(184) Feasibility of the OMOP Common Data Model for German claims data

Background: There is a growing need to harmonize observational data using common data models (CDMs) to improve interoperability. A CDM stores data from different sources and formats in a common structure, standardizing the content and, in some cases, the vocabulary. One widely used CDM is the Observational Medical Outcomes Partnership (OMOP) CDM.

Objectives: To evaluate the feasibility of transforming German claims data to the OMOP CDM.

Methods: First, we tested the ETL (Extract Transform Load) process from the German Pharmacoepidemiological Research Database (GePaRD; claims data of ~20% of the German population) to the OMOP CDM. We assessed which information of GePaRD can be converted to the data tables of the OMOP CDM and to what extent the German coding systems are already mapped to the standard concepts for the purpose of semantic data harmonization. Next, we assessed whether the meaning of the transformed data is preserved in comparison to the original source data.

Results: The detailed descriptions and the open-source tools available for OMOP were useful to simplify the ETL and the mapping process. Almost all data from GePaRD could be transferred to the table structure of the OMOP CDM, although it required some compromises and thus a loss of information. Another challenge is the semantic harmonization of GePaRD data. So far, not all German coding systems are fully integrated in the OMOP vocabulary. It turned out that up to now two billion codes of the German coding systems have to be mapped to standard concepts. Furthermore, mapping of one code to one standard concept can lead to loss of information since codes can have more than one meaning. For example, either frailty or immobility of a patient can be derived by the same code indicating a home-visit by the physician. Although the OMOP CDM includes data tables for derived variables (so called ERA tables), the used algorithms do not consider the specific characteristics of the German health care system. For instance, relying on only one diagnosis of diabetes in German claims data leads to an overestimation of the diabetes prevalence, as has often been shown.

Conclusions: Advantages of standardizing to the OMOP CDM include that users do not need to understand all database-specific details allowing interoperability. So far, ICD codes are already fully integrated in the OMOP vocabulary. Nevertheless, standardization of claims data using the OMOP CDM carries the risk of loss of information. OMOP does currently not take into account the specifics of the German health care system. This poses a great risk for non-experts to draw incorrect conclusions from the data.