Nivel Primary Care Database (Nivel-PCD) contains longitudinal, pseudonymized, linkable extracts from electronic health record (EHR) data of general practitioners (GPs), out-of-hours (OOH) GP services, physiotherapy practices, exercise therapy practices, speech therapy practices, and dietitian practices in the Netherlands.
These primary care providers periodically supply Nivel-PCD (through their EHR software provider) with extracts of EHR data, including contacts, health problems, treatment, measurements, and referrals to secondary care.
The coverage in 2023 was 1.8 million persons registered at 409 GPs, 12.3 million people living in a catchment area of 26 OOH-GP services, 970 000 patients who were treated in 508 physiotherapy practices, 25 000 patients in 116 exercise therapy practices, 54 000 patients in 175 speech therapy practices, and 68 000 patients in 83 dietitian practices.
Data are available at contact levels from 2011 onwards, can be aggregated to any other overarching level (e.g. individual, practice, region), and are linkable with other data sources such as secondary care, social care, or social economic data.
The unique combination of integrated data from various primary care disciplines in the Netherlands provides unique opportunities for research. Data are accessible upon request for bona fide researchers worldwide; researchers can file a data request through nivel.nl/en/our-databases-and-panels/nivel-primary-care-database.
Data resource basics
Rationale
Large amounts of health data are recorded routinely in healthcare settings. In many countries, primary care facilities use an electronic health record (EHR) system. The data that are recorded in these systems for patient care and administrative purposes are increasingly used for other purposes as well, known as “secondary” or “further” use [1]. The purposes of the secondary use of primary care data include epidemiology, incidence and prevalence of (chronic) diseases, disease surveillance, and analysis of patterns of healthcare utilization. Moreover, primary care data are used to construct and monitor quality of care and to evaluate policy measures. Primary care data are increasingly linked to other data sources, to form a more complete picture of patient groups, their health problems, and their use of care, and to follow patients on their journey through the health and social care system [2].
Nivel, the Netherlands Institute for Health Services Research, started a paper-based sentinel network of general practices in 1970 in order to gain insight into the epidemiology of a number of illnesses and conditions that were presented to the general practitioner (GP) [3]. This network infrastructure transitioned in the 1990s from paper-based data collection into extractions from EHRs and broadened to include GPs as well as other primary healthcare providers; it is currently known as Nivel Primary Care Database (Nivel-PCD).
Geographic coverage
In the Netherlands, the GP acts as the first access point to the healthcare system and almost every citizen is registered with a GP [4]. Medical care during office hours is provided by GPs. The provision of urgent out-of-hours (OOH) primary care is managed collaboratively in regional cooperatives (OOH-GP services).
Practices that are participating in Nivel-PCD are located across the country, as shown in Figure 1. The Netherlands is a small country (i.e. 41 543 km2) with 17.8 million citizens [5]. The three largest cities (Amsterdam, Rotterdam, and the Hague; each with >500 000 citizens) are located in the west of the country. The northern, eastern, and southern regions have middle-sized cities and more rural areas.
Coverage of participation in Nivel Primary Care Database, the Netherlands, 2023. (a) General practices. (b) Out-of-hours general practitioner services (coverage yes/no). (c) Physiotherapy practices. (d) Exercise therapy practices. (e) Speech therapy practices. (f) Dietitian practices. Darker shades indicate higher coverage. Areas with the lightest shade have no coverage.
In 2023, Nivel-PCD for GPs had a nationally representative coverage of 10% of the Dutch population (i.e. 1.8 million) in 409 participating GPs out of a national total of 4900 [6]. The catchment area of the OOH-GP services in Nivel-PCD represented 69% of the Dutch population (i.e. 12.3 million). From the 52 OOH-GP services in the Netherlands, 26 (i.e. 50%) participated in 2023. Coverage of the yearly provided care by the participating physiotherapy and exercise therapy practices was 17% and 13%, respectively, based on the number of therapists in Nivel-PCD compared with the total number of therapists in the Netherlands in 2023. Coverage of the yearly provided care by dietitian and speech therapy practices was 16% and 19%, respectively, based on the number of patients in Nivel-PCD compared with the number treated according to the Dutch Healthcare Authority [7]. Patients were registered at these practices when they had at least one contact with the allied health professional. The representativeness of practices that are participating in Nivel-PCD is shown in more detail in the Supplementary material.
