Roland Eils’s research while affiliated with Berlin Institute of Health at Charité - Universitätsmedizin Berlin and other places

Electronic Health Records (EHRs) offer rich potential for clinical prediction, yet their inherent complexity and heterogeneity pose significant challenges for traditional machine learning approaches. Domain-specific EHR foundation models trained on large collections of unlabeled EHR data have demonstrated promising improvements in predictive accuracy and generalization; however, their training is constrained by limited access to diverse, high-quality datasets and inconsistencies in coding standards and healthcare practices. In this study, we explore the possibility of using general-purpose Large Language Models (LLMs) based embedding methods as EHR encoders. By serializing patient records into structured Markdown text, transforming codes into human-readable descriptors, we leverage the extensive generalization capabilities of LLMs pretrained on vast public corpora, thereby bypassing the need for proprietary medical datasets. We systematically evaluate two state-of-the-art LLM-embedding models, GTE-Qwen2-7B-Instruct and LLM2Vec-Llama3.1-8B-Instruct, across 15 diverse clinical prediction tasks from the EHRSHOT benchmark, comparing their performance to an EHRspecific foundation model, CLIMBR-T-Base, and traditional machine learning baselines. Our results demonstrate that LLM-based embeddings frequently match or exceed the performance of specialized models, even in few-shot settings, and that their effectiveness scales with the size of the underlying LLM and the available context window. Overall, our findings demonstrate that repurposing LLMs for EHR encoding offers a scalable and effective approach for clinical prediction, capable of overcoming the limitations of traditional EHR modeling and facilitating more interoperable and generalizable healthcare applications.

The COVID-19 pandemic exposed a global deficiency of systematic, data-driven guidance to identify high-risk individuals. Here, we illustrate the utility of routinely recorded medical history to predict the risk for 1741 diseases across clinical specialties and support the rapid response to emerging health threats such as COVID-19. We developed a neural network to learn from health records of 502,489 UK Biobank participants. Importantly, we observed discriminative improvements over basic demographic predictors for 1546 (88.8%) endpoints. After transferring the unmodified risk models to the All of US cohort, we replicated these improvements for 1115 (78.9%) of 1414 investigated endpoints, demonstrating generalizability across healthcare systems and historically underrepresented groups. Ultimately, we showed how this approach could have been used to identify individuals vulnerable to severe COVID-19. Our study demonstrates the potential of medical history to support guidance for emerging pandemics by systematically estimating risk for thousands of diseases at once at minimal cost.

With progressive digitalization of healthcare systems worldwide, large-scale collection of electronic health records (EHRs) has become commonplace. However, an extensible framework for comprehensive exploratory analysis that accounts for data heterogeneity is missing. Here we introduce ehrapy, a modular open-source Python framework designed for exploratory analysis of heterogeneous epidemiology and EHR data. ehrapy incorporates a series of analytical steps, from data extraction and quality control to the generation of low-dimensional representations. Complemented by rich statistical modules, ehrapy facilitates associating patients with disease states, differential comparison between patient clusters, survival analysis, trajectory inference, causal inference and more. Leveraging ontologies, ehrapy further enables data sharing and training EHR deep learning models, paving the way for foundational models in biomedical research. We demonstrate ehrapy’s features in six distinct examples. We applied ehrapy to stratify patients affected by unspecified pneumonia into finer-grained phenotypes. Furthermore, we reveal biomarkers for significant differences in survival among these groups. Additionally, we quantify medication-class effects of pneumonia medications on length of stay. We further leveraged ehrapy to analyze cardiovascular risks across different data modalities. We reconstructed disease state trajectories in patients with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) based on imaging data. Finally, we conducted a case study to demonstrate how ehrapy can detect and mitigate biases in EHR data. ehrapy, thus, provides a framework that we envision will standardize analysis pipelines on EHR data and serve as a cornerstone for the community.

