gms | German Medical Science

Introduction: Medical experts continuously seek to cope with the rapid development of new effective therapy plans. This goal becomes an increasing challenge when targeting chronic diseases like cancer, where many factors should be observed simultaneously to identify the most effective treatments for each case. Besides, healthcare insurance companies seek to examine the clinical evidence of cancer treatments before accepting medical claims. Therefore, a medical information system is needed to provide personalised clinical decision support for Molecular Tumor Board (MTB) experts in cancer care [1] and generating new medical evidence for new treatment options. In this work, we present a technical implementation for a data integration solution to realize such a system using Extract-Transform-Load(ETL) processes to integrate clinically valuable and diverse, multi-source data of cancer patients, including their therapy plans, into a single data model to support an evidence-based clinical decision making.

Methods: Our solution starts with defining the clinical data to be integrated (input) and the target schema (output). First, the data model for both input and output schemas should be defined carefully, and then a mapping from input to output schemas is documented accordingly. Regarding input, we receive both clinical and genetic patient data documented by the Institute of Pathology of Heidelberg University Hospital and exported from the Onkostar system [2] in xml format. Afterwards, the ETL pipelines are built in Talend data integration platform to integrate input data into an output repository in json format following the target schema. Afterwards, the integrated data of all patients can be queried to provide the required evidence-based clinical decision support.

Results: The ETL pipelines are already developed, tested, and set as productive for the regular processing and integration of oncology patients' data for the ZPM project [3], [4] with a complex structure of 200+ data elements distributed among 30 data objects with multiple levels of data hierarchy (up to 3 levels of nested loops). This project seeks to establish a network between regional university hospitals for an aggregated collection of oncology data in the state of Baden-Württemberg, which is referred to as bwHealthCloud (bwHC) [3]. As for April 2023, oncological data for > 300 patients has been already integrated into bwHC. This integrated data is available for therapy planning, evidence-based communication between clinics, and research studies.

Discussion: The implemented data integration solution for investigating evidence-based medical treatments and conducting clinical research is technically feasible, and can be reproduced and transferred to other clinics that use the Onkostar documentation system. However, this solution is limited to the Onkostar-specific XML input schema. In the future, we plan to integrate data from different input schemas of other documentation systems by mapping the data first into an intermediary staging database, where data can be stored in a tabular structured format. Afterwards, these structured data can be mapped to the target schema. The data platform is currently extended from a regional (bwHC) to a nationwide (DNPM:dip) level [5]. Besides oncology, this approach can be also generalized to advance evidence-based personalized medicine in other medical use cases.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.