Around 35 general practices, so-called GP sentinel practices, act as a nationally representative subgroup of ±1% of general practices in the Netherlands. The GP sentinel practices collect data through ICPC-triggered questionnaires concerning e.g. influenza, eating disorders, or suicidality [8] and have a coverage of ∼135 000 registered persons (i.e. 0.8% of the Dutch population). For disease surveillance purposes, ∼100 of GP sentinel practices, with a coverage of ∼500 000 registered persons (i.e. 3% of the Dutch population), collect oro-/nasopharyngeal swabs from a selection of patients who present with influenza-like illness. A selection of GP sentinel practices provide questionnaires and collect swabs (Figure 2). The data collected from the GP sentinel practices are not requestable, and are therefore not mentioned under “Data collected.”
Geographical spread (at the level of municipalities) of sentinel general practices participating in Nivel Primary Care Database, the Netherlands, 2024. Black circle: municipalities with sentinel practices participating in questionnaire data and oro-/nasopharyngeal swab collection. Gray: municipalities with sentinel practices participating only in oro-/nasopharyngeal swab collection. Black triangle: municipalities with sentinel practices participating in all types of activities.
Data governance and privacy
Under Dutch law, the use of EHRs data for research purposes is allowed under certain conditions without individual explicit or non-explicit consent. When these conditions are fulfilled, neither informed consent nor medical ethics committee approval are needed [1, 9]. Data may be used for research for the common good if consent is not practically feasible and if all possible measures have been taken, technically and organizationally, to avoid privacy risks.
First, patients are informed appropriately about the existence of Nivel-PCD and an opt-out is offered. Patients are informed of this option with (digital) posters in the waiting rooms of participating primary care providers.
Second, patient data are pseudonymized by a Trusted Third Party before leaving the EHR system of participating primary care practices. The pseudonymized data are subsequently sent to Nivel. Thus, directly identifiable information does not leave the EHR system while the pseudonyms allow data linkage between databases. So, pseudonymized EHR data are created in the practices, sent to Nivel, and stored in Nivel-PCD [9].
Third, the governance includes steering, scientific, privacy, and advisory committees that altogether safeguard Nivel-PCD activities. The steering committee consists of Nivel management and chairs of the scientific committees. The scientific committees consist of representatives of umbrella organizations of professions that are participating in Nivel-PCD. The privacy committee consists of two data protection officers and a patient representative. Stakeholders are members of the advisory committee. The funder (Dutch Ministry of Health, Welfare and Sport) is excluded from all committees to guarantee that Nivel-PCD is an independent data source.
Fourth, all the safety measures described ensure adherence to the General Data Protection Regulation (GDPR) that took effect in the EU in 2018 (art. 24 GDPR Implementation Act in conjunction with art. 9.2 sub j GDPR). Nivel-PCD has been externally evaluated and is recertified every 3 years according to national standards for Information Security Management Systems (NEN7510- based on ISO/IEC 27001). Nivel research procedures have been ISO-9001 (Quality Management System) certified since 1999.
Data collected
The contents of Nivel-PCD are shown in Table 1 and include patient characteristics, e.g. sex, age, and level of urbanization (based on the first four digits of the postal code). For GPs, the database includes health problems recorded using the International Classification of Primary Care (ICPC-1) [10], consultations and interventions based on claims codes, drug prescriptions recorded using Anatomical Therapeutic Chemical (ATC) codes [11], laboratory and clinical measurements using a Logical Observation Identifiers Names and Codes (LOINC) [12] compatible Dutch coding system, and referrals to medical specialist care including indication and specialism by using a national coding system.
Contents of Nivel Primary Care Database per type of healthcare provider, the Netherlands, 2023
| Primary care provider | No. of practices | No. of patients; % coveragea | Contentsb | Available time period |
|---|---|---|---|---|
| General practices | 409 | 1 760 300 (registered); 10% | A–F | 2011–ongoing |
| Out-of-hours general practitioner services | 26 primary care cooperativesc | 12 268 600 (catchment area); 69% | A–D, G (E + F as of 2024) | 2012–ongoing |
| Physiotherapy practices | 508 | 971 200 (treated); 17% | A–C, E, H, I | 2013–ongoing |
| Exercise therapy practices | 116 | 25 100 (treated); 13% | A–C, E, H, I | 2013–ongoing |
| Speech therapy practices | 175d | 54 400 (treated); 19% | A–C, H, I | 2012–22 |
| Dietitian practices | 83 | 68 200 (treated); 16% | A–C, E, H, I | 2013–ongoing |
| Primary care provider | No. of practices | No. of patients; % coveragea | Contentsb | Available time period |
|---|---|---|---|---|
| General practices | 409 | 1 760 300 (registered); 10% | A–F | 2011–ongoing |
| Out-of-hours general practitioner services | 26 primary care cooperativesc | 12 268 600 (catchment area); 69% | A–D, G (E + F as of 2024) | 2012–ongoing |
| Physiotherapy practices | 508 | 971 200 (treated); 17% | A–C, E, H, I | 2013–ongoing |
| Exercise therapy practices | 116 | 25 100 (treated); 13% | A–C, E, H, I | 2013–ongoing |
| Speech therapy practices | 175d | 54 400 (treated); 19% | A–C, H, I | 2012–22 |
| Dietitian practices | 83 | 68 200 (treated); 16% | A–C, E, H, I | 2013–ongoing |
Based on percentage of Dutch population for general practitioner and out-of-hours general practitioner services, on number of therapists in Nivel Primary Care Database versus nationally for physiotherapy and exercise therapy practices, and on number of treated patients in Nivel Primary Care Database versus nationally for speech therapy and dietitian practices.