The COVID-19 pandemic exposed a global deficiency of systematic, data-driven guidance to identify high-risk individuals. Here, we illustrate the utility of routinely recorded medical history to predict the risk for 1883 diseases across clinical specialties and support the rapid response to emerging health threats such as COVID-19. We developed a neural network to learn from health records of 502,460 UK Biobank. Importantly, we observed discriminative improvements over basic demographic predictors for 1774 (94.3%) endpoints. After transferring the unmodified risk models to the All of US cohort, we replicated these improvements for 1347 (89.8%) of 1500 investigated endpoints, demonstrating generalizability across healthcare systems and historically underrepresented groups. Ultimately, we showed how this approach could have been used to identify individuals vulnerable to severe COVID-19. Our study demonstrates the potential of medical history to support guidance for emerging pandemics by systematically estimating risk for thousands of diseases at once at minimal cost.

Early detection of high-risk individuals is crucial for healthcare systems to cope with changing demographics and an ever-increasing patient population. Images of the retinal fundus are a non- invasive, low-cost examination routinely collected and potentially scalable beyond ophthalmology. Prior work demonstrated the potential of retinal images for risk assessment for common cardiometabolic diseases, but it remains unclear whether this potential extends to a broader range of human diseases. Here, we extended a retinal foundation model (RETFound) to systematically explore the predictive potential of retinal images as a low-cost screening strategy for disease onset across >750 incident diseases in >60,000 individuals. For more than a third (n=308) of the diseases, we demonstrated improved discriminative performance compared to readily available patient characteristics. This included 281 diseases outside of ophthalmology, such as type 2 diabetes (Delta C-Index: UK Biobank +0.073 (0.068, 0.079)) or chronic obstructive pulmonary disease (Delta C-Index: UK Biobank +0.047 (0.039, 0.054)), showcasing the potential of retinal images to complement screening strategies more widely. Moreover, we externally validated these findings in 7,248 individuals from the EPIC-Norfolk Eye Study. Notably, retinal information did not improve the prediction for the onset of cardiovascular diseases compared to established primary prevention scores, demonstrating the need for rigorous benchmarking and disease-agnostic efforts to design cost-efficient screening strategies to improve population health. We demonstrated that predictive improvements were attributable to retinal vascularisation patterns and less obvious features, such as eye colour or lens morphology, by extracting image attributions from risk models and performing genome-wide association studies, respectively. Genetic findings further highlighted commonalities between eye-derived risk estimates and complex disorders, including novel loci, such as IMAP1, for iron homeostasis. In conclusion, we present the first comprehensive evaluation of predictive information derived from retinal fundus photographs, illustrating the potential and limitations of easily accessible and low-cost retinal images for risk assessment across common and rare diseases.

With progressive digitalization of healthcare systems worldwide, large-scale collection of electronic health records (EHRs) has become commonplace. However, an extensible framework for comprehensive exploratory analysis that accounts for data heterogeneity is missing. Here, we introduce ehrapy, a modular open-source Python framework designed for exploratory end-to-end analysis of heterogeneous epidemiology and electronic health record data. Ehrapy incorporates a series of analytical steps, from data extraction and quality control to the generation of low-dimensional representations. Complemented by rich statistical modules, ehrapy facilitates associating patients with disease states, differential comparison between patient clusters, survival analysis, trajectory inference, causal inference, and more. Leveraging ontologies, ehrapy further enables data sharing and training EHR deep learning models paving the way for foundational models in biomedical research. We demonstrated ehrapys features in five distinct examples: We first applied ehrapy to stratify patients affected by unspecified pneumonia into finer-grained phenotypes. Furthermore, we revealed biomarkers for significant differences in survival among these groups. Additionally, we quantify medication-class effects of pneumonia medications on length of stay. We further leveraged ehrapy to analyze cardiovascular risks across different data modalities. Finally, we reconstructed disease state trajectories in SARS-CoV-2 patients based on imaging data. Ehrapy thus provides a framework that we envision will standardize analysis pipelines on EHR data and serve as a cornerstone for the community.

