Technical Concepts for Cross-Institutional Data Sharing in Neonatology

Zeitschrift für Geburtshilfe und Neonatologie, vol. 230, pp. S111

Abstract

Objective: Research on neonatal data is often constrained by high inter-patient variability and small patient numbers even in large centres. Especially the development of algorithms for complication detection or prediction, and the use of artificial intelligence benefit from large sets of recorded data. The accumulation of such data is challenging. Monocentric data is limited to a particular cohort and likely to suffer from an institutional bias. The possibility to share neonatal data between institutions in an effective, data-secure and legal way is therefore a topic of interest. We propose possible technical implementations of such a shared database, which address these challenges. Methods: The core idea of a shared, neonatal database is a fixed data scheme, which allows for the extraction and processing of cross-institutional data. Such a data scheme needs to be fitted to the needs of the collaborating partners, while also giving the option of future adjustments. The architecture should therefore be open for unforeseeable demands. Data distribution could generally be achieved in two different ways: One approach would be a centralised database, to which all collaborating partners synchronise and fit their data into the determined data scheme. All partners then request data from one database instance, in which all data is stored in one scheme. This option comes with easier and faster data queries, and less dependency on local servers. The second option is a decentralised data hosting, in which a central server handles queries to the different databases of the collaborating partners. Requests would be performed in a standardised format and then transformed, if possible, for each database to extract the correct information in the determined scheme. This approach offers a higher flexibility. Depending on the differences in the databases, the performance of the queries is less efficient and more complex. Nevertheless, the flexibility of this approach would allow for changes in local databases and the so called “loose coupling” of the different databases. Both options should benefit from existing schemes, projects (e.g. MII), and query technologies, thus building on established work and allowing for potential interfaces to other applications in the future. The compliance with the European data protection law by pseudonymisation or anonymisation and the establishment of suitable committees in early stages of sharing can promote transnational collaboration.

Authors 6

1. Mike Grüne Aachen Klinik für Kinder- und Jugendmedizin

   Universitätsklinikum Aachen

   Affiliation as printed

   Uniklinik RWTH Aachen, Klinik für Kinder- und Jugendmedizin, Sektion Neonatologie und Pädiatrische Intensivmedizin, Aachen, Germany

2. Lena Olivier Aachen Klinik für Kinder- und Jugendmedizin

   Universitätsklinikum Aachen

   Affiliation as printed

   Uniklinik RWTH Aachen, Klinik für Kinder- und Jugendmedizin, Sektion Neonatologie und Pädiatrische Intensivmedizin, Aachen, Germany

3. C Oprea Aachen Lehrstuhl für Embedded Software (Informatik 11)

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen, Lehrstuhl für Embedded Software (Informatik 11), Aachen, Germany

4. T Orlikowsky Aachen Klinik für Kinder- und Jugendmedizin

   Universitätsklinikum Aachen

   Affiliation as printed

   Uniklinik RWTH Aachen, Klinik für Kinder- und Jugendmedizin, Sektion Neonatologie und Pädiatrische Intensivmedizin, Aachen, Germany

5. A Stollenwerk Aachen Lehrstuhl für Embedded Software (Informatik 11)

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen, Lehrstuhl für Embedded Software (Informatik 11), Aachen, Germany

6. Mark Schoberer Aachen Klinik für Kinder- und Jugendmedizin

   Universitätsklinikum Aachen

   Affiliation as printed

   Uniklinik RWTH Aachen, Klinik für Kinder- und Jugendmedizin, Sektion Neonatologie und Pädiatrische Intensivmedizin, Aachen, Germany