A: Patient characteristics; B: Diagnoses/symptoms; C: Consultations; D: Prescriptions; E: Measurements; F: Referrals; G: Triage following a national triage protocol; H: Additional information on the start and evaluation of the treatment trajectory for the specific diagnosis; I: Referrer.
A primary care cooperative consists of at least one out-of-hours general practitioner service. Patient numbers are rounded to the closest multiple of 100.
In 2022.
Contents of Nivel Primary Care Database per type of healthcare provider, the Netherlands, 2023
| Primary care provider | No. of practices | No. of patients; % coveragea | Contentsb | Available time period |
|---|---|---|---|---|
| General practices | 409 | 1 760 300 (registered); 10% | A–F | 2011–ongoing |
| Out-of-hours general practitioner services | 26 primary care cooperativesc | 12 268 600 (catchment area); 69% | A–D, G (E + F as of 2024) | 2012–ongoing |
| Physiotherapy practices | 508 | 971 200 (treated); 17% | A–C, E, H, I | 2013–ongoing |
| Exercise therapy practices | 116 | 25 100 (treated); 13% | A–C, E, H, I | 2013–ongoing |
| Speech therapy practices | 175d | 54 400 (treated); 19% | A–C, H, I | 2012–22 |
| Dietitian practices | 83 | 68 200 (treated); 16% | A–C, E, H, I | 2013–ongoing |
| Primary care provider | No. of practices | No. of patients; % coveragea | Contentsb | Available time period |
|---|---|---|---|---|
| General practices | 409 | 1 760 300 (registered); 10% | A–F | 2011–ongoing |
| Out-of-hours general practitioner services | 26 primary care cooperativesc | 12 268 600 (catchment area); 69% | A–D, G (E + F as of 2024) | 2012–ongoing |
| Physiotherapy practices | 508 | 971 200 (treated); 17% | A–C, E, H, I | 2013–ongoing |
| Exercise therapy practices | 116 | 25 100 (treated); 13% | A–C, E, H, I | 2013–ongoing |
| Speech therapy practices | 175d | 54 400 (treated); 19% | A–C, H, I | 2012–22 |
| Dietitian practices | 83 | 68 200 (treated); 16% | A–C, E, H, I | 2013–ongoing |
Based on percentage of Dutch population for general practitioner and out-of-hours general practitioner services, on number of therapists in Nivel Primary Care Database versus nationally for physiotherapy and exercise therapy practices, and on number of treated patients in Nivel Primary Care Database versus nationally for speech therapy and dietitian practices.
A: Patient characteristics; B: Diagnoses/symptoms; C: Consultations; D: Prescriptions; E: Measurements; F: Referrals; G: Triage following a national triage protocol; H: Additional information on the start and evaluation of the treatment trajectory for the specific diagnosis; I: Referrer.
A primary care cooperative consists of at least one out-of-hours general practitioner service. Patient numbers are rounded to the closest multiple of 100.
In 2022.
OOH-GP services data include information on triage outcomes, including the urgency level of the presented health problem (e.g. critical, urgent, non-urgent), consultations based on claims data (type of consultation), health problems, and prescriptions.
For physiotherapy practices and exercise therapy practices, data include health problems based on Diagnosis Code Systematic Paramedical Help codes [13], consultations based on claims data and information on the treatment trajectory regarding the start of the treatment (e.g. prognosis, referrer), and evaluation of the treatment (e.g. outcome measures, result).
The data from speech therapists and dietitians include health problems based on a national diagnosis coding system [14] and additionally include information on prognosis, consultations based on claims data, patient measurements, treatment (type/result), and referrer (i.e. GP/specialist/other/none).
Data resource use
From each primary care database, yearly reports are published on care uptake and registered health problems. These reports, as well as a weekly “Nivel Surveillance” bulletin, are published at https://www.nivel.nl/nl/zorg-en-ziekte-in-cijfers/actuele-cijfers-ziekten-per-week. Methods for our surveillance system are published in [15]. To give some examples of the unique possibilities for research, some publications based on Nivel-PCD data are highlighted below; this is not an exhaustive list and is rather meant to be indicative.
