Open data in healthcare: a complete guide to understand, source, and leverage official data

 

Each year, billions of health data points are generated by hospitals, regulatory agencies, epidemiological registries, and reimbursement systems. Yet the vast majority of this data remains untapped, not because of access barriers, but due to a lack of structuring. This guide answers the fundamental questions about open data in healthcare: what it is, where to find it, why it is difficult to exploit, and how to transform it into decision-making intelligence.

Q1 - What is health data?

Health data is any information relating to a person's physical or mental health, their medical care, or their consumption of healthcare services. Under European law (Article 4, paragraph 15 — EU Regulation 2016/679), it constitutes sensitive personal data, subject to enhanced protection.

 

Two main categories can be distinguished:

• Individual health data : medical records, prescriptions, lab results, imaging. Strictly regulated by GDPR and HDS regulations in France.
• Population health data : epidemiology, hospital statistics, aggregated real-world data. This is the data that forms the core of open data in healthcare.

 

Q2 - What is open data? And open data in healthcare?

Open data refers to data made available to the public freely, openly, and reusably, without major technical or legal restrictions. In healthcare, open data covers all medical, epidemiological, regulatory, and hospital data published by official bodies (WHO, EMA, HAS, Eurostat, national health agencies) and freely accessible.

 

According to the Open Definition v2.1 from the Open Knowledge Foundation, data is considered open if "anyone is free to access, use, modify, and share it". In practice, this implies four concrete requirements:

• Open license or public domain : no legal restriction on reuse
• Free access : available free of charge and downloadable online
• Machine readability : in a format processable by a computer
• Open format : with no technical or monetary restrictions on its use

 

Q3 - What is the difference between raw data and structured data?

Raw data is data in its original, untransformed state: an HAS report in PDF format, a CSV file exported from a hospital registry without standardized headers, free text from a medical report. This data exists, but using it directly is tedious and costly.

 

Structured data, by contrast, is organized according to a defined logical model and stored in directly queryable formats: SQL relational databases, standardized REST APIs, documented data warehouses.

 

In practice, transforming a raw epidemiological report into usable structured data can represent several weeks of data engineering work. It is precisely to eliminate this cost that Tekkare has spent 10 years structuring and normalizing more than 500 official sources in OIP Databank®, so that your teams can directly access ready-to-use data, with no preliminary processing pipeline required.

 

Q4 - What is health data structuring?

Data structuring consists of transforming raw, heterogeneous information into an organized, documented, and directly usable corpus. In practice, structuring health data involves several successive steps:

• Collection and ingestion : automated retrieval from primary sources (official APIs, public files, registries)
• Cleaning : detection and correction of errors, removal of duplicates, handling of missing values
• Semantic normalization : unification of nomenclatures so that the same medication, the same disease, or the same country carries the same identifier across all sources
• Modeling : organization into relational tables with documented metadata
• Continuous updating : regular synchronization with primary sources to maintain data freshness

 

Q5 - What are the main sources of open data in healthcare?

The landscape of open health data sources is fragmented across international organizations, national regulatory agencies, and specialized databases. Here is a structured overview.

International sources

• WHO Global Health Observatory : global health statistics, mortality, morbidity, health systems in 200+ countries
• EMA (European Medicines Agency) : marketing authorizations, scientific assessments, pharmacovigilance data
• FDA (USA) : U.S. marketing authorizations, clinical trial data, pharmacovigilance databases
• ClinicalTrials.gov : global registry of clinical trials, 400,000+ registered studies
• EUCT / CTIS : European clinical trials registry
• PubMed : 35+ million indexed biomedical publications, free access via API
• Eurostat : harmonized European health statistics (hospitalizations, causes of death, healthcare spending)
• OECD Health Statistics : performance indicators for member countries' health systems

 

French national sources

• SNDS (French National Health Data System) : reimbursement database of the French national health insurance, care pathways for 67 million people
• PMSI (Hospital Information Systems Medicalization Program) : hospital activity data, diagnoses, and procedures
• HAS (French National Health Authority) : medico-economic assessments, transparency opinions, recommendations
• Santé Publique France (French Public Health Agency) : epidemiological surveillance, disease registries
• ANSM (French National Agency for Medicines Safety) : pharmacovigilance, clinical trial authorizations in France

Note: Tekkare aggregates and structures all of these sources in OIP Databank®, making them directly usable via API, each source normalized, documented, and certified HDS + ISO 27001.

 

Q6 - Why is health data difficult to exploit?

It is one of the sector's major paradoxes: health data is abundant, yet exploiting it remains complex for four structural reasons:

 

Fragmentation of sources

Data is scattered across dozens of organizations with no common format or unified access point. A single epidemiological indicator can exist in 12 different formats depending on the source.

 

Heterogeneity of formats

Non-standardized PDF, Excel, CSV files, free clinical text: each source speaks its own technical language. Cross-referencing a PMSI file with EMA data and PubMed publications without dedicated infrastructure takes weeks of data engineering work.

