Work

Data resource profile: the Intego-II primary care database

Year: 2025

Type: article

Abstract: Intego-II builds on three decades of primary care data collection in Flanders, Belgium, incorporating substantial advancements in the database’s structure, operations, and accessibility. Established in 1994, the database collects pseudonymized electronic medical record (EMR) data from participating general practices. Its integration with Healthdata.be provides scalable linkage to mortality, environmental, and disease-specific datasets at the national level. A robust two-step Extract-Transform-Load (ETL) process ensures data security, privacy, tidiness, and quality. To enhance international research interoperability, the database is aligned with the OMOP Common Data Model. Intego-II is organized into three key modules: Patient Information, Medical History, and Clinical Encounters, enabling longitudinal analyses across diverse healthcare domains covering, among others, demographic variables, diagnoses, prescriptions, and laboratory test results. Structured quarterly releases with unique Digital Object Identifiers (DOIs) and detailed metadata ensure findability and reusability. Researchers can access Intego-II via a secure research environment provided by Healthdata.be, following submission and approval of a study protocol. The Data Access Process is detailed at www.intego.be. As of December 2024, 180 practices have contributed data (of which 135 active practices). The data included 973 444 unique individuals (of which 512 093 active patients). Currently, active patients represent 7.46% of the Flemish population. The Intego project, initiated by the Academic Centre for General Practice (ACHG) at KU Leuven, was established to create a comprehensive primary care morbidity database in Flanders, Belgium. Starting with semi-automated data collection in 1994, the project evolved into Belgium’s first fully automated general practice registration system by 2004 [1]. Coordinated by KU Leuven’s ACHG, primarily funded by the Department of Care of the Flemish government and backed by data collected through software provided by Corilus (Corilus NV, Ghent, Belgium) [2], the success and evolution of the Intego project can be attributed to the effective collaboration between governmental services, academic institutions, and software companies. Involvement of these three main actors has been noted as the most effective strategy to build and sustain long-term, wide-ranging data collection projects [3]. In this Data Resource Profile, we describe the release of Intego-II, which builds on three decades of primary care data collection, incorporating substantial advancements in the database’s structure, operations, and accessibility. In Belgium, nearly the entire population (99%) is covered by government-funded compulsory health insurance [4–6]. Patients have the freedom to choose their general practitioner (GP), who typically act as the first point of contact for healthcare and refer patients to specialists when needed [6]. Patients can request their chosen GP to maintain a global medical file (GMF)—an electronic health record that consolidates their medical history, promoting continuity of care [5]. As of 2022, 89.9% of the Flemish population had a GMF, and 87% had at least one GP consultation [7]. Loyalty to GPs is high, with 80% of patients consistently visiting the same practice, a trend that has remained stable over time [7]. The Intego database, initially described by Truyers et al. in 2014, has undergone major updates. In 2018, participating GPs transitioned from the obsolete Medidoc® EMR to the more advanced CareConnect® system. Over time, the database has been broadened by adding more practices to improve geographic coverage and enriched by linking patient-level data to other sources. To reinforce the role of Intego-II as a vital resource for advancing health research, driving innovation in primary care practice, and informing policy development, substantial advancements have been made to ensure full compliance with FAIR data principles—Findable, Accessible, Interoperable, and Reusable [8]. Practices are recruited through outreach by the ACHG, with the goal of achieving broad geographic coverage across Flanders, allowing Intego to expand its network effectively. Figure 1 illustrates the database’s growth, showcasing the increase in the number of patients who visited participating practices within a given year—referred to as the yearly contact group (YCG). Additionally, Table 1 presents the main characteristics of the Intego-II population of December 2024 [9]. Growth of Intego database over time showing the evolution of the Yearly Contact Group (YCG). Metadata of the Intego-II database as of December 2024 Adapted from Casper et al. (2025). Active cohort: non-deceased patients from the acceptable cohort with at least one registered encounter within 2 years before the version cutoff date. Acceptable cohort: patients with at least two registered encounters and with unambiguous gender and birth date. Patient follow-up time: time between first encounter and death date for deceased patients; time between first encounter and last encounter for non-deceased, non-active patients; time between first encounter and version cutoff date for active patients. Metadata of the Intego-II database as of December 2024 Adapted from Casper et al. (2025). Active cohort: non-deceased patients from the acceptable cohort with at least one registered encounter within 2 years before the version cutoff date. Acceptable cohort: patients with at least