Het modelleren van opportunistisch gesampelde eerstelijnsregistergegevens voor inzichten in de volksgezondheid
Year: 2026
Type: dissertation
Abstract: During routine clinical practice, general practitioners record clinical decisions in electronic medical records. When aggregated across practices, these records form continuously updated primary care registries that reflect population health as it unfolds in routine clinical practice. Such data offer opportunities for epidemiological research and public health surveillance, capturing large populations over extended periods and across a wide range of conditions. However, these data are not collected for research purposes. They arise from routine healthcare delivery, reflecting not only underlying disease processes but also patterns of healthcare utilisation and clinical recording. In practice, such registries consist of voluntarily participating practices and include data on patients who consult general practitioners within these practices. As a result, they represent a subset of the population, those seeking care from participating practices, rather than a systematically sampled population. This fundamental characteristic raises a central question addressed in this thesis: how can opportunistically sampled primary care registry data be used to generate public health insights? The thesis is organised around three complementary components: the development of a research-ready data infrastructure, applied studies addressing public health questions, and the evaluation of methodological challenges inherent to opportunistically sampled primary care data. The work is based on the Intego primary care registry in Flanders, Belgium, and its evolution into the Intego-II database. Building on more than three decades of data collection, Intego already provided a rich source of longitudinal primary care data and supported linkage with external data sources such as mortality and environmental data. In this thesis, the development of Intego-II (Chapter 2) focused on improving the organisation, consistency, and reusability of these data for research, including structured data processing and initial steps towards standardization and harmonization in line with FAIR principles and the OMOP Common Data Model. Chapter 3 demonstrates the value of primary care data for public health research through a study on pneumococcal vaccination, showing that coverage in Flanders remained suboptimal and that socioeconomic inequalities persisted in both uptake and adherence to recommended schedules. Chapters 4 and 5 examine how standard statistical models can be applied to primary care data to investigate how data characteristics influence effect estimates. In Chapter 4, Bayesian hierarchical models were used to estimate geographic variation in disease risk while accounting for clustering within practices and spatial dependence. A case and simulation study showed that, despite spatial imbalance and practice variation, meaningful spatial patterns could be recovered, particularly under favourable conditions such as lower practice-level variation, a strong spatial signal, and more balanced spatial coverage. Chapter 5 examines how characteristics of routine primary care data affect statistical inference in applied analyses. It demonstrates that accounting for clustering within practices is essential, as ignoring the data collection entity can bias estimates for covariates. The study further showed that the choice of analytical entity can substantially alter effect estimates and can be used to detect potential biases. While exposure to black carbon showed consistent positive associations with respiratory diagnoses across analytical entities, other associations varied across models, highlighting their sensitivity to modelling choices. In such cases, differences between outcome definitions complicate interpretation, making it unclear whether observed associations reflect true differences in disease occurrence or underlying patterns in healthcare utilisation and recording. Chapter 6 combines complementary data sources to study spatial inequalities in cardiovascular health. While both data sources capture different aspects, with claims-based indicators partly reflecting healthcare utilisation and GP data better capturing diagnosed disease, socioeconomic vulnerability was consistently associated with higher ASCVD burden across both, with similar geographic patterns such as higher prevalence in coastal and deprived areas. Overall, this dissertation positions routine primary care data as a structured and valuable source of information for public health research. It demonstrates their use for surveillance and applied epidemiology, while highlighting the need to account for and be aware of the challenges inherent to opportunistic data collection and the characteristics of the registry.
Source: Lirias
Author: Janssens, Arne
Language: English
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7201933054