A computer framework for analysing networks of locations and topics in a clinical trials database

Abstract

In this article, we aim to extract information from a large online clinical research database; demonstrate the pitfalls of data quality; and create a computer framework for the visualization of the results. Data were downloaded from the clinicaltrials.gov website and loaded into a relational database after cleaning. Network analysis methods were applied to find hierarchical relations and inspect temporal connections in the database. A web-based, open source application "h-Vis" (Hierarchical Data Visualization) was developed to visualize the results. Our results highlight some data quality issues and show what steps were necessary for the cleaning and analysis of the data. Despite the lack of uniform data quality, we were able to extract information from the database and present some statistical overview and visualization examples. Institution-level examination was impossible due to a lot of missing or inaccurate data. City-level connections between the research sites were extracted. Using the visualization tool, it is possible to identify cities with experience in certain disease or medical intervention, or to extract collaboration between medical centres. The extracted information can be useful for researchers to see the latest trends in medical research reaching clinical study phase. In addition, they provide a good orientation for the general audience (e.g. decision makers, investors) about the focus of clinical problems investigated in certain countries or cities. Our custom, publicly available framework might enable to perform further analysis on the data.