two registered encounters and with unambiguous gender and birth date. Patient follow-up time: time between first encounter and death date for deceased patients; time between first encounter and last encounter for non-deceased, non-active patients; time between first encounter and version cutoff date for active patients. On 31 December 2024, the database included 512 093 active patients (7.46% of the Flemish population). To validate the distribution of key population demographics, Fig. 2A demonstrates that Intego-II is representative of the Flemish population in terms of gender and age distribution. Figure 2B shows geographical variations in the population coverage, with an overrepresentation in Antwerp and Flemish Brabant and an underrepresentation in West Flanders, East Flanders, and Limburg. Comparison of the Intego-II population with the general Flemish population. This figure compares the characteristics of the active Intego-II population on 31 December 2023, with the general population of Flanders in 2023, obtained from the Belgian statistical office. (A) Distribution of key demographics, illustrating the representativeness of Intego-II in terms of gender and age distribution. (B) Geographical distribution of the active Intego-II population across the five Flemish provinces, showing deviations from the expected distribution if the Intego-II population mirrored the general population of Flanders. Compared to the expected distribution, there is an overrepresentation in Antwerp and Flemish-Brabant, and an underrepresentation in West-Flanders, East-Flanders, and Limburg. Intego-II data are enriched through linking patient-level data with additional external sources, providing a more comprehensive resource for research. The currently linked external data sources include: Mortality data: To ensure more comprehensive data and accurate research studies, Intego-II links patient data with mortality records from the National Register of Natural Persons (RRNP) to obtain the dates of death of deceased patients on a weekly basis [10]. This continuous linkage started in 2021 and included historic mortality data. Environmental data: Since 2017, Intego has been commissioned by the Flemish Department of Care to monitor the impact of environmental pollution on public health [11]. Environmental data are provided by the Flemish Institute for Technological Research (VITO), an independent Flemish research organization, and include annual averages of noise, traffic, air quality, and climate indicators, along with daily measurements of temperature and relative humidity. These indicators are linked to patients’ residential locations at various spatial resolutions. Diabetes Data Cell project: The Diabetes Data Cell Project seeks to create a comprehensive database to monitor diabetes trends and improve care quality. Launched in 2024, the project received approval from the information security council to link two major databases with Intego; the Initiative for Quality Improvement and Epidemiology in Diabetes (IQED) database, and data from the Inter-Mutualistic Agency (IMA), a non-profit organization founded by Belgian health insurance funds [12–14]. The creation of the Intego-II database involved three distinct steps: data acquisition, transformation, and release (Fig. 3). Overview of Intego-II data process. This figure provides an overview of the Intego-II data process, consisting of three main steps. Data acquisition, initiated in 1994 in collaboration with Corilus [the electronic medical record (EMR) software provider], has continued uninterrupted, using Medidoc® until 2018 and CareConnect® thereafter. Practices across Flanders are recruited by KU Leuven’s ACHG, with data collection encompassing both historical records—dating back to 1924—and weekly updates. External data sources, including mortality records from the National Register of Natural Persons (RRNP), environmental data from the Flemish Institute for Technological Research (VITO), data related to diabetes from the IQED and the IMA databases, are linked via pseudonymized Social Security Numbers (SSNs) and secured through end-to-end encryption using Belgium’s eHealth services. The data undergoes two ETL (Extract, Transform, Load) cycles: the first, managed by Healthdata.be, focuses on encryption, pseudonymization, and integration of external datasets, while the second, managed by the Intego team, emphasizes restructuring, standardization, and adjustments to accurately define the patient population for epidemiological studies. Quarterly data releases are assigned unique DOIs, accompanied by metadata to ensure reproducibility, with access provided to researchers via two-factor authentication in a secure environment. ACHG, Academic Centre for General Practice; DOI, digital object identifier; ETL, extract, transform, load; EMR, electronic medical record; IQED, initiative for quality improvement and epidemiology in diabetes; IMA, InterMutualistic Agency; RRNP, national register of natural persons; SSN, social security number; VITO, Flemish Institute for Technological Research. Data collection for the Intego-II database began in 1994 and has continued uninterrupted. All participating practices (data providers) use the EMR software provided by Corilus. Until 2018, the software used was Medidoc®, which was later replaced by CareConnect®, featuring a more robust technological infrastructure that supports weekly automated data collection and additional coded health data. In participating practices, patients are informed of the practice’s participation through a document provided