The COVID-19 pandemic has caused strains on health systems worldwide disrupting routine hospital services for all non-COVID patients. Within this retrospective study, we analyzed inpatient hospital admissions across 18 German university hospitals during the 2020 lockdown period compared to 2018. Patients admitted to hospital between January 1 and May 31, 2020 and the corresponding periods in 2018 and 2019 were included in this study. Data derived from electronic health records were collected and analyzed using the data integration center infrastructure implemented in the university hospitals that are part of the four consortia funded by the German Medical Informatics Initiative. Admissions were grouped and counted by ICD 10 chapters and specific reasons for treatment at each site. Pooled aggregated data were centrally analyzed with descriptive statistics to compare absolute and relative differences between time periods of different years. The results illustrate how care process adoptions depended on the COVID-19 epidemiological situation and the criticality of the disease. Overall inpatient hospital admissions decreased by 35% in weeks 1 to 4 and by 30.3% in weeks 5 to 8 after the lockdown announcement compared to 2018. Even hospital admissions for critical care conditions such as malignant cancer treatments were reduced. We also noted a high reduction of emergency admissions such as myocardial infarction (38.7%), whereas the reduction in stroke admissions was smaller (19.6%). In contrast, we observed a considerable reduction in admissions for non-critical clinical situations, such as hysterectomies for benign tumors (78.8%) and hip replacements due to arthrosis (82.4%). In summary, our study shows that the university hospital admission rates in Germany were substantially reduced following the national COVID-19 lockdown. These included critical care or emergency conditions in which deferral is expected to impair clinical outcomes. Future studies are needed to delineate how appropriate medical care of critically ill patients can be maintained during a pandemic.

Appropriate reference intervals are essential when using laboratory test results to guide medical decisions. Conventional approaches for the establishment of reference intervals rely on large samples from healthy and homogenous reference populations. However, this approach is associated with substantial financial and logistic challenges, subject to ethical restrictions in children, and limited in older individuals due to the high prevalence of chronic morbidities and medication. We implemented an indirect method for reference interval estimation, which uses mixed physiological and abnormal test results from clinical information systems, to overcome these restrictions. The algorithm minimizes the difference between an estimated parametrical distribution and a truncated part of the observed distribution, specifically, the Kolmogorov-Smirnov-distance between a hypothetical Gaussian distribution and the observed distribution of test results after Box-Cox-transformation. Simulations of common laboratory tests with increasing proportions of abnormal test results show reliable reference interval estimations even in challenging simulation scenarios, when <20% test results are abnormal. Additionally, reference intervals generated using samples from a university hospital’s laboratory information system, with a gradually increasing proportion of abnormal test results remained stable, even if samples from units with a substantial prevalence of pathologies were included. A high-performance open-source C++ implementation is available athttps://gitlab.miracum.org/kosmic.

To take full advantage of decision support, machine learning, and patient-level prediction models, it is important that models are not only created, but also deployed in a clinical setting. The KETOS platform demonstrated in this work implements a tool for researchers allowing them to perform statistical analyses and deploy resulting models in a secure environment. The proposed system uses Docker virtualization to provide researchers with reproducible data analysis and development environments, accessible via Jupyter Notebook, to perform statistical analysis and develop, train and deploy models based on standardized input data. The platform is built in a modular fashion and interfaces with web services using the Health Level 7 (HL7) Fast Healthcare Interoperability Resources (FHIR) standard to access patient data. In our prototypical implementation we use an OMOP common data model (OMOP-CDM) database. The architecture supports the entire research lifecycle from creating a data analysis environment, retrieving data, and training to final deployment in a hospital setting. We evaluated the platform by establishing and deploying an analysis and end user application for hemoglobin reference intervals within the University Hospital Erlangen. To demonstrate the potential of the system to deploy arbitrary models, we loaded a colorectal cancer dataset into an OMOP database and built machine learning models to predict patient outcomes and made them available via a web service. We demonstrated both the integration with FHIR as well as an example end user application. Finally, we integrated the platform with the open source DataSHIELD architecture to allow for distributed privacy preserving data analysis and training across networks of hospitals. The KETOS platform takes a novel approach to data analysis, training and deploying decision support models in a hospital or healthcare setting. It does so in a secure and privacy-preserving manner, combining the flexibility of Docker virtualization with the advantages of standardized vocabularies, a widely applied database schema (OMOP-CDM), and a standardized way to exchange medical data (FHIR).

