Biostatistics

Our experts build and maintain robust research infrastructure based on vast knowledge on real-world data, common data models, and analytical pipelines for large-scale and fast-paced safety studies on vaccines and medicines with regulatory purpose. These capabilities allow researchers, regulators, and industry partners to efficiently conduct multi-center studies, generate reproducible evidence, and accelerate the translation of real-world data into actionable measures. Through these efforts, UMC Utrecht provides partners with innovative, privacy-conscious, and scalable solutions to unlock the full potential of real-world healthcare data.

Our experts build and maintain robust research infrastructure, integrating common data models, analytical pipelines, and scalable platforms for rapid, multi-center safety studies. These capabilities enable researchers, regulators, and industry partners to conduct reproducible studies efficiently, accelerate evidence generation, and translate real-world data into actionable insights.

Experts: Oisín Ryan, Albert Cid Royo, Emmy Manders, Mattia Cinelli, Yinan Mao, Linda Nab

Missing data are pervasive in observational studies, and classic approaches often fall short in today’s increasingly complex data structures and analyses. Specifically for settings involving longitudinal and time-to-event outcomes, we develop Bayesian modeling strategies that handle missingness in a principled way while preserving relationships between variables. Implemented in the R package JointAI, we provide a unified customizable framework for Bayesian regression models with incomplete covariates, applicable to applications ranging from simple linear regression to multivariate joint models.

Beyond addressing incomplete data, the Bayesian framework facilitates the implementation of rich, application-specific models, tailored to the specific structure and needs of each application.

Experts: Nicole Erler

Understanding causal relations in observational data is critical for generating reliable evidence when randomized trials are not feasible. Our work focuses on advanced causal inference methods, including target trial emulation, which applies the principles of randomized trials to real-world data to reduce bias and improve validity. Another line of research is the use of (causal) prediction models for supporting treatment decisions, particularly how to develop, evaluate and monitor such prediction models.

Applications include vaccine safety studies, comparative effectiveness research, and evaluating interventions using routinely collected health data. These approaches help bridge the gap between observational research and clinical decision-making by providing robust, transparent estimates of causal effects.

Experts: Wouter van Amsterdam, Oisín Ryan, Linda Nab

For our expertise on advanced clinical trial methodology, please see [link to new Clinical Trials website].

Experts: Peter van de Ven, Rob Kessels, Rutger van de Bor

High-quality data is the foundation of reliable real-world evidence. Our Data Quality team ensures that healthcare data from electronic health records, registries, and other real-world sources are accurate, complete, and fit for purpose once they have been standardized to common data models. We develop and apply rigorous methods and tools such as the INSIGHT tool to evaluate data suitability for specific research questions.

Our team collaborates closely with RWE researchers across Europe through diverse research networks (VAC4EU, SIGMA, EU PE&PV). Their expertise ensures that real-world datasets meet the highest scientific and regulatory standards, supporting the application of desired study design and obtaining meaningful conclusions after a thorough understanding of the origin data. Through these efforts, we help partners unlock the full potential of real-world data,

Experts: Vjola Hoxhaj, Judit Riera Arnau, Constanza Andaur Navarro

AI and machine learning models are increasingly used in clinical decision-making, but their development and validation demand rigorous methodology. Our work addresses key challenges in building and validating AI and machine learning models for healthcare, including proper handling of bias, overfitting, and transportability across populations. We contribute to international standards like TRIPOD+AI and PROBAST+AI and develop frameworks for assessing model performance, calibration, and clinical utility.

A central focus is understanding how prediction models behave over time and under changing conditions, and developing strategies for monitoring, updating, and maintaining their clinical usefulness. Beyond traditional prediction tasks, we also validate AI systems such as large language models (LLMs) for healthcare applications—for example, assessing the accuracy and safety of AI-generated discharge letters.

By combining methodological innovation with transparent reporting, we aim to enhance the reliability and impact of AI-driven tools in practice.

Experts: Maarten van Smeden, Ewout Steyerberg

Text data in healthcare (e.g. clinical notes, discharge letters, and patient communications) contain rich information but are challenging to automatically process. Our research develops and evaluates NLP and LLM-based methods to unlock this potential while ensuring accuracy, safety, and transparency.