in the waiting room. Data collection includes recent weekly data and all historical records stored in the practice’s EMR system, but only if the patient has not opted out of data sharing. This approach accounts for the extensive time span of the collected data, with the earliest records in Intego-II dating back to 1924—70 years before the official start of data collection in 1994—and covering nearly a century of data up to the present. Intego-II undergoes two Extract-Transform-Load (ETL) cycles, each serving a specific purpose: the first ensures data security and privacy, while the second focuses on data tidiness and quality. The first ETL cycle is managed by Healthdata.be, a Belgian platform developed by Sciensano, a third-party organization responsible for the secure transfer of data from participating practices. The data are loaded into a secure data warehouse, which provides controlled access to researchers via two-factor authentication [15]. The trusted third-party process has been in place since 2012. A key transformation in this cycle is the implementation of end-to-end (E2E) encryption to protect patient information during transmission. To safeguard patient privacy while enabling data linkage across different sources, Healthdata.be employs Belgium’s eHealth services to receive pseudonymized national patient identifiers (Social Security Numbers or SSN) [15]. Through the SSN identifier, data from the same patient can be linked across different databases without revealing their identity, thereby maintaining confidentiality. However, in Intego data that were collected using Medidoc software before 2015, SSN identifiers were not included and hence patients in this historic subset are ineligible for linkage to external data sources, such as mortality or environmental data. In addition to encrypting patient identifiers, a Small Cell Risk Assessment (SCRA) was conducted by a third party to evaluate and mitigate the risk of patient re-identification through indirect identifiers within the database [16]. Measures such as removing and pseudonymizing potential identifiers were implemented, successfully reducing the risk of disclosure while preserving the data’s utility for research purposes. Once the first ETL cycle is completed, the pseudonymized data are loaded into a secure research environment. At this stage, the Intego data management team conducts the second ETL cycle, which focuses on ensuring data tidiness and quality using R scripts. The database from Healthdata.be is restructured to improve usability for researchers. Originally designed for clinical consultations, its format is not optimal for research. For example, diagnosis codes, which are collected in two separate tables, are merged into one. The diagnosis data are then divided based on the nature of diagnoses: chronic conditions, which persist over a patient’s lifetime, are placed in a dedicated table, while other diagnoses related to short- or medium-term illnesses remain in a separate table. The classification of chronic diseases follows the methodology of Nielen et al. to categorize the International Classification of Primary Care (ICPC-2) codes, adapted for the Belgian healthcare system [17]. This reorganization enhances the database’s structure, making analysis more efficient for researchers. To assess the data quality (DQ) of Intego-II, we applied the harmonized data quality assessment framework tailored for the secondary use of EMR data as outlined by Kahn et al [18]. A key measure introduced is the concept of “acceptable” patients, which filters data at patient level to improve completeness and accuracy. Acceptable patients must have unique, non-null gender and birth year values. Additionally, acceptable patients must have more than one registered clinical encounter over 2 years before the release date to remove transient cases. Another key objective of the second ETL cycle is to ensure the database’s compliance with international standards for semantic interoperability. To achieve this, we adopted the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) v5.4, an open community data standard designed to harmonize the structure and content of observational healthcare data from diverse sources [19]. By aligning Intego-II with the OMOP CDM, we enable the database’s consistent and standardized representation, facilitating interoperability and comparative research across global databases. For data structuring, we developed an ETL script in R to transform Intego-II data into the OMOP CDM format. Additionally, we developed a centralized mapping table to serve as a comprehensive glossary, which aligns both source and standardized codes and vocabularies used across different tables. For epidemiological studies using EMR data, defining the patient denominator is crucial to ensure accurate surveillance. In the Intego-II database, this involves two steps: selecting eligible practices based on the quality of their morbidity data registration and estimating their population size. First, practices are evaluated to minimize recording bias by ensuring they meet specific criteria: recording, on average, more than one new diagnosis per patient per year and keeping the proportion of uncoded diagnoses below 20%. This eligibility is reassessed with each data release to update the list of practices eligible for epidemiological studies. Second, the practice population is estimated for eligible practices. In Belgium, unlike in the Netherlands or the UK, patients are not registered with a specific practice but tend to visit the same practice over