¹Institute of Medical Statistics, Computer and Data Sciences (IMSID), Jena University Hospital – Friedrich-Schiller University Jena, Jena, Germany
²Department of Clinical Pharmacy, Institute of Pharmacy, University of Bonn, Bonn, Germany
³Interdisciplinary Centre for Bioinformatics, Leipzig University, Leipzig, Germany
⁴Institute for Medical Informatics, Statistics and Epidemiology (IMISE), Leipzig University, Leipzig, Germany
⁵Deutsches Biomasseforschungszentrum gGmbH, Leipzig, Germany

Text

Introduction: Within the use-case “POLypharmacy, drug interActions and Risks” (POLAR_MI) of the Medical Informatics Initiative Germany (MII), we aimed at assessing medication-related health risks in adult inpatients by analysing routine health care data from German university hospitals [1], [2].

Methods: Relying on the methods and processes of the MII, we developed and applied a two-step, privacy-preserving distributed analysis approach. Within this approach, a local data retrieval and analysis at each university hospital’s data integration centre (DIC) [3] was followed by a random-effects meta-analysis conducted centrally by the analysis team. During development, implementation and application on data within the analysis interval from 2018 to 2021 at ten university hospitals on nearly 800,000 encounters from about 500,000 patients, we encountered several challenges. These challenges are summarised here as our lessons learned.

Results: Through the MII, a general database structure at the DIC, the “core data set” (CDS), is specified [4]. Allowing data integration from different hospital information systems (HIS) at the university hospitals, the guidelines for filling the CDS items are relatively flexible in format and content (depth). This presents a challenge itself for the researchers. In addition, the prohibited direct access to the patient data aggravated the writing of the R modules for the local data retrieval and analysis within POLAR_MI. Nevertheless, we were able to find solutions and workarounds through simplifications and assumptions to handle challenges related to missing information and the integration of documented medications, diagnoses and laboratory values. Examples for sources of missing information were missing information in the HIS or induced through missing links between different resources of a patient in the CDS. For medications, the availability depended on the source system (connected to the CDS) and both intake and prescription had to be handled. Diagnoses were frequently coded after patient’s discharge for reimbursement purposes. The interplay of the laboratory value’s metric value and provided unit had to be accounted for. Additionally, determination of laboratory value measurements depended on the disease (severity), the medical discipline and the suspected pathology. In contrast to the before mentioned challenges, the integration of the chronology of events (e.g., of medication documentation and laboratory results transmission) remained elusive.

Conclusion: Based on our experiences, there are four essential prerequisites for planning such large-scale, multicentre distributed analyses. (1) The multidisciplinary analysis team must be willing to observe rigorously and to question every detail taken for granted. (2) An in-depth knowledge of both the technical characteristics and the clinical relevance of the data is essential. (3) Close collaboration and exchange with the DIC at the university hospitals is crucial. (4) A sophisticated, extensive logging and an implemented option for controlled abortion of the local data retrieval and analysis support the identification of bugs in the R modules and of rare, unexpected, implausible or erroneous coincidences in the data. Given the ongoing progress of the MII, it is important to repeat our investigations with more recent data to monitor improvements in data completeness and quality, as we plan to do within the ongoing project INTERPOLAR (INTERventional POLypharmacy-Drug interActions-Risks) [5].

The authors declare that they have no competing interests.

The authors declare that a positive ethics committee vote has been obtained.

References

[1] Scherag A, Andrikyan W, Dreischulte T, Dürr P, Fromm MF, Gewehr J, et al. POLAR – „POLypharmazie, Arzneimittelwechselwirkungen und Risiken“ – wie können Daten aus der stationären Krankenversorgung zur Beurteilung beitragen? Prävention und Gesundheitsförderung. 2022. DOI: 10.1007/s11553-022-00976-8
[2] Semler SC, Wissing F, Heyder R. German Medical Informatics Initiative. Methods of Information in Medicine. 2018;57(S 01):e50-e6. DOI: 10.3414/ME18-03-0003.
[3] Albashiti F, Thasler R, Wendt T, Bathelt F, Reinecke I, Schreiweis B. Die Datenintegrationszentren – Von der Konzeption in der Medizininformatik-Initiative zur lokalen Umsetzung in einem Netzwerk Universitätsmedizin. Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz. 2024;67(6):629-36. DOI: 10.1007/s00103-024-03879-5
[4] Ammon D, Kurscheidt M, Buckow K, Kirsten T, Lobe M, Meineke F, et al. Arbeitsgruppe Interoperabilität: Kerndatensatz und Informationssysteme für Integration und Austausch von Daten in der Medizininformatik-Initiative. Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz. 2024;67(6):656-67. DOI: 10.1007/s00103-024-03888-4
[5] Loeffler M, Maas R, Neumann D, Scherag A. INTERPOLAR – prospektive, interventionelle Studien im Rahmen der Medizininformatik-Initiative zur Verbesserung der Arzneimitteltherapiesicherheit in der Krankenversorgung. Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz. 2024;67(6):676-84. DOI: 10.1007/s00103-024-03890-w