We work on extracting structured information from unstructured text, modeling temporal relationships between clinical events, and validating generative AI systems for real-world applications. Projects include assessing the quality and reliability of AI-generated discharge letters and designing frameworks to detect omissions, hallucinations, and bias in LLM outputs.

Beyond validation, we explore innovative architectures and adaptation strategies for LLMs, enabling efficient and domain-specific text generation and refinement. These efforts aim to make language technologies robust, interpretable, and clinically useful.

Experts: Ruurd Kuiper

Healthcare data are often collected repeatedly over time, offering opportunities to capture patient trajectories rather than relying on static snapshots. Our research develops statistical methods that integrate longitudinal measurements — often with time-to-event outcomes through joint modeling — enabling more accurate and clinically meaningful predictions.

A key focus is dynamic prediction, where risk estimates are continuously updated as new information becomes available. This approach supports personalized monitoring and adaptive decision-making in settings such as hospital wards and home-based care. We also address challenges of complex data structures, including multi-level designs and continuous monitoring streams.

To make these methods accessible, we develop open-source software. Applications range from optimizing follow-up strategies to improving early warning systems for patient deterioration.

Experts: Nicole Erler

Real-World Evidence (RWE) group at University Medical Center Utrecht is a global leader in generating robust, actionable evidence from real-world data (RWD. RWE refers to clinical evidence derived from data collected during routine care delivery, such as electronic health records, registries, or claims data. Unlike clinical trials, which often include carefully selected patient populations under controlled conditions, RWE allows us to study how treatments work in routine clinical practice, in diverse populations, and in situations not captured in trials such as pregnancy.

We apply cutting-edge methods for data quality assessment, causal inference, and large-scale analytical pipelines, ensuring that research outcomes are reliable and actionable. Our work supports a learning healthcare system, where insights from clinical practice feed directly into regulatory decisions, clinical guidance, and patient care.

We collaborate closely with industry, regulators, and public-private consortia across Europe and globally. Our team coordinates and contributes to high-profile international initiatives, including projects for the European Medicines Agency (EMA) through EU PE&PV, post-authorization safety studies via VAC4EU and SIGMA, and the IMI-ConcePTION project. Through these collaborations, we provide partners with methodological expertise, analytical tools, and scientific leadership, producing evidence that meets the highest standards for regulatory, clinical, and public health impact.

Expert: Miriam Sturkenboom

Our team specializes in evaluating the safety and effectiveness of vaccines and medicines using real-world data. Using advanced pharmacoepidemiology and pharmacovigilance methods, we assess both common and rare outcomes, providing actionable insights to clinicians, regulators, and industry partners. This research ensures interventions are safe, effective, and aligned with public health needs.

We monitor adverse events, evaluate treatment patterns, and apply rigorous study designs to understand how medicines and vaccines perform in routine care. Many of our studies are conducted in large international collaborations, including EMA initiatives and public-private consortia, generating high-quality evidence that informs regulatory decision-making, public health policies, and clinical practice.

Experts: Miriam Sturkenboom, Vjola Hoxhaj, Constanza Andaur Navarro, Hilda de Jong, Odette de Bruin, Oisín Ryan, Jungyeon Choi, Sebastian Mildiner Moraga, Linda Nab.

Pregnant women are often excluded from clinical trials, making real-world evidence essential to understanding the safety and effectiveness of vaccines and medicines during pregnancy and lactation. Our team holds a comprehensive understanding on how data on pregnancy, lactation and mother-child dyad is available in Europe and uses robust pharmacoepidemiological study designs to assess benefits and risks in this special population, providing regulators, clinicians, and industry partners with high-quality, actionable evidence.

Through international collaborations, including EMA projects and consortia such as CONSIGN and IMI-ConcePTION, we generate evidence that directly informs treatment decisions, clinical guidance, and public health policy. Our work supports safer and more equitable healthcare for mothers and children.

Experts: Miriam Sturkenboom, Vjola Hoxhaj, Constanza Andaur Navarro, Odette de Bruin