From Numbers to Insights: Developing a Visual Cohort Explorer for Feasibility Requests.

Abstract

Introduction: The FDPG (German Portal for Medical Research Data) feasibility portal provides the number of patients who meet the inclusion and exclusion criteria at a national level. In addition, it is possible to perform local installation of the Feasibility Portal, with a view to addressing queries relating to local feasibility. In order to facilitate the accurate interpretation of the cohort count and enhance the comprehensibility of the data at a local level, efforts have been undertaken to develop a visual representation of the local FDPG feasibility portal cohort. This paper aims to address the challenge of providing insights into the cohort while preserving the anonymity of the data for the local version.

Method: In order to be able to visualize the cohort, it was necessary to ascertain a method for extracting the patient data from the cohort definition established in the FDPG. The present study employed the available Medical Informatics Initiative (MII) tools in conjunction with the local feasibility portal to achieve the objective of visualizing the cohort. Subsequently, an investigation was conducted to determine the most efficient approach within the context of the local environment. An interactive R Shiny dashboard was implemented using fhircrackr, echarts4r, and plotly to visualize gender, diagnoses (ICD-10-GM), labs (LOINC), procedures (OPS), and medication (ATC).

Results: Two variants have been developed for the extraction of patient data from our database. The first variant is based on FHIR, while the second is based on SQL. Both pipelines successfully visualized cohort data. The developed Shiny app delivered interactive visualizations validated by clinical experts.

Conclusion: The SQL approach outperformed FHIR in processing time, especially at large scale, while FHIR allows flexible deployment across sites. The implementation is suitable for local deployment. However, implementation on a national scale would require considerable additional effort, data protection and significant improvements to the infrastructure.