Improved analysis of treatment data in medical research
New software makes FHIR resources accessible for statistical software
The HL7 FHIR standard data format defined by the Medical Informatics Initiative (MII) is ideally suited for interoperable data exchange and storage. The possibilities for statistical analysis, on the other hand, are limited. To ensure improved analysis of treatment data, the fhircrackr software was developed as part of the MII Use Case POLAR by the University of Leipzig and the University Hospital Jena, which are members of the SMITH Consortium.
The Medical Informatics Initiative (MII) is working with numerous partners to make routinely collected treatment data from healthcare usable for medical research. For this purpose, the MII has agreed on the data format HL7 FHIR Standard for the implementation of the core data set. This is a format that processes heterogeneous data in an interoperable manner, i.e. seamlessly merges them. FHIR data are optimally adapted for the interoperable exchange and storage of medical data. However, they are less suitable for statistical analysis. The FHIR standard stores data in individual, nested and interlinked resources, rather than in tabular structures. Common statistical methods rely precisely on these structures.
To solve the problem, the software fhircrackr was developed as part of the MII Use Case POLAR by the University of Leipzig and the University Hospital Jena, which are members of the SMITH Consortium. fhircrackr allows the user to access a FHIR server using the statistical software and programming language R and to download data from there. The downloaded data can then be converted into tabular form and processed for statistical analysis. In the Use Case POLAR, which deals with health risks in patients with polymedication, medication data available in FHIR can thus be derived from the Data Integration Centers (DIC) and converted into tables by fhircrackr. On this basis, arbitrary aggregations of the data are possible.
Fhircrackr was first published on CRAN (Comprehensive R Archive Network) in July 2020. CRAN is a public online archive for R packages that tests and makes package code freely available. In addition, fhircrackr is being developed open source under the software management service Github. As one of the first solutions for transforming FHIR data into tables, this has been very well received throughout the MII and will be used equally in SMITH use cases such as HELP and PheP and cross-cutting use cases such as POLAR and CORD.