New study highlights opportunities, risks and the need for policy action on the use of AI.

On 15 October, the study “The Use of Data in Public Health” was presented and discussed at a meeting of the European Parliament's Committee on Public Health (SANT). Commissioned by the Committee, the study explores the use of artificial intelligence (AI) in healthcare, its opportunities and risks, and the role of health data. This marked the first time the committee had explicitly addressed the use of AI in healthcare.

Opportunities for healthcare

The study highlights the potential of AI to improve healthcare and reduce the workload of healthcare professionals. Possible applications include diagnosis, early detection of diseases and treatment tailored to patients' needs.

For example, AI can analyse medical images to help identify abnormalities at an early stage. It can also help predict the risk of developing conditions such as diabetes or cardiovascular disease. In addition, AI applications can simplify administrative tasks, including medical documentation and appointment scheduling. Analysing health data can also help identify potential healthcare needs at an early stage. This could help ensure that healthcare staff and other resources are used more effectively. AI could also support public health by helping to detect disease outbreaks and identify links between environmental factors and health.

Risks and challenges of using AI

Despite these opportunities, several challenges remain. A lack of data makes it difficult to assess how widely AI is currently used in healthcare across the EU. Technical problems, such as limited interoperability between IT systems, and concerns about the reliability of AI applications also make their wider use more difficult.

Another challenge is the quality of health data. If certain population groups are insufficiently represented in the development of AI systems, this can lead to inaccurate results and reinforce existing inequalities in healthcare. AI systems may also generate false or fabricated information that initially appears convincing. It is therefore essential that healthcare professionals check AI-generated results.

It also remains unclear who is responsible when mistakes occur in treatment involving AI. Healthcare professionals may face additional liability risks when using unreliable systems. The study also warns that relying too heavily on AI could weaken clinical skills over time.

Health Data as the foundation for AI

Access to high-quality and representative health data is essential for the development of reliable AI systems. The European Health Data Space (EHDS) is intended to make it easier to reuse such data for research, innovation and health policy decisions. For this secondary use, health data should generally be provided in anonymised form. Under certain conditions, pseudonymised data may also be used in secure processing environments.

However, the study identifies legal uncertainties regarding how the EHDS Regulation interacts with the General Data Protection Regulation (GDPR), the Data Governance Act and the AI Act. In particular, the conditions under which health data can be used for scientific research remain unclear in some cases, especially when commercial interests are involved. Differences in interpretation between Member States could also make cross-border data sharing more difficult.

Recommendations of the study

To make the most of AI's potential in healthcare, the author recommends clearer and more consistent EU-wide rules on the use of health data. Reducing legal uncertainty and administrative barriers would help facilitate data sharing and the voluntary provision of data for research.

The study also emphasises the importance of protecting sensitive patient data. Such data should remain confidential and be protected against inappropriate commercial exploitation. Patients should also be able to receive information about significant findings concerning their health that emerge from the analysis of their data.

To support the safe use of AI, the author recommends dedicated training for healthcare professionals and a central European register of verified AI applications. These measures are intended to help verify AI-generated results and identify reliable systems more easily. Ultimately, the use of AI should be guided by its benefits for patients, healthcare professionals and society as a whole.

Outlook

Following the presentation, several members of the SANT Committee expressed concerns about the reliability of AI-generated results, the quality of health data and potential risks relating to data protection and liability. They warned in particular about incorrect diagnoses and a possible loss of clinical skills, emphasising the need for human oversight.

The presentation of the study is unlikely to be the last time the Committee on Public Health discusses AI. The study highlights the complexity of the interaction between different EU legal frameworks and the regulatory challenges associated with the use of AI in healthcare.