[FAIRification of real world data for health research]

Abstract

Background: The provision of real-world data according to the FAIR principles is prerequisite for an efficient exploitation of the potential of health data for prevention and care.

Objectives: To discuss the opportunities and limitations of reuse and record linkage of health data in Germany.

Materials and methods: Initiatives to establish an improved research data infrastructure are presented and the limitations that hinder record linkage of personal health data are illustrated using an example.

Results: In general, health data in Germany do not meet the requirements of the FAIR principles. Their findability already fails because either no metadata are available or they are not posted in searchable repositories in a standardized way. Record linkage of personal health data is extremely limited by restrictive data protection regulations and the lack of a so-called unique identifier. Privacy-compliant solutions for linking health data, which are successfully practiced in neighboring European countries, could serve as a model here.

Conclusions: The establishment of a National Research Data Infrastructure (NFDI), especially for personal health data (NFDI4Health), can only be realized with considerable efforts and legislative changes. Already existing structures and standards that have been for instance developed by the Medical Informatics Initiative and the Netzwerk Universitätsmedizin (English: University Medicine Network), and international initiatives such as the European Open Science Cloud should be taken into consideration.

Keywords: Data protection; FAIR principles; Linkage of person-related health data; National Research Data Infrastructure; Unique identifier.

Abb. 1

Datenfluss für die Durchführung beider Linkage-Verfahren (PID projektspezifische Identifikationsnummer, Pseud. KVN pseudonymisierte Krankenversicherungsnummer, KVN Krankenversicherungsnummer, Klartext Klartextmerkmale [Geburtsmonat und -jahr, Geschlecht, PLZ], KN Kontrollnummer, RegN registerinterne Nummer, Gewicht Übereinstimmungsgewicht, VarKR Variablen aus Krebsregister [Tumordiagnose, Datumsangaben zur Diagnose, TNM-Schlüssel und UICC-Stadium, Informationen zur Todesursache], VarKR2 Variablen aus Krebsregister [Tumordiagnose, Datumsangaben zur Diagnose, TNM-Schlüssel und UICC-Stadium, Informationen zur Todesursache, GKZ, Geburtsjahr, Geschlecht], EpiVar Variablen aus der Forschungsdatenbank [Geburtsjahr, Geschlecht, Komobiditäten und Komedikation]). (Quelle: [22]. Die Abbildung ist von dieser Open-Access-Lizenz ausgenommen. Jede weitere Verwendung muss erneut und gesondert bei permission@thieme.de angefragt werden.)