To evaluate whether neural networks can distinguish between seropositive RA, seronegative RA, and PsA based on inflammatory patterns from hand MRIs and to test how psoriasis patients with subclinical inflammation fit into such patterns. ResNet neural networks were utilized to compare seropositive RA vs PsA, seronegative RA vs PsA, and seropositive vs seronegative RA with respect to hand MRI data. Results from T1 coronal, T2 coronal, T1 coronal and axial fat-suppressed contrast-enhanced (CE), and T2 fat-suppressed axial sequences were used. The performance of such trained networks was analysed by the area under the receiver operating characteristics curve (AUROC) with and without presentation of demographic and clinical parameters. Additionally, the trained networks were applied to psoriasis patients without clinical arthritis. MRI scans from 649 patients (135 seronegative RA, 190 seropositive RA, 177 PsA, 147 psoriasis) were fed into ResNet neural networks. The AUROC was 75% for seropositive RA vs PsA, 74% for seronegative RA vs PsA, and 67% for seropositive vs seronegative RA. All MRI sequences were relevant for classification, however, when deleting contrast agent–based sequences the loss of performance was only marginal. The addition of demographic and clinical data to the networks did not provide significant improvements for classification. Psoriasis patients were mostly assigned to PsA by the neural networks, suggesting that a PsA-like MRI pattern may be present early in the course of psoriatic disease. Neural networks can be successfully trained to distinguish MRI inflammation related to seropositive RA, seronegative RA, and PsA.

Background Many research initiatives aim at using data from electronic health records (EHRs) in observational studies. Participating sites of the German Medical Informatics Initiative (MII) established data integration centers to integrate EHR data within research data repositories to support local and federated analyses. To address concerns regarding possible data quality (DQ) issues of hospital routine data compared with data specifically collected for scientific purposes, we have previously presented a data quality assessment (DQA) tool providing a standardized approach to assess DQ of the research data repositories at the MIRACUM consortium's partner sites. Objectives Major limitations of the former approach included manual interpretation of the results and hard coding of analyses, making their expansion to new data elements and databases time-consuming and error prone. We here present an enhanced version of the DQA tool by linking it to common data element definitions stored in a metadata repository (MDR), adopting the harmonized DQA framework from Kahn et al and its application within the MIRACUM consortium. Methods Data quality checks were consequently aligned to a harmonized DQA terminology. Database-specific information were systematically identified and represented in an MDR. Furthermore, a structured representation of logical relations between data elements was developed to model plausibility-statements in the MDR. Results The MIRACUM DQA tool was linked to data element definitions stored in a consortium-wide MDR. Additional databases used within MIRACUM were linked to the DQ checks by extending the respective data elements in the MDR with the required information. The evaluation of DQ checks was automated. An adaptable software implementation is provided with the R package DQAstats. Conclusion The enhancements of the DQA tool facilitate the future integration of new data elements and make the tool scalable to other databases and data models. It has been provided to all ten MIRACUM partners and was successfully deployed and integrated into their respective data integration center infrastructure.

Data quality assessments (DQA) are necessary to ensure valid research results. Despite the growing availability of tools of relevance for DQA in the R language, a systematic comparison of their functionalities is missing. Therefore, we review R packages related to data quality (DQ) and assess their scope against a DQ framework for observational health studies. Based on a systematic search, we screened more than 140 R packages related to DQA in the Comprehensive R Archive Network. From these, we selected packages which target at least three of the four DQ dimensions (integrity, completeness, consistency, accuracy) in a reference framework. We evaluated the resulting 27 packages for general features (e.g., usability, metadata handling, output types, descriptive statistics) and the possible assessment’s breadth. To facilitate comparisons, we applied all packages to a publicly available dataset from a cohort study. We found that the packages’ scope varies considerably regarding functionalities and usability. Only three packages follow a DQ concept, and some offer an extensive rule-based issue analysis. However, the reference framework does not include a few implemented functionalities, and it should be broadened accordingly. Improved use of metadata to empower DQA and user-friendliness enhancement, such as GUIs and reports that grade the severity of DQ issues, stand out as the main directions for future developments.