Health data: asset or responsibility?

As Chinese hospitals begin listing clinical datasets on data exchanges to feed the AI industry, the question for Vietnam is not whether hospitals should sell data, but whether they are ready to govern data as a responsible asset.

A new wave across the border

According to Nikkei (as summarised by the Vietnam Digital Health Network), a number of public hospitals in China have listed de-identified clinical datasets on data exchanges, from liver disease and carotid stenting records to cancer imaging and fetal ultrasound. Buyers include AI companies, pharmaceutical and medical device firms, and research institutes. State policy names healthcare as a priority sector for unlocking the economic value of data.

Yet the most telling detail is a paradox: raw data is of little value to buyers, while cleaning, standardising and labelling it is costly for hospitals. One industry insider compared it to building rows of half-finished apartments.

Vietnam: the legal infrastructure exists, with clear limits

Vietnam’s 2025 Law on Personal Data Protection classifies health information as sensitive personal data and prohibits buying and selling personal data unless another law provides otherwise. In parallel, the Law on Data and new rules on data exchanges open the way for a market in data products, provided that personal data is not exposed.

In other words, the door is opening, but only for organisations with genuine data governance capability.

The real bottleneck: quality, not the market

Most hospital data today is generated for care delivery and billing, not for research. It is scattered across systems, largely free text, and lacks consistent coding. Without investment in data governance, we too will end up with “half-finished apartments”.

A data standardisation strategy: fix it at the source

Data standardisation is not a one-off technology project but a governance capability built over time. I see this strategy as having four layers.

• Governance – who is accountable for data: establish a data governance council bringing together clinical leadership, IT, legal and quality; give every data domain a clear owner and maintain a hospital-wide data dictionary.
• Standards – speaking one language: code diagnoses in ICD-10, moving towards ICD-11; laboratory results in LOINC, imaging in DICOM, and system-to-system exchange in HL7 FHIR. This is what allows hospital data to connect with national systems and international partners.
• Capture – standardise at the point of entry: cleaning data later always costs more than capturing it correctly. Electronic records need structured templates for priority specialties, with less free text. AI can help turn clinical notes into structured data, but qualified professionals must always review the result.
• Quality – what gets measured gets improved: track completeness, accuracy and timeliness regularly, and include data quality in the hospital’s quality indicators alongside patient safety measures.

In practice, do not try to standardise everything at once. Start with one or two priority datasets where the hospital has clinical strength and stable processes, such as cardiology or sleep medicine. Do it well, measure the results, then scale. A small but clean dataset is worth far more than a large repository no one can use.

Four principles for Vietnamese hospitals

• Partnership, not outright sale: co-develop AI products with partners, share intellectual property and benefits, with ethics committee approval for each project.
• Data stays in the hospital: favour approaches such as federated learning or controlled analysis environments, bringing the algorithm to the data rather than moving the data out.
• Transparency and consent from the start: patients should know how their data may contribute to research, and have a choice.
• Value must return to patients: benefits from data should be reinvested in quality of care and the community.

Trust is the greatest asset

Before it is an asset, health data is a responsibility: a responsibility to every patient who has entrusted the hospital with their health information.

Health data can be a valuable resource for innovation. But once patient trust is lost, no data deal can make up for it.