This paper describes prototyping experiences in a population health use case of the draft Health DCATAP specification for health data catalogues under the European Health Data Spaces Regulation (EHDS). Using data catalogues to support data-driven health planning like this is an important use case. Our work included the development of a data catalogue metadata model, catalogue record creation via direct data entry and scraping of open data, and development of record quality and feasibility reports. It was found necessary to extend the catalogue with new classes and properties for this use case, some of which were from the Data Privacy Vocabulary (DPV), and a number of limitations in the current Health DCAT-AP specification draft were discovered. Stakeholders were generally positive in their assessment of the contribution of this novel structured approach to health data indicator discovery and assessment. This shows the potential for the semantic data governance infrastructure specified by the European Health Data Spaces Regulation to influence future data-driven decision making at all levels of European health services. The catalogue metadata model, report queries and data scraping code are all made available as open source resources for reuse by others. One new property has been added to DPV as a result of this work and it will feed into the Health DCAT-AP standardisation process in the ETSI/TC Data. This paper describes a population health use case based on defining a health and wellbeing profile for older adults, data catalogue competency questions for this use case, a metadata model for the catalogue that meets these requirements, and a data quality feasibility and assessment reporting workflow along with stakeholder feedback.
Many countries still struggle with health data management [
Hence there is a need for a structured, repeatable approach to health indicator and dataset
search and data quality appraisal, for example to feed into a wider indicator prioritisation process
[
The research question studied in this paper is “To what extent can a EHDS-compliant Health DCAT-AP data catalogue support population health indicator identification and quality assessment? ”. The technical approach is to first develop a RDF-based model of health indicators and their source datasets using Health DCAT-AP as a basis. Then open data sources were scraped to populate the indicator and dataset catalogue. Finally a prototype semantic web toolchain was developed to query the catalogues to generate a quality assessment and feasibility report as input to the expertbased health indicator prioritisation and selection process. This work was carried out in collaboration with the Irish National Clinical Programme for Older People, the National Health Service Improvement Department and the National Health Intelligence Unit in the Irish Health Service Executive1 (the body responsible for delivering health services nationally). This highlighted a number of limitations in the current Health DCAT-AP draft standard for this use case and some potential extensions.
The contributions of this paper are: i) the first documented development of a large scale secondary use application for Health DCAT-AP; ii) a set of reusable competency questions for population health indicator quality and feasibility assessment; iii) iv) a set of lessons learned from a large scale application of Health DCAT-AP in a National Health Service; and v) a set of open source scripts, RML mappings and SPARQL queries for our reporting toolchain.
The rest of this paper is structured as follows: §2 describes our use case, §3 gives an overview of related work, §4 describes our data catalogue model for population health indicators, §5 describes our case study-based evaluation and §6 provides brief conclusions.
the poverty line and who lack basic necessities. Consistent poverty is a broader measure that considers both income poverty and the experience of deprivation (inability to afford basic necessities).
Data Source
Data Source
This work was carried out in the context of the need to develop a national Older Adult Health and
Wellbeing Profile for Ireland to enable population-based planning at national, regional and local
areas called integrated health areas (IHAs). The profile would consist of a set of health indicators
1 https://about.hse.ie/
(metrics), typically with a name, definition, numerator and denominator, measurement unit, data
source, time frame, rationale and limitation (see Table 1 for a brief example). Each numerator and
denominator could have separate data sources and there are a large number of candidate datasets
in Ireland from the Central Statistics Office, The Irish Longitudinal Study on Ageing (TILDA) a
large-scale, nationally representative, longitudinal study on ageing in Ireland, the HSE National
Health Intelligence Unit Core Indicator List, the Irish Hospital In-Patient Enquiry (HIPE) system,
charities, and international sources such as the OECD or EuroStat. A seven step health indicator
prioritisation process was developed to enable review and input from experts, patients, data
publishers, international best practice, planners and policy-makers [
However the wide variety of data sources and lack of established data catalogues or unified data governance processes meant that information was siloed on questions of data source quality and feasibility, for example: • Does this dataset cover the appropriate population? (i.e. people who are aged 65+) • Does this dataset support appropriate spatial and temporal granularity for this use case? • Is this data updated frequently enough to fit the indicator? • Is this data accurate and complete enough for population-based planning? • How easy will it be to find and use this data? • Are there data protection concerns for using this dataset? If it was available, then a well maintained data catalogue could answer many of these questions which are orthogonal to the issue of the clinical suitability of a given indicator.