Background: The reuse of data from electronic health records (EHRs) for research purposes promises to improve the data foundation for clinical trials and may even support to enable them. Nevertheless, EHRs are characterized by both, heterogeneous structure and semantics. To standardize this data for research, the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) standard has recently seen an increase in use. However, the conversion of these EHRs into the OMOP CDM requires complex and resource intensive Extract Transform and Load (ETL) processes. This hampers the reuse of clinical data for research. To solve the issues of heterogeneity of EHRs and the lack of semantic precision on the care site, the openEHR standard has recently seen wider adoption. A standardized process to integrate openEHR records into the CDM potentially lowers the barriers of making EHRs accessible for research. Yet, a comprehensive approach about the integration of openEHR records into the OMOP CDM has not yet been made. Methods: We analysed both standards and compared their models to identify possible mappings. Based on this, we defined the necessary processes to transform openEHR records into CDM tables. We also discuss the limitation of openEHR with its unspecific demographics model and propose two possible solutions. Results: We developed the OMOP Conversion Language (OMOCL) which enabled us to define a declarative openEHR archetype-to-CDM mapping language. Using OMOCL, it was possible to define a set of mappings. As a proof-of-concept, we implemented the Eos tool, which uses the OMOCL-files to successfully automatize the ETL from real-world and sample EHRs into the OMOP CDM. Discussion: Both Eos and OMOCL provide a way to define generic mappings for an integration of openEHR records into OMOP. Thus, it represents a significant step towards achieving interoperability between the clinical and the research data domains. However, the transformation of openEHR data into the less expressive OMOP CDM leads to a loss of semantics.

Making health data available for secondary use enables innovative data-driven medical research. Since modern machine learning (ML) methods and precision medicine require extensive amounts of data covering most of the standard and edge cases, it is essential to initially acquire large datasets. This can typically only be achieved by integrating different datasets from various sources and sharing data across sites. To obtain a unified dataset from heterogeneous sources, standard representations and Common Data Models (CDM) are needed. The process of mapping data into these standardized representations is usually very tedious and requires many manual configuration and refinement steps. A potential way to reduce these efforts is to use ML methods not only for data analysis, but also for the integration of health data on the syntactic, structural, and semantic level. However, research on ML-based medical data integration is still in its infancy. In this article, we describe the current state of the literature and present selected methods that appear to have a particularly high potential to improve medical data integration. Moreover, we discuss open issues and possible future research directions.

The availability of high-throughput molecular diagnostics builds the foundation for Molecular Tumor Boards (MTBs). Although more fine-grained data is expected to support decision making of oncologists, assessment of data is complex and time-consuming slowing down the implementation of MTBs, e.g., due to retrieval of the latest medical publications, assessment of clinical evidence, or linkage to the latest clinical guidelines. We share our findings from analysis of existing tumor board processes and defininion of clinical processes for the adoption of MTBs. Building on our findings, we have developed a real-world software prototype together with oncologists and medical professionals, which supports the preparation and conduct of MTBs and enables collaboration between medical experts by sharing medical knowledge even across the hospital locations. We worked in interdisciplinary teams of clinicians, oncologists, medical experts, medical informaticians, and software engineers using design thinking methodology. With their input, we identified challenges and limitations of the current MTB approaches, derived clinical process models using Business Process and Modeling Notation (BMPN), and defined personas, functional and non-functional requirements for software tool support. Based on it, we developed software prototypes and evaluated them with clinical experts from major university hospitals across Germany. We extended the Kanban methodology enabling holistic tracking of patient cases from “backlog” to “follow-up” in our app. The feedback from interviewed medical professionals showed that our clinical process models and software prototype provide suitable process support for the preparation and conduction of molecular tumor boards. The combination of oncology knowledge across hospitals and the documentation of treatment decision can be used to form a unique medical knowledge base by oncologists for oncologists. Due to the high heterogeneity of tumor diseases and the spread of the latest medical knowledge, a cooperative decision-making process including insights from similar patient cases was considered as a very valuable feature. The ability to transform prepared case data into a screen presentation was recognized as an essential feature speeding up the preparation process. Oncologists require special software tool support to incorporate and assess molecular data for the decision-making process. In particular, the need for linkage to the latest medical knowledge, clinical evidence, and collaborative tools to discuss individual cases were named to be of importance. With the experiences from the COVID-19 pandemic, the acceptance of online tools and collaborative working is expected to grow. Our virtual multi-site approach proved to allow a collaborative decision-making process for the first time, which we consider to have a positive impact on the overall treatment quality.