time. Therefore, we use a method developed by Bartholomeeusen et al., which combines the Yearly Contact Group (YCG) with correction factors from reimbursement claims data. These factors adjust for individuals who did not visit a GP, stratified by demographics [20]. However, as these corrections have a 2-year delay, researchers should use the “active patient” population—patients with at least one recorded contact in the past 2 years and no death date—for recent data. To enhance reproducibility across multiple data extractions, the Intego-II database adopts a structured quarterly release schedule. Each version is assigned a unique Digital Object Identifier accompanied by a detailed metadata table that the database’s key characteristics and its across Researchers are to the specific version they ensuring and in data This approach aligns with the findability and of the FAIR data promoting and consistent use of the database [8]. The Intego-II database is organized into three key modules: Patient Information, Medical History, and Clinical (Fig. These are designed to longitudinal enabling researchers to link data across all using unique patient and practice identifiers, and the date with each data structure of This figure illustrates the three of the Intego-II database, their and key The in the patient encounters between first and last registration dates illustrating three from the Patient Yearly Contact Group and Active based on the between last registration date and release The Patient as the of the Intego-II The Medical provides a of a patient’s medical including both coded and uncoded data. other recorded dates the historical than a clinical The Clinical provides detailed records of patient diagnoses, prescriptions, laboratory test prescriptions, and the International Classification of the International Classification of Primary and are registered using the Belgian National codes, managed by the of of the a system for by Corilus and used in by and Over the data from Intego have the of over research to into multiple of primary care research. A of these studies has on epidemiological research, mapping the and trends of diseases from chronic and diabetes to and health Another group of studies has healthcare in test use coverage and that clinical practices. in data and have from the Intego a group of studies has contributed to public health and health policy research, and and the of major public health A full list of can be on the Intego is to that these research domains are not and the Intego-II potential for in other research domains such as and other The Intego-II database from its of coverage, and representativeness to the Flemish population. is by an two-step ETL which not only privacy standards but enhances the quality, standardization, and usability of the data. Additionally, the robust integration with the Healthdata.be infrastructure provides an scalable for Intego-II with datasets at the Belgian level. This for linkage the database to new and analyses across various domains of population these scalable linkage the of Intego-II with the OMOP CDM interoperability and comparative studies on an international the potential for these the database’s as a resource for advancing health research and informing these is the on a secure research environment for data this can on for or quality minimize as and uncoded and advanced statistical or not patient registration with a general practice, accurately defining and practice can be more tailored In Intego-II a data source for primary care research, but must for these to fully its Researchers in the Intego-II database should visit the Intego for comprehensive on the process. A Data Access must be the and The Data Access the access process based on the and compliance of the secure research ensuring that all to privacy through a Data Access that data and terms of data use is to the specific study described in the with additional or a new can be to The Intego-II database of pseudonymized EMR data collected during primary care encounters across participating practices in Flanders. The use of pseudonymized health data for the Intego project was by the KU number and the Belgian Security the Social Security and in their number with the most recent update in number December The approval a comprehensive to ensure compliance with data The the participating primary care practices for their in providing access to data, which the of the Intego-II are to the patients data were included in this ensuring to the and privacy we all of the Intego project for their to advancing healthcare research. and the and and developed the designed the database and the Intego-II database and contributed to the R for data and as the data of Intego-II and the quarterly data is the of and is the data acquisition, the Intego database by GP practices, and maintaining network and the and provided to the and as of and provided the is the on the Intego environmental project to assess impact of environmental pollution on public contributed to on the of the database infrastructure and potential All and the To maintain and data across and data and are at the Intego-II The includes on the secure research as as a and the mapping table across vocabularies with The Intego project is primarily funded by the Department of Care of the Flemish from the Research council of the via number The had no role in study data collection, to or of the Data resource access was used as a to improve the and of the The full over the the of all and full for the of the

Source: International Journal of Epidemiology

DOI: 10.1093/ije/dyaf200