A data catalogue is a repository and metadata management tool that provides an organised and
searchable inventory of an organisation’s data assets. It is a fundamental enabler of data
governance in or between organisations. Data catalogues empower users to discover, understand,
and leverage data for analytical purposes, reporting, and informed decision-making [
A key development for all health data sharing is the European Health Data Space (EHDS)
Regulation2 which came into force in March 2025. This will result in primary legislation supporting
health data sharing for primary and secondary uses by 2030. However the EHDS outcomes are
mainly legal, ethical and regulatory. Deployment relies on technical details based on the
recommendations of the Data Spaces Support Center (DSSC)3 which coordinates between many
Data Space initiatives and projects such as Gaia-X4. Use of linked data knowledge graphs to
organise machine readable data is central to the DSSC plans, as are DCAT (Data Catalog
Vocabulary)-based data catalogues in dataspaces protocol specification of the International Data
Spaces Association (IDSA)5 referenced by the DSSC. HealthDCAT-AP (Application Profile) is being
developed by the EU Health Data Spaces Pilot project as use of DCAT-AP is recommended by them
for the EHDS. Given the sensitivity of health data there is a crucial role for security in the EHDS
and there is emerging work on how this may be applied to sharing machine-readable knowledge
models [
Since data protection concerns are central to sensitive heath data processing, part of the
extensions provided by Health DCAT-AP are additional fields to cover this. These additional fields
2 https://health.ec.europa.eu/ehealth-digital-health-and-care/european-health-data-space_en
3 https://dssc.eu/
4 https://gaia-x.eu
5 https://github.com/International-Data-Spaces-Association/ids-specification/releases/tag/2024-1
are taken from the Data Privacy Vocabulary (DPV) [
that is not controlled by the HSE? (i.e. needs data sharing agreement)
A set of requirements were developed for the data catalogue based on the public Health DCAT-AP
draft, a series of stakeholder workshops from January to May 2025 and examining the literature
and public metadata for existing Older Adult Health and Wellbeing Profiles such as the UK
National Health Service (NHS) Fingertips6. The workshops consisted of over 30 individual
contributors from the public health professionals, population health experts, patient
representatives, clinicians from the National Clinical Programme for Older People, health service
providers, knowledge engineers and data governance experts. One face to face workshop was held
in January 2025, and three smaller online groups met in March, April and May. Documenting the
overall use cases for the Older Adult Health and Wellbeing Profiles and their reporting
requirements resulted in a set of competency questions shown in Table 1. This broadly followed
the NeOn methodology [
A data catalogue metadata model (Fig. 1) was then developed with the guiding principles of: i) using the Health DCAT-AP sub-profile for sensitive data as a foundation; and ii) including minimal additional fields to answer the competency questions and stakeholder concerns. This resulted in 40 data fields being used for datasets. A set of application profile recommendations were also defined as per Health DCAT-AP with each field being marked mandatory, recommended or optional. This started with the Health DCAT-AP profile for sensitive data as a baseline for the profile constraints 6 https://fingertips.phe.org.uk/ of metadata fields. This worked well for the dataset record but some mandatory fields are currently impossible to complete such as the identity of the Irish Health Data Access Body (a required entity under EHDS) has not yet been specified in Irish law so had to be ignored. In general the fields we added were made recommended or optional. All fields that were necessary to answer our competency questions were made mandatory for this profile as the purpose of this exercise was to enable generation of a quality and feasibility report from the data catalogue. Having local conformance profile that is higher than the standard one will not decrease interoperability with other EHDS data catalogues. However some applications like this pilot study may also choose to relax conformance requirements compared to Health DCAT-AP simply due the limitations of resources and lack of metadata availability. An example of a field treated this way for this study was the Health DCAT-AP required field “sample” which provides a sample distribution of data from the dataset. It is already only mandatory in the Health DCAT-AP “non-public” or sensitive profile and this suggests some variability is expected. For indicators as opposed to datasets the conformance profile has less guidance and our approach was to make mandatary the fundamental elements (e.g. name, numerator) that are required to display the indicator in a health profile and other elements like rationale for the indicator were classified based on the clinical members of the team’s experience and their perceived importance for the final health profiling work.
