap between clinical data and spatial analysis, allowing for the visualization of health outcomes on a map and the identification of geographically concentrated health issues.
  • Identifying Regional Disparities: The tool's ability to highlight regional variations in disease incidence and mortality is crucial for understanding and addressing health inequities.

Key Results

The paper demonstrates AEGIS's applicability and quality through several case studies:

  • Cancer Incidence in Korea: AEGIS-generated spatial distributions of major cancer incidences in Korea were consistent with previous reports. For instance, the incidence of liver cancer in women showed a high Moran's I value (0.44; p < 0.001), indicating significant spatial autocorrelation.
  • Malaria Clusters in Korea: AEGIS successfully identified a primary malaria cluster in Paju-si, Korea (p < 0.001), aligning with known endemic areas.
  • Heart Disease Clusters in the US: When applied to US data, AEGIS identified a significant cluster for heart disease, with Kansas showing the highest annual incidence and a primary cluster identified in the surrounding region (p < 0.001).
  • Validation: The study compared AEGIS outputs with published reports and found that estimated cancer incidences fell within the 95% credible intervals of reported data, validating the tool's accuracy.

Limitations and Future Work

The authors acknowledge several limitations:

  • Data Standardization Requirement: AEGIS requires observational health data to be standardized into the OMOP-CDM format. While the OHDSI community is growing, this remains a prerequisite.
  • Privacy Concerns: AEGIS does not provide latitude and longitude coordinate-based analysis to protect sensitive personal information.
  • False Alarms in Clustering: Disease clustering can sometimes generate false alarms (false positives), necessitating careful interpretation of trends and clusters.
  • Limited Geographic Scope (for this study): The proof-of-concept study primarily used data from South Korea and synthetic data from the United States. The authors suggest that further validation with additional real-world databases from other countries is needed.

Future work could involve expanding the range of diseases and health outcomes analyzed, incorporating more diverse environmental and socioeconomic factors as covariates in the Bayesian models, and further refining the clustering algorithms. The open-source nature of AEGIS encourages community contributions to address these areas.

Summary updated

Abstract

As an open-source, cross-country, spatial analytics solution, AEGIS may globally assess the differences in geographical distribution of health outcomes through the use of standardized geocode and observational health databases.
BibTeX
@article{cho2020application,
  author = {Jaehyeong Cho and Seng Chan You and Seongwon Lee and DongSu Park and Bumhee Park and George Hripcsak and Rae Woong Park},
  title = {{Application of Epidemiological Geographic Information System: An Open-Source Spatial Analysis Tool Based on the OMOP Common Data Model}},
  journal = {International Journal of Environmental Research and Public Health},
  year = {2020},
  doi = {10.3390/ijerph17217824},
}
Figure