 

Lack of semantic normalization

The same medication can be referenced under 5 different names depending on the country and the agency. Without a common reference framework, data cross-referencing becomes impossible or generates silent errors.

 

Variable update frequency 

Some sources are updated daily (clinical trials, etc.), others annually (demographic statistics). Maintaining a consistent, synchronized, up-to-date corpus requires a continuous collection infrastructure that few teams can build in-house.

 

Q7 - Why is open data strategic in healthcare?

Open data in healthcare is not only a question of access to information; it is a direct competitive lever for pharmaceutical, biotech, and institutional players.

 

For pharmaceutical decision-making, open data enables data triangulation: a drug launch strategy relies simultaneously on epidemiological data (market potential), regulatory data (marketing authorizations, HTA dossiers), hospital data (patient pathways), and competitive data (pipelines of other manufacturers). Without structured open data, each team works in a silo with a partial view.

 

For artificial intelligence in healthcare, official open data has become essential. AI models require massive volumes of clean, labeled, normalized data. Structured official open data is the only source that meets these criteria without selection bias or representativeness issues.

 

Q8 - What is the difference between open data and proprietary data?

In terms of access, open data is free and unrestricted whereas proprietary data is paid and contract-bound. In terms of reliability, open data comes from official, traceable, and verifiable sources, while the quality of proprietary data varies by provider. Regarding structuring, proprietary data often arrives pre-normalized, whereas open data is frequently provided in raw form and requires preliminary processing work. Regarding scientific replicability, finally, open data offers a decisive advantage: analyses can be verified and audited by third parties since the sources are public, which is impossible with non-reproducible proprietary data.

 

Tekkare combines the advantages of both: data from official open data sources (reliability, traceability, replicability) structured and normalized like proprietary data (ready-to-use, interoperable, kept up to date).

 

Q9 - How is open health data updated?

Update frequency varies considerably from one source to another, it is one of the least documented criteria, yet one of the most critical for the reliability of analyses.

• Daily : clinical trials (ClinicalTrials.gov, EUCT), new marketing authorizations (EMA, FDA), pharmacovigilance alerts
• Monthly : partial reimbursement data (SNDS), national epidemiological bulletins
• Quarterly : financial results and pipelines of pharmaceutical manufacturers
• Annually : consolidated hospital statistics (PMSI), demographic data (Eurostat, OECD), cancer registries

Cross-referencing sources with different cycles without explicit handling of temporal consistency introduces silent inconsistencies in the results. 

 

This is why OIP Databank® integrates a continuous update system that synchronizes each source according to its own cadence, while ensuring overall consistency of the data warehouse.

 

Q10 - What is health data interoperability?

Interoperability is the ability of heterogeneous systems and datasets to communicate, exchange, and use information consistently, without loss of meaning or need for manual adaptation.

 

In healthcare, technical interoperability relies on international standards:

• ICD-10 / ICD-11 : International Classification of Diseases (WHO)
• ATC : Anatomical, Therapeutic, and Chemical classification of medicines
• Orphanet : rare disease reference framework
• HUGO : international nomenclature of human genes

Without interoperability, cross-referencing French hospital data with European registries and marketing authorizations is a project of several months. With a normalized data warehouse, it is a single query.

 

Q11 - What is multi-source health data normalization?

Normalization consists of harmonizing data from heterogeneous sources according to a unified model, in order to make them comparable and usable together. 

In practice, normalizing multi-source health data involves several dimensions:

• Unification of nomenclatures: the same medication, the same disease, the same country must carry the same identifier throughout the system, regardless of the name used by each source
• Format standardization : dates, units of measure, currencies, and geographic codes must be uniform to enable temporal and geographic comparisons
• Resolution of semantic conflicts: when two sources provide different values for the same indicator, an explicit rule determines which source prevails, and that rule is documented
• Metadata creation: each data point is accompanied by its source, collection date, and calculation method, essential for traceability and auditability considering regulatory requirements

This is precisely the work that distinguishes OIP Databank® from a simple link aggregator: each data point is cleaned, normalized, documented, and ready to directly feed your analyses, dashboards, or artificial intelligence models.

 

Source reference:

Règlement (UE) 2016/679 du Parlement européen, article 4, eur-lex.europa.eu

Defining Open in Open Data, Open Content and Open Knowledge, Open Definition v2.1 https://opendefinition.org/od/2.1/en/

OECD (2019), Health in the 21st Century: Putting Data to Work for Stronger Health Systems, OECD Health Policy Studies, OECD Publishing, Paris, https://doi.org/10.1787/e3b23f8e-en.

Ethics and governance of artificial intelligence for health. Guidance on large multi-modal models. Geneva: World Health Organization; 2024. Licence: CC BY-NC-SA 3.0 IGO. https://iris.who.int/server/api/core/bitstreams/e9e62c65-6045-481e-bd04-20e206bc5039/content