Table 3 New Metadata Fields Added to Health DCAT-AP for this use case
Field
indicator prioritisation process. Values: Include in Catalogue, Exclude from Catalogue, Under Consideration, Exclude from Profile, Include in Profile
5.1. Deployment Context
The data catalogue model described above was tested by deploying in support of the use case
described in section 2. The goal was to provide a structured way to record information about the
large number of datasets (n=24) and indicators (n=1146) being considered in the process of defining
the final set of indicators for the Older Adult Health and Wellbeing Profile. This was a seven step
process, see McGlacken et al. [
To populate the catalogue a workflow (Fig. 2) was defined using web scraping scripts to CSV files, some manual data entry and semi-automated valuation before uplift using R2RML. The RDF-based data catalogue had SPARQL scripts created to answer the competency questions in Table 2. All are provided as open source in our public code repository8. The dataset and indicator records were defined as a CSV file which included the metadata field mappings to standard RDF terms from the Health DCAT AP and DPV, application profile cardinality constraints (Mandatory, Recommended and Optional), and documentation on all categorical field values to assist domain experts in completing or validating the metadata. A spreadsheet-based solution for data entry was adopted as the fastest and most familiar interface for health service staff in this pilot study due to the current lack of standardised tooling for Health DCAT AP deployed within the Irish Health Service. This also facilitated the development of web scraping scripts for metadata open source datasets by web engineers unfamiliar with Semantic Web technology. Version control and managing the master spreadsheet must be carefully approached in this case as only disciplined use of these tools will efficiently enable data processing. The merged master spreadsheet needed some semi-manual data curation to fix obvious errors like duplication and missing identifiers. Dataset and indicator identifier management became another challenge as many data sources do not have web-based Linked Open Data-style identifiers available and so an identifier creation and governance process was needed. Finally the master data was processed by a set of R2RML mappings to produce standarised RDF-based metadata which could be queried for the quality assessment process. There were 2 aspects to this: i) quality assessment of the metadata itself (e.g. completeness i.e. adherence to the application profile we had defined for mandatory fields) for self assessment of our progress in creating the data catalogue and ii) answering the competency questions to generate a dataset (and indicator) quality and feasibility report for input into the health profile prioritisation process. Figure 1 illustrates how the input of domain experts were necessary at most stages of the population health metadata ingestion, validation, assessment and reporting pipeline we built. 5.2. Evaluation and Self Reflection This work is still ongoing and so definitive findings on the effectiveness of this approach will be published subsequently. Therefore we focus on our findings from applying Health DCAT-AP as the basis for modelling the health indicator datasets and developing the metadata and population health indicator data quality report pipeline described above. Thus this reflection focuses on the vocabulary or schema aspects of this system. 8 https://github.com/junli-liang-johnny/hse-scripts The following Health DCAT-AP issues were identified: 1. Complexity. Health DCAT AP builds on many other specifications and data stores for defining the contents of several properties. Discussion: This takes a long to follow all the sources, even for someone who is very familiar with Linked Data. 2. Insufficient attention is provided to data protection as can be seen from our additional fields. In some cases the reuse of DPV properties used by Health DCAT-AP seem to assumes a string can be used as the range of the property when a range class is defined in DPV and these cases should be made compliant with the DPV specification.Discussion: This is important as use/reuse of datasets is often critically dependent on understanding the data protection status, questions like iis this personal data?’, ‘who is the data controller?’ are critical to making data governance decisions. 3. Spatial resolution of datasets is identified in metres. It would be much more useful for population health to be able to specify national, county, NUTS2 regions and also new regional areas. Discussion: most statistical datasets are colleced with a spatial component but it is with reference to standard polygons for counties or statistical regions rather than raw spatial measurements like meters. 4. Spatial coverage is limited to regions that are modelled in Geonames. This does not include regional subdivisions like the IHAs defined by the Irish health service.Discussion: It is not a sustainable solution to have coverage definitions managed by a 3rd party private organisation so other spatial spatial region definition authorities like National Mapping Agencies should be allowed by the specification. 5. In many cases Health DCAT-AP defers to WikiData for the definition of categorical field codes. This is a fine pragmatic solution but it should not be the only source allowed e.g. National authorities should be able to publish their own IDs. Discussion: many authoritative data sources are not linked to WikiData and they should be allowed. 6. No Publisher codes were defined by Health DCAT-AP. A suggested set of codes we found by searching the EHDS text was: National Public Health Institute, National Mapping Agency, Statistical Agency, Hospitals and Healthcare Providers, Universities and Research Centers. Health Departments. Community-based and Clinical Care Organisations.
This work has shown that the EHDS gives a large opportunity to improve the data governance infrastructure for all healthcare data governance. In particular the Health DCAT-AP draft specification gives a strong basis for implementation but will need to be refined further for deployment, especially for secondary use cases. The results of this will be fed into the standardisation process both nationally and at ETSI.
This research was conducted with the financial support of the Health Service Executive National Clinical Programme for Older People Research Award 2024 (NCPOP RA3/2024) and the Research Ireland ADAPT Research Centre in the RI Research Centres Programme (Grant 13/RC/2106_P2), For the purpose of Open Access, the authors have applied a CC-BY public copyright license to any author accepted manuscript version arising from this submission.
We would like to extend our gratitude and appreciation to the stakeholders for their feedback and engagement with this work. Special thanks to our colleagues in the HSE for their invaluable contributions to this paper: Thereese McGlacken, Stephen Barrett, Jacinta Mulroe, Teresa Bennett, Declan McKeown, Gerardine Sayers, Mary Browne, Aparna Keegan, Graham Hughes. The authors have not employed any Generative AI tools.