Epidemiological studies include: improving morbidity estimates by combining GP data with claims data from secondary care providers and mortality statistics [16] and improvement of risk equalization for Dutch healthcare insurance companies [17]. Furthermore, studies on influenza [18], respiratory syncytial virus (RSV) [19], knee and hip osteoarthritis [2], socioeconomic inequalities in OOH-GP services use [20], and the impact of COVID-19 on healthcare utilization [21] have been published. Prescriptions data were used in studies on e.g. asthma medication use and outcomes [22], non-adherence to drugs prescribed by GPs [23], adoption of new medicines in primary care [24], and opioid prescriptions [25]. Quality of care was investigated in studies on e.g. the adherence to cancer treatment guidelines [26], pediatric respiratory tract infections [27], and co-prescription of laxatives with opioids [28]. Treatment outcomes are discussed for e.g. care for burn victims [29], shoulder pain [30], chronic obstructive pulmonary disease [31], and weight loss achievement [32].
Strengths and limitations
The main strengths of Nivel-PCD is that it contains data of GPs, OOH-GP services, physiotherapy practices, exercise therapy practices, speech therapy practices, and dietitian practices. The unique combination of longitudinal, integrated data from diverse primary care disciplines in the Netherlands provides unique opportunities for research. The GP part of Nivel-PCD concerns both a large and a representative sample of patients according to sex, age, and degree of urbanization whereas the OOH-GP services cover the larger part of the Netherlands. Data on representativeness per primary care provider are available in the Supplementary material. The use of pseudonymized patient data makes it possible to link Nivel-PCD data between the different primary care disciplines and socioeconomic characteristics, but also with data that are generated in the rest of the healthcare system, e.g. claims data from healthcare insurers and secondary care (hospital) data.
Nevertheless, some limitations should be addressed. As a database with extracts from EHRs, Nivel-PCD is reliant on the registration habits from primary care providers for data quality. Changes in the healthcare system or coding systems (e.g. ICPC-3 [33]) pose continuous potential threats to the continuous utility of the data [34]. Nivel-PCD is dependent on a considerable number of software vendors; to ensure good quality, great effort is taken to stay in touch with new and existing vendors. Extracting data from EHRs is complex and challenging due to the existence of multiple EHR systems that are often designed differently. Many physio- and exercise therapy practices compose the data extractions themselves, which is a demanding task. Currently, all content within Nivel-PCD comes from coded data; free text is not extracted from the EHRs. However, steps are being taken to enrich Nivel-PCD by exploiting the potential of data science and artificial intelligence.
Data resource access
Any bona fide researcher can apply for data by filling out a request form at https://www.nivel.nl/en/our-databases-and-panels/nivel-primary-care-database. The Nivel-PCD team is a group of researchers with multidisciplinary expertise and handles on average 45 data requests per year; all handled data requests are published on our website (nivel.nl). Data requests are granted on the basis of the Nivel-PCD governance. A fee is charged for data handling, which may also include analysis by one of our researchers (upon request). The governance protocol can be accessed (in Dutch) on our website (nivel.nl). First, data requests are reviewed for need-to-know and proportionality. Then, the scientific committee reviews the feasibility. As stated earlier, the scientific committee consists of representatives of professional associations of primary care providers. In the case of data linkage, the privacy committee reviews the request. When there are no objections from the committees, a data-sharing agreement is signed. Data are made available within a secure data environment. Finally, the researcher publicly publishes the results. The latter is a requirement of the data-sharing agreement.
Ethics approval
Nivel-PCD contains observational data; therefore, research and reporting are not subject to the Medical Research Involving Human Subjects Act and do not require approval from a medical research ethical committee. This is described further under “Data governance and privacy.”
Acknowledgements
We want to thank all primary care providers and their patients who are contributing to Nivel-PCD and the steering, scientific, privacy, and advisory committees. We are also grateful for the support received from all members of the Nivel-PCD team, past and present.
Author contributions
J.W.V. and L.I.O. drafted and finished the text. R.A.V. was responsible for the overall design and implementation of the Nivel-PCD infrastructure. All authors provided information for the content and critically reviewed the manuscript.
Supplementary data
Supplementary data are available at IJE online.
Use of artificial intelligence (AI) tools
AI tools were not used in the collecting or analysing of the data, in producing images or graphical elements, or in the writing of this paper.
Conflict of interest
None declared.
Funding
This work was supported by the Dutch Ministry of Health, Welfare and Sport, who has provided funding to Nivel in order to maintain Nivel-PCD since 2014 for the production of key figures that are deemed important for evaluation of (public) health policy in the Netherlands.
Data availability
See “